The majority’s adopted opinion1 finds that the rule announced in Halbert v. Michigan, 545 U.S. 605, 125 S.Ct. 2582, *452162 L.Ed.2d 552 (2005), does not apply retroactively to Simmons's case because it is a “new rule” under the Teague analysis. Teague v. Lane, 489 U.S. 288, 301, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1988). As I stated in the original majority opinion in this case, retroactivity analysis is an inexact science. Apparently the en banc majority thinks that any inexactitude is an opportunity for reversal2 and has created with its adopted opinion an arbitrary and unfair result that denies a significant group of people the opportunity to vindicate their constitutional rights. The new opinion misstates Halbert’s rule and finds that it turns on the discretionary nature of first-tier appeals from plea-based convictions. This proposition goes against the plain words of the Supreme Court, as well as the rationale behind the Halbert decision. Halbert’s rule is dictated by, and inextricably linked .to, the longstanding rule that a state must provide counsel for indigent defendants in a first-level appeal from a criminal conviction. This rule was established long ago in Douglas v. California, 372 U.S. 353, 83 S.Ct. 814, 9 L.Ed.2d 811 (1963). The majority’s reliance on the precedent set by Ross v. Mojfit is misplaced; Ross, too, reiterated that a state cannot adopt procedures that leave an indigent defendant “entirely cut off from any appeal at all” by virtue of his indigency. 417 U.S. 600, 612, 94 S.Ct. 2437, 41 L.Ed.2d 341 (1974) (citing Lane v. Brown, 372 U.S. 477, 481, 83 S.Ct. 768, 9 L.Ed.2d 892 (1963)). The rule is thus not new. Moreover, even if considered “new,” it is clearly a watershed rule of criminal procedure because it addresses the fundamental right to counsel, which is necessary to provide the accuracy and fundamental fairness that must underlie a conviction. Teague, 489 U.S. at 315, 109 S.Ct. 1060; see also Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963).
I.
The majority creates an arbitrary window in time during which acknowledged constitutional rights continue to be denied. A bit of background illuminates the problem. In 1994, Michigan amended its constitution to restrict appeals from pleas of guilty or nolo contendere in an effort to reduce the workload of the Michigan Court of Appeals. The purpose of the amendment, however, was never to deny poor defendants the right to counsel. Indeed, the Michigan State Bar Task force, which was charged with recommending solutions to the Appellate Court backlog, stated in its recommendations that “the rights to counsel and to a transcript of the proceedings would remain” in both guilty plea and nolo contendere appeals. Robert B. Webster, Introduction to the Report of the Task Force on Appellate Courts, 7 Mich. B.J. 895 (1993), cited in Note, Limiting Michigan’s Guilty and Nolo Contendere Plea Appeals, 73 U. Det. Mercy L.R. 431, 448 (1996). Recognizing that some trial judges and legislators were reading the amendment as abolishing the right to appointed counsel for defendants who sought to appeal their guilty or nolo contendere pleas, the State Appellate Defender’s Office Director, James R. Neuhard, wrote a letter to Chief Justice Cavanagh on the Michigan Supreme Court stating that “the language of the amendment very obviously does not change the language regarding the right to counsel on appeal.” Letter *453from the State Appellate Defender Office to Chief Justice Michael F. Cavanagh, at 2 (Dec. 22, 1994) (on file with University of Detroit Mercy Law Review), cited in 73 U. Det. Mercy L.R. at 449. The Michigan Supreme Court later reiterated this understanding. People v. Bulger, 462 Mich. 495, 614 N.W.2d 103, 107 (Mich.2000) (“Proposal B did not alter the provision of art 1, § 20 that grants defendants” “as provided by law, when the trial court so orders, ... such reasonable assistance as may be necessary to perfect and prosecute an appeal.”). On December 30, 1994, the Michigan Supreme Court adopted an interim rule, MCR 6.425(F)(1)(c), which provided that “in a case involving a conviction following a plea of guilty or nolo contendere the court should liberally grant the request [for appointed counsel] if it is filed within 42 days after sentencing.” Id. at 505, 614 N.W.2d 103. The court later extended the rule indefinitely, citing the need “to protect an indigent defendant’s ability to have the assistance of counsel in pursuing appellate remedies.” Id., MCR 6.425(G)(1)(c).
The unfortunate window opened with the help of a law providing that those who plead guilty generally “shall not have appellate counsel appointed for review of the defendant’s conviction or sentence,” Mich. Comp. Laws Ann. § 770.3a (2000), a law this Court found unconstitutional in Tesmer v. Granholm, 333 F.3d 683 (6th Cir.2003) (en banc), rev’d on standing, Kowalski v. Tesmer, 543 U.S. 125, 125 S.Ct. 564, 160 L.Ed.2d 519 (2004). The window finally closed with Halbert in 2005, which guaranteed the right to counsel on first-level appeals in plea-based convictions. For the limited group of people who pleaded while the window was open, however, the majority’s adopted opinion arbitrarily and unfairly ensures that the unconstitutional law lives on.
II.
In 1963, the Supreme Court established the rule that a state must provide counsel for indigent defendants in a first-level appeal from a criminal conviction. See Douglas v. California, 372 U.S. 353, 83 S.Ct. 814, 9 L.Ed.2d 811. Halbert is dictated by, and inextricably linked to, this rule and is thus not new. The Supreme Court in Teague v. Lane explained what amounts to a “new rule” of criminal procedure as follows: “a case announces a new rule when it breaks new ground or imposes a new obligation on the states or the Federal Government. To put it differently, a case announces a new rule if the result was not dictated by precedent existing at the time the defendant’s conviction became final.” 489 U.S. at 301, 109 S.Ct. 1060 (internal citations omitted). Justice Harlan himself noted long before Teague that a decision does not announce a new rule where it “simply applie[s] a well-established constitutional principle to govern a case which is closely analogous to those which have been previously considered in the prior case law.” Mackey v. United States, 401 U.S. 667, 695, 91 S.Ct. 1160, 28 L.Ed.2d 404 (1971) (Harlan, J., concurring in part and dissenting in part); see also Penry v. Lynaugh, 492 U.S. 302, 314, 109 S.Ct. 2934, 106 L.Ed.2d 256 (1989) (citing with approval Justice Harlan’s approach to retroactivity).
Under this approach, Halbert did not announce a new rule but merely applied the forty-year-old rule from Douglas that in a first-level appeal from a criminal conviction, a state must provide appointed counsel for indigent defendants. This conclusion is apparent not only in the language of Halbert itself, but also in the rationale behind Ross v. Moffit, in the error-correcting function of the appellate *454courts in Michigan, and in our own Court’s prior treatment of this topic.
I turn first to the language of Halbert and the adopted majority’s contention that the discretionary nature of the appeals at issue made its ruling “new.” The Supreme Court itself squarely and explicitly rejected this analysis, stating instead that it was the error-correction role of the court that made the difference:
Michigan urges that review in the Court of Appeals following a plea-based conviction is as “discretionary” as review in the Michigan Supreme Court because both require an application for leave to appeal.... Therefore, Michigan maintains, Ross is dispositive of this case. The Court in Ross, however, recognized that leave-granting determinations by North Carolina’s Supreme Court turned on considerations other than the commission of error by a lower court.... By contrast, the Michigan Court of Appeals, because it is an error-correction instance, is guided in responding to leave to appeal applications by the merits of the particular defendant’s claims, not by the general importance of the questions presented.
Halbeyt, 545 U.S. at 618, 125 S.Ct. 2582 (internal citations omitted). There, the Court recognized that potential appellants were being “disarmed in their endeavor to gain first-tier review,” the exact situation presented in Douglas. Halbert, 545 U.S. at 619, 125 S.Ct. 2582. The Court thus plainly states that “Douglas provides the controlling instruction.” 545 U.S. 605 at 616-617, 125 S.Ct. 2582, 162 L.Ed.2d 552; see also id. at 610, 125 S.Ct. 2582 (“[we] conclude that Halbert’s case is properly ranked with Douglas rather than Ross ”). The adopted majority opinion in this case does an admirable job of trying to parse the words of the Halbert Court and of other courts in subsequent decisions, but the Supreme Court itself has instructed us that the presence of dissenting opinions does not construct an absolute barrier to a determination that one of its decisions did not create a new rule under Teague. See Beard v. Banks, 542 U.S. 406, 416 n. 5, 124 S.Ct. 2504, 159 L.Ed.2d 494 (2004); Williams v. Taylor, 529 U.S. 362, 410, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000).
Next, I examine the language and rationale behind Ross v. Moffit, 417 U.S. 600, 94 S.Ct. 2437. The adopted majority opinion tries to place Halbert in a no-man’s land between Douglas and Ross, suggesting that any rule that depends on both opinions would necessarily be a “new rule.” This is a mischaracterization both of the opinions themselves and Halbert’s reliance on Douglas. Looking at the opinions, Douglas required counsel in first-tier appeals, holding that an indigent’s “one and only appeal” must not be decided without the benefit of counsel. 372 U.S. at 357, 83 S.Ct. 814. The Ross Court forcefully restated this rule: “the State cannot adopt procedures which leave an indigent defendant entirely cut off from any appeal at all by virtue of his indigency.” 417 U.S. at 612, 94 S.Ct. 2437 (internal citations omitted). Moreover, the opinion in Ross emphasized that its holding was distinguishable from Douglas. The adopted majority opinion claims that this was because Ross, like Simmons’s case, involved discretionary appeals as opposed to appeals of right. This is simply not true. The defining feature of Ross was that it involved a second level of appellate review. This second level had two features: the nature of the cases and the ability to appeal effectively. With regard to the first point, the Ross Court emphasized that the appeal in question there was to be heard by a court that was primarily concerned with whether cases involve matters of significant public interest or legal principles of major significance, not to correct errors at the trial *455court level. See Ross, 417 U.S. at 615, 94 S.Ct. 2437 (“The Supreme Court [of North Carolina] may deny certiorari even though it believes that the decision of the Court of Appeals was incorrect.”).
Second, and more importantly, the Ross Court explained that a defendant seeking leave to appeal in the North Carolina Supreme Court had already “received the benefit of counsel in examining the record of his trial and in preparing an appellate brief on his behalf for the state Court of Appeals, [and t]hus prior to his seeking discretionary review in the State Supreme Court, his claims had ‘once been presented by a lawyer and passed upon by an appellate court.’ ” Ross, 417 U.S. at 614, 94 S.Ct. 2437 (quoting Douglas, 372 U.S. at 365, 83 S.Ct. 814). The Ross decision assumed that a petitioner seeking leave to appeal to a second appellate court “will have, at the very least, a transcript or other record of trial proceedings, a brief on his behalf in the Court of Appeals setting forth his claims of error, and in many cases an opinion by the Court of Appeals disposing of his case.” 417 U.S. at 615, 94 S.Ct. 2437. It thus held that subsequent appeals did not require counsel. But that analysis cannot be applied satisfactorily to a plea-based conviction, because the defendant will likely not have any sort of opinion “disposing of his case” from which to work if he attempts to withdraw his plea or move for resentencing citing a purported error. It is simply unrealistic to expect an appellant to develop a record sufficient to pass the high bar for a certificate of ap-pealability without the assistance of appointed appellate counsel in this situation.
It is the “error correction” role of the Michigan Court of Appeals that dictated the result in Halbert — not whether the appeal in question was discretionary. See Halbert, 545 U.S. at 618, 125 S.Ct. 2582 (rejecting Michigan’s argument that the discretionary nature of the appeal was controlling). As the Supreme Court held, because the Michigan Court of Appeals acts to correct errors, even if it first exercises its discretion in deciding which potential errors to address, the result in Halbert is an application of the “old rule” from Douglas. When the Halbert Court examined the combined precedential value of Douglas and Ross, it followed the same distinction that the Ross Court did, recognizing a stronger claim for appointed appellate counsel in first-tier appellate review than in second-tier review. Under this view, it was simply a tangentially-related circumstance that most first-level appeals in state courts happen to be as of right, while second-level appeals tend to be discretionary. It necessarily follows that Halbert was dictated by Douglas, and therefore does not create a new rule under Teague.
Notably, before the Supreme Court’s decision in Halbert, this Court addressed en banc the very same question presented here and concluded that Halbert was dictated by Douglas. See Tesmer v. Granholm, 333 F.3d 683, 701 (6th Cir.2003) (en banc). There, we declared Michigan’s denial of appointed counsel for first-level applications of leave to appeal unconstitutional:
Michigan’s statute creates unequal access even to the first part of the appellate system. Though the judge-appellants argue that any distinctions in Michigan’s appellate system stem from the fact the indigent pleads guilty, or that the appeal is merely discretionary, the effect is to create a different opportunity for access to the appellate system based upon indigency. As applied, the statute violates the due process provision of the Fourteenth Amendment to the United States Constitution, and is thus unconstitutional.
*456That decision was later reversed by the Supreme Court, which held, without reaching the merits, that the plaintiffs did not have standing to bring suit. Kowalski v. Tesmer, 543 U.S. 125, 125 S.Ct. 564, 160 L.Ed.2d 519 (2004). Although the Supreme Court’s reversal rendered this Court’s decision unenforceable and without precedential value, the Tesmer opinion, which relied on Douglas, offers persuasive support for the proposition that the decision in Halbert was dictated by Douglas and that Halbert therefore did not announce a new rule. At a minimum, a majority of the active judges of this Court — who, unlike the Supreme Court, lacked the authority to extend Douglas— believed that the result in Halbert was commanded by Douglas. The Halbert decision later confirmed this reading of Douglas in a case where standing was deemed appropriate. While non-binding for present purposes, our pre-HalbeH conclusion that HalbeH was dictated by Douglas reinforces the logic that no new rule was created.
III.
Even assuming, arguendo, that HalbeH did establish a new rule, this rule amounts to an extension of the right to counsel after a plea-based conviction and should be applied retroactively under Teague because it pertains to the right to counsel and has a significant impact on the accuracy of a conviction.
Under Teague, even if a rule of criminal procedure is “new,” it still may apply retroactively on collateral review if it meets one of two exceptions. 489 U.S. at 311, 109 S.Ct. 1060. The first exception calls for a rule to be “applied retroactively if it places ‘certain kinds of primary, private individual conduct beyond the power of the criminal law-making authority to proscribe.’ ” Id. This exception is not relevant here. The second exception covers what Teague termed “watershed rules of criminal procedure,” which the Court described as “those new procedures without which the likelihood of an accurate conviction is seriously diminished.” Id. at 311, 313, 109 S.Ct. 1060. Significantly, in defining these exceptions to the general rule against retroactivity, the Teague Court quoted Justice Harlan’s earlier commentary, pointing to the right to counsel at trial as a primary example of a “watershed rule” that seriously impacts the accuracy of a conviction. Id. at 311, 109 S.Ct. 1060. The Supreme Court has recently noted that rules falling under the second exception are rare. See Beard 542 U.S. at 41, 124 S.Ct. 25047. Even so, there is a test for a reason, and that reason is not simply to tease appellants and the lower courts by creating a standard that can never be met.
The Supreme Court has continued to emphasize that decisions involving the right to counsel are the paradigmatic examples of watershed rules. Id. at 417, 124 S.Ct. 2504 (“[i]n providing guidance as to what might fall within this exception, we have repeatedly referred to the rule of Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963) (right to counsel), and only to this rule.”). At its core, the HalbeH decision is about the right to counsel, which brings it within the ambit of the watershed exception. In 2002, the Eleventh Circuit applied retroactively the rule from Alabama v. Shelton, 535 U.S. 654, 122 S.Ct. 1764, 152 L.Ed.2d 888 (2002) (prohibiting the use of prior uncounseled state court convictions in a defendant’s criminal history for sentencing purposes) based on the reasoning that Shelton extended the right to counsel. Howard v. United States, 374 F.3d 1068,-*4571080 (11th Cir.2004).3 Although the Eleventh Circuit recognized how narrow the Teague watershed exception is, it added that “[t]he exception that proves the exception, however, is a new Gideon-related rule” and that “[a]t the risk of oversimplification, for purposes of the second Teague exception there are new rules, and then there are new Gideon-extension rules.” 374 F.3d at 1080. The Howard court cata-logued numerous pre-Teague cases where the Supreme Court had applied new rules regarding the right to counsel retroactively, including the holding of Gideon itself, 372 U.S. at 344-45, 83 S.Ct. 792, which was judged to be retroactively applicable in Kitchens v. Smith, 401 U.S. 847, 847, 91 S.Ct. 1089, 28 L.Ed.2d 519 (1971). The right to counsel at plea hearings, recognized in White v. Maryland, 373 U.S. 59, 83 S.Ct. 1050, 10 L.Ed.2d 193 (1963), was held to be retroactively applicable in Arsenault v. Massachusetts, 393 U.S. 5, 6, 89 S.Ct. 35, 21 L.Ed.2d 5 (1968). The right to counsel at probation revocation hearings, announced in Mempa v. Rhay, 389 U.S. 128, 88 S.Ct. 254, 19 L.Ed.2d 336 (1967), was held to be retroactively applicable in McConnell v. Rhay, 393 U.S. 2, 3-4, 89 S.Ct. 32, 21 L.Ed.2d 2 (1968). The right to counsel in any prosecution leading to actual imprisonment announced in Argersinger v. Hamlin, 407 U.S. 25, 40, 92 S.Ct. 2006, 32 L.Ed.2d 530 (1972), was deemed retroactively applicable in Berry v. City of Cincinnati, 414 U.S. 29, 29-30, 94 S.Ct. 193, 38 L.Ed.2d 187 (1973). Finally, the right to counsel on first-tier appeals, recognized in Douglas, has also been retroactively applied, see McConnell, 393 U.S. at 3, 89 S.Ct. 32. This point bears restating: the case that was most similar to Halbert (and dictated the result there, arguably preventing it from being a new rule in the first place) was Douglas, which the Supreme Court noted was applied retroactively. 393 U.S. at 3, 89 S.Ct. 32. As the Howard court said, “A score that is perfect packs punch in any analysis.” 374 F.3d at 1077-1078.
The adopted majority opinion argues against the proposition that all fundamental right-to-counsel cases fall within the Teague exception, citing instead the need to guarantee fundamental fairness “at trial.” This interpretation ignores the purpose of such a guarantee. The reason that trial guarantees are so important is that without them, “the likelihood of an accurate conviction is seriously diminished.” Teague, 489 U.S. at 313, 109 S.Ct. 1060. Appeals after guilty pleas, too, directly implicate a procedure without which the accuracy of a conviction cannot be assured. See Halbert, 545 U.S. at 617, 125 S.Ct. 2582 (“Of critical importance, the tribunal to which he addresses his application, the Michigan Court of Appeals, unlike the Michigan Supreme Court, sits as an error-correction instance.”). In 2002, ninety-five percent of state-court felony convictions and ninety-six percent of federal convictions were obtained by guilty pleas. See Matthew R. Durose & Patrick A. Langan, Felony Sentences in State Courts, 2002, U.S. Dep’t of Justice, Bureau of Justice Statistics, Bull., No. 206916, at 1 (2004). Studies documenting abbreviated investigations, sentencing disparities, and incorrect but uncorrected presentence report calculations show that the efficiency gained by our increasingly plea bargain-based system is compromised by its tremendous risks. See, e.g., Nancy J. King, Judicial Oversight of Negotiated Sentences in a World of Bargained Punishment, 58 Stan. L. Rev. 293 (2005); see also Nancy J. King and Michael E. O’Neill, Appeal Waivers and the Future of Sentencing Policy, 55 Duke L.J. 209 (2005) (detailing how in*458creased use of stipulations, combined with a lack of review, raises the risk that sentences not in compliance with the law can proliferate without scrutiny). In Michigan at the time of the 1994 constitutional amendment, the State Appellate Defender’s Office in Michigan estimated that of the guilty pleas that were heard on appeal, forty-seven percent were given some degree of relief. Limiting Michigan’s Guilty and Nolo Contendere Plea Appeals, 73 U. Det. Mercy L.R. at 440-41. The forty-seven percent relief rate suggests that “a significant number of errors in sentencing are being made at the trial court level and that an appeal as a matter of right is a check against these mistakes and results in savings to the State by eliminating unnecessarily long periods of incarceration.” Id. In light of the astonishing number of cases processed through appeals and the significant risks that this system entails, it is no stretch to say that appeals have become the new focus-point in accurate determinations of innocence or guilt. See Teague, 489 U.S. at 313, 109 S.Ct. 1060.
The Halbert decision gives significant emphasis to the accuracy and fairness that appointed counsel would add to first-tier appeals. 125 S.Ct. at 2592-93 (“A first-tier review applicant, forced to act pro se, will face a record unreviewed by appellate counsel, and will be equipped with no attorney’s brief prepared for, or reasoned opinion by, a court of review.... Persons in Halbert’s situation are particularly handicapped as self-representatives.... Navigating the appellate process without a lawyer’s assistance is a perilous endeavor for laypersons.”). The emphasis on fairness and accuracy in Halbert further suggests that if it is in fact “new,” it would qualify for the watershed exception because in our present system, appeals are often the only way to ensure accuracy of convictions. Teague, 489 U.S. at 313, 109 S.Ct. 1060.
IV.
The Federal Constitution imposes on the States no obligation to provide appellate review of criminal convictions. McKane v. Durston, 153 U.S. 684, 687, 14 S.Ct. 913, 38 L.Ed. 867 (1894). Having provided such an avenue, however, a State may not “bolt the door to equal justice” to indigent defendants. Griffin v. Illinois, 351 U.S. 12, 24, 76 S.Ct. 585, 100 L.Ed. 891 (1956) (Frankfurter, J., concurring); see id., at 23, 76 S.Ct. 585, (“[W]hen a State deems it wise and just that convictions be susceptible to review by an appellate court, it cannot by force of its exactions draw a line which precludes convicted indigent persons ... from securing such ... review.”). The State cannot adopt procedures that leave an indigent defendant “entirely cut off from any appeal at all,” by virtue of his indigency, Lane v. Brown, 372 U.S. at 481, 83 S.Ct. 768, or extend to such indigent defendants merely a “meaningless ritual” while others in better economic circumstances have a “meaningful appeal.” Douglas v. California, 372 U.S. at 358, 83 S.Ct. 814. Michigan had its own protection against this inequality before 1994, and in 2005 the Halbert decision ensured that no one would again be left out. Yet for a small group of people arbitrarily caught between Michigan’s own protections and the protection offered by Hal-bert, the “meaningless ritual” of indigent appeals continues to be a harsh and unjust reality thanks to the adopted majority’s opinion. Therefore, I respectfully dissent.
. The en banc court merely adopted the dissent of the prior panel’s opinion, Simmons v. Kapture, 474 F.3d 869 (6th Cir.2007).
. I reiterate the concerns expressed by Judge Moore in her dissent in Bell v. Bell, 512 F.3d 223, 250 (6th Cir.2008), in which she notes that under both the Federal Rules of Appellate Procedure and our own circuit rule, en banc review may not be used to exercise plenary review over panel decisions. See Fed R.App. P. 35(a); 6 Cir. R. 35(c). I reiterate those concerns, and feel that yet again in our Circuit, en banc review was improperly granted.
. Howard involved collateral review of a federal conviction under 28 U.S.C. § 2255.