concurring:
I join Judge Wardlaw’s excellent opinion in full and parts III, IV and V of Judge Trott’s fine concurrence. I write separate*1212ly to confess my befuddlement that we’re not unanimous.
I can’t help scratching my head at my dissenting colleagues’ dogged insistence that they’ve found a plausible way to reach the result they prefer. (Befuddlement and head-scratching often cause me to be melodramatic and use heated rhetoric, so delicate souls are cautioned to continue reading only under strict medical supervision.)
While the dissent tries to swaddle the case in all sorts of meta-considerations about the titanic struggle between the judicial and executive branches over the soul of the criminal justice process, see, e.g., Kleinfeld Dissent at 1229, 1231-1232, 1232, 1233, 1236, 1238, 1239, 1240-1241, the question presented is narrow: Does a district court have sua sponte authority to vacate a previously entered and accepted guilty plea? The dissent points to nothing that confers such authority on district courts — not in the Federal Rules of Criminal Procedure, not anywhere else.
Yet, this is precisely the kind of authority one would expect to be granted expressly, if at all. A guilty plea is a singular event in the course of a criminal prosecution, quite different from routine matters such as the setting of a trial date or a briefing schedule. A guilty plea is an adjudication of guilt, a waiver of defendant’s rights to a trial, to a jury, to stand mute and to have the prosecution prove its ease beyond a reasonable doubt. See Kercheval v. United States, 274 U.S. 220, 223, 47 S.Ct. 582, 71 L.Ed. 1009 (1927) (“A plea of guilty differs in purpose and effect from a mere admission or an extra-judicial confession; it is itself a conviction. Like a verdict of a jury it is conclusive. More is not required; the court has nothing to do but give judgment and sentence.”). Not surprisingly, the Federal Rules of Criminal Procedure devote considerable attention to the process, specifying how the plea is to be taken, when and under what circumstances it may be withdrawn and what use may be made of the various communications relating to pleas and plea bargains. Fed.R.Crim.P. 11. Conspicuously absent from these comprehensive procedures is anything authorizing the district court to vacate a properly accepted guilty plea without defendant’s consent. See United States v. Partida-Parra, 859 F.2d 629, 631-33 (9th Cir.1988).
The dissent concedes this absence of express authority, but argues that we must infer such authority in order to harmonize Rule 11 with Rule 32 and United States Sentencing Commission Guidelines Manual § 6B1. The argument holds no water. The interplay between Rule 11 and Rule 32 leads to precisely the opposite result from that urged by the dissent. Rule 32 provides that the district court may not, absent defendant’s written consent, review the presentence report before defendant has pleaded guilty (or nolo) or been convicted at trial. Fed.R.Crim.P. 32(e)(1). This means that, unless defendant agrees, the district court may not consider the contents of the report in deciding whether to accept the plea. Yet, if the district court reviews the presentence report and then vacates the plea based on information in the report, it will have done precisely what Rule 32(e)(1) prohibits — seen the presentence report of a defendant whose guilt has not yet been adjudicated. Based on this straight-forward logic, three circuits, including our own, have held that the district court may not vacate a guilty plea based on information in the presentence report. See Partida-Parra, 859 F.2d at 632-33; United States v. Cruz, 709 F.2d 111, 114-15 (1st Cir.1983), overruled in irrelevant part by United States v. Santiago Soto, 825 F.2d 616 (1st Cir.1987); United States v. Blackwell, 694 F.2d 1325, 1339 (D.C.Cir.1982). Far from supporting the dissent’s argument, the interplay between Rules 11 and 32 sinks it.
*1213In a paragraph unencumbered by citation of authority, the dissent offers three reasons we needn’t worry about Rule 32’s categorical prohibition. First, in the dissent’s view, “defendant’s protection is not so absolute,” because, if defendant chooses to withdraw his plea after the district court rejects the plea bargain, defendant will then be tried before a judge who has read the report. Kleinfeld Dissent at 1234-1235. Nice try. That the judge may preside over a criminal trial after he has read the presentence report in circumstances authorized by the rule is no argument whatsoever for ignoring the rule’s prohibition in circumstances where it is expressly applicable. That the rule treats the two situations differently calls on us to respect the distinction, not ignore it.
The dissent’s other reasons are less persuasive still: “Second, most criminal cases are tried to juries, and the jury does not see the presentence report. Third, if the judge feels that his ability to be impartial has been compromised by his knowledge of the presentence report, he must recuse himself.” Id. at 1234-1235. These so-called reasons run smack-dab into the Supreme Court’s ruling in Gregg v. United States, 394 U.S. 489, 89 S.Ct. 1134, 22 L.Ed.2d 442 (1969). The Court there said:
Rule 32 is explicit. It asserts that the “report shall not be submitted to the court ... unless the defendant has pleaded guilty or has been found guilty.” This language clearly permits the preparation of a presentence report before guilty plea or conviction but it is equally clear that the report must not, under any circumstances, be “submitted to the court” before the defendant pleads guilty or is convicted. Submission of the report to the court before that point constitutes error of the clearest kind.
Moreover, the rule must not be taken lightly. Presentence reports are documents which the rule does not make available to the defendant as a matter of right. There are no formal limitations on their contents, and they may rest on hearsay and contain information bearing no relation whatever to the crime with which the defendant is charged. To permit the ex parie introduction of this sort of material to the judge who will pronounce the defendant’s guilt or innocence or who will preside over a jury trial would seriously contravene the rule’s purpose of preventing possible prejudice from premature submission of the presentence report. No trial judge, therefore, should examine the report while the jury is deliberating since he may be called upon to give further instructions or answer inquiries from the jury, in which event there would be the possibility of prejudice which Rule 32 intended to avoid. Although the judge may have that information at his disposal in order to give a defendant a sentence suited to his particular character and potential for rehabilitation, there is no reason for him to see the document until the occasion to sentence arises, and under the rule he must not do so.
Id. at 491-92, 89 S.Ct. 1134 (footnote omitted) (emphasis added). Gregg holds quite clearly that Rule 32’s prohibition applies even when the case is tried to a jury and explains precisely why. Moreover, the Court emphasized that non-compliance with Rule 32 is “error of the clearest kind,” id. at 492, 89 S.Ct. 1134; it did not rely, as does the dissent, on some nonspecific duty of the judge to recuse himself if he no longer feels impartial.
Nor does Guidelines Manual § 6B1 help the dissent. This is so for three independent reasons, each sufficient to refute the dissent’s position. The dissent’s Guidelines Manual argument goes like this: Section 6B1 directs district courts to withhold approval of charge bargains — plea bargains where the prosecution agrees to drop certain charges in exchange for de*1214fendant’s guilty plea on other charges— unless “the remaining charges adequately reflect the seriousness of the actual offense behavior,” id., which the district judge might only know for sure after he reads the presentence report. But Rule 32 precludes the district court from seeing the report, unless defendant first pleads guilty. According to the dissent, this creates a catch-22: The judge must first accept a guilty plea in order to see the report (to satisfy Rule 32), but he then can’t reject the charge bargain based on what he reads in the report (as instructed by Guidelines Manual § 6B1) — unless he can vacate the plea.
The first fallacy of this reasoning should be apparent from the text of Rule 32. As the dissent recognizes, accepting a guilty plea is not the only way the district court can see the presentence report; the court may also see it with defendant’s written consent. Fed.R.Crim.P. 32(e)(1). Nothing precludes the court from withholding its approval of the plea, and the plea bargain, until it first sees the presentence report. Defendant has no right to have the plea accepted. Santobello v. New York, 404 U.S. 257, 262, 92 S.Ct. 495, 30 L.Ed.2d 427 (1971). If the district court insists on seeing the presentence report before accepting the plea, defendant can choose whether to agree in writing or have the court reject it outright. If defendant doesn’t agree, the court can reject the plea and the plea bargain; if he agrees, the court can then satisfy Rules 11 and 32, as well as Guidelines Manual § 6B1. The district court can also insist that, as part of its recitation supporting the plea, the government disclose the nuts and bolts of its case. Based on this information, the district court can decide whether to accept the plea at once or defer it until after the presentence report is prepared. The short of it is that there’s no conflict between these provisions, no catch 22 and no need to arrogate the power to vacate a plea that the dissent agitates for.
But even if there were a conflict between the Federal Rules of Criminal Procedure and Guidelines Manual § 6B1, the Rules of Criminal Procedure would trump. As the dissent grudgingly recognizes, section 6B1 is not a sentencing guideline; it is a free-standing policy statement. Only actual guidelines, and their applicable commentary and policy statements, are binding. See Stinson v. United States, 508 U.S. 36, 42-43, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993); Williams v. United States, 503 U.S. 193, 201, 112 S.Ct. 1112, 117 L.Ed.2d 341 (1992). Section 6B1 is in Chapter Six of the Guidelines Manual. That chapter is made up entirely of policy statements and their commentary; it contains no guidelines. While there is no case law dealing with the status of Chapter Six policy statements, no fewer than ten circuits, including our own, have held that the highly analogous policy statements in Chapter Seven (dealing with violations of probation and supervised release) were not guidelines and therefore were not binding. See United States v. Forrester, 19 F.3d 482, 484 (9th Cir.1994), superseded by statute; United States v. Sparks, 19 F.3d 1099, 1101-02 & n. 3 (6th Cir.1994); United States v. Anderson, 15 F.3d 278, 283-84 (2d Cir.1994); United States v. O’Neil, 11 F.3d 292, 301 n. 11 (1st Cir.1993); United States v. Levi, 2 F.3d 842, 845 (8th Cir.1993); United States v. Hooker, 993 F.2d 898, 900-01 (D.C.Cir.1993); United States v. Thompson, 976 F.2d 1380, 1381 (11th Cir.1992) (per curiam); United States v. Headrick, 963 F.2d 777, 782 (5th Cir.1992); United States v. Lee, 957 F.2d 770, 773 (10th Cir.1992); United States v. Blackston, 940 F.2d 877, 893 (3d Cir.1991); see also United States v. Cade, 279 F.3d 265, 270 n. 2 (5th Cir.2002) (“We have held ... that some policy statements are advisory only.”). There is no plausible argument *1215that Chapter Six policy statements are binding, and the dissent doesn’t even try to make one. Because section 6B1 is hortatory, while Rules 11 and 32 are mandatory, any conflict between the rules and 6B1 would have to be resolved in favor of the rules.
Finally, Guidelines Manual § 6B1 can’t even be used as an aid to interpreting Rule 11 for the obvious reason that it came along many years after the rule, and was drafted by a wholly different body. Interpretation, to be worthy of the name, must be a sincere effort to ascertain the meaning that the drafters of the text were seeking to capture, not an expense-paid shopping-spree through the bazaar of all conceivable meanings, rummaging for the one we like best. Our inquiry is generally limited to the document’s language though, on occasion, we do take a peek at drafting history and other such contemporaneous materials. Under no stretch of the imagination, however, can we figure out what the document means by looking at things said and written by a different entity many years later. To allow for this possibility would mean that Rule 11 would have had one meaning from 1974 to 1987, and then changed meanings abruptly upon adoption of Guidelines Manual § 6B1, as if by induction. Strict construction this ain’t. The most one could say about the interplay between Rule 11 and Policy Statement 6B1 is that, when the Sentencing Commission drafted its Manual, it may not have focused on Rule 32 as interpreted by Gregg, and thus encouraged district courts to do something that’s a bit awkward under the Rules of Criminal Procedure. Certainly, no one can claim with a straight face that Rules 11 and 32 were drafted to conform with the not-yet-in-existence Guidelines Manual § 6B1.
It is thus perfectly clear that section 6B1 cannot do the heavy lifting the dissent asks of it, first because it does not contra-diet Rules 11 and 32 and, even if it did, because it would have to yield to the logic of those provisions, which (for reasons already explained) precludes the district court from vacating a guilty plea based on a presentence report it was not authorized to see before approving the plea.
In the end, nothing is left of the dissent except strongly voiced policy preferences that judges, not prosecutors, have the final say in charge bargains. In explicating this view, the dissent unfairly impugns the prosecutor’s courage and professionalism by insinuating he was too chicken or lazy to go to trial. Kleinfeld Dissent at 1240-1241. The government’s sentencing memorandum paints the very different picture of a realistic prosecutor, aware of the strengths and weaknesses of his own case and alive to the subtle nuances of a tragic and difficult situation:
In fact, there is much about the crime that is abhorrent. The victim, Donald Ray Barker, is the epitome of the unfortunate soul in the wrong place at the wrong time. As far as anyone knows, he did nothing more on March 5, 1999, than pick up the defendant as the last fare of his life. Minutes later, he was shot three times in the back of the head.
On the other hand, there is much about the defendant that also is fairly characterized as most unfortunate. At the time that he shot Mr. Barker, he was approximately 16 1/2 years old. He lived in an environment with minimal, if any, structure. For many years before the murder, he had had no meaningful contact with his father who lives in the East Coast. Moreover, he received virtually no supervision from his mother who, it appears, was aware of his drug use, and, indeed, aware of his possession of the handgun that was used in the shooting.
*1216Neither the defendant nor any of his associates attended school regularly and most didn’t attend at all. Their days were indistinguishable, one from the next, and consisted primarily of sleeping, eating junk food, hanging out at one apartment or another, playing video games, using alcohol and drugs, engaging in sexual activities, and, perhaps most significantly, attempting to be “tough” or “cool.” Thus, for weeks before the shooting, the defendant carried in his jacket pocket, and proudly displayed to one and all, the handgun which his friend Deborah Galvan had purchased for him. To add to his “mystique,” the defendant repeatedly bragged about his criminal prowess, often claiming to have committed crimes for which he was not responsible.
Murder is a heinous crime, and this one is, in many ways, particularly troubling. Even acknowledging this, however, there are logical reasons for the Court to accept the Plea Agreement in this instance and to sentence the defendant to the stipulated term of imprisonment agreed to by the parties. First, the defendant was quite young at the time of the offense and his criminal record before this act was limited to a single violent crime, a residential burglary committed approximately 1 1/2 years earlier. Many other second degree murder convictions involve defendants who are significantly older with substantially more serious criminal records. Thus, a sentence in the middle of the range is not unreasonable under the circumstances. Second, the evidence is consistent with the crime of second degree murder. There are no eye witnesses to the actual shooting and no clear picture of what occurred inside the taxicab. Nor is there clear evidence that on the day of the shooting, the defendant was actually planning the criminal conduct.1 Instead, witnesses recalled talking to the defendant during the day about a birthday party which they urged him to attend at a Lakewood house in the neighborhood from which the defendant and his mother had recently moved away. When the defendant eventually arrived at the party that night, his accounts of the shooting were quite varied. For example, he told one friend that he had panicked when he thought the driver had locked him in the taxicab. If believed, this explanation is arguably inconsistent with the concept of premeditation. On another occasion, the defendant claimed that he shot the taxicab driver after the driver had laughed at him. Perhaps all that can be said about the various accounts of the shooting is that they show a very mixed up youth, perhaps influenced by drugs.2
Third, the defendant has demonstrated great remorse for his conduct. In the letter which he wrote to the Court, he stated in part, “from the bottom of my heart I want to apologize to Mr. Barkers family for taking apart (sic) of their life.” In the same letter, the defendant expressed hope that because he is still quite young, he may have the opportunity to turn his life around, avoid drugs, and get on the right track. Although *1217the stipulated sentence would result in the defendant spending a number of years in prison, it gives him the opportunity to someday walk out and make something of his life.
Gov’t Sentencing Mem. at 2-3 (No. CR99-5386).
In arguing the matter in open court, the Assistant United States Attorney shows, once again, what a Mensch he is:
The Plea Agreement, Your Honor, is one that was arrived at after considerable discussion between the defense and the United States. The interest of both parties was to arrive at a fair, just result. We recognize that the plea was a plea to second degree murder. We recognize that within the state system a plea to second degree murder would result in a sentence that approximates the sentence that is the subject of this Plea Agreement. Within the federal system the sentence that we have proposed in the Plea Agreement is above the guideline range for second degree murder without any adjustments. That range ■would be 97 to 121 months. This Plea Agreement calls for a sentence of 132 months. It is in the approximate middle of the range with the adjustment for vulnerable victim. The Plea Agreement takes into account the various circumstances and evidentiary matters that are present in the case. They’re part of the fabric of this case. We have to accept that and work within those parameters in attempting to structure the Plea Agreement. And we’ve done so. Both sides have.
The Defendant has accepted responsibility for the crime. He’s written the Court a moving letter, a letter that certainly appears to be a genuine statement of remorse, a genuine acknowledgment that he has committed the most serious of crimes. And that ... he wishes that he could go back and undo it.
He was a 16-and-a-half year old young man at the time of the shooting. This Plea Agreement gives him some hope. The alternative, if the Court does not accept the plea, if the matter proceeds to trial and he’s convicted, is a mandatory life sentence. The Court has no discretion. A mandatory life sentence, if the Court doesn’t accept the plea.
The Plea Agreement brings some finality to this matter. The family of the victim has had to live with this uncertainty for more than two years because it has taken that extended period of time to work through the juvenile process in the federal court system. Not only have they had to live through that period of uncertainty, but they have had to exist in a state of unknowing frustration during the bulk of that time because the United States was not able to share with them much of the information as to what was going on. They were not able to attend court sessions. They were not able to know the basics about the Defendant and the charge against the Defendant. Now they have some certainty. We have met with them. We have reviewed the facts and circumstances with them. Some of them are here in Court today. And I believe that I can speak for them in saying that they too share the government’s view that this Plea Agreement is under the circumstances fair and just in part because it brings finality to this matter.
Rep. Tr. at 4-6 (No. CR99-5386) (Apr. 17, 2001).
The dissent works hard to create the illusion that the presentence report brings to light all manner of telling details omitted by the plea agreement: “About all the stipulation in the plea bargain resolved was that Ellis got a ride in the cab, and the driver turned up dead a few minutes later.” Kleinfeld Dissent at 1238. This is *1218far from the truth. The agreement states quite clearly that defendant shot and killed the victim: “The defendant, who was the sole passenger in Mr. Barker’s taxi cab at the time of the shooting, did, with malice aforethought, unlawfully kill Donald Ray Barker by shooting him three times in the back of the head with a firearm.” Presen-tence Rep. at 4 (No. CR99-05386JET-001) (quoting Plea Agreement). Thus, the plea agreement discloses all of the observable facts that would support either first- or second-degree murder — in particular that defendant killed victim by shooting him in the head. Most of what the dissent culls from the presentence report, though inflammatory, proves nothing more than what defendant already confessed to. The facts that matter, the facts that differentiate first from second-degree murder, are those bearing on defendant’s mens rea— whether he deliberated about the killing before pulling the trigger, or acted from impulse or intoxication. On that point, the presentence report discloses nothing conclusive. See pages 1218-1219 infra.
The dissent ruminates that “[s]ometimes a lawyer offers a good deal because crucial evidence (or a crucial witness) disappears. The prosecutor has never suggested that anything like that happened in this case.” Kleinfeld Dissent at 1241. But surely a prosecutor need not lose a key witness in order to doubt whether he can prove his case to the satisfaction of a jury. The prosecutor here repeatedly alluded to the problems he would face in proving premeditation and deliberation. In his sentencing memorandum, quoted above, the prosecutor states: “There are no eye witnesses to the actual shooting and no clear picture of what occurred inside the taxicab. Nor is there clear evidence that on the day of the shooting, the defendant was actually planning the criminal conduct.” Sentencing Mem. at 2-3 (footnote omitted). In his statement in open court, also quoted above, the prosecutor says: “The Plea Agreement takes into account the various circumstances and evidentiary matters that are present in the case. They’re part of the fabric of this case. We have to accept that and work within those parameters in attempting to structure the Plea Agreement.” Rep. Tr. at 4 (Apr. 17, 2001). Later during the same hearing, the prosecutor expresses very clearly his concern about the strength of his evidence:
We have looked at a variety of cases in preparation for this hearing, and we understand the seriousness of this offense. We also are aware of certain evidentiary problems that exist. They are real. They won’t go away.
We need to view the case, the entire mosaic of this process, in light of those evidentiary concerns, and we have done so.
Rep. Tr. at 4 (No. CR99-5386JET) (Apr. 27, 2001). And, again:
We don’t disagree with the court. This is a heinous crime. We don’t disagree with the court that under some view of the evidence, first degree murder could in fact be the appropriate finding of the trier of fact. But there are many other facts which the court and the trier of fact will ultimately have to hear, which may not support that ultimate conclusion.
Id. at 10-11. And yet again:
We also have the obligation to the public to evaluate the evidence and determine whether, in the final analysis, that result is reasonably certain, reasonably uncertain, or lies somewhere between those two extremes.
We have done that. Based upon that analysis, we have concluded that the second degree disposition is appropriate.
Id. at 11-12.
The dissent brushes aside these concerns, quoting selected portions of the pre-*1219sentence report. The point of this is unclear: No one claims the prosecutor had no evidence whatever of premeditation; he must have had some, else defendant would have had no incentive to cop a plea to second-degree murder. The question is whether the government’s ease was so open-and-shut that the prosecutor must have been a coward or a laggard to accept a plea to a lesser charge, so the district court had to step in to prevent an injustice — as the dissent vehemently argues. It was not.
The dissent quotes with fanfare a taped conversation between defendant and one of his friends, where Ellis admits the killing and claims to have taken $2300, which he used to buy drugs. Kleinfeld Dissent at 1237. But the taped conversation is hardly conclusive of premeditation; defendant says nothing about planning the crime in advance. The part about robbing the victim appears to be a lie. The presentence report does not list robbery as part of the offense conduct, and the probation officer admits there is no proof of robbery: “It should be noted that there is no evidence that [defendant] actually robbed, or attempted to rob the murdered victim. Therefore, the motive for the murder is known only to the defendant.” Sentencing Rec. at 2.
Moreover, the taped conversation is only one of several inconsistent accounts defendant gave as to why he committed the murder: He panicked when he thought the taxi driver had locked him in the cab; he shot the driver because the latter laughed at him. There is also solid evidence that defendant was a daily drug user, and was picked up by the cab outside a bar. At trial, the jury would have to decide, based on all this evidence, whether defendant committed the murder with deliberation and premeditation, as a result of an impulse or because of alcohol- or drug-induced diminished capacity. See Kane v. United States, 399 F.2d 730, 736 (9th Cir.1968).
Looking at the full texture of the evidence, considering the prosecution’s heavy burden in a criminal case and taking into account the sympathy a jury might feel for defendant because of his youth and tough life, conviction of first-degree murder was far from a foregone conclusion, and voluntary man-slaughter, if the jury believed defendant acted “[u]pon a sudden quarrel or heat of passion,” 18 U.S.C. § 1112(a), was a risk. Even an outright acquittal on grounds of self-defense — if the jury believed defendant’s “the-taxi-driver-locked-me-in-the-cab” story — was at least a theoretical possibility. These are the kinds of considerations the prosecutor, who knew the strengths and weaknesses of his own case much better than the district court ever could, clearly took into account.
Far from providing a vivid example of why judges need to have a veto over pros-ecutorial decisions to accept a guilty plea to a lesser charge, as the dissent argues, this is the poster case for why judges should probably not have that power. We have here a splendid example of cooperation between a meticulous and thoughtful prosecutor — a model of what prosecutors should be — and experienced and dedicated defense counsel. Working together, they forged an agreement that avoided the risk, delay and cost of a trial, comported with the wishes of the victim’s family, and gave a troubled and misguided youth who committed this heinous crime when barely old enough to get a driver’s license, the hope of salvaging a piece of his life.
Into this carefully arranged glass menagerie burst the district judge with the force and finesse of a cannon ball. Armed only with information summarized in a few paragraphs of a report prepared by non-lawyers; having seen or heard none of the witnesses; having talked to no member of *1220the victim’s family; betraying no appreciation of the evidentiary objections and defenses that could be interposed to the prosecution’s case, the judge forced the United States to go to trial on a crime it did not believe it should or could prove. As Judge Trott correctly observes, the district judge was “defensive, inflexible, and intractable, not to mention uninformed and obdurate.” Trott Concurrence at 1229. If this is what judicial supervision over charge bargains looks like, count me out.
A final note about the dissent’s policy argument. The dissent refers repeatedly to an alleged injustice in this case, Klein-feld Dissent at 1229, 1236, 1238; it quotes the district court’s statement that “justice ... hasn’t been done in this case,” and adds, emphatically, “He was right.” Id. at 1229. But what is this terrible injustice the dissent is complaining about? Putting aside the government’s risk that it might not be able to prove first- or even second-degree murder, what exactly does the public gain by forcing an unwilling prosecutor and defendant to trial on first-degree murder?
If convicted of first-degree murder, defendant would be subject to a mandatory life sentence. 18 U.S.C. § 1111(b). The maximum sentence for second-degree murder is also life, though not mandatory, id., so it’s subject to the Sentencing Guidelines regime. Yet, even under the Sentencing Guidelines, the district judge could consider un-charged conduct, see U.S.S.G. § 1B1.3, to wit, premeditation and deliberation, which need not even be proven beyond a reasonable doubt, see United States v. Johansson, 249 F.3d 848, 853-54 (9th Cir.2001). If the district judge were to find that defendant acted with premeditation and deliberation, that would easily take the case out of the heartland of second-degree murder cases. In such circumstances, the Supreme Court has instructed that we must defer to departures by district judges, Koon v. United States, 518 U.S. 81, 98, 116 S.Ct. 2035, 135 L.Ed.2d 392 (1996); if the judge here chose to give a life sentence as an exercise of his sentencing discretion, he would most likely be upheld. Thus, the only practical effect of the charge bargain is that defendant is spared the often harsh effects of a mandatory minimum life sentence for a crime he committed as a teenager.
While reasonable minds differ on the merits of the Sentencing Guidelines, compare G. Thomas Eisele, The Sentencing Guidelines System? No. Sentencing Guidelines? Yes., 55 Fed. Probation 16, 16 (1991) (opposing Guidelines), with Andrew J. Kleinfeld, The Sentencing Guidelines Promote Truth and Justice, 55 Fed. Probation 16, 17 (1991) (“I have sometimes felt compelled by the guidelines to impose a sentence which seemed much too long, and this has been a very painful event.”), I am aware of no respectable support for mandatory mínimums, and certainly none for mandatory life sentences for crimes' committed by children. In fact, our most distinguished jurists and commentators have spoken out against the Procrustean regime of mandatory minimum sentences, and in favor of sentences that reflect the informed discretion of the trial judge. See, e.g., The Honorable Anthony M. Kennedy, Speech at the American Bar Association Annual Meeting (Aug. 9, 2003); Harris v. United States, 536 U.S. 545, 122 S.Ct. 2406, 2420, 153 L.Ed.2d 524 (2002) (Breyer, J., concurring in part and concurring in the judgment); Judicial Conference of the United States, Long Range Plan for the Federal Courts, at 60 (Dec.1995) (expressing official position of federal judiciary that “Congress should be encouraged not to prescribe mandatory minimum sentences”); Stephen J. Schulhofer, Rethinking Mandatory Minimums, 28 Wake Forest L.Rev. 199 (1993). Can we really say *1221that a grave injustice has been committed, that the public was denied its pound of flesh, because the district judge would have to exercise informed discretion before it could cast defendant into the slammer and throw away the key? My dissenting colleagues seem to think so, but I just don’t get it.
Of course, this is all beside the point because the Federal Rules of Criminal Procedure provide an easy means for district courts to do precisely what the dissent wishes them to do: They can refuse to accept a plea until and unless defendant consents in writing to the preparation of a presentence report, to be reviewed by the district court prior to acceptance of the plea. Moreover, if it really turns out to be a big problem that district courts lack authority to vacate guilty pleas, the Federal Rules of Criminal Procedure can be tweaked to give them such power. Since its adoption in 1944, Rule 11 has been amended in each of the following years: 1966, 1974, 1975, 1979, 1982, 1988, 1985, 1987, 1988, 1989, 1999 and 2002; Rule 32 has been amended even more frequently. The Sentencing Commission, too, can alter the process easily enough by changing the Chapter Six policy statements into guidelines. And, of course, Congress can always act to make the Chapter Six policy statements binding, as it did with respect to those in Chapter Seven. See Pub.L. No. 103-322, § 280001, 108 Stat. 1796, 2096 (1994); see also United States v. Plunkett, 94 F.3d 517, 519 (9th Cir.1996). The dissent’s cataclysmic predictions about a seismic shift of power from the judiciary to the executive branch are greatly overstated. This is a single mistake by a single judge in a single case, not the sacking of Rome by the Visigoths.
But there is a more fundamental issue here, one having to do with our own responsibility in interpreting and applying the law. Are we to be driven by our policy preferences to twist and bend the rules of construction in order to achieve the result that pleases us — and gives judges more power? Or are we to interpret rules and precedents according to neutral principles, following our policy preferences only where these principles do not yield a clear result? My understanding of judicial restraint is that we must do the latter, not the former, and if neutral application of the rule of law leads to a result we don’t much like, we must trust the relevant political or administrative processes to make a change. I am sorry to learn that my dissenting colleagues’ view as to the proper role of the judiciary differs so markedly from my own.
TROTT, Circuit Judge, concurring in part and dissenting in part, with whom KOZINSKI, Circuit Judge, joins in parts III, IV, and V:
I respectfully disagree with my colleagues’ understanding of this case because I read the record differently than they do, starting with the first issue of whether the district court ever accepted Ellis’s proffered plea of guilty. Judge Wardlaw, Judge Kozinski, and Judge Kle-infeld in dissent adopt as a premise for their conclusions the notion that the district court accepted Ellis’s plea on December 8, 2000. Although the record is not free from arguable ambiguity on this issue, I disagree. The record does not demonstrate — implicitly or otherwise — that the district court ever' accepted Ellis’s plea to second degree murder. Thus, my colleagues see this case as involving what it takes to vacate a plea once that plea has been accepted. I do not, for two reasons. First, on this record I am not prepared to accept the proposition that a district court judge has misrepresented to us what he did; and second, and more importantly, that same record — read in the light of the governing rules and caselaw — does provide *1222ample support for the judge’s explanation and characterization of his actions. The record shows that the district court, confronted with a Rule 11(c)(1)(A) and (C) plea agreement containing a charge bargain as well as a sentencing agreement, did exactly what the law “required.” Rule 11(c)(3)(A) says,
To the extent the plea agreement is of the type specified in Rule 11(c)(1)(A) or (C), the court may accept the agreement, reject it, or defer a decision until the court has reviewed the presentence report.
As the Supreme Court noted in Hyde, 520 U.S. 670, 675 n. 2, 117 S.Ct. 1630, 137 L.Ed.2d 935 (1997), more than two years before Ellis’s proffered plea,
Under the Sentencing Guidelines, a district court is required to defer its decision about whether to accept a type A or type C agreement until after it has reviewed the presentence report, unless the court believes that a presentence report is not required. United States Sentencing Commission, Guidelines Manual § 6Bl.l(c) (Nov.1995) (USSG).
Granted, it would have been better practice for the district court to have said when Ellis offered his plea as part of the plea bargain that it was deferring acceptance of both the plea and the plea bargain until later, but it didn’t. The court simply put the case over pending the preparation of a PSR; but deferral is not acceptance. United States v. Shaker, 279 F.3d 494 (7th Cir.2002).
I
Judge Wardlaw asserts that “[tjhere can be no dispute that the district court accepted Ellis’s guilty plea to second degree murder and deferred acceptance of the plea agreement.” No dispute? To begin with, the judge himself says that he never accepted Ellis’s plea. He says so in his Response to the Petition for a Writ of Mandamus filed on December 12, 2002, and he said it definitively on April 17, 2000, in a hearing on this very issue in the district court.
Let me begin with the judge’s explanation in his Response of what happened on December 8, 2000, to Ellis’s proffered plea, which was embedded in the Rule 11(e)(1)(A) and (C) plea agreement tendered to the court. In his Response, the judge categorically denies ever accepting the proffered plea.
The judge points out in this regard (1) that a defendant does not have a constitutional right to have his guilty plea accepted by the court, citing Lynch v. Overholser, 369 U.S. 705, 719, 82 S.Ct. 1063, 8 L.Ed.2d 211 (1962) and North Carolina v. Alford, 400 U.S. 25, 38 at n. 11, 91 S.Ct. 160, 27 L.Ed.2d 162(1970); (2) that a district court has discretion to accept or reject proffered pleas, citing United States v. Miller, 722 F.2d 562 (9th Cir.1983); and (3) that he did not find that the plea was supported by a factual basis or say that he had accepted it, ever. What he did do on December 8, 2000, he represents, was to put everything regarding the plea agreement, which included the plea, over to April 17, 2001, when he would be in a position to decide— based on consideration of the PSR— whether (1) to accept the plea, and (2) to accept the plea agreement with respect to sentencing. In footnote # 1 of his Response, he offers the PSR itself as an exhibit in support of his position, which he says he relied upon in rejecting both the plea and the sentencing agreement.
Turning to April 17, 2001, and the cold record, the session opened with a statement by Ellis’s counsel that “the defendant has previously appeared before this court and entered a plea of guilty pursuant to a plea agreement. This is the time for sentencing.” The court responded to counsel as follows: “I think I should tell you now, I’m not going to accept it.” “It” clearly *1223means both the plea and the plea agreement of which, and I repeat, the plea to a lesser offense was an integral part. We find this meaning of “it” in the court’s explanatory statement, “I think the matter should go to a jury, period. So the ball is back in the government’s court.” At this point, both the plea and the sentencing agreement were dust, and Ellis’s counsel immediately responded by saying, “Your Honor, we are prepared to proceed. The court must arraign the defendant. He has not been arraigned on the indictment and that is step number one. Step number two, I believe is the trial.” As requested by counsel, the court then arraigned Ellis on the charge in the indictment of first degree murder and set the matter for trial. Ellis’s counsel said nothing about withdrawing his plea pursuant to Rule 11(e)(4). Why? Because he understood that the court had never accepted it.
This understanding of counsel’s understanding becomes manifest ten days later on April 27, 2001, when counsel presented a new motion to the court pursuant to Rule 11(e)(4) in an attempt to resurrect for his client — with the support of the government — the benefit of the failed bargain:
MR. WESTINGHOUSE: Your Honor, the court began the proceedings on April 17th with the statement, “I think I should tell you now, I am not going to accept it. I have read the government’s sentencing memorandum and the probation’s recommendation, I can’t accept it.” The court, I suspect, is saying that it did not accept the plea, as contrasted to saying it did not accept the plea agreement.
(emphasis added).
Counsel then argued from our unpublished holding in United States v. Nelson, 2 FedAppx. 709 (9th Cir.2001), that whether the court recognized it or not, he was stuck with the plea because the court’s handling of the preliminary Rule 11 proceedings on December 8, 2000, was tantamount to accepting it. The court disagreed saying:
THE COURT: It is the court’s position, and the record reflects, that I never intended to accept the plea agreement in this case, nor did I accept the plea in this case.
As far as this court is concerned, the question of the rejection of the plea agreement is not an issue. I never accepted it.
(emphasis added).
Given the Supreme Court’s statements in Hyde in 1997(1) that a district court has discretion to accept a plea or not, and (2) that the court is required to defer a decision on a Rule 11 plea agreement until after reviewing the PSR, I do not understand how the majority can finesse this aspect of the record and claim there can be “no dispute” that the district court “impliedly” accepted Ellis’s plea. Based on this flawed understanding of what happened, the majority then embark on an irrelevant inquiry: what it takes to unravel or to vacate a plea already accepted. Unfortunately, both Judge Kleinfeld, writing first for the panel and now in dissent, and Judge Kozinski make the same mistake.
I see no grounds here to disregard what the district court said, either here or in the district court, with respect to how it treated Ellis’s proffered plea of guilty. Counsel understood what the court was doing, and so did the government. Both disagreed with the court’s decisions, but both understood: the plea as well as the plea agreement had not been accepted.
When I say that the government understood that the court had refused to accept Ellis’s plea, as contrasted with my colleagues’ view that he first accepted and then rejected it, I need look no deeper than the government’s response to Ellis’s *1224petition for a writ of mandamus. Here is what the government says:
It is the position of the United States that the district court erred in refusing to accept the defendant’s guilty plea to murder in the second degree, and that ... the district court should be directed to formally accept the defendant’s guilty plea ....
(emphasis added).
Later in its brief, the government pointedly disagrees with Ellis’s contention in his petition that the district court in fact accepted his guilty plea on December 8, 2000:
The United States disagrees with defendant’s characterization of [the Rule 11] events. The district court made it abundantly clear during the sentencing proceeding and again during the defendant’s motion to comply with Rule 11(e)(4) that it never intended to accept the guilty plea because it fundamentally disagreed with the proposed disposition.
The government goes on to argue not that the district court erred in rejecting the plea after having accepted it, but that the district court abused its discretion in refusing to accept the plea in the first place:
It is the position of the United States that the district court’s discretion to accept or reject a guilty plea once the requirements of Rule 11 have been satisfied is quite narrow, and that under the circumstances presented in this case, the district court abused its limited discretion to reject the defendant’s guilty plea.
The government then cites a whole host of cases limiting a district court’s discretion in this area. In this regard, the government directs our attention primarily to United States v. Miller, 722 F.2d 562 (9th Cir.1983) and the following language:
... Charging decisions are genei-ally with the prosecutor’s exclusive domain .... Prosecutors — representatives of the executive branch of the government — are not mere servants of the judiciary. The tradition of prosecutorial independence is recognized both by case law ... and the Federal Rules of Criminal Procedure, see Fed.R.Crim.P. 48(a).
Although courts are free to accept or reject individual charge bargains, they should avoid creating broad rules that limit traditional prosecutorial independence. Generally, courts should be wary of second-guessing prosecutorial choices.
Id. 722 F.2d at 565-66 (footnotes and internal citations omitted).
Why do we now say the district court did something it says it did not do? The court never said, as Ellis admits, “I accept the plea,” or, “I find the defendant guilty,” or even, “I find as required by Rule 11(b)(3) that there is a factual basis for the plea.” Ellis’s attempt to brush this lacuna off as a simple failure by the court “to utter the formula” is unpersuasive. The court said it had not accepted the plea, period. When the court said so, neither the prosecution nor the defense argued that it had already accepted the plea. The defense said, okay, set the matter for trial. It was not until 10 days later that the defense claimed the court had accepted the plea without realizing it, a claim now disputed by the government. I’m afraid we have locked into an issue that is not presented by the record, and we seem intent on addressing that issue whether it is part of this case, or not. At no point did the court indicate it was vacating a plea already accepted.
The cases and the Federal Rules of Criminal Procedure make it clear that there are two basic and separate parts to the conversion of a plea of guilty into a judgment of guilty. The first step is for the defendant to enter his plea in compliance with the requirements of Rule 11. *1225The second step is for the court to accept, or to reject, the plea. Confirmation of the need for the district court independently to accept the defendant’s plea of guilty appears over and over in the Rule itself. Rule 11(b) is entitled “Considering and Accepting a Guilty or Nolo Contendere Plea.” Rule 11(b)(1) says, “Before the court accepts a plea of guilty.... ” Rule 11(b)(2) repeats the same “before accepting” language; and Rule 11(b)(3) says, “Before entering judgment on a guilty plea, the court must determine that there is a factual basis for the plea.” Moreover, the Supreme Court said in Santobello v. New York, 404 U.S. 257, 262, 92 S.Ct. 495, 30 L.Ed.2d 427 (1971),
There is, of course, no absolute right to have a guilty plea accepted. Lynch v. Overholser, 369 U.S. 705, 82 S.Ct. 1063, 8 L.Ed.2d 211 ... (1962); Fed. Rule Crim. Proc. 11. A court may reject a plea in exercise of sound discretion.
An example of the application of this principle appears in Maxwell v. United States, 368 F.2d 735 (9th Cir.1966) where a defendant indicted for first degree murder claimed a right to plead guilty to murder in the second degree. After discussing the matter with the defendant, the court rejected this attempt, and the jury returned a verdict of guilty of murder in the first degree. Maxwell’s sentence was to imprisonment for life. He appealed, claiming he had a right to plead guilty to a lesser charge authorized by the statute under which he was charged. We disagreed with Maxwell’s argument, and what we said, although probably in dicta, supports the district court’s conduct in this case:
In our opinion the district court did not abuse its discretion in refusing to accept the plea of guilty. This is especially true since Maxwell did not offer to plead guilty to the crime charged — first degree murder — but to the lesser charge of second degree murder. Even if the Government had consented to accept a plea of guilty to a lesser charge, the court ivould still not have been obliged to accept it.
Id. at 739 (emphasis added).
Footnote # 3 in Maxwell is equally on point in connection with Ellis’s mistaken assertion that the court found a factual basis for his plea:
... We do not hold, however, that because Maxwell was unable to recollect the transaction and so could not personally vouch for his guilt, the trial court was obliged to reject his plea of guilty to second degree murder. The offer to plead guilty came at the close of the Government’s case, when a factual basis for a plea of guilty to at least second degree murder had been established, or so the trial court might have found. Under these circumstances, the court, if it saw fit, could have accepted Maxwell’s plea.
In this connection it may be noted that Rule 11, Federal Rules of Criminal Procedure has recently been amended to add, among other things, a sentence at the end reading: “The court shall not enter a judgment upon a plea of guilty unless it is satisfied that there is a factual basis for the plea.”
Id. at n. 3.
This is precisely what the court identifies as a critical step he had not yet taken: find that there was a factual basis for the plea.
In addition, we have the Supreme Court’s statement in Hyde that a court has discretion when it comes to whether or not to accept a plea. Id. 520 U.S. at 674, 117 S.Ct. 1630.
The Seventh Circuit has approved the deferral by a court of the acceptance of a plea of guilty offered during a thorough Rule 11 proceeding until after the PSR is reviewed: “In United States v. Ellison, *1226835 F.2d 687, 689-908 nn., 4-5 (7th Cir.1987), ... we explicitly encouraged district courts to defer acceptance of a guilty plea pending review of the PSR.” United States v. Shaker, 279 F.3d at 497:
We view a guilty plea as a process involving both the defendant and the district court, and culminating in the court’s acceptance of the plea. The ten- or of the Supreme Court’s opinion in United States v. Hyde, 520 U.S. 670, 117 S.Ct. 1630, 137 L.Ed.2d 935 (1997), supports our understanding that acceptance is a crucial part of this process. See id. at 674, 520 U.S. 670, 117 S.Ct. 1630, 137 L.Ed.2d 935 (Rule 11 lists “steps a district court must take ‘[b]efore accepting a plea of guilty,’ and without which it ‘shall not accept a plea of guilty.’ Based on this language, we conclude that once the court has taken these steps, it may, in its discretion, accept a defendant’s guilty plea.”).
Id.
Finally, Sentencing Guideline § 6Bl.l(e) (2000) states that a district court “shall defer” a decision to accept or reject a plea agreement pursuant to Rule 11(c)(1)(A) or (C) “until there has been an opportunity to consider the presentence report.”
II
The real mischief that might emerge from this case stems from Judge Kozin-ski’s view that a district court somehow is precluded from reading the PSR before deciding whether to accept a defendant’s Rule 11 plea. Judge Kozinski claims that
Rule 32 provides that the district court may not, absent defendant’s written consent, review the presentence report before defendant has pleaded guilty (or nolo) or been convicted at trial. Fed. R.Crim.P. 32(e)(1). This means that, unless defendant agrees, the district court may not consider the contents of the report in deciding whether to accept the plea.
With all respect to Judge Kozinski, his view of Rule 32(e) as it applies to this Rule 11 case, is mistaken. First, Ellis did plead guilty. Therefore, it was perfectly appropriate for the court to have read the PSR. Second, the plea was part of a Rule 11(c)(1)(A) and (C) plea agreement which Ellis asked the court to approve. Rule 11(c)(3) says that “[t]o the extent that the plea agreement is of the type specified in Rule 11(c)(1)(A) or (C), the court may accept the agreement, reject it, or defer a decision until the court has reviewed the 'presentence report. ” (emphasis added).
What sense does it make to give a district court discretion to accept or to reject an offered plea of guilty but at the same time to tell the court, you cannot have access to the best source for rendering such a decision — the PSR — until after you make the decision to accept it. As Judge Kleinfeld recognizes, Rule 11(c)(1)(A) agreements are always charge bargains. Formal consent here is inapposite; consent is implied in the submission of the agreement to the court for approval and from the controlling rules of the Rule 11 process. Judge Kozinski’s fírst-the-decision, then-the-evidence analysis is unconvincing, and thus, I conclude he is wrong-in his Rule 32 claim. Once a defendant has done his part in a Rule 11 proceeding and entered a plea of guilty, it is pointless to tell a district court judge it is improper to look at the PSR, especially when looking at the PSR is called for under Rule 11(c)(3), and especially in a case like this where the plea is an unseverable part of a plea agreement which requires the court’s approval. I repeat what I said earlier: Rule 11(c)(3)(A) says, “[t]o the extent the plea agreement is of the type specified in Rule 11(c)(1)(A) or (C), the court may accept the agreement, reject it, or defer a *1227decision until after the court has reviewed the presentence report.” (emphasis added).
Ill
In my view, the real question presented by Ellis’s petition boils down to whether the district court abused its discretion in refusing to accept his plea, either (1) independently, or (2) pursuant to Rule 11(e)(4), which permits a defendant to persist in his plea in the face of a rejected plea agreement of which the plea is part. The law on this issue is clear: a district court is empowered with discretion to reject charge bargains that are part of Rule 11 plea bargains, but because this power “implicates executive discretion with respect to charging decisions,” a court must be hesitant “before second-guessing prosecutorial choices.” United States v. Robertson, 45 F.3d 1423, 1438 (10th Cir.1995). As we said in United States v. Miller,
Generally, courts should be wary of second-guessing prosecutorial choices. Courts do not know which charges are best initiated at which time, United States v. Lovasco, 431 U.S. 783, 793-94, 97 S.Ct. 2044, 2050-51, 52 L.Ed.2d 752 (1977), which allocation of prosecutorial resources is most efficient, United States v. Ammidown, 497 F.2d 615-621 (D.C.Cir.1973), or the relative strengths of various cases and charges. See Vorenberg, Decent Restraint of Prosecutorial Power, 94 Harv.L.Rev. 1521, 1547 (1981).
722 F.2d at 565.
Accordingly, we established a protocol a district court must follow in making a decision to reject a charge bargain:
... Although the general rule governing plea bargaining grants courts broad discretion, Fed.R.Crim.P. 11, the specific rule governing prosecutorial charging decisions gives courts only a limited supervisory power over such decisions. Fed.R.Crim.P. 48(a). That rule requires courts to grant prosecutors leave to dismiss charges unless dismissal is “clearly contrary to manifest public interest.” Rinaldi v. United States, 434 U.S. 22, 30, 98 S.Ct. 81, 85, 54 L.Ed.2d 207 (1977) (per curiam). Many of the policies underlying Rule 48 are equally applicable to judicial consideration of charge bargains. Although Rule 48 antedates Rule 11, we should not refuse its guidance when we interpret the Federal Rules. To assure that judicial discretion is exercised with due regard for prosecutorial independence, we hold that courts must review individually every charge bargain placed before them. They must set forth, on the record, both the prosecutor’s reasons for framing the bargain as he did and the court’s justification for rejecting the bargain. See United States v. Ammidown, 497 F.2d 615, 623 (D.C.Cir.1973) (trial judge must state reasons on the record for rejecting plea bargain).
By requiring that rejection of a charge bargain be accompanied by a more complete trial court record, we uphold the separation of powers in two ways. First, we guarantee that the trial court is aware of and gives due deference to the prosecutorial choices reflected in a particular plea bargain. Second, we facilitate appellate review of rejected plea bargains. If the prosecutorial decisions reflected in specific charge bargains deserve broad deference, the discretion of the trial court to reject these bargains is fairly narrow. By requiring a more complete statement of the trial court’s basis for rejecting a bargain, we make it possible to apply more careful appellate review.
722 F.2d at 565-66.
IV
So now we get to the end game. Did the court fully comply with Miller in refus*1228ing, once the plea agreement was out of the way, to accept Ellis’s plea to second degree murder, a plea urged by the prosecutor, and a plea with no sentencing strings attached? After reviewing the record and rereading Miller, I answer this question in the negative. All the court said was, “justice in my opinion hasn’t been done in this case.” The court made no attempt to distinguish between the sentence agreement and the plea itself. Although the record, as arrayed by Judge Kleinfeld, might provide general support for the court’s conclusion, and although the government admitted to the court that a first degree conviction might well be appropriate, the court failed to explain itself in a manner sufficient for us to review the decision in any meaningful way. The court’s unelaborated “opinion” per se is not enough to resolve an issue that has its roots in the constitutional doctrine of separation of powers. I note here that both the government and Ellis’s attorney pleaded with the court to make a record on this issue, but it refused to do so. Consequently, as we did in Miller, ordinarily I would remand this case to the district court from whence it came with orders to comply with Miller and to set forth on the record the reasons for its decision.
V
Ordinarily, as I said, I would not hesitate to return a matter such as this to the same judge from whence it came. However, I have read and reread Judge Tanner’s strong-arm responses to Ellis’s counsel’s entreaties during the post-plea agreement rejection proceedings and to the government’s eloquent and well-grounded request to accept Ellis’s proffered straight up plea to second-degree murder and to make an adequate record, and I find the court’s responses — even when read in a light most favorable and respectful to the court — to be defensive, inflexible, and intractable, not to mention uninformed and obdurate. It is nothing short of ironic that Judge Tanner now relies on our opinion in Miller to justify his handling of the post-plea agreement rejection proceedings. Had he followed Miller in the first place, we probably would not be in this unfortunate hole, a hole that has unnecessarily delayed justice for more than two years. Therefore, recognizing that such a move will erase most of what we have been debating, I concur in the majority’s decision to remand this case to a new judge for a fresh start. Justice and the rule of law so require.
KLEINFELD, Circuit Judge, with whom Circuit Judge GOULD joins, dissenting:
I agree with the majority’s holding that the judge in this case accepted Ellis’s plea. Where we differ is that I would hold that after reading the presentence report, a judge has the authority to strike the plea because he has rejected the plea bargain of which the plea to a lesser offense was an essential part. The Federal Rules of Criminal Procedure are silent on whether the judge can do so. The Sentencing Guidelines suggest that he can and must. For these reasons, and those elaborated below, I respectfully dissent.
Today’s decision marks a substantial shift from judicial to executive control over much of the criminal law process. Traditionally, plea bargains have been subject to judicial approval. Plea bargains include sentence bargains and charge bargains. In sentence bargains, the prosecution and defense agree both on a guilty plea to the initial charges and on a sentence or sentence recommendation. In charge bargains, the parties may agree on a guilty plea to new, lesser charges rather than the original charges. Both have customarily been subject to judicial approval. Today we have cut the judiciary out of the *1229charge-bargaining process. In so doing, we have put ourselves in conflict with the Fifth Circuit decision of United States v. Foy,1 and we have created a novel judicial gloss that requires judges to disregard one of the policy statements in the Sentencing Guidelines concerning the acceptance of a plea agreement.2
Plea bargaining has gone on for a long time and follows a well-established pattern. Often, pursuant to a plea agreement, the defendant pleads to lesser charges, and the judge orders a presen-tence report. Then, after the court pronounces judgment at the end of sentencing, the prosecutor moves in open court pursuant to Rule 48(a) to dismiss the original charges, and the judge grants the motion. The prosecutor may, but rarely does, file a Rule 48(a) motion earlier in the process to dismiss the original charges. More often, the prosecutor keeps the original indictment available in case the defendant backs out of the plea agreement.
Ellis was indicted for first-degree murder, but he entered into a charge bargain to plead guilty to second-degree murder. The judge accepted the plea and ordered a presentence report. But after reviewing the report, he determined that justice would not be served in this case if he accepted the charge bargain. The majority claims that allowing the judge to vacate Ellis’s plea after the court has accepted the plea, but before the plea agreement had been accepted, usurps the prosecutor’s power to make the charging decision. This is incorrect. The prosecutor retains charging power, should he choose to exercise it. If he believes he cannot prove the greater charge, or that it is excessive for any reason, he can file a Rule 48(a) motion to dismiss the first-degree murder indictment at any time. The district court’s discretion to deny the motion to dismiss would be narrow (though extant).3 The case would then proceed as though the grand jury had never indicted Ellis for first-degree murder.
Here the prosecutor seeks the power without the responsibility. He never stepped up to the plate and moved to dismiss the first-degree murder indictment. Had the prosecutor chosen to take responsibility for prosecuting this killing (which appears in the presentence report to be a premeditated, cold-blooded thrill killing) as nothing more than second-degree murder, he could have done so, by filing a motion under Rule 48(a) to dismiss the first-degree murder indictment. That way he would have avoided the judge’s broad discretion to reject plea bargains and shifted the case over to the judge’s very constricted discretion to reject Rule 48(a) dismissals. This case would have gone the way the prosecutor wanted in district court, and the way the majority opinion today makes it go, without gutting judicial control over charge bargaining. But the prosecutor did not make the Rule 48(a) motion. Because the grand jury indicted Ellis for first-degree murder and the charge was not dismissed, the district court had a responsibility to the public to assure that the grand jury’s determination was not bargained away in a manner that disserved the public interest.4
Rule 11(c)(1)(A) speaks to a charge bargain, part of the plea agreement made in this case. It says, “If the defendant *1230pleads guilty or nolo contendere to either a charged offense or a lesser or related offense, the plea agreement may specify that an attorney for the government (A) will not bring or will move to dismiss other charges.” Rule 11(c)(5) speaks to what happens if the court rejects a plea of guilty to lesser or related charges. It says that if the court rejects a plea agreement containing provisions of the type specified in Rule 11(c)(1)(A) or (C), the court must do the following: inform the parties that the court rejects the plea agreement; advise the defendant personally that the court is not required to follow the plea agreement and give the defendant an opportunity to withdraw the plea; and advise the defendant personally that if the plea is not withdrawn the court may dispose of the case less favorably toward the defendant than the plea agreement contemplated.
Several things are plain from these provisions. First, plea agreements include both charge bargains and sentence bargains. Rule 11(c)(1)(A) agreements are always charge bargains, because they are agreements that the government will not bring or will move to dismiss other charges, and the introductory language refers to plea agreements for pleading guilty to a lesser or related offense. Second, the court has the authority to reject such a charge bargain. Subsections (c)(1) and (c)(5) contemplate a situation where a judge accepts a plea and then subsequently rejects the charge bargain. But, under the majority’s reading of Rule 11, a judge’s authority to reject a plea agreement containing a charge bargain loses all practical meaning because he is unable to vacate a plea. Third, the defendant may withdraw his plea upon a court’s rejection of a charge bargain.
The majority infers from the language stating that the court has to “give the defendant an opportunity to withdraw his plea” that his plea stands if he doesn’t withdraw it. That is not a literal interpretation of the rule. It is an inference treating the express provision for the defendant to withdraw his plea as an implication that the court cannot strike bis plea. The majority cannot construe Rule 11 literally to reach its result because the rule does not say one way or the other what to do with the plea if the defendant doesn’t withdraw it, but the judge disapproves of the charge bargain. Neither the vigorous attacks in the majority nor the melodramatic rhetoric of Judge Kozinski’s concurring opinion can substitute for words in the rule. The words are not there. Our task is to fill the gap left by silence in Rule 11.
Though from a purely grammatical point of view, the inference the majority makes from Rule ll’s silence is permissible, it is not compelling. In the overall context of Rule 11, Rule 32, and § 6B1.2 of the Sentencing Guidelines, the majority’s inference of a negative pregnant is not reasonable. Reasonableness here is not a matter of wrenching the rule into a judicial policy preference. It is a matter of reading Rule 11 so that it is coherent and consistent with other rules and so that our interpretation is consistent with the better-reasoned views of our sister circuits.
Both the majority and we in the minority are interpreting a silence in the rules. There is language in the majority opinion suggesting that the rules command its result, and overheated rhetoric in Judge Ko-zinski’s concurring opinion suggesting that no one but a fool could think otherwise. But in fact, the rules do not say whether a judge can strike a plea to a lesser charge made as part of a charge bargain that the judge, in his discretion, chooses to reject. Rule 11 says only that the defendant can withdraw his plea. We are disagreeing, with more heat than makes any sense to me, about whether there is a negative pregnant in that provision.
*1231The Rule 11(d) limitations speak to when the defendant can withdraw a plea, not to when the court can strike it. Rule 11(c)(5) states what the defendant’s rights are — that the court must “give the defendant an opportunity to withdraw the plea” — not what the court’s authority is if the defendant does not exercise that right. Although this speaks to the defendant’s rights, the majority infers that it restricts the court’s authority to set aside the plea if the defendant does not withdraw the plea. Rather than draw this inference from silence in Rule 11(c)(5), it would be more reasonable to draw an inference as to judicial authority from the subsection that speaks to judicial authority, Rule 11(e).
Rule 11(e) controls “Finality of a Guilty or Nolo Contendere Plea.” Unlike Rule 11(c)(5), it speaks to both a defendant’s rights and the court’s authority. It says that a plea may be set aside “only on direct appeal or collateral attack,” but conditions this restriction to the time “[a]fter the court imposes sentence.” The condition in 11(e), “[ajfter the court imposes sentence,” implies that the court’s authority to set aside the plea before the court imposes sentence is not so restricted. My inference, like the majority’s, is from a silence, and like the majority’s, reads a negative pregnant into the silence, but we differ in that I am looking at the subsection of the rule speaking to judicial authority and the majority is not. My inference, as is explained below, provides coherence and consistency with Rule 32 and the Sentencing Guidelines, requiring the district court to obtain a presentence report after taking the plea and to review the presen-tence report before accepting the plea agreement and settling upon a sentence.
Those of us who have done criminal defense have long been familiar with courts striking pleas where charge bargains were rejected. Though initially it may be puzzling that the long-established practice is not reflected in the authorities, that is probably explained by the tremendous change in the legitimacy accorded to the plea bargaining process in the last thirty years. When many of us began practice, plea bargaining was done in the shadows and created awkward moments when the judge asked the defendant, “Have you been promised anything in return for your plea?” The Task Force on the Administration of Justice in the 1960’s criticized, not plea bargaining itself, but the shadowy way in which it was done:
New practices in the system of criminal justice created a greater sense of unease and suspicion than the negotiated plea of guilty.... The system usually operates in an informal invisible manner. There is ordinarily no formal recognition that the defendant has been offered an inducement to plead guilty. Although the participants and frequently the judge know that negotiation is taking place, the prosecutor and defendant must ordinarily go through a courtroom ritual in which they deny that the guilty plea is the result of any threat or promise. As a result, there is no judicial review of the propriety of the bargain — no check on the amount of pressure put on the defendant to plead guilty. The judge, the public, and sometimes the defendant himself cannot know for certain who got what from whom in exchange for what. The process comes to look less rational, more subject to chance factors, to undue pressures, and sometimes to the hint of corruption.5
The commission recommended bringing the plea bargaining process out into the *1232open and assuring among other things that “the prosecutor did not agree to an inadequate sentence for a serious offender.”6
Gradually, Rule 11 and comparable state rale reforms implemented the recommendations of the commission, so what we have is a relatively new (as law goes) judicial recognition and supervision of plea bargaining, superimposed on a very old criminal-law system. What we run across in this case is an instance where the traditional and well-established practices of the courts operate in the context of silence in the rules, and our court has now decided to turn away from its prior precedent and from our sister circuits to interpret that silence in a way that reduces judicial authority to supervise the plea bargaining-process.
The majority misreads Rule 11 as denying a judge the power to revisit acceptance of a plea to a lesser charge when, after reading the presentence report, the judge determines that the charge bargain contained in the plea agreement does not reflect the seriousness of the charged behavior. Rule 11 does not say that the district court is stuck with the plea to a lesser charge if the defendant elects. The majority infers this proposition because the rale says that the defendant is entitled to withdraw his plea and is silent about whether the judge can strike it if the defendant elects not to withdraw it. In so interpreting the silence in Rule 11, the majority impermissibly isolates Rule 11 from the context of the judge’s role in the plea bargain process as governed by Rule 32 and the Sentencing Guidelines. Rule 11 is not the only rule; it is but part of a coherent set of rules. That’s why it’s called “11.” And Rule 32 and the Guidelines fill in the silence.
The majority holds that if the defendant chooses not to withdraw his plea to a lesser offense, the judge cannot reject his charge bargain.7 Once the judge rejected the charge bargain, the majority claims that “it became Ellis’s choice” whether to take a sentence limited to what the court had available for second-degree murder, or to withdraw his plea. First, this is an ipse dixit based on an inference from silence. The majority can cite to no prohibition on the judge’s authority to vacate a plea. Second, this “choice” is inconsistent with § 6B1.2(a) of the Sentencing Guidelines which directs a judge to reject a charge bargain unless “the remaining charges adequately reflect the seriousness of the actual offense behavior.”8 The majority’s misreading of Rule 11 enables the parties to force the charge bargain down the judge’s throat against the court’s judgment. This misreading conflicts with the Sentencing Guidelines, and it makes no sense. Why have a judge independently review the charge bargain to assure that it adequately reflects the seriousness of the crime, if the judge is just a rubber stamp *1233without the power to do anything if the deal is too lenient?
Rule 32 requires, generally, that a probation officer conduct a presentence investigation and submit a presentence report to the court.9 A report cannot be submitted to the court until the defendant has pleaded guilty or been found guilty, unless the defendant consents in writing.10 Thus the judge cannot read the presentence report and learn about the details until after the plea has been accepted. Under § 6B1.2 of the Sentencing Guidelines, a court “may accept the agreement[involving a charge bargain under Rule 11(c)(1)(A)] if the court determines, for reasons stated on the record, that the remaining charges adequately reflect the seriousness of the actual offense behavior.”11 The Commentary on that provision states: “[W]hen the dismissal of charges or agreement not to pursue potential charges is contingent on acceptance of a plea agreement, the court’s authority to adjudicate guilt and impose sentence is implicated, and the court is to determine whether or not dismissal of charges will undermine the sentencing guidelines.”12
The Guidelines also recognize that the presentence report has to be obtained before the court can accept a plea agreement, which includes both charge and sentence bargains. Sentencing Guideline § 6Bl.l(c) states that a court “shall defer” a decision to accept or reject a plea agreement pursuant to Rule 11(c)(1)(A) or 11(c)(1)(C) “until there has been an opportunity to consider the presentence report.” 13 The Commentary to § 6B1.1 explains that “[s]ince a presentence report normally will be prepared, the court must defer acceptance of the plea agreement until the court has had an opportunity to consider the presentence report.”14
It is logically absurd, and unjust, to bind a court to a plea to lesser charges under Rule 11, in the face of the § 6Bl.l(c) policy that the court’s acceptance of the charge bargain not take place when the plea is accepted, and the § 6B1.2(a) requirement that the charge bargain be accepted only if the remaining charges adequately reflect the seriousness of the actual offense behavior. The public interest is put at risk, because the judge cannot, when the plea is taken, know enough to evaluate the charge bargain. Adequate knowledge requires the presentence report, which is why the Guidelines provide that the court has to *1234defer decision until after reading the pre-sentence report.
Judge Kozinski would have us disregard these provisions of the Sentencing Guidelines because, as policy statements, they are arguably not binding. We need not decide whether § 6B1.2 is binding to use it as part of the context in which to construe Rule 11. Of course if it is binding, the majority’s construction is impermissible.
Judge Kozinski also appears to suggest, amidst all the heated rhetoric, that the district judge somehow violated Rule 32 by looking at the presentenee report, and would have us lard up change of plea proceedings by requiring a defendant’s consent to examine the presentence report where the court might reject a charge bargain after reading it. There is no problem needing a solution. The district judge complied with the Rule 32(e)l requirement that the court not obtain the presentence report, in the absence of consent, “until the defendant has pleaded guilty or nolo contendere.” Ellis pleaded guilty to second-degree murder, and then the court obtained the presentence report and decided to set aside the plea because second-degree murder charge did not comprehend the vileness of the crime. Judge Kozinski, on policy grounds (drawn in part from cases preceding the Sentencing Guidelines and current policies on judicial evaluation of plea bargains) would have the judge remain blind to the facts and accept a pig in a poke until the plea could not be withdrawn. That argument runs contrary to 18 U.S.C. § 3553(a)(2)(A), which requires a judge to consider whether the sentence imposed reflects the seriousness of the offense, and § 6B1.2 of the Sentencing Guidelines, which directs a judge to defer acceptance of a plea agreement until after reviewing the presentence report.
Judge Kozinski also suggests that if the district judge has doubts about a charge bargain, the judge should defer acceptance of the plea until reading the presentence report and condition going forward with the plea bargain on the defendant’s consent to a presentence report in the absence of a plea. That may work when the judge knows enough to have doubts and the defendant’s bargaining position is weak enough so that he consents. But ordinarily, at the time of the plea, the judge does not know anything but what the prosecutor and defendant disclose — not enough to know at that time whether to disapprove. That is why the Guidelines provide that the court has to defer decision until after reading the presentence report.
Judge Kozinski bolsters his argument with the contention that striking a plea after reading the presentence report, and setting the case for trial, is inconsistent with the defendant’s Rule 32 protection against the judge seeing the presentence report in a case that will go to trial. The argument errs for three reasons. First, the defendant’s protection is not so absolute as Judge Kozinski suggests. If the court rejects an 11(c)(1)(C) agreement and the defendant withdraws his plea, the defendant ordinarily goes to trial before the judge who has read the presentence report, so there is no reason why the same result should not follow when the court rejects a charge bargain and strikes a guilty plea. Second, most criminal cases are tried to juries, and the jury does not see the presentence report. Third, if the judge feels that his ability to be impartial has been compromised by his knowledge of the presentence report, he must recuse himself.
We held in United States v. Cordova-Perez15 that a judge has the authority that the majority takes away from him today. *1235In that case, we held that a judge who had accepted a guilty plea to a lesser offense, and then, after reading the presentence report, decided that the plea agreement was too lenient, properly struck the plea and set the case for trial.16 Obviously Cordova-Perez controls the case at bar unless it is overruled. The majority does not purport to overrule Cordova-Perez itself, but says that Cordova-Perez is “no longer good law” after United States v. Hyde,17 and suggests that under Hyde, Rule 11 is the only rule that matters to the outcome.
Hyde cannot carry the weight the majority gives it. True, it says we were mistaken in holding that acceptance of the plea and acceptance of the plea bargain were “inextricably bound up together.” But the Supreme Court said this in the context of rejecting our holding that a defendant could withdraw a plea “for any reason or for no reason” if the plea was entered pursuant to a plea bargain that the judge had not yet accepted.18 Hyde held that we had mistakenly construed Rule 11 to enable a defendant to withdraw his plea “simply on a lark” up until the point when the court accepted the plea agreement.19 Our mistake, the Court explained, was in not reconciling Rule 11 with the “fair and just reason” requirement in what is now Rule 11(d)(4).20 Today’s majority makes a similar mistake in reading Rule 11 in isolation without reconciling it with Rule 32 and the Guidelines.
The better reading of Hyde is that it leaves intact the basic Cordova-Perez principle.21 This principle is that, while under Hyde the defendant is bound by his accepted plea, the court is not bound by it when it is part of a plea agreement that the court has not yet accepted (and cannot accept until the presentence report is submitted to the court). The district court in Hyde had not yet decided whether to approve the plea bargain when the defendant moved to withdraw his plea. Had the judge rejected the plea agreement, then, as the Court explained, the defendant would have been able to “back out of his promised performance (the guilty plea), just as a binding contractual duty may be extinguished by the nonoccurrence of a condition subsequent.”22 Guilty pleas both before and after the “fair and just reason” was added to what is now Rule 11(d) “were [ ] already subject to this sort of condition subsequent.”23
The majority takes Hyde’s holding that a defendant’s plea is not so bound up with *1236the plea agreement that he can withdraw his plea “on a lark” and runs away with it. The majority treats Hyde as though it held that the plea and plea agreement are entirely independent, so that a prosecutor and defendant can hold the court to a defendant’s plea to a lesser charge despite the court’s disapproval of the plea bargain. That does not follow from Hyde. In fact, in Hyde, the Court noted that our Cordovar-Perez statement that “ ‘[t]he plea agreement and the [guilty] plea are inextricably bound up together’ ... on its own, is not necessarily incorrect.”24
Hyde does not overrule Cordova-Perez, and nothing in Hyde suggests that it does. The majority’s decision today is inconsistent with Cordova-Perez. Instead of expressly overruling it and giving a justification for doing so, the majority misreads Hyde to say that the Supreme Court had done what it carefully avoided doing. The consequence is to take a decision that strengthened the district court’s hand in plea bargaining, Hyde, and misread it as a decision that eliminated judicial control over one important kind of plea bargaining. The judge’s proper and heretofore well-understood role in plea bargaining is as a neutral stranger to the agreement, not a party to it, whose task it is to make sure that the agreement adequately reflects the public interest in sufficient punishment for the crimes committed.25
Ellis’s plea to second-degree murder was part of the plea agreement, an exchange of a guilty plea to a lesser charge for dismissal of the greater charge in the indictment. As characterized in Hyde, the agreement was subject to a condition subsequent, judicial approval, which condition failed to materialize, terminating the agreement and extinguishing all duties thereunder. Thus Ellis was no longer bound to plead guilty to anything. The prosecution was no longer bound to recommend anything. And the court certainly should not have been bound. The court had before it an unresolved criminal case ready to be set for trial (which is what the court quite properly did).
Because the court could not approve the plea agreement until obtaining the presen-tence report and evaluating whether a second-degree murder conviction would “adequately reflect the seriousness of the actual offense behavior,”26 yet the plea to a lesser offense is held to bind the court, the majority effectively deprives the court of the power to disapprove charge bargains. A judge’s authority to reject a plea agreement with a charge bargain for a plea to a lesser offense is emptied of all meaning if the defendant and prosecutor can force the lesser charge on the court.
The facts of this case highlight the injustice that may be perpetrated on the public because the majority denies a judge the ability to vacate the plea after determining that the plea agreement is improper.27 In this case, the judge accepted Ellis’s plea to second-degree murder, ordered a presen-tence report, and set a date for sentencing. The facts in the presentence report, which *1237the judge properly relied upon in his evaluation of the plea agreement,28 tell a story that the judge could not have known from the plea agreement.29
Marciano Ellis asked a high-school Mend to come over to Ellis’s mother’s apartment to “help him plan the robbery and killing of a cab driver.” His Mend came over, and Ellis showed off his new gun (which his older girlfriend had bought for him illegally). Although his Mend wanted nothing to do with it, Ellis went ahead with the murder.
On the evening of March 5, 1999, Ellis called a taxicab from a payphone in front of a doughnut shop, and was picked up a few minutes later by Donald Ray Barker, a 46-year-old cabdriver. Ellis directed the taxi driver to take him to Fort Lewis, a nearby military reservation, and then shot Barker three times in the back of the head.
After the murder, Ellis bragged about it to the Mend he’d tried to enlist. He showed the Mend the cab driver’s taxi license card on a chain, and he told the friend that he stole $2,300 in cash, which he later used to buy drugs. The Mend called the FBI, and agreed to wear a wire. The friend met with Ellis and asked him about the murder of the cab driver, and Ellis said “Yeah, that was me.” Ellis went on, in the recording, to explain in detail just how he had done it. A search of Ellis’s home turned up a receipt for the murder weapon and a brass shell casing matching the ballistics of the bullets buried in the cabdriver’s skull. The FBI later *1238bought the gun for $1,000 from another hoodlum to whom it had been given, and it matched the brass in Ellis’s home and the bullet in the taxi driver’s skull.
With these facts before him, the judge plainly had reason to exercise his discretion and prevent this case from being “pleaded down.” About all the stipulation in the plea bargain resolved was that Ellis got a ride in the cab, and the driver turned up dead a few minutes later. The judge said, with respect to the second-degree murder charge, that “justice in my opinion hasn’t been done in this case.”30 He was right.31
Judges, not prosecutors, are supposed to decide whether to accept the terms of the plea agreement. The court cannot be a mere rubber stamp for contractual agreements between prosecutors and defense attorneys. Sometimes a neutral judge has to say “No!” to an injustice that would be perpetrated on the public by a plea bargain, as this judge did.
In United States v. Foy,32 as here, the district court accepted a plea to a lesser charge, and then, after getting a presen-tence report, vacated the plea because of inadequacy of the plea agreement. On appeal, the Fifth Circuit held that it was not reversible error for a judge who had already accepted the defendant’s guilty plea to reject the plea agreement after he reviewed the presentence report.33 Like the judge here, the judge in Foy determined that “the circumstances in the pre-sentence investigation justified one heck of a lot more sentence than the maximum [under the plea agreement].”34 The Fifth Circuit held that because § 6B1.1 of the Sentencing Guidelines requires a judge to defer acceptance of the plea agreement until after he reviews the presentenee report, a plea accepted pursuant to a Rule 11 plea agreement is contingent upon that review.35
Similarly, United States v. Carrigan36 was a petition to the Tenth Circuit Court *1239of Appeals for a writ of mandamus commanding a district court to accept a charge bargain. The government had charged a corporation and its employee with submitting false claims to the government, and the prosecutor agreed to settle a related civil action against the corporation, take a guilty plea from the corporation, and dismiss the charges against the employee. The district judge rejected the deal because “every individual ... connected with this fraud will walk free.”37 Though Car-rigan differs from this case in that the corporation had not yet entered a plea (the corporation refused to enter its plea unless the charges against its employee were dismissed), that difference was not material to the ratio decidendi of the issue of judicial authority to reject a charge bargain. As in the case at bar, the prosecution argued that executive discretion over charging required the court to accept the charge bargain. Following Fifth Circuit authority, the Tenth Circuit reasoned that limiting the judge’s discretion to reject a charge bargain improperly limited judicial control over sentencing, and that Rule 48(a), governing dismissals, did not apply where the prosecution did not move for dismissal under Rule 48(a).38 Moreover, since “the authority to accept or reject a plea bargain agreement under[Rule 11(c)] is confined almost entirely to the trial court’s discretion,”39 a litigant’s claimed right to force the charge bargain on the court could not justify a writ of mandamus because it was not “ ‘clear and indisputable.’ ”40
Judge Kozinski cites to another panel opinion of ours, and to cases from two other circuits, in his critique of this dissent. The citations are inapposite. The other panel opinion of ours, United States v. Partidas-Parra,41 holds that the rules “cannot be read to authorize the court to vacate a plea on the basis of a discrepancy in the parties’ understanding of the plea agreement,”42 where the government (not the court, as here) had moved to vacate the plea because the prosecutor who made the plea bargain should not have done so on the terms the defendant agreed to. That case does not speak to the court’s authority to vacate a plea to ensure the adequacy of the plea agreement.
The First Circuit case, United States v. Cruz43 is distinguishable in at least two respects. First, the court notes that “the judge unqualifiedly accepted the plea bargain,”44 while here, the judge made it quite clear that he accepted the plea, because it was voluntary, the defendant was satisfactorily counseled, there was a basis *1240in fact, and so forth, but that he was not yet prepared to accept the plea bargain. Second, Cruz came down in 1983, four years before the Sentencing Guidelines came into effect and established the policy that the court “shall” defer a decision on whether to accept a plea agreement “until there has been an opportunity to consider the presentence report.”45
Scrounging even further for authority, Judge Kozinski cites another pre-Guide-lines case from the District of Columbia Circuit, United States v. Blackwell.46 In Blackwell, the court held that the district judge had erred by warning the codefen-dant (Blackwell’s girlfriend who pleaded guilty and was not an appellant) that she could be prosecuted on the charges dismissed pursuant to her plea agreement if her testimony at Blackwell’s trial established that she was guilty of them.47 But, unlike the case here, the judge had accepted the plea bargain as well as the plea,48 and, in any event, he did not rescind acceptance of either. Therefore, the question not only was distinguishable from the one in this case because the Guidelines did not yet exist, but also because the court did not strike the plea.
Thus, none of the cases cited in Judge Kozinski’s concurring opinion contradict a reading of Rule 11 that allows a judge who has accepted a guilty plea, but not yet accepted the plea agreement, to vacate the plea upon rejection of the plea agreement. Only the Fifth Circuit, in Foy, addressed the same issue of how to interpret Rule ll’s silence when a judge is presented with material in the presentence report that persuades him that he cannot accept the plea agreement after he has accepted the plea. And Foy resolved the issue as I think we should.
Because the judge did not err in vacating Ellis’s plea, mandamus is unjustifiable in the case at bar. The district court acted in compliance with § 6Bl.l(c) and followed our holding in Cordovar-Perez. Not only was the judge’s decision in Ellis’s case not the sort of plain and irremediable error required for mandamus,49 it was not error at all. Because today’s opinion is novel, contrary to our previous precedent in Cor-dovar-Perez, unsupported by the literal application of the rules and guidelines, and surprising in light of the established customs of plea bargaining and judicial acceptance, the case cannot possibly satisfy the Bauman factor that it be “clearly erroneous.” 50 It also does not satisfy the Bau-man factor that the harm be not correctable on an appeal from a final judgment.51
As for assignment to a different district judge, the law of recusal is plain that a judge need not recuse himself because of what he has learned from judicial proceedings in the same case,52 and no ground for *1241bias is claimed except for what the district judge learned from this very case. That the district judge developed a plainly stated opinion that a second-degree murder charge did not “adequately reflect the seriousness of the actual offense behavior” is merely to say that he performed a judgment in accord with his duty, not that he is biased. Rule 11(d)(2)(A) expressly contemplates a case where a judge accepts a plea, reviews the presentence report, rejects the plea agreement, and the defendant withdraws his plea and goes to trial before the very same judge. There is no justification for taking this case away from the district judge to whom it was assigned.
There is a reason courts need to retain control over plea bargains. A court is not a mere market for guilty pleas and sentences, to be dispensed solely according to the market judgments of defense counsel and prosecutors. “Plea bargaining is not a private contractual arrangement unaffected by the public interest. The public interest does not favor fictional minimization of crimes actually committed for purposes of sentencing. Under the guidelines, a strong judicial hand is necessary to protect the public interest.”53 The individual participants in a governmental process sometimes have individual interests and incentives that differ from the purported missions of them agencies. It is a myth that prosecutors and police are always overzealous. They may be overzealous, underzealous, or just-right zealous. For a good lawyer, a jury trial, though extremely exciting and punctuated with moments of great pleasure and satisfaction, is also tremendously stressful, time consuming, physically and mentally exhausting, and risky. One does not always feel up to a gunfight at the O.K. Corral, which is what a jury trial may feel like to a lawyer. Life is much easier, though duller, for prosecutors and defense attorneys, if cases can be bargained to a judgment without the need for a jury trial.
We simply cannot know why the prosecutor was so lenient in this case. Sometimes a lawyer offers a good deal because crucial evidence (or a crucial witness) disappears. The prosecutor has never suggested that anything like that happened in this case, but a prosecutor may be reluctant to disclose this information out of concern that a lenient plea bargain will have to become even more lenient. The likelihood of such things happening is why judges are restrained in rejecting plea agreements. But the reasons for prosecu-torial leniency are not always this good. Sometimes prosecutors take lenient pleas to avoid the work and stress of trial. Though second-degree murder with a long sentence may look adequate to a prosecutor despite his having a plain and simple first-degree murder case, the public interest requires that judges retain final authority to approve or disapprove.
By depriving the judge of power to reject charge bargains that are inadequate in light of the seriousness of the crime committed, today’s opinion deregulates the market between prosecutors and defendants. We cannot and should not allow this kind of market, which has the potential to set the value of innocent human life too low.
. To support a first degree murder charge the United States must prove the defendant acted with premeditation.
. On the other hand, there is some limited evidence that defendant had talked about robbing taxi-cab drivers before this crime was committed. For example, on one earlier occasion, defendant supposedly invited a friend to join him in a taxicab robbery, and promised to call him later. However, defendant never called and nothing apparently came from this conversation.
. U.S. v. Foy, 28 F.3d 464 (5th Cir.1994).
. U.S. Sentencing Guideline Manual § 6B 1.2(a) (2000).
. A district court may deny a Rule 48(a) motion to dismiss an indictment only when "the motion was clearly contrary to manifest public interest.” United States v. Garcia-Valenzuela, 232 F.3d 1003, 1008 (9th Cir.2000) (citation omitted).
.See U.S. Sentencing Guidelines Manual § 6B 1.2(a) (2000).
. Task Force on Administration of Justice, The President's Commission on Law Enforcement and Administration of Justice, Task Force Report: The Courts 9-10 (1967) (internal citations omitted).
. Id. at 13.
. The majority states that the dissent “flatly mischaracterizes” the majority's holding because the judge may still "dispose of the case less favorably toward the defendant than the plea agreement contemplated” under Rule 11(c)(5). Maj. Op. at 1208 n. 13. I wish that were so. If a defendant pleads to lesser charges with a lesser maximum but open sentencing, and the judge disapproves of the charge bargain after reading the presentence report, the judge is limited, under the majority opinion, to the inadequate conviction, and, if the maximum is too low, the inadequate sentence. Under the majority opinion, if the prosecutor and defense agreed to plead first-degree murder down to a misdemeanor such as careless use of firearms, and, knowing nothing but the stipulation of facts in tire plea bargain, the court accepted the plea, there would be nothing the court could do about it when the presentence report revealed what had occurred.
.U.S. Sentencing Guidelines Manual § 6B 1.2(a) (2000).
. Fed.R.Crim.P. 32(c)(1)(A).
. Fed.R.Crim.P. 32(e)(1).
. U.S. Sentencing Guidelines Manual § 6B 1.2(a) (2000).
. The full paragraph of Commentary for § 6B1.2 lays out the heretofore well understood balancing of executive charging and judicial control over sentencing:
The court may accept an agreement calling for dismissal of charges or an agreement not to pursue potential charges if the remaining charges reflect the seriousness of the actual offense behavior. This requirement does not authorize judges to intrude upon the charging discretion of the prosecutor. If the government's motion to dismiss charges or statement that potential charges will not be pursued is not contingent on the disposition of the remaining charges, the judge should defer to the government's position except under extraordinary circumstances. Rule 48(a), Fed. R.Crim.P. However, when the dismissal of charges or agreement not to pursue potential charges is contingent on acceptance of a plea agreement, the court's authority to adjudicate guilt and impose sentence is implicated, and the court is to determine whether or not dismissal of charges will undermine the sentencing guidelines.
. U.S. Sentencing Guidelines Manual § 6B 1.1(c) (2000) (emphasis added).
. U.S. Sentencing Guidelines Manual § 6B1.1, cmt. ¶2 (2000) (emphasis added).
. U.S. v. Cordova-Perez, 65 F.3d 1552 (9thf Cir.1995).
. Id. at 1557.
. United States v. Hyde, 520 U.S. 670, 117 S.Ct. 1630, 137 L.Ed.2d 935 (1997).
. Hyde, 520 U.S. at 677, 117 S.Ct. 1630.
. Id. at 676, 117 S.Ct. 1630.
. Id. at 673-74, 117 S.Ct. 1630.
. LaFave sets out, consistently with my view, what was, before today’s decision, the general understanding of Hyde's rejection of the “bound up together” concept:
Although such reasoning was rejected by the Supreme Court when used to support a quite different contention, namely, that the defendant may withdraw his plea as a matter of right any time prior to the judge's decision as to the plea agreement, see the Hyde case, ... that does not put into question the balance of the Cordova-Perez analysis in the text immediately following.
5 Wayne R. LaFave, et al., Criminal Procedure § 21.4(g) n. 201 (2d ed.1999).
The "text immediately following” is our holding that the "deferment of the decision whether to accept the plea agreement carried with it postponement of the decision whether to accept the plea ... even though the court explicitly stated it accepted Cordova-Perez's plea.” Cordova-Perez, 65 F.3d at 1556.
. Hyde, 520 U.S. at 677-78, 117 S.Ct. 1630.
. Id. at 679, 117 S.Ct. 1630.
. Id. at 677, 117 S.Ct. 1630 (alteration in original).
. See U.S. v. Lewis, 979 F.2d 1372, 1375 (9th Cir.1992) ("Although both the government and the defendant are expected to comply with the terms of the plea agreement, the court is not a party to the agreement and may reject it.").
. U.S. Sentencing Guidelines Manual § 6B 1.2(a) (2000).
.The majority points out that these “facts” have not been proved. Of course not. A trial is needed to find out whether the facts are this egregious. That is what the district court correctly provided for. Under Sentencing Guidelines § 6B1.2, the facts as described in the presentence report, not just those stipulated to in the plea agreement, are supposed to furnish the predicate for judicial approval or disapproval.
. See Rule 11(c)(3)(A); U.S. Sentencing Guidelines Manual § 6B 1.1(c) (2002).
. The plea agreement included this sparse record of facts:
8. Statement of Facts. The parties stipulate and agree that the following facts, which defendant admits, provide the factual basis for defendant's plea of guilty to the Superseding Information:
a. Defendant was born on September 9, 1982. On March 5, 1999, the defendant resided with his mother at The Palisades Apartment Complex, 14701 South C Street, #61, Tacoma, (Spanaway) Washington, located directly behind the shopping center where a Winchell's Donut Shop and the Strap Tavern are located.
b. On Friday, March 5, 1999, at approximately 7:45 p.m., the defendant called Tacoma Yellow Cab from a pay telephone outside the Winchell's Donut Shop located at 15012 Pacific Avenue South, Spanaway, Washington, requesting a pick-up at the Strap Tavern located nearby. Donald Ray Barker, a 46-year old male who was driving Tacoma Yellow Cab number 60 that night, was dispatched to the Strap Tavern where he picked up the defendant shortly after 8:00 p.m.
c. At approximately 8:20 p.m., a passerby was driving southbound on North Gate Road on Fort Lewis when he saw cab 60 pulled off to the right side of the road with the lights on, in a shallow ditch. The passerby stopped, noticed that the engine was running, and found the driver, Donald Ray Barker, laying [sic] on the front seat. Mr. Barker had been shot in the back of the head. Emergency medical assistance was summoned and Mr. Barker was transported to the Madigan Army Hospital where he was treated for multiple gunshot wounds to the back of his head.
d. An autopsy was performed by the Pierce County Medical Examiner who determined that Mr. Barker was killed by three gunshot wounds to the back of his head.
e. The location of Mr. Barker's cab, when it was found by the passerby to the side of the road with its lights on and its engine running, was approximately one-half mile inside the perimeter of Fort Lewis, which is within the special maritime and territorial jurisdiction of the United States, as defined by Title 18, United States Code, Section 7(3).
f. The defendant, who was the sole passenger in Mr. Barker's taxi cab at the time of the shooting, did, with malice aforethought, unlawfully kill Donald Ray Barker by shooting him three times in the back of the head with a firearm.
g. The events described above occurred within the Western District of Washington.
. Judge Trott relies on United States v. Miller, 722 F.2d 562, 565-66 (9th Cir.1983), for the proposition that the district judge did not adequately set forth its justifications for rejecting the charge bargain. Miller is a 1983 case, pre-Guidelines. In light of the subsequent changes in the Rules and Sentencing Guidelines, it is questionable authority. In the context of this case where the judge’s remarks must be read in light of the presen-tence report to which the prosecutor and defendant, as well as the court, were responding, the court's justification is sufficiently set out.
. Judge Kozinski is evidently impressed with what he calls Ellis’s "tragic and difficult situation,” e.g., that, as the prosecutor characterized it, Ellis's "days were indistinguishable, one from the next, and consisted primarily of sleeping, eating junk food, hanging out at one apartment or another, playing video games, using alcohol and drugs, engaging in sexual activities, and, perhaps most significantly, attempting to be 'tough’ or 'cool.' ” The district judge had the discretion to decide whether this "tragic and difficult situation” justified lenience, and was well within his discretion in deciding that it did not.
. United States v. Foy, 28 F.3d 464, 468 (5th Cir.1994) (remanding for reconsideration of the plea agreement on other grounds).
. Id. at 471.
. Id. at 468.
. Id. at 471 ("We conclude that section 6B 1.1(c) makes a district court's acceptance of a guilty plea contingent upon the court's review of the PSR [presentence report].”).
. United States v. Carrigan, 778 F.2d 1454 (10th Cir.1985); see also United States v. Martin, 287 F.3d 609, 623 (7th Cir.2002) (holding that a district court’s rejection at sentencing of a plea agreement involving a charge bargain did not usurp executive power, explaining that had the U.S. Attorney been so inclined, he could have effectuated the same result the plea agreement sought by moving to dismiss the other counts with prejudice under Rule 48, which the district court would have *1239had to grant, absent a specific finding that the dismissal would be manifestly against the public interest).
. Carrigan, 778 F.2d at 1458.
. See United States v. Bean, 564 F.2d 700 (5th Cir.1977), cited by Carrigan, 778 F.2d at 1463-64.
. Carrigan, 778 F.2d at 1467.
. Id. (quoting Allied Chem. Corp. v. Daiflon, Inc., 449 U.S. 33, 36, 101 S.Ct. 188, 66 L.Ed.2d 193 (1980) (per curiam)).
. United States v. Partida-Parra, 859 F.2d 629 (9th Cir.1988).
. Id. at 631.
. United States v. Cruz, 709 F.2d 111 (1st Cir.1983). Cruz says that under the facts of that case jeopardy attached when the plea was accepted. The majority and concurring opinions do not argue that double jeopardy bars Ellis’s trial for first degree murder, no doubt because the Cruz position could no longer be maintained in the face of the Supreme Court decision a year later in Ohio v. Johnson, 467 U.S. 493, 501, 104 S.Ct. 2536, 81 L.Ed.2d 425 (1984). In cases that end by a guilty plea, jeopardy does not attach where no judgment of conviction has been entered. 5 Wayne R. LaFave, et al., Criminal Procedure § 25.1(d) (2d ed.1999).
. Cruz, 709 at 112.
. U.S. Sentencing Guidelines Manual § 6B 1.1(c) (2000).
. United States v. Blackwell, 694 F.2d 1325 (D.C.Cir.1982).
. Id. at 1337.
. Id. at 1338.
. "[I]t is clear that only exceptional circumstances amounting to a judicial usurpation of power will justify the invocation of this extraordinary remedy.” Will v. United States, 389 U.S. 90, 95, 88 S.Ct. 269, 19 L.Ed.2d 305 (1967) (citations and internal quotation marks omitted).
. Bauman v. United States District Court, 557 F.2d 650, 654 (9th Cir.1977)
. Id.
. " 'Impartiality is not gullibility. Disinterestedness does not mean child-like innocence. If the judge did not form judgments of the actors in these court-house dramas called trials, he could never render decisions.' ” Liteky v. United States, 510 U.S. 540, 551, 114 S.Ct. 1147, 127 L.Ed.2d 474 (1994) (quoting In re J.P. Lindhan, Inc., 138 F.2d 650, 654 (2d Cir.1943)).
. United States v. Fine, 975 F.2d 596, 604 (9th Cir.1992) (en banc).