with whom Justice Thomas joins, and with whom The Chief Justice joins as to all but Part IV, dissenting.
“If a plea bargain has been offered, a defendant has the right to effective assistance of counsel in considering whether to accept it. If that right is denied, prejudice can be shown if loss of the plea opportunity led to a trial resulting in a conviction on more serious charges or the imposition of a more severe sentence. ” Ante, at 168.
“The inquiry then becomes how to define the duty and responsibilities of defense counsel in the plea bargain process. This is a difficult question.... Bargaining is, by its nature, defined to a substantial degree by personal style.... This case presents neither the necessity nor the occasion to define the duties of defense counsel in those respects . . . .” Missouri v. Frye, ante, at 144-145.
With those words from this and the companion case, the Court today opens a whole new field of constitutionalized criminal procedure: plea-bargaining law. .The ordinary criminal process has become too long, too expensive, and unpredictable, in no small part as a consequence of an intricate federal Code of Criminal Procedure imposed on the States by this Court in pursuit of perfect justice. See Friendly, The Bill of Rights as a Code of Criminal Procedure, 53 Cal. L. Rev. 929 (1965). The Court now moves to bring perfection to the alternative in which prosecutors and defendants have sought relief. Today’s opinions deal with only two aspects of counsel’s plea-bargaining inadequacy, and leave *176other aspects (who knows what they might be?) to be worked out in further constitutional litigation that will burden the criminal process. And it would be foolish to think that “constitutional” rules governing counsel’s behavior will not be followed by rules governing the prosecution’s behavior in the plea-bargaining process that the Court today announces “ ‘is the criminal justice system/ ” Frye, ante, at 144 (quoting approvingly from Scott & Stuntz, Plea Bargaining as Contract, 101 Yale L. J. 1909, 1912 (1992) (hereinafter Scott)). Is it constitutional, for example, for the prosecution to withdraw a plea offer that, has already been accepted? Or to withdraw an offer before the defense has had adequate time to consider and accept it? Or to make no plea offer at all, even though its case is weak — thereby excluding the defendant from “the criminal justice system”?
Anthony Cooper received a Ml and fair trial, was found guilty of all charges by a unanimous jury, and was given the sentence that the law prescribed. The Court nonetheless concludes that Cooper is entitled to some sort of habeas corpus relief (perhaps) because his attorney’s allegedly incompetent advice regarding a plea offer caused him to receive a full and fair trial. That conclusion is foreclosed by our precedents. Even if it were not foreclosed, the constitutional right to effective plea bargainers that it establishes is at least a new rule of law, which does not undermine the Michigan Court of Appeals’ decision and therefore cannot serve as the basis for habeas relief. And the remedy the Court announces — namely, whatever the state trial court in its discretion prescribes, down to and including no remedy at all — is unheard of and quite absurd for violation of a constitutional right. I respectfully dissent.
I
This case and its. companion, Missouri v. Frye, ante, p. 134, raise relatively straightforward questions about the scope of the right to effective assistance of counsel. Our case law *177originally derived that right from the Due Process Clause, and its guarantee of a fair trial, see United States v. Gonzalez-Lopez, 548 U. S. 140, 147 (2006), but the seminal case of Strickland v. Washington, 466 U. S. 668 (1984), located the right within the Sixth Amendment. As the Court notes, ante, at 164-165, the right to counsel does not begin at trial. It extends to “any stage of the prosecution, formal or informal, in court or out, where counsel’s absence might derogate from the accused’s right to a fair trial.” United States v. Wade, 388 U. S. 218, 226 (1967). Applying that principle, we held that the “entry of a guilty plea, whether to a misdemeanor or a felony charge, ranks as a ‘critical stage’ at which the right to counsel adheres.” Iowa v. Tovar, 541 U. S. 77, 81 (2004); see also Hill v. Lockhart, 474 U. S. 52, 58 (1985). And it follows from this that acceptance of a plea offer is a critical stage. That, and nothing more, is the point of the Court’s observation in Padilla v. Kentucky, 559 U. S. 356, 373 (2010), that “the negotiation of a plea bargain is a critical phase of litigation for purposes of the Sixth Amendment right to effective assistance of counsel.” The defendant in Padilla had accepted the plea bargain and pleaded guilty, abandoning his right to a fair trial; he was entitled to advice of competent counsel before he did so. The Court has never held that the rule articulated in Padilla, Tovar, and Hill extends to all aspects of plea negotiations, requiring not just advice of competent counsel before the defendant accepts a plea bargain and pleads guilty, but also the advice of competent counsel before the defendant rejects a plea bargain and stands on his constitutional right to a fair trial. The latter is a vast departure from our past. cases, protecting not just the constitutionally prescribed right to a fair adjudication of guilt and punishment, but a judicially invented right to effective plea bargaining.
It is also apparent from Strickland that bad plea bargaining has nothing to do with ineffective assistance of counsel in the constitutional sense. Strickland explained that “[i]n *178giving meaning to the requirement [of effective assistance], ... we must take its purpose — to ensure a fair trial — as the guide.” 466 U. S., at 686. Since “the right to the effective assistance of counsel is recognized not for its own sake, but because of the effect it has on the ability of the accused to receive a fair trial,” United States v. Cronic, 466 U. S. 648, 658 (1984), the “benchmark” inquiry in evaluating any claim of ineffective assistance is whether counsel’s performance “so .undermined the proper functioning of the adversarial process” that it failéd to produce a reliably “just result.” Strickland, 466 U. S., at 686. That is what Strickland’s requirement of “prejudice” consists of: Because the right to effective assistance has as its purpose the assurance of a fair trial, the right is not infringed unless counsel’s mistakes call into question the basic justice of a defendant’s conviction or sentence. That has been, until today, entirely clear. A defendant must show “that counsel’s errors were so serious as to deprive the defendant of a fair trial, a trial whose result is reliable.” Id., at 687. See also Gonzalez-Lopez, supra, at 147. Impairment of fair trial is how we distinguish between unfortunate attorney error and error of constitutional significance.1
*179To be sure, Strickland stated a rule of thumb for measuring prejudice which, applied blindly and out of context, could support the Court’s holding today: “The defendant must show that there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” 466 U. S., at 694. Strickland itself cautioned, however, that its test was not to be applied in a mechanical fashion, and that courts were not to divert their “ultimate focus” from “the fundamental fairness of the proceeding whose result is being challenged.” Id., at 696. And until today we have followed that course.
In Lockhart v. Fretwell, 506 U. S. 364 (1993), the deficient performance at issue was the failure of counsel for a defendant who had been sentenced to death to make an objection that would have produced a sentence of life imprisonment instead. The objection was fully supported by then-extant Circuit law, so that the sentencing court would have been compelled to sustain it, producing a life sentence that principles of double jeopardy would likely make final. See id., at 383-385 (Stevens, J., dissenting); Bullington v. Missouri, 451 U. S. 430 (1981). By the time Fretwell’s claim came before us, however, the Circuit law had been overruled in light of one of our cases. We determined that a prejudice analysis “focusing solely on mere outcome determination, without attention to whether the result of the proceeding was fundamentally unfair or unreliable,” would be defective. Fretwell, 506 U. S., at 369. Because counsel’s error did not *180“deprive the defendant of any substantive or procedural right to which the law entitles him,” the defendant’s sentencing proceeding was fair and its result was reliable, even though counsel’s error may have affected its outcome. Id., at 372. In Williams v. Taylor, 529 U. S. 362, 391-393 (2000), we explained that even though Fretwell did not mechanically apply an outcome-based test for prejudice, its reasoning was perfectly consistent with Strickland. “Fretwell’s counsel had not deprived him of any substantive or procedural right to which the law entitled him.” 529 U. S., at 392.2
Those precedents leave no doubt about the answer to the question presented here. As the Court itself observes, a criminal defendant has no right to a plea bargain. Ante, at 168. “[T]here is no constitutional right to plea bargain; the prosecutor need not do so if he prefers to go to trial.” Weatherford v. Bursey, 429 U. S. 545, 561 (1977). Counsel’s mistakes in this case thus did not “deprive the defendant of *181a substantive or procedural right to which the law entitles him,” Williams, supra, at 393. Far from being “beside the point,” ante, at 168, that is critical to correct application of our precedents. Like Fretwell, this case “concerns the unusual circumstance where the defendant attempts to demonstrate prejudice based on considerations that, as a matter of law, ought not inform the inquiry,” 506 U. S., at 373 (O’Connor, J., concurring); he claims “that he might have been denied ‘a right the law simply does not recognize,’ ” id., at 375 (same). Strickland, Fretwell, and Williams all instruct that the pure outcome-based test on which the Court relies is an erroneous measure of cognizable prejudice. In ignoring Strickland’s “ultimate focus ... on the fundamental fairness of the proceeding whose result is being challenged,” 466 U. S., at 696, the Court has lost the forest for the trees, leading it to accept what we have previously rejected, the “novel argument that constitutional rights are infringed by trying the defendant rather than accepting his plea of guilty,” Weatherford, supra, at 561.
II
Novelty alone is the second, independent reason why the Court’s decision is wrong. This case arises on federal ha-beas, and hence is governed by the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA). Since, as the Court acknowledges, the Michigan Court of Appeals adjudicated Cooper’s ineffective-assistance claim on the merits, AEDPA bars federal courts from granting habeas relief unless that court’s decision was “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U. S. C. § 2254(d)(1). Yet the Court concludes that § 2254(d)(1) does not bar relief here, because “[b]y failing to apply Strickland to assess the ineffective-assistance-of-counsel claim respondent raised, the state court’s adjudication was contrary to clearly established federal law.” Ante, at 173. That is not so.
*182The relevant portion of the Michigan Court of Appeals decision reads as follows:
“To establish ineffective assistance, the defendant must demonstrate that his counsel’s performance fell below an objective standard of reasonableness and that counsel’s representation so prejudiced the defendant that he was deprived of a fair trial. With respect to the prejudice aspect of the test, the defendant must demonstrate a reasonable probability that, but for counsel’s errors, the result of the proceedings would have been different, and that the attendant proceedings were fundamentally unfair and unreliable.
“Defendant challenges the trial court’s finding after a Ginther hearing that defense counsel provided effective assistance to defendant during the plea bargaining process. He contends that defense counsel failed to convey the benefits of the plea offer to him and ignored his desire to plead guilty, and that these failures led him to reject a plea offer that he now wishes to accept. However, the record shows that defendant knowingly and intelligently rejected two plea offers and chose to go to trial. The record fails to support defendant’s contentions that defense counsel’s representation was ineffective because he rejected a defense based on [a] claim of self-defense and because he did not obtain a more favorable plea bargain for defendant.” People v. Cooper, No. 250583 (Mar. 15, 2005), App. to Pet. for Cert. 45a, 2005 WL 599740, *1 (per curiam) (footnote and citations omitted).
The first paragraph above, far from ignoring Strickland, recites its standard with a good deal more accuracy than the Court’s opinion. The second paragraph, which is presumably an application of the standard recited in the first, says that “defendant knowingly and intelligently rejected two plea offers and chose to go to trial.” This can be regarded *183as a denial that there was anything “fundamentally unfair” about Cooper’s conviction and sentence, so that no Strickland prejudice had been shown. On the other hand, the entire second paragraph can be regarded as a contention that Cooper’s claims of inadequate representation were unsupported by the record. The state court’s analysis was admittedly not a model of clarity, but federal habeas corpus is a “guard against extreme malfunctions in the state criminal justice systems,” not a license to penalize a state court for its opinion-writing technique. Harrington v. Richter, 562 U. S. 86, 102 (2011) (internal quotation marks omitted). The Court’s readiness to find error in the Michigan court’s opinion is “inconsistent with the presumption that state courts know and follow the law,” Woodford v. Visciotti, 537 U.S. 19, 24 (2002) (per curiam), a presumption borne out here by the state court’s recitation of the correct legal standard.
Since it is ambiguous whether the state court’s holding was based on a lack of prejudice or rather the court’s factual determination that there had been no deficient performance, to provide relief under AEDPA this Court must conclude that both holdings would have been unreasonable applications of clearly established law. See Premo v. Moore, 562 U. S. 115, 121 (2011). The first is impossible of doing, since this Court has never held that a defendant in Cooper’s position can establish Strickland prejudice. The Sixth Circuit thus violated AEDPA in granting habeas relief, and the Court now does the same.
Ill
It is impossible to conclude discussion of today’s extraordinary opinion without commenting upon the remedy it provides for. the unconstitutional conviction. It is a remedy unheard of in American jurisprudence — and, I would be willing to bet, in the jurisprudence of any other country.
The Court requires Michigan to “reoffer the plea agreement” that was rejected because of bad advice from counsel. Ante, at 174. That would indeed be a powerful remedy— *184but for the fact that Cooper’s acceptance of that reoffered agreement is not conclusive. Astoundingly, “the state trial court can then exercise its discretion in determining whether to vacate the convictions and resentence respondent pursuant to the plea agreement, to vacate only some of the convictions and resentence respondent accordingly, or to leave the convictions and sentence from trial undisturbed.” Ibid, (emphasis added).
Why, one might ask, require a “reoffer” of the plea agreement, and its acceptance by the defendant? If the District Court finds (as a necessary element, supposedly, of Strickland prejudice) that Cooper would have accepted the original offer, and would thereby have avoided trial and conviction, why not skip the reoffer-and-reacceptance minuet and simply leave it to the discretion of the state trial court what the remedy shall be? The answer, of course, is camouflage. Trial courts, after all, regularly accept or reject plea agreements, so there seems to be nothing extraordinary about their accepting or rejecting the new one mandated by today’s decision. But the acceptance or rejection of a plea agreement that has no status whatever under the United States Constitution is worlds apart from what this is: “discretionary” specification of a remedy for an unconstitutional criminal conviction.
To be sure, the Court asserts that there are “factors” which bear upon (and presumably limit) exercise of this discretion — factors that it is not prepared to specify in full, much less assign some determinative weight. “Principles elaborated over time in decisions of state and federal courts, and in statutes and rules” will (in the Court’s rosy view) sort all that out. Ante, at 171. I find it extraordinary that “statutes and rules” can specify the remedy for a criminal defendant’s unconstitutional conviction. Or that the remedy for an unconstitutional conviction should ever be subject at all to a trial judge’s discretion. Or, finally, that the remedy could ever include no remedy at all.
*185I suspect that the Court’s squeamishness in fashioning a remedy, and the incoherence of what it comes up with, is attributable to its realization, deep down, that there is no real constitutional violation here anyway. The defendant has been fairly tried, lawfully convicted, and properly sentenced, and any “remedy” provided for this will do nothing but undo the just results of a fair adversarial process.
IV
In many — perhaps most — countries of the world, American-style plea bargaining is forbidden in cases as serious as this one, even for the purpose of obtaining testimony that enables conviction of a greater malefactor, much less for the purpose of sparing the expense of trial. See, e. g., World Plea Bargaining 344, 363-366 (S. Thaman ed. 2010). In Europe, many countries adhere to what they aptly call the “legality principle” by requiring prosecutors to charge all prosecutable offenses, which is typically incompatible with the practice of charge bargaining. See, e. g., id., at xxii; Langbein, Land Without Plea Bargaining: How the Germans Do It, 78 Mich. L. Rev. 204, 210-211 (1979) (describing the “Legalitátsprinzip,” or rule of compulsory prosecution, in Germany). Such a system reflects an admirable belief that the law is the law, and those who break it should pay the penalty provided.
In the United States, we have plea bargaining aplenty, but until today it has been regarded as a necessary evil. It presents grave risks of prosecutorial overcharging that effectively compels an innocent defendant to avoid massive risk by pleading guilty to a lesser offense; and for guilty defendants it often — perhaps usually — results in a sentence well below what the law prescribes for the actual crime. But even so, we accept plea bargaining because many believe that without it our long and expensive process of criminal trial could not sustain the burden imposed on it, and our system of criminal justice would grind to a halt. See, e. g., *186Alschuler, Plea Bargaining and Its History, 79 Colum. L. Rev. 1, 38 (1979).
Today, however, the Supreme Court of the United States elevates plea bargaining from a necessary evil to a constitutional entitlement. It is no longer a somewhat embarrassing adjunct to our criminal justice system; rather, as the Court announces in the companion case to this one, “ ‘it is the criminal justice system.’” Frye, ante, at 144 (quoting approvingly from Scott 1912). Thus, even though there is no doubt that the respondent here is guilty of the offense with which he was charged; even though he has received the exorbitant gold standard of American justice — a full-dress criminal trial with its innumerable constitutional and statutory limitations upon the evidence that the prosecution can bring forward, and (in Michigan as in most States3) the requirement of a unanimous guilty verdict by impartial jurors; the Court says that his conviction is invalid because he was deprived of his constitutional entitlement to plea bargain.
I am less saddened by the outcome of this ease than I am by what it says about this Court’s attitude toward criminal justice. The Court today embraces the sporting-chance theory of criminal law, in which the State functions like a consci-. entious casino operator, giving each player a fair chance to beat the house, that is, to serve less time than the law says he deserves. And when a player is excluded from the tables, his constitutional rights have been violated. I do not subscribe to that theory. No one should, least of all the Justices of the Supreme Court.
* ⅝ *
Today’s decision upends decades of our cases, violates a federal statute, and opens a whole new boutique of constitutional jurisprudence (“plea-bargaining law”) without even *187specifying the remedies the boutique offers. The result in the present case is the undoing of an adjudicatory process that worked exactly as it is supposed to. Released felon Anthony Cooper, who shot repeatedly and gravely injured a woman named Kali Mundy, was tried and convicted for his 'crimes by a jury of his peers, and given a punishment that Michigan’s elected representatives have deemed appropriate. Nothing about that result is unfair or unconstitutional. To the contrary, it is wonderfully just, and infinitely superior to the trial-by-bargain that today’s opinion affords constitutional status. I respectfully dissent.
Rather than addressing the constitutional origins of the right to effective counsel, the Court responds to the broader claim (raised by no one) that “the sole purpose of the Sixth Amendment is to protect the right to a fair trial.” Ante, at 164 (emphasis added). Cf. Brief for United States as Amicus Curiae 10-12 (arguing that the “purpose of the Sixth Amendment right to counsel is to secure a fair trial” (emphasis added)); Brief for Petitioner 12-21 (same). To destroy that straw man, the Court cites cases in which violations of rights other than the right to effective counsel— and, perplexingly, even rights found outside the Sixth Amendment and the Constitution entirely — were not cured by a subsequent trial. Vasquez v. Hillery, 474 U. S. 254 (1986) (violation of equal protection in grand jury selection); Ballard v. United States, 329 U. S. 187 (1946) (violation of statutory scheme providing that women serve on juries); Stirone v. United States, 361 U. S. 212 (1960) (violation of Fifth Amendment right to indictment by grand jury). Unlike the right to effective counsel, no showing of prejudice is required to make violations of the rights at issue in Vasquez, *179Ballard, and Stirone complete. See Vasquez, supra, at 263-264 (“[D]is-erimination in the' grand jury undermines the structural integrity of the criminal tribunal itself, and is not amenable to harmless-error review”); Ballard, supra, at 195 (“[R]eversible error does not depend on a showing of prejudice in an individual case”); Stirone, supra, at 217 (“Deprivation of such a basic right is far too serious to be treated as nothing more than a variance and then dismissed as harmless error”). Those cases are thus irrelevant to the question presented here, which is whether a defendant can establish prejudice under Strickland v. Washington, 466 U. S. 668 (1984), while conceding the fairness of his conviction, sentence, and appeal.
Kimmelman v. Morrison, 477 U. S. 365 (1986), cited by the Court, ante, at 169, does not contradict this principle. That case, which predated Fret-well and Williams, considered whether our holding that Fourth Amendment claims fully litigated in state court cannot be raised in federal habeas “should be extended to Sixth Amendment claims of ineffective assistance of counsel where the principal allegation and manifestation of inadequate representation is counsel’s failure to file a timely motion to suppress evidence allegedly obtained in violation of the Fourth Amendment.” 477 U. S., at 368. Our negative answer to that question had nothing to do with the issue here. The parties in Kimmelman had not raised the question “whether the admission of illegally seized but reliable evidence can ever constitute ‘prejudice’ under Strickland” — a question similar to the one presented here — and the Court therefore did not address it. Id., at 391 (Powell, J., concurring in judgment); see also id., at 380. Kimmelman made clear, however, how the answer to that question is to be determined: “The essence of an ineffective-assistance claim is that counsel’s unprofessional errors so upset the adversarial balance between defense and prosecution that the trial was rendered unfair and the verdict rendered suspect,” id., at 374 (emphasis added). “Only those habeas petitioners who can prove under Strickland that they have been denied a fair trial . . . will be granted the writ,” id., at 382 (emphasis added). In short, Kimmel-man’s only relevance is to prove the Court’s opinion wrong.
See People v. Cooks, 446 Mich. 503, 510, 521 N. W. 2d 275, 278 (1994); 6 W. LaFave, J. Israel, N. King, & 0. Kerr, Criminal Procedure § 22.1(e) (3d ed. 2007 and Supp. 2011-2012).