dissenting from denial of rehearing en banc:
Judge Wallace’s dissent ably explains why the court errs in holding that the new rule established in Crawford v. Washington, 541 U.S. 36, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004), applies retroactively. See Bockting v. Bayer, 399 F.3d 1010, 1024 (9th Cir. Feb.22, 2005) (Wallace, J., concurring in part and dissenting in part). I write only to add a few additional reasons why I believe the majority’s holding— which conflicts with the conclusion of all five other circuits to have reached the issue, see infra at 10406-07 — is in serious tension with the retroactivity jurisprudence of the Supreme Court as well as our own court. With respect, I believe that we have erred in failing to rehear this case en banc.
I
A
The last time the Supreme Court had occasion to reverse this circuit’s holding that a new rule of criminal procedure applied retroactively, it instructed us as follows: “That a new procedural rule is ‘fundamental’ in some abstract sense is not enough; the rule must be one without which the likelihood of an accurate conviction is seriously diminished.” Schriro v. Summerlin, 542 U.S. 348, 124 S.Ct. 2519, 2523, 159 L.Ed.2d 442 (2004) (internal quotation marks, brackets, and citation omitted). It seems to me that the majority ignores the emphasis that the Court itself placed on the word “seriously.” Almost any new rule will work to a criminal defendant’s advantage in some circumstances. The question, though, is whether Crawford’s new rule is of the magnitude of the one ruling that the Court has told us would apply retroactively — namely, the one articulated in Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963), that a defendant has the right to be represented by counsel.
Yet Crawford hardly bears comparison with Gideon. To deny a criminal defendant charged with a serious crime the benefit of counsel is to put him awash in a sea of doctrines, deadlines, and technicalities, in which his ability to defend his own interest is highly unlikely to survive. While we value individual autonomy enough to permit defendants to represent themselves when they so insist, see Faretta v. California, 422 U.S. 806, 95 S.Ct. 2525, 45 L.Ed.2d 562 (1975), we do so knowing that such autonomy comes at a serious cost to our confidence in the ultimate verdict. Id. at 833. It is thus reasonable to say that, without counsel, “the likelihood of an accurate conviction is seriously diminished.”
The Crawford rule simply does not approach this magnitude. It did not establish ex nihilo the right to confrontation, as Gideon established the right to counsel; it merely reshaped the contours of that right. The difference between the pre-Crawford regime of Ohio v. Roberts, 448 U.S. 56, 100 S.Ct. 2531, 65 L.Ed.2d 597 (1980), in which out-of-court statements not subject to cross-examination were admissible if they bore adequate indicia of reliability, and the new regime in which they are per se inadmissible, is small in comparison to the difference between giving a defendant competent counsel and giving him none at all. Cf. Sawyer v. Smith, 497 U.S. 227, 244, 110 S.Ct. 2822, 111 L.Ed.2d 193 (1990) (“But given that [the rule of Caldwell v. *1057Mississippi, 472 U.S. 320, 105 S.Ct. 2633, 86 L.Ed.2d 231 (1985),1] was added to an existing guarantee of due process protection against fundamental unfairness, we cannot say this systemic rule enhancing reliability is an ‘absolute prerequisite to fundamental fairness’ of the type that may come within Teague’s second exception.” (citation omitted)). It is thus apparent that the “narrow right ... that [Crawford] affords to defendants in a limited class of ... cases ... possesses little of the ‘watershed’ character envisioned by Teague’s second exception.” O’Dell v. Netherland, 521 U.S. 151, 167, 117 S.Ct. 1969, 138 L.Ed.2d 351 (1997).
B
Indeed, Crawford’s rule does less to decrease the chance of an inaccurate conviction than many rules that have been held not to apply retroactively. Most recently, in Schardt v. Payne, 414 F.3d 1025, (9th Cir.2005), we refused to give retroactive effect to Blakely v. Washington, 542 U.S. 296, 124 S.Ct. 2531, 159 L.Ed.2d 403 (2004), which invalidated state sentencing guidelines that increased a defendant’s sentence based on facts found by a judge by a mere preponderance of the evidence. Id. at 2537-38. The application of a mere preponderance standard instead of the reasonable-doubt standard required by Blakely surely increases the likelihood of inaccurate criminal punishment more than the admission of evidence under the Roberts test did.2 Cf. Ivan V. v. City of New York, 407 U.S. 203, 204, 92 S.Ct. 1951, 32 L.Ed.2d 659 (1972) (per curiam) (“[T]he reasonable-doubt standard is a prime instrument for reducing the risk of convictions resting on factual error.”) (quoting In re Winship, 397 U.S. 358, 363-64, 90 S.Ct. 1068, 25 L.Ed.2d 368 (1970)); id. at 205, 92 S.Ct. 1951 (“[T]he major purpose of the constitutional standard of proof beyond a reasonable doubt announced in Winship was to overcome an aspect of a criminal trial that substantially impairs the truth-finding function, and Winship is thus to be given complete retroactive effect.”). If even the standard-of-proof aspect of Blakely does not satisfy the Teague test, I do not see how Craivford can do so.
C
Also instructive is Gilmore v. Taylor, 508 U.S. 333, 113 S.Ct. 2112, 124 L.Ed.2d 306 (1993), in which the Supreme Court considered the retroactivity of the Seventh Circuit’s holding that certain jury instructions in Illinois violated due process because they “allowed the jury to return a verdict of murder even if the jury made findings that should have resulted in a verdict of voluntary manslaughter,” Fal*1058coner v. Lane, 905 F.2d 1129, 1130 (7th Cir.1990). Under the challenged instructions, the Seventh Circuit had noted, “[n]o matter how clearly either the State or the defense proved the existence of the mitigating ‘manslaughter defenses,’ the jury could nevertheless return a murder verdict.” Id. at 1136. Applying Teague, the Supreme Court noted that “the Falconer court expressed concern that the jury might have been confused by the instructions in question,” but nevertheless refused to apply the rule retroactively because it did not “fall[ ] into that small core of rules requiring observance of those procedures that ... are implicit in the concept of ordered liberty.” Gilmore, 508 U.S. at 345, 113 S.Ct. 2112 (quoting Graham v. Collins, 506 U.S. 461, 478, 113 S.Ct. 892, 122 L.Ed.2d 260 (1993)).
Thus even a rule forbidding a jury instruction that concededly permits the jury to convict the defendant of a crime he did not commit is insufficiently fundamental and accuracy-enhancing to warrant retroactive application under Teague. Again, the Bockting majority’s holding is in serious tension with that of the Supreme Court. The only evidence admissible before Crawford but now excluded consists of out-of-court testimonial statements that trial and appellate courts have explicitly found to bear adequate indicia of reliability. It is difficult to see how the introduction of such evidence could be more likely to lead to the conviction of an innocent defendant than a set of jury instructions that significantly misdefine the substance of the crime.
D
The Bockting majority points out, of course, that the Crawford Court severely criticized the “indicia of reliability” test that held sway under Roberts. And so it did. See Crawford, 541 U.S. at 62-65, 124 S.Ct. 1354. The Court’s criticism, however, is primarily of the test’s doctrinal un-manageability, see id. at 63, 124 S.Ct. 1354 (criticizing the Roberts test as “unpredictable” and “amorphous”), and its incompatibility with the Framers’ intentions, see id. (“The unpardonable vice of the Roberts test [is] its demonstrated capacity to admit core testimonial statements that the Confrontation Clause plainly meant to exclude.”); id. at 65, 124 S.Ct. 1354 (conceding that “most of the usual safeguards of the adversary process attend the [admitted out-of-court] statement” but noting that “the single safeguard missing is the one the Confrontation Clause demands.”) Neither of those criticisms goes directly to the crucial question, which is whether “the likelihood of an accurate conviction is seriously diminished” when evidence is admitted under the Roberts test.
The test may have been unpredictable at the margins, as almost any balancing test will be to one degree or another, but nothing in the Crawford opinion suggests that trial and appellate judges were likely to admit clearly unreliable evidence in anything but the exceptional case. Indeed, the Court stated that the vagueness of the Roberts test “might be a small concern in run-of-the-mill ... prosecutions,” even if it could leave defendants unprotected in “great state trials” in “politically charged cases like [Sir Walter] Raleigh’s.” Id. at 68, 100 S.Ct. 2531.
Nor, of course, can the Court’s emphasis on the stark incompatibility between the Roberts test and the Framers’ understanding of the Confrontation Clause be taken to imply that the Roberts test dramatically increases the likelihood of an inaccurate conviction. “That the Framers made a particular judgment about the best way to ensure the reliability of testimony does not mean that any rule other than the one they envisioned creates an impermissibly high *1059risk of inaccurate conviction.” Bockting, 399 F.3d at 1029 (Wallace, J., concurring in part and dissenting in part); cf. Crawford, 541 U.S. at 61, 124 S.Ct. 1354 (“[The Confrontation Clause] commands, not that evidence be reliable, but that reliability be assessed in a particular manner.”).
E
In that respect and others, Craiqford’s rule resembles nothing so much as the last new rule we held to apply retroactively, only to be quickly reversed by the Supreme Court. In Summerlin v. Stewart, 341 F.3d 1082 (9th Cir.2003), we gave retroactive application to the Court’s holding in Ring v. Arizona, 536 U.S. 584, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002), that juries and not judges must determine the existence of any aggravating factor necessary for imposition of the death penalty. See Summerlin, 341 F.3d at 1108-21. In Ring, the Court had emphasized that “[t]he guarantees of jury trial in the [Constitution] reflect a profound judgment about the way in which law should be enforced and justice administered.” 536 U.S. at 609, 122 S.Ct. 2428 (quoting Duncan v. Louisiana, 391 U.S. 145, 155-56, 88 S.Ct. 1444, 20 L.Ed.2d 491 (1968)). The Court had thus held that the Constitution reflects the Founders’ insistence that the fairness and accuracy of criminal prosecutions (and imposition of the death penalty) are best guaranteed by giving the defendant the ability to insist that relevant facts be decided by a jury. See id.; Apprendi v. New Jersey, 530 U.S. 466, 477, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000) (noting that jury factfinding is necessary “[t]o guard against a spirit of oppression and tyranny on the part of rulers.”).
Nevertheless, when we held that Ring applied retroactively because it was a watershed decision of criminal procedure without which the fairness and accuracy of a death sentence were seriously diminished, see Summerlin, 341 F.3d at 1108-21, the Supreme Court quickly reversed us. See Summerlin, 124 S.Ct. at 2524-26. Even though the Constitution demands factfinding by juries — and even though juries may, in fact, be more accurate factfin-ders than judges' — the Court held that there is not sufficient evidence to demonstrate that “judicial factfinding so seriously diminishes accuracy that there is an impermissibly large risk of punishing conduct that the law does not reach.” Id. at 2525 (internal quotation marks and brackets omitted).
A parallel principle governs this case: even though the Confrontation Clause demands the exclusion of out-of-court testimony — and even though blanket exclusion of such testimony may, in fact, be more accurate than the more nuanced rule of Roberts — there is little reason to think that judicial determination of reliability so seriously diminishes accuracy as to make likely the conviction of the innocent. If Apprendi, Ring, and Blakely, with their massive implications striking to the core of our system of criminal justice, were not watershed rules with retrospective application, then surely the relatively minor— though still quite significant — change wrought by Crawford is not either.3
*1060II
And such has been precisely the conclusion of every other circuit to have considered the question. See Mungo v. Duncan, 393 F.3d 327, 336 (2d Cir.2004); Dorchy v. Jones, 398 F.3d 783, 788 (6th Cir. Feb.23, 2005); Murillo v. Frank, 402 F.3d 786, 790 (7th Cir. Apr.1, 2005); Bintz v. Bertrand, 403 F.3d 859, 867 (7th Cir. Apr.7, 2005); Brown v. Uphoff, 381 F.3d 1219, 1227 (10th Cir.2004); see also Evans v. Luebbers, 371 F.3d 438, 444-45 (8th Cir.2004) (strongly suggesting that Crawford does not apply retroactively). It was the conclusion reached by Judge Wallace in his convincing dissent. See Bockting, 399 F.3d at 1024 (Wallace, J., concurring in part and dissenting in part). It was the unanimous conclusion of three judges of this court in an earlier, unpublished disposition. See Hiracheta v. Att’y Gen’l, 105 Fed.Appx. 937, 938 (9th Cir.2004) (unpublished memorandum disposition).4
Even the Supreme Court itself has indirectly suggested that the Crawford rule is not retroactive. In its opinion in Summer-lin — issued well after Crawford and written by Justice Scalia, who also authored the opinion in Crawford — the Court stated that the class of retroactively applicable rules “is extremely narrow, and it is unlikely that any has yet to emerge.” Summerlin, 124 S.Ct. at 2523 (internal quotation marks, ellipsis, and brackets omitted) (emphasis added) (quoting Tyler v. Cain, 533 U.S. 656, 667 n. 7, 121 S.Ct. 2478, 150 L.Ed.2d 632 (2001)). The Bockting majority argues that this pronouncement from the Court “offer[s] discouragement but no guidance.” Bockting, 399 F.3d at 1016. But we treat even Supreme Court dicta with “due deference,” United States v. Baird, 85 F.3d 450, 453 (9th Cir.1996), and the Court’s statement suggests that our holding that Cratvford applies retroactively is likely to meet the same fate as our similar holding in Summerlin with regard to Ring — namely, speedy reversal.
The two-judge Bockting majority thus stands alone in its conviction that Crawford applies retroactively. Its holding will have serious consequences: it will open the door for a slew of habeas petitions (and, for federal prisoners, motions under 28 U.S.C. § 2255) from prisoners whose convictions were based, even partially, on out-of-court testimonial statements. Concerns about taxing the state and federal governments’ resources to retry convicted criminals should not prevent us from granting writs of habeas corpus when the Constitution requires it. But those concerns certainly counsel us to consider very carefully any precedent that will lead to the granting of an unknowable — but likely large— number of such writs. See Beard v. Banks, 542 U.S. 406, 124 S.Ct. 2504, 2511, 159 L.Ed.2d 494 (“In many ways the application of new rules to cases on collateral review may be more intrusive than the enjoining of criminal prosecutions, for it continually forces the States to marshal resources in order to keep in prison defendants whose trials and appeals conformed to then-existing constitutional standards.” *1061(quoting Teague v. Lane, 489 U.S. 288, 310, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989)) (citations omitted)).
Ill
Because Bockting conflicts with the decision of every other circuit to have considered the retroactivity of Craivford; because it conflicts with our own decision in Hiracheta; and, most of all, because it was wrongly decided, I respectfully dissent from our order denying rehearing en banc.
. Caldwell held that "it is constitutionally impermissible to rest a death sentence on a determination made by a sentencer who has been led to believe that the responsibility for determining the appropriateness of the defendant's death rests elsewhere." 472 U.S. at 328-29, 105 S.Ct. 2633.
. Of course, Blakely relates to the accuracy of sentences, not underlying convictions. See United States v. Sanchez-Cervantes, 282 F.3d 664, 671 (9th Cir.2002) (relying, in part, on that difference in holding Apprendi not to apply retroactively). I do not see how the difference can be material, though, when the point of Blakely and the entire line of jurisprudence stemming from Apprendi is precisely that sentencing factors must be treated as elements of a crime when they increase the defendant's maximum sentence. Moreover, the Supreme Court has not distinguished between sentences and convictions when applying Teague; rather, it has implied that a watershed rule could be retroactive under Teague if it "seriously diminish[ed] the likelihood of obtaining an accurate determination in [a] sentencing proceeding.” Graham v. Collins, 506 U.S. 461, 113 S.Ct. 892, 122 L.Ed.2d 260 (1993) (first alteration in original) (internal quotation marks omitted).
. The Seventh and Tenth Circuits have each argued that the watershed status of the Crawford rule is further belied by the fact that violations of the Confrontation Clause are subject to harmless-error review and thus are not structural error on par with the denial of counsel. See Brown, 381 F.3d at 1226-27; Murillo, 402 F.3d at 791. The Bockting majority rejects this reasoning, arguing that even non-structural constitutional rules can constitute watershed, bedrock rules of procedure. See Bockting, 399 F.3d at 1020. Whether or not the majority is correct, though, its holding flatly contradicts our holding and reasoning in United States v. Sanchez-Cervantes, 282 F.3d 664 (9th Cir.2002), which held that Ap-*1060prendi did not apply retroactively on collateral review. The Sanchez-Cervantes panel concluded that "[b]y applying harmless error analysis ... to Apprendi claims, we have necessarily held that Apprendi errors do not render a trial fundamentally unfair. Therefore, it would seem illogical to hold that such an error is a watershed rule." Id. at 670. The Bockting majority does not even cite Sanchez-Cervantes, and — whether or not the rationale in that case was correct — the conflict between the two opinions provides another reason to rehear Bockting en banc.
. 1 cite Hiracheta not as precedent, of course, but because a conflict with a previously issued memorandum disposition is a factor weighing in favor of rehearing en banc. See Ninth Circuit Rule 36-3(b)(iii).