Michael Miller was convicted in Indiana of three counts of child molestation and sentenced to three consecutive 40-year terms. The sexual abuse, including anal intercourse, began when the victim was nine and continued for six years. When imposing the lengthy term (effectively life in prison), the state judge relied not only on the nature of Miller’s- conduct but also on his four prior convictions, his failure to reform after stretches of imprisonment, and the absence of any mitigating factors.
The convictions were affirmed on direct appeal, see Miller v. State, No. 34A02-0307-CR-563, 811 N.E.2d 496 (Ind.App. June 8, 2004). Miller then filed a collateral attáck, contending that his appellate lawyer furnished ineffective assistance by-failing to contest' the length of his sentence. (Miller made many other arguments, but all except the sentencing issue have been abandoned.) The state’s court of appeals eventually concluded that appellate counsel should have raised this issue, but that its omission did not result in prejudice under the standard of Strickland v. Washington, 466 U.S. 668, 694-96, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), and equivalent state decisions, which ask whether the petitioner, has demonstrated a “reasonable probability” that the outcome of the direct appeal would have been different.
The court observed that the substantive standard for appellate review in Indiana is whether the sentence is “inappropriate in light of the nature of the offense and the character of the offender.” Ind. App. R. 7(B). (Miller does not contend that his sentence violates the Cruel and Unusual Punishments Clause or any other rule of federal law.) After reviewing the evidence, the appellate court held “that Miller has not established that his 120-year aggregate sentence is inappropriate in light of the nature of the offense and the character of the offender.” Miller v. State, 2013 Ind.App. Unpub. Lexis 377 (Mar. 16, 2013) at *19.
This means that, if Miller’s appellate lawyer had contested the sentence, the argument would have failed on the merits. Because, in the state court’s view, the chance of success was zero, it necessarily followed that Miller had not shown a “reasonable probability” that a better appellate lawyer could have obtained a lower sentence for- him. A federal district judge then denied Miller’s petition for a writ of habeas corpus under 28 U.S.C, § 2254. Miller v. Zatecky, No. 1:13-cv-913-SEB-TAB (S.D.Ind. Mar. 26, 2015).
Because Indiana addressed on the merits the question whether Miller has established prejudice, we must decide whether the state’s decision “was contrary *277to, 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).
Miller recognizes that Indiana’s judiciary articulated the legal standard ¡the same way the Supreme Court does, so he contends that the state court’s decision was an “unreasonable application” of the governing standard. But, as far as we can see, the decision was not based on federal law at all, let alone federal law “clearly established” by the Supreme Court., It rests on a- conclusion that, as a matter of state law, it would have been futile, to contest the sentence’s length on appeal, because .a 120-year sentence is not “inappropriate in light of the nature of tljie offense and the character of the offender.” A federal court' cannot disagree with a state court’s resolution of an issue of state law. See, e.g., Bradshaw v. Richey, 546 U.S. 74, 126 S.Ct. 602, 163 L.Ed.2d 407 (2005); Estelle v. McGuire, 502 U.S. 62, 112 S.Ct. 475, 116 L.Ed.2d 385 (1991).
Shaw v. Wilson, 721 F.3d 908, 914-15 (7th Cir.2013), shows that a defendant may use ineffective-assistance doctrine to gain the benefit of state law when a lawyer’s error prevented the state judiciary from recognizing the force of a potential state-law defense (or other advantage secured by state law). Doing that does not use § 2254 to override a state court’s conclusion that state law does not provide the petitioner with' the benefit he sought. But that’s exactly what Miller needs. A lawyer’s blunder has not led Indiana’s judiciary to overlook-the possibility that Miller may have been entitled to a lower sentence. Instead-the state’s court of appeals addressed that subject directly and concluded that appellate review of his sentence in 2004 would have done him no good — as a matter of state law. That’s the sort of decision § 2254 leaves to the state judiciary.
Miller maintains that the state court’s decision was “unreasonable” because, when considering dispositions of similar cases, the court of appeals did not discuss any opinion issued after June 8, 2004, the date Miller’s direct appeal was decided. Later decisions, according to Miller, look more favorably on contentions that sentences in sex-offense cases are too long, and had the court of appeals used them as comparisons this would have demonstrated a “reasonable probability” of appellate success! But the court of appeals considered only decisions that were “available as precedent during Miller’s direct appeal” (Miller v. State at *17).
Section 2254(d)(1) does not ask whether a decision was reasonable in the abstract. It asks whether the state court reasonably applied rules clearly established by the Supreme Court of the United States. Miller does not point to any decision by that Court clearly establishing that a state must give a petitioner the benefit of state-law precedent that comes after his conviction’s finality. That would amount to saying, as a matter of federal law, that all state-law decisions must apply retroactively. That isn’t how federal courts apply their own decisions. See Welch v. United States, No. 15-6418 (U.S. Apr. 18, 2016) (explaining federal retroactivity doctrine). The Supreme Court has never questioned the states’ ability to choose whether to apply their own case law retroactively. State courts must apply federal decisions retroactively when federal doctrine so requires (if they entertain collateral attacks at all), see Montgomery v. Louisiana, — U.S. -, 136 S.Ct. 718, 727-32, 193 L.Ed.2d 599 (2016), but they are free to decide when their own doctrinal changes apply to cases that have become final.
*278And why should a state be required, as a matter of either state or federal law, to give Miller the benefit of decisions released after the conclusion of his direct appeal? Miller does not question the state judiciary’s conclusion that, if his lawyer had contested the length of his sentence in 2004, he would have lost. The goal of ineffective-assistance doctrine is to give criminal defendants the benefit of the counsel to which the Sixth Amendment says they are entitled. We know from Indiana’s decision that, even if appellate counsel had done exactly what Miller says counsel should have done, this would not have helped him. It would be weird to say that a defendant is better off having a lawyer who omits a losing issue than if that lawyer had performed exactly as a zealous and capable counsel should.
Nonetheless, Miller insists that Lockhart v. Fretwell, 506 U.S. 364, 113 S.Ct. 838, 122 L.Ed.2d 180 (1993), entitles him to the benefit of. hindsight. That isn’t what Lockhart holds. It did not concern § 2254(d)(1) — not only because it predates by three years the Antiterrorism and Effective Death Penalty Act, which rewrote § 2254 and insulated many state decisions that might have led to relief in earlier years, see Cullen v. Pinholster, 563 U.S. 170, 180-90, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011), but also because Lockhart does not deal with an asserted change of state law. It does not make defendants with poor lawyers better off than defendants with good ones.
Fretwell was convicted of a capital crime in Arkansas and sentenced to death. The jury instructions permitted the jurors to consider, as an aggravating factor, that the crime had been committed for pecuniary gain. A few months before Fretwell’s sentencing, a panel of the Eighth Circuit had held that the Eighth Amendment forbids a state to use as an aggravating factor any element of the substantive crime. Collins v. Lockhart, 754 F.2d 258 (8th Cir.1985). Fretwell’s lawyer did not ask the Arkansas judge to apply Collins. By the time Fret-well filed a collateral attack in federal court, Collins had been overruled as inconsistent with Lowenfield v. Phelps, 484 U.S. 231, 108 S.Ct. 546, 98 L.Ed.2d 568 (1988). See Perry v. Lockhart, 871 F.2d 1384 (8th Cir.1989). Another panel of the Eighth Circuit held that Fretwell was entitled to the benefit of the holding in Collins, even though that decision was wrong, but the Supreme Court reversed. It held that the Sixth Amendment does not entitle litigants to gain from judicial errors.
Miller concludes from this that anyone who advances an ineffective-assistance claim is entitled to the benefit of hindsight. Yet that was not the Court’s point. The case stands for the proposition that a person seeking federal collateral review must show that the state court committed an error of federal law — not that it took a step mistakenly (and temporarily) thought to- be an error, but that the state judiciary really was in error. Given Lowenfield and Perry, the instructions at Fretwell’s sentencing trial were constitutionally valid; he had no legitimate beef about them, and at any new sentencing hearing the instructions would have been repeated verbatim.
Later decisions have emphasized that Lockhart does not redefine Strickland’s “prejudice” component. See, e.g., Lafler v. Cooper, — U.S. -, 132 S.Ct. 1376, 1386-87, 182 L.Ed.2d 398 (2012); Glover v. United States, 531 U.S. 198, 202-04, 121 S.Ct. 696, 148 L.Ed.2d 604 (2001). Indeed, the Court has understood Lockhart as an anii-hindsight decision, warning against the “natural tendency to speculate as to whether a different ... strategy might have been more successful.” Maryland v. Kulbicki, — U.S. -, 136 S.Ct. 2, 4, — *279L.Ed.2d - (2015), quoting from Lockhart, 506 U.S. at 372, 113 S.Ct. 838.
Miller observes that our opinion in Shaw remarks in passing that “hindsight is permissible.” 721 F.3d at 918. That’s so in the sense that, under Lockhart, federal law must favor the petitioner when the collateral attack is resolved; as well as when the state’s decision became final. Nothing in Lockhart justifies any general resort to hindsight. Immediately after the langue we have quoted, our opinion in Shaw cites 506 U.S. at 372, 113 S.Ct. 838, where Lock-hart condemns resort to hindsight to disparage the performance of counsel. And we cannot forget that Strickland asks whether 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, 104 S.Ct. 2052. That has the same temporal focus as the state’s court of appeals in Miller’s case, asking whether a better performance by counsel could have affected the outcome then and there.
If Shaw meant more by its remark, it still cannot control the outcome in a pro:-ceeding under § 2254(d)(1).. That statute limits the federal courts’ role to- applying law clearly established by the Supreme Court of the United States. The Justices insist that district and circuit judges disregard their own decisions. See, e.g., Lopez v. Smith, — U.S. -, 135 S.Ct. 1, 4, 190 L.Ed.2d 1 (2014) (“Circuit precedent cannot ‘refine or sharpen a general principle of Supreme Court jurisprudence into a specific legal rule that this Court has not announced.’”); Marshall v. Rodgers, — U.S. -, 133 S.Ct. 1446, 1450, 185 L.Ed.2d 540 (2013). Lockhart itself does not “clearly establish” that state courts must consider state-law decisions that postdate counsel’s deficient performance; it has nothing to say about how state courts determine ■ the temporal scope of their own decisions. And Kulbicki holds federal law does not require the use of hindsight in ineffective-assistance cases.
Miller has not shown that the state judiciary made an error of federal law, so he is riot entitled to collateral relief.
Affirmed.