dissenting.
I disagree with the majority’s analysis and would reverse the grant of habeas corpus. I therefore respectfully dissent. The Michigan Court of Appeals decided that, in this case, a failure to inform a criminal defendant, before a guilty plea, that the prosecution may have a right to appeal his being sentenced as a juvenile, does not constitute ineffective assistance of counsel. To succeed on a claim of ineffective assistance of counsel, the petitioner must show (1) deficient performance by the attorney, and (2) prejudice. Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Legitimate trial tactics are expressly excluded from the definition of deficient performance. Bell v. Cone, — U.S. -, 122 S.Ct. 1843, 1854, 152 L.Ed.2d 914 (2002). To establish prejudice after having pled guilty, the pleader must show a reasonable probability that he would not have pled, had he received competent advice. Hill v. Lockhart, 474 U.S. 52, 106 S.Ct. 366, 88 L.Ed.2d 203 (1985). In this case, the petitioner has established neither element of Strickland.
At trial, Lyons faced near-certain conviction. The defenses that he would have asserted at trial were, in essence, vague hopes for jury nullification. His defense counsel knew, however, that there was a very real hope that he could be sentenced as a juvenile if he pled guilty, because of the sentencing habits of the particular judge in the case. Counsel’s performance was not deficient; indeed, the idea for the gamble sprang from defense counsel’s superior and intimate knowledge of the court. Lyons’s counsel took extra steps to ensure that his client’s decision was well informed. He explained extensively the possible consequences of pleading to first degree murder. He arranged for Lyons’s father to be involved in the decisionmaking process. The decision to plead was calculated, voluntary, and informed; this is precisely the sort of competent representation that is excluded from the Strickland definition of deficient performance. Strickland, 466 U.S. at 689-90, 104 S.Ct. 2052; Cone, 122 S.Ct. at 1854.
As for the prejudice element of Strickland, it seems clear that even with knowledge of the possibility of prosecutorial appeal, Lyons would still have pled guilty. Lyons made no attempt to withdraw his guilty plea after the prosecution appealed; rather, he argued his case on the merits before the Court of Appeals. This makes sense: even with the possibility of prosecu-torial appeal, the chance to be sentenced as a juvenile was petitioner’s only option short of trial. Lyons wanted a juvenile sentence, not a new trial. Lyons, through counsel, gamed the system and attempted to withdraw his guilty plea only after his juvenile sentence was rejected by the Court of Appeals. The entire point of the post-conviction strategy was to get two bites at the apple; that Lyons now argues he would have given up his first (and ex ante more likely) bite because of the possibility of prosecutorial appeal defies credence.
Lyons was very aware of the possible consequences of pleading guilty to first *604degree murder. Judge Roberson told him, before he pled, that he could possibly be sentenced as an adult, and therefore be subject to a mandatory life sentence, as a result of his guilty plea. That is exactly what happened, after the prosecutorial appeal. Lyons knew the risk, and took his chances. The fact that he did not correctly assess whether or not he actually would be sentenced as an adult is wholly irrelevant. The plea was made voluntarily and with full knowledge of the possible consequences.
Moreover, AEDPA imposes additional constraints on the ability of federal courts to grant a writ of habeas coi'pus. In order to meet the gatekeeper standards set forth by AEDPA, 28 U.S.C. § 2254(d), for the issuance of writs of habeas corpus, the state court adjudication of the claim that is challenged on collateral appeal must have resulted in a decision that was “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” Ibid; Cone, 122 S.Ct. at 1850.
The Supreme Court held in Williams v. Taylor, 529 U.S. 362, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000), that “[t]he most important point is that an unreasonable application of federal law is different from an incorrect application of federal law.” Id. at 410, 120 S.Ct. 1495 (emphasis in original). A state-court decision can involve an unreasonable application of United States Supreme Court precedent in only two ways. First, a state court might identify the correct legal rule, but unreasonably apply it to the facts. Second, a state-court decision might unreasonably extend a Supreme Court legal precedent to a new context where it should not apply or unreasonably refuse to extend that principle to a new context where it should apply. Id. at 406, 120 S.Ct. 1495. The Supreme Court recently reiterated that the Williams objective unreasonableness standard applies to the ineffective assistance of counsel setting. Cone, 122 S.Ct. at 1850. In Cone, the Court held that when a state court correctly identifies the Strickland standard as applicable, and makes a reasonable judgment regarding attorney performance and prejudice under the correct standard, a federal court lacks the power to grant a writ of habeas corpus contravening that determination. Id. at 1853-54. I would not hold that the Michigan court was objectively unreasonable in its judgment, even were I to disagree with it, as the majority does.
Judge Gilman’s concurrence makes a number of interesting points that deserve a brief response. Contrary to the statement at page 601, I had thought that I did note, in the first paragraph of this dissent, and did not dispute the obvious facts in the record, that counsel did not inform their clients of the possibility of a government appeal. All of the counsel involved noted that they did not contemplate that such an appeal could occur or that it would be successful. In the parallel case before us, Miller’s counsel, as well as his client, was in court when the prosecutor said that the state would appeal, and specifically testified that he felt that the chances of such an appeal succeeding (one had never been tried in Michigan, to this point), were sufficiently low that it did not affect their judgment. No. 00-2150, 299 F.3d 570 (6th Cir.2002).
Judge Gilman’s interesting mathematical analysis at page 602 begins to address the correct inquiry, but does not go far enough. The relevant question is whether the ultimate advice was a strategy within the broad bounds of professional competence. Relatedly, we ask whether the decrease in the probability of a successful *605outcome due to prosecutorial appeal renders credible petitioners’ assertions that they would have given up their chance at a juvenile sentence before Judge Roberson. The one counsel who testified specifically thought that the chances of the state succeeding on appeal were fairly small, and we have no basis for discounting that judgment. Thus, if counsel calculated the chances of Judge Roberson’s sentencing the defendants as juveniles at 90% (when the reality turned out to be 100%), and believed the chance of success on appeal by the state to be 10%, contingent on the juvenile sentencing below, then the chances of ultimate success would have shrunk from 90% to 81%. This might still look extremely good, as opposed to a very large probability of conviction of first degree murder and a sentence of life without parole.
While my assigning of values is just as speculative as Judge Gilman’s, it does emphasize that the proper inquiry for effective assistance of counsel should be the overall strategy employed. It does not appear to me that we can properly say, under the AEDPA standard, that the state courts unreasonably applied Strickland.
Finally, I do not agree with Judge Gil-man that the failure of the state court to discuss the prejudice prong of Strickland means that it did not correctly identify and apply Strickland as the governing federal rule. But even if we were to parse the two parts of Strickland in this way, we would not conduct a full de novo review, but rather a “independent” review as laid out in our recent case of Harris v. Stovall, 212 F.3d 940, 943 (6th Cir.2000). There, we said, in a similar case: “That independent review, however, is not a full, de novo review of the claims, but remains deferential because the court cannot grant relief unless the state court’s result is not in keeping with the strictures of the AED-PA.” Ibid. Thus, even in conducting an independent review we must determine more than simply that we believe the state court was wrong in its ultimate outcome.
As the above analysis of Strickland, Williams, and Cone shows, we can by no means term the judgment of the Michigan Court of Appeals unreasonable. The judgment is perhaps debatable; if we ignore Strickland’s clear requirements, we might say it was incorrect; under no circumstances can we say it was unreasonable.