dissenting:
Reading the majority opinion, one could hardly discern that this appeal is about a case of egregious prosecutorial misconduct — threatening a witness with execu*1151tion if he testifies favorably to the defendant. Instead, the majority engages in a remarkable and creative act of hypothetical appellate fact-finding and, on the basis of the testimony it believes might have been given, concludes that the witness’s testimony and that of others, whatever that testimony might have been, would be of no consequence. Given that the threatened witness was the only person other than the petitioner who had direct knowledge of the critical aspects of the events of the evening in question and the only person other than the petitioner who was capable of testifying to them, it is not surprising that the prosecutor failed to anticipate that any court, let alone an en banc appellate court, would determine without his testimony having been adduced that whatever he said would be of no significance, and that regardless of what his testimony might be an evidentiary hearing could not have produced any material facts. Had the prosecutor realized that our court could have reached such an unprecedented result, he most surely would not have felt compelled to engage in so extreme an act of prosecutorial misconduct. He would have let the witness testify instead of threatening to have him executed.
I.
In this case, Smith’s co-defendant, in order to avoid murder charges and obtain
a light sentence, turned state’s evidence and agreed to swear that Smith, an accomplice to the robbery, was the person who killed the victim. The facts, however, showed that the co-defendant, himself, was more likely the killer.1 Nevertheless, the co-defendant’s plan succeeded and he was out of prison within three years of his plea agreement, while Smith, who, according to the evidence, most likely assaulted no one, pled no contest to felony murder in order to avoid trial on a capital murder charge. Following his plea, Smith received a life sentence.
The plea arrangement in this case was not unique. The participant in criminal activity who confesses first and fingers his fellow wrongdoer frequently receives a far lighter sentence, even though he may have been the primary culprit and committed the more egregious act. As a result, his hapless associate who played a lesser role often receives the harsher punishment— sometimes even a death sentence.2 See, e.g., Coleman v. McCormick, 874 F.2d 1280 (9th Cir.1989) (en banc).
What is unique about this case is that when Edmonds, the probable killer, attempted, seven years later, to recant and to testify that Smith was not the murderer, the state prosecutor succeeded in preventing an evidentiary hearing by threatening to seek the death penalty against Edmonds if he testified favorably to Smith. *1152At the time of the threat, Smith had filed his federal habeas petition asserting that he had not committed a violent act against anyone and that, under Oregon law, he was actually innocent of felony murder. Accordingly, he asserted, he was entitled under Schlup v. Delo, 513 U.S. 298, 115 S.Ct. 851, 130 L.Ed.2d 808 (1995), to bypass his state procedural default and present his claim that his conviction of felony murder was obtained in violation of his constitutional rights.3 In seeking to bypass the procedural default, Smith simply sought to have the federal court examine those constitutional claims on the merits.
Not surprisingly, in response to the state’s threat to have him executed, Ed-monds, the likely killer, invoked his Fifth Amendment privilege and refused to testify. As a result, the district court found that a Schlup evidentiary hearing would be a “pointless exercise” and concluded that, without the admitted killer’s4 exculpatory testimony, Smith could not establish actual innocence in a manner sufficient to overcome his procedural default and to allow him to proceed on the merits of his constitutional claims.
In short, the effect of the state’s prose-cutorial misconduct5 — threatening to seek the death penalty against Smith’s key witness if he testified favorably to Smith— was to deny Smith the opportunity to develop the facts necessary to establish his Schlup claim. Because of the state’s misconduct, Smith could not overcome the procedural barrier to his receiving a hearing on the merits of the alleged constitutional violations. Put differently, the state’s threat to have Edmonds executed eliminated the crucial — indeed, the only— evidentiary process that would have allowed Smith to pass through the Schlup gateway and, as a result, present his constitutional claims in federal court.
II.
The primary question before us is thus simply: what does a federal court do when the state, by an act of gross prosecutorial misconduct, interferes with a habeas petitioner’s ability to make a showing of his actual innocence in a manner that would allow the federal court to excuse his state procedural default? It is not a question we have faced before, nor a situation we are likely to face often. The majority *1153seeks to avoid answering the question by asserting that even had an evidentiary hearing been held, Smith could not have proved his innocence. Aside from the fact that this is somewhat like saying, “We don’t need a trial because we know that the defendant is guilty,” there is a fundamental flaw in the majority’s approach. My colleagues do not evaluate Smith’s claim on the basis of what evidence could have been adduced at an evidentiary hearing. Instead, they base their decision on what Edmonds said years ago, shortly after the killing, when he was trying to shift the blame to Smith so that he would not be subject to a lengthy term of imprisonment himself. The majority’s mode of analysis obviously ignores the basic point that Ed-monds has filed affidavits retracting his earlier testimony and is now, in offering to exonerate Smith, purporting to tell the truth for the first time. Whether he did so shortly after the killing or whether his current story is true can be determined only when he is placed under oath and is subject to cross-examination by both sides — in short, at an evidentiary hearing.
Let me emphasize that the record is clear. As the district court found, “The Linn County prosecutor’s office warned Edmond’s [sic] counsel that the district attorney would seek the death penalty if Edmonds testified ...” (emphásis added). There can be no doubt about the prosecution’s goal in making that threat. Critically, the prosecution did not threaten to seek the death penalty against Ed-monds if he turned out to be the true killer. Rather, it threatened Edmonds with execution only if he testified at Smith’s evidentiary hearing. That the state dropped its threat against Edmonds as soon as he declined to testify — that it never investigated his statements or sought to determine whether his recantation might be true — makes crystal-clear the state’s purpose in threatening Ed-monds with the death penalty: to prevent Edmonds from testifying at Smith’s hearing and exculpating Smith of felony murder.
The question we now face, as I have explained above, is how should we consider Smith’s Schlup claim given that the state’s misconduct has rendered an evidentiary hearing a “pointless exercise.” The majority’s solution is to imagine a hypothetical evidentiary hearing at which Edmonds would testify, having been granted use immunity by the state, “that he, rather than Smith, killed Mr. Konzelman.” Maj. op. at 1137. The majority then determines that Edmonds’s hypothetical testimony that Smith was not the true killer would not be credible and that, even if it were credible, it would not benefit Smith because Smith would still have to prove that he himself never held the crowbar nor had a reasonable ground to believe that Edmonds had it. Relying almost entirely on Edmonds’s prior self-serving and ambiguous statements made when he was trying to shift the blame to Smith to avoid capital murder charges himself, the majority holds that Smith could not meet his burden of proving Schlup actual innocence because at least one reasonable juror would still find Smith guilty of felony murder.
The principal fallacy in the majority’s approach is apparent on its face. It disregards the historic and constitutionally guaranteed system under which American courts determine the facts. A full and fair hearing at which Edmonds was permitted to testify is the only way that Smith could have established the factual elements required to make out his defense to felony murder, and thus his Schlup actual innocence claim. By conducting an evidentiary hearing we could have determined whether Edmonds’s initial story or his current affidavits were true. In the absence of that hearing, it is impossible to determine what Edmonds might have said or what Smith might have established. Had a hearing *1154been held, Smith’s lawyer, who presumably understood the elements of the affirmative defense that Smith was required to establish, could have explored those elements fully in questioning Edmonds. Smith, presumably, would have also testified at the hearing. His testimony, like Edmonds’s, would likely have provided significant details regarding the events that took place in the garage, the timing of the two men’s entrance into the Konzelman home, and the subsequent events inside the home, all of which could have established definitively whether Smith had held the crowbar or, if he had not, whether he had reason to believe that Edmonds was armed with it. Both men would have been subject to rigorous cross-examination. Perhaps their testimony on direct would not have held up. Perhaps it would not have been sufficient to allow Smith to meet the five-part test necessary to establish an affirmative defense to felony murder under Oregon law. At least a fair and full hearing would have been held and a judge could have ruled on the two defendants’ credibility and assessed the facts. We will never know what the results of a Schlup hearing would have been, however, because the prosecution’s blatant misconduct rendered such a hearing impossible.
The majority’s unprecedented attempt to construct a hypothetical evidentiary hearing as a substitute for an actual hearing — the one the prosecution prevented from being held — fails to fill the gaping holes in the record created by the state’s unconscionable interference. In the majority’s mythical hearing, Edmonds states one thing only: that he, not Smith, was the actual killer. In that construct, Edmonds does not provide details about the night in question. He does not say whether Smith ever held the crowbar nor does he provide facts that might indicate whether Smith had reason to believe that Edmonds had the crowbar. In sum, the majority’s construct assumes that Edmonds’s affidavits — which do not address Smith’s affirmative defense but state only that Smith was not the killer- — comprise the full scope of what Edmonds’s testimony would have been had an evidentiary hearing taken place.
Such an assumption is entirely unreasonable. The purpose of Edmonds’s affidavits was not to establish all the elements of Smith’s affirmative defense. Edmonds sought only to inform the interested parties, including the state, that his statements inculpating Smith seven years earlier, made in exchange for lenient treatment, were false. The affidavits do not purport to set forth any of the factual circumstances surrounding the killing or to describe fully the events of the evening in question.6 Accordingly, it makes little sense to assume, as the majority’s construct does, that, had an evidentiary hearing been held, Edmonds would have merely repeated the statements in his affidavits, and nothing more. Rather, in the absence of specific knowledge as to what Edmonds and Smith would have said had the state’s misconduct not rendered an evidentiary hearing futile, an approach far more consistent with our constitutional principles of due process is for the court to presume that the evidence precluded by the state’s misconduct would have materially benefitted Smith with respect to his Schlup claim, and, accordingly, to allow him to present his constitutional claims on the merits.
*1155Such a presumption is hardly unusual in civil litigation.7 “Generally, a trier of fact may draw an adverse inference from the destruction of evidence relevant to a case.” Akiona v. United States, 938 F.2d 158, 161 (9th Cir.1991); see also 2 Kenneth S. Broun et al., McCormick on Evidence § 265 (6th ed.2006). As we stated in Akio-na,
The adverse inference is based on two rationales, one evidentiary and one not. The evidentiary rationale is nothing more than the common sense observation that a party who has notice that a document is relevant to litigation and who proceeds to destroy the document is more likely to have been threatened by the document than is a party in the same position who does not destroy the document....
The other rationale for the inference has to do with its prophylactic and punitive effects. Allowing the trier of fact to draw the inference presumably deters parties from destroying relevant evidence before it can be introduced at trial.
Akiona, 938 F.2d at 161 (quoting Nation-Wide Check Coup., Inc. v. Forest Hills Distribs., Inc., 692 F.2d 214, 218 (1st Cir.1982)). Both rationales apply equally strongly to the state’s willful intimidation of a critical witness with the result that he refuses to testify at a habeas petitioner’s Schlup evidentiary hearing and thus renders the hearing, in the words of the district court, a “pointless exercise.” First, the state’s interference with Edmonds’s willingness to testify raises the obvious inference that the state was “likely to have been threatened by,” Akiona, 938 F.2d at 161 (quoting Nation-Wide, 692 F.2d at 218), Edmonds’s testimony — why else would the prosecution have gone to such unconscionable lengths to prevent him from testifying? Put differently, if, as the majority suggests, Edmonds’s testimony would have been of no assistance to Smith in a Schlup hearing because it would not establish Smith’s innocence of felony murder, why was the state so desperate to prevent his testimony? Why would it have engaged in egregious prosecutorial misconduct by threatening to have Edmonds executed if he testified? Thus the inference is rational and fairly drawn. Equally important, if we presume that Edmonds’s testimony would have been favorable to Smith, our action would have a most beneficial effect. It would deter the state from unlawfully interfering with the introduction of relevant evidence in future habeas and other proceedings.
The deterrence rationale takes on added importance when considered in light of the purposes of the Schlup doctrine. The Schlup gateway is an outgrowth of the Supreme Court’s recognition that “the conviction of one innocent of the crime” for which he was convicted represents a “fundamental miscarriage of justice.” McCleskey v. Zant, 499 U.S. 467, 494, 111 S.Ct. 1454, 113 L.Ed.2d 517 (1991); see also Schlup, 513 U.S. at 314-15, 115 S.Ct. 851. Meeting the Schlup standard does not by itself prove that such a miscarriage has taken place,8 but doing so raises a significant possibility that the petitioner may not have been guilty of the offense for which he was convicted. This possibility warrants an exception to the procedural default rule to allow the federal court to *1156make absolutely certain that no miscarriage of justice occurs. See Schlup, 513 U.S. at 316, 115 S.Ct. 851. When the state, through an act of flagrant prosecuto-rial misconduct, precludes the introduction of evidence that, had it been admitted, would undermine the court’s confidence in the outcome of a criminal proceeding, it substantially increases the possibility that a fundamental miscarriage of justice — the conviction of an innocent individual — has occurred. Even more so if the state by that egregious misconduct precludes the holding of the evidentiary hearing itself. This is certainly behavior that federal courts, charged with upholding constitutional protections, should seek to deter. For this reason, it is especially appropriate in the context of a habeas petitioner’s Schlup actual innocence claim to apply the standard presumption that the precluded evidence was unfavorable to the party that caused its preclusion and, here, that the precluded evidentiary hearing would have resulted in an outcome favorable to the other party.
III.
I will not here engage in the extraordinary process of appellate factfinding that the majority does when it concludes that a reasonable juror would more likely than not find that Smith had failed to meet at least one element of his affirmative defense. Such factual determinations are a matter for the district court in the first instance. They are to be made after a thorough review of facts adduced at a ha-beas petitioner’s evidentiary hearing. In this case, however, such a hearing was precluded by the state’s deliberate and egregious misconduct, and the district court was unable to make any findings. As a result, the majority does so on its own, relying almost entirely on Edmonds’s past statements — which are in serious doubt in light of his recantations — as well as on hypothetical testimony at a hearing that never took place. Such creative prestidigitation is no substitute for an actual evidentiary hearing. Moreover, the majority makes credibility findings as to Ed-monds’s hypothetical testimony at its hypothetical hearing. Needless to say, such credibility findings are a question for the district court and cannot be made on appellate review, especially in the absence of any testimony on which to base those findings.9 The majority’s analyses of elements (b) and (e) of the affirmative de*1157fense — whether Smith killed Konzelman and whether he was himself armed with the crowbar — as well as its analysis of element (d) — whether Smith had reason to believe Edmonds was armed — are based almost entirely on hypothetical, non-existent evidence. Because such determinations are wholly invalid in the absence of an evidentiary hearing, I will not discuss each of them individually here.
To the extent that the majority’s analysis of element (d) relies on evidence other than Edmonds’s past statements,10 such evidence is trivial and fails to prove that Smith had reason to believe that Edmonds was armed with the crowbar. First, there is Mrs. Konzelman’s testimony. Contrary to the majority’s assertion, there is little reason to believe on the basis of her testimony that Smith saw the crowbar on the bedroom floor when he paused briefly in the doorway. Mrs. Konzelman stated that she herself did not see the crowbar until the attacker picked it up from the ground just prior to assaulting her husband. She also stated that the second burglar’s pause at the doorway was brief — just long enough for her to see his bandana — which is consistent with Edmonds’s affidavit stating that Smith “never entered the Konzle-man’s[sic] bedroom.” These facts strongly suggest that Smith did not have the occasion during his brief pause at the doorway to search the room visibly and observe the crowbar, which was on the floor and out of plain sight.
Second, three of the majority’s primary pieces of evidence prove nothing more than that Smith learned at some point, either during, or more likely after, the attack, that Edmonds was armed with the crowbar.11 It would be patently unreason *1158able to find that Smith “[h]ad ... reason to believe that [Edmonds] was armed with a dangerous or deadly weapon” for the purposes of Oregon Revised Statute § 163.115(3)(d) if he discovered that Ed-monds had possession of the crowbar only while Edmonds was committing the attack or after it had taken place. The purpose of an affirmative defense statute, after all, is to ensure that only persons who were aware beforehand that a death could occur during the felony, and therefore might have acted to prevent it, may be held liable for felony murder. These points are thus wholly irrelevant to the question whether Smith had reasonable ground to believe prior to the attack that Edmonds was armed.
The majority’s attempt to bolster its analysis through Smith’s 1992 post-conviction deposition testimony is also unavailing. Although Smith “consulted with his counsel” about other defenses his trial counsel could have raised, we have no idea whether Smith’s post-conviction counsel told him that not having reason to know that Edmonds had a crowbar constituted such a defense, or whether Smith understood him if he had. (We could, of course, have learned the answer to these questions had the evidentiary hearing to which Smith was entitled been held.) Thus, the fact that Smith never explicitly stated during that deposition that he was unaware Edmonds had the crowbar tells us nothing. Moreover, the majority interprets Smith’s deposition statement that he “didn’t know that [Edmonds] was going to[kill Konzel-man]” as meaning that Smith was “unaware Edmonds had the intent to use the crowbar.” Maj. op. at 1145 (emphasis added). Smith’s statement, however, could just as easily mean that Smith was unaware that Edmonds had the crowbar. Indeed, that may be the more reasonable interpretation of Smith’s vague utterance — after all, Smith may have been unaware that Edmonds was going to kill Konzelman precisely because he never saw Edmonds holding the crowbar.
In sum, the majority’s evidentiary analysis depends almost entirely on credibility determinations of Edmonds’s never-given testimony at a hearing that never occurred; the few crumbs of evidence that do not fall in that category fail to support the majority’s conclusion, and, if they did, they would hardly be sufficient.12
*1159IV.
Finally, I note briefly that, although the majority does not advocate holding a hearing, because it believes that its hypothetical construct dispenses with any need for doing so, it also appears to suggest that granting Edmonds use immunity could effectively remedy the prosecutorial misconduct. This is also erroneous. As the panel opinion explained, granting Edmonds use immunity
would not effectively counter the threat of execution, as the state would be free to seek the death penalty even if barred from relying on Edmonds’s testimony. If Edmonds did decide to testify, there would be no way to ensure that the looming prosecutorial threat of execution would not significantly influence his testimony.... Any testimony that contradicted his affidavits would be of doubtful reliability.
Smith, 466 F.3d at 826-27. Because the state’s threat to have Edmonds executed would taint any testimony that Edmonds might give so long as the threat of execution loomed, we could not now send the case back for an evidentiary hearing at which Edmonds testifies under use immunity. On the other hand, granting Ed-monds “transactional immunity” — or total immunity from prosecution on the subject matter of his testimony — would allow Ed-monds to testify more readily but would provide too much of a reward for testimony that may be untruthful, and such immunity might encourage Edmonds to lie with impunity. Thus, allowing Smith to proceed to a hearing on the merits of his constitutional claims provides the least intrusive solution; such a remedy would effectively cure the harm wrought by the prosecution’s unconstitutional threat that it would seek to have a witness executed if he testified in Smith’s favor, without providing too much or too little immunity to that witness. This remedy would, of course, grant no substantive relief to Smith, who would still be required to establish on the merits his claims of constitutional violations. In any event, the majority affirms the denial of an evidentiary hearing and does not remand for the grant of use immunity. So, its discussion of that remedy is as worthless as its construct of a non-existent hearing.
V.
In the end, it is difficult to believe that the Court took this case en banc to decide the factual question whether the evidence does or does not support an affirmative defense to felony murder under a particular provision of Oregon law.13 We do not *1160go en banc to sort out questions of fact or state law, or to create mythical records for hearings that never were. See Fed. R.App. P. 35(a) (“An en banc ... rehearing is not favored and ordinarily will not be ordered unless: (1) en banc consideration is necessary to secure or maintain uniformity of the court’s decisions; or (2) the proceeding involves a question of exceptional importance.”). It is even more unfathomable that, by its fallacious evaluation of non-existent facts, the majority reaches a result the correctness of which is impossible to evaluate, all in order to avoid considering the state’s flagrant prosecuto-rial misconduct and the consequences that should attach to it.
I regret that this court did not hold, as it should have, that threatening a witness with execution if he testifies that a person convicted of a crime is not guilty is in direct contravention of the Due Process Clause of the Constitution. We have an obligation to label such egregious prosecu-torial misconduct for what it is — a gross deprivation of the petitioner’s rights and a reprehensible abuse of the prosecutor’s authority. Here, it is evident from the district court’s finding that the effect of the state’s misconduct was to deny a habeas petitioner the very evidentiary hearing necessary to establish his affirmative defense. Yet, the majority decides against Smith’s Schlup claim almost entirely on the basis of the earlier ambiguous testimony of the probable killer who at the time was attempting to pin the murder on his partner in the robbery in order to obtain a lighter sentence for himself, and who has since recanted his story. Because, unlike the majority, I believe we cannot countenance the state’s blatant violation of constitutional processes, and because, in the absence of an evidentiary hearing, I cannot deem the violation harmless on the basis of a story the threatened witness has disavowed, I respectfully dissent.
. See Smith v. Baldwin, 466 F.3d 805, 814-17 (9th Cir.2006), reh’g en banc granted 482 F.3d 1156 (9th Cir.2007).
. The Supreme Court has recognized the inherent unreliability of an accusation made by a defendant against his co-defendant. As the Court explained in Lilly v. Virginia, 527 U.S. 116, 119 S.Ct. 1887, 144 L.Ed.2d 117 (1999),
when one person accuses another of a crime under circumstances in which the declarant stands to gain by inculpating another, the accusation is presumptively suspect and must be subjected to the scrutiny of cross-examination.... 'Due to his strong motivation to implicate the defendant and to exonerate himself, a codefendant’s statements about what the defendant said or did are less credible than ordinary hearsay evidence.’
Id. at 132, 119 S.Ct. 1887 (quoting Bruton v. United States, 391 U.S. 123, 141, 88 S.Ct. 1620, 20 L.Ed.2d 476 (1968) (White, J., dissenting)). Despite this admonition from our highest court, criminal convictions are still obtained on the basis of co-defendant accusations with great frequency.
. Smith also pled guilty to the robbery, but in his habeas petition he did not deny his guilt of that offense and did not seek to set aside that conviction or his sentence on that count.
. Although Edmonds never formally confessed to having killed Konzelman, he twice stated affirmatively, in his affidavits, that Smith was not the killer. Given that Smith and Edmonds were the only two intruders in the house when the attack occurred, the only reasonable inference from Edmonds's absolution of Smith is that Edmonds himself was the murderer. Indeed, as the state itself acknowledged, "Edmonds has implicitly admitted that he, rather than [Smith], was the killer, [even though] he has never directly admitted as much."
.There can be little doubt that the state's actions qualify as prosecutorial misconduct. As the panel opinion explained:
Threatening a potential witness for the defense with execution constitutes prosecuto-rial misconduct far more coercive than that present in any reported case of which we are aware. The cases in which courts have considered the prosecution's threats to charge witnesses with perjury or other criminal offenses, have all involved the possibility of far less serious punishment.... Here, the prosecution’s unprecedented threat to seek the death penalty against Ed-monds if he testified that Smith was not the killer was unquestionably coercive and constituted substantial interference with Ed-monds’s decision whether to testify.
Smith, 466 F.3d at 823 (citations omitted; emphasis in original).
. Indeed, the majority’s criticism of those affidavits for their incompleteness, see maj. op. at 1143-44 (stating that "[njeither recantation addresses th[e] issue” whether Smith and Ed-monds entered the house at the same time); id. at 1144 (noting that neither of Edmonds’s affidavits addresses whether Smith could have seen the crowbar when he paused briefly at the bedroom door), only underscores this point.
. A petition for habeas corpus, of course, initiates a civil proceeding. Mayle v. Felix, 545 U.S. 644, 654 n. 4, 125 S.Ct. 2562, 162 L.Ed.2d 582 (2005).
. This is in contrast to prevailing on a freestanding actual innocence claim, which alone is grounds for granting habeas relief but imposes a much more difficult — indeed, possibly insurmountable — burden on the petitioner. See Herrera v. Collins, 506 U.S. 390, 113 S.Ct. 853, 122 L.Ed.2d 203 (1993).
. The majority strangely suggests that we are "being inconsistent,” maj. op. at 1142 n. 11, in criticizing its hypothetical credibility determinations because an en banc decision that we joined, Carriger v. Stewart, 132 F.3d 463 (9th Cir.1997) (en banc), concluded that a state court’s credibility determination was “not fairly supported by the record as a whole” and therefore was "not entitled to a presumption of correctness,” id. at 475-76. The majority's comparison of its own approach to that of Carriger is, to say the least, bizarre. In Carriger, the state court had found that a deceased witness's confession, which it had not itself witnessed, was not consistent with the physical evidence and therefore not credible. Id. at 473. The en banc court, reviewing the entire record, concluded that there was "virtually no ... support for the state court’s rejection of[the witness's] confession” and, consequently, declined to defer to the state court’s credibility finding, pursuant to 28 U.S.C. § 2254(d)(8) (1994). Id. at 475; see also 28 U.S.C. § 2254(d)(8) (1994) ("[A] determination ... of a factual issue, made by a State court of competent jurisdiction ..., shall be presumed to be correct, unless ... the Federal court on a consideration of ... the record as a whole concludes that such factual determination is not fairly supported by the record.”). The Carriger court's appellate review, pursuant to statute, of a credibility determination made by the state court — that the state court was in no better position to make — is a far cry from this majority’s credibility assessment of hypothetical testimony that has never been given, let alone passed upon by a lower court.
. The linchpin of the majority's reasoning as to element (d) is Edmonds’s prior statements indicating that he and Smith entered the Kon-zelmans’ residence "at or around the same time.” Maj. op. at 1144. Without Edmonds’s statements, which were, in any event, vague and imprecise, two of the majority's primary factual bases — that the crowbar is three-feet long (and therefore difficult to conceal) and that there was sufficient light in the garage to see objects — would be wholly irrelevant. It would make little difference that the crowbar was long or that the garage was sufficiently lit if Smith and Edmonds did not enter the house together. If Edmonds entered sometime after Smith (or perhaps even vice versa) he could have easily removed the crowbar from the garage and carried it into the house without Smith's knowledge.
. These pieces are: (1) the ultimate locations of the rope and crowbar, (2) Smith's guilty plea on the robbery charge, and (3) the majority's theory that Smith may have witnessed Edmonds committing the attack from another room.
First, the majority asserts that "the ultimate placement of the [rope and crowbar] permits the reasonable inference that Smith and Ed-monds crossed paths while Edmonds had the crowbar and Smith the rope[]....” Maj. op. at 1144. To the extent that the objects’ placement implies that Smith and Edmonds "crossed paths,” however, it does not imply that they did so prior to the beating. To the contrary, given that Smith never entered the bedroom before the beating, the only reasonable inference that can be drawn is that, if the two crossed paths at all, allowing Smith to see Edmonds holding the crowbar, it was only after the attack had taken place.
Second, the fact that Smith pled to the robbery charge that alleged that the defendants used a crowbar proves nothing more than that he had learned at some time that Edmonds had a crowbar. This is hardly a surprising admission, given that Smith would certainly have learned of that fact when he discovered that Konzelman had been murdered, if only from being so informed by the police.
Finally, the majority’s conjecture that "it is entirely plausible that Smith returned to the bedroom doorway when he heard the beating, or that he moved to a vantage point in the hallway from which he was not visible to Mrs. Kozelman [sic] but could still see the attack,” maj. op. at 1144, if it may even be believed, proves only that Smith discovered that Ed-monds had the crowbar while the attack was taking place. It says nothing about whether Smith had reason to believe prior to the attack *1158that Edmonds had picked up the crowbar— the only time frame that is relevant to element (d) of the affirmative defense.
. Judge Fisher’s approach, while rightly rejecting the majority's reliance on Edmonds's hypothetical testimony, see Fisher op. at 1150, ultimately fails for the same reason the majority’s does. Judge Fisher notes that Ed-monds's past statements concerning "Smith's and Edmonds’ respective comings-and-goings into the Konzelmans’ house” were highly inconsistent, Fisher op. at 1149, and concludes on that basis that a reasonable juror would most likely find that Edmonds and Smith were together at least sometime during the evening and that Smith would have seen Ed-monds carrying the crowbar at that point, Fisher op. at 1150. Like the majority, Judge Fisher believes that an evidentiary hearing would have made no difference because Smith could not have proven his actual innocence regardless of what Edmonds might have said. But the fact that Edmonds changed his story so many times when he was attempting to shift the responsibility to Smith surely cannot defeat Smith’s Schlup claim. That a defendant trying to pin the blame on someone else was unable to keep his story straight does not provide reason to credit one of those stories over the others, as Judge Fisher suggests; if anything, it is reason to discount all of Edmonds’s past stories as fabricated in an attempt to avoid culpability himself. At the very least, Edmonds's failed past inconsistent statements are not a reason to deprive Smith of an evidentiary hearing. Because the existing evidence in the record is so utterly unreliable, as Judge Fisher himself recognizes, we cannot assume that what Ed-monds might say at an evidentiary hearing at which he finally accepts responsibility for the crime would have no bearing on whether Smith could prove his affirmative defense.
*1159Given that Judge Fisher also appears to recognize the egregious nature of the prosecu-torial misconduct, I am puzzled as to why he would simply assume that fact for the purpose of the opinion rather than stating expressly that he, like the dissenters, would hold that the prosecutor's behavior was grossly improper and unethical. Although I disagree with the majority for all the reasons expressed in this dissent, I find its failure to adopt such a holding and to condemn expressly the prosecutor’s conduct particularly inexcusable. I would hope that its unwillingness to do so would not encourage other prosecutors to believe that they may engage in similar threats to seek the death penalty against witnesses who might testify to facts that are contrary to the prosecution's theory of the case.
. Despite the majority's insistence, this case is not about construing Schlup. What a habe-as petitioner must prove to establish actual innocence under Schlup is not at issue in this case. Rather, as I explained, earlier, the unique problem we face is determining what a court should do when the state, by threatening to execute a habeas petitioner’s key witness if he testifies at an evidentiary hearing, makes it impossible for the petitioner to establish his Schlup claim. The majority evades that question, however, and instead decides this case based entirely on an unprecedented process of appellate factfinding and credibility assessing of hypothetical testimony at a hearing-that-never-was.