Gerald Winfield v. Stephanie Dorethy

In the United States Court of Appeals For the Seventh Circuit ____________________ Nos. 19-1441 & 19-1547 GERALD WINFIELD, Petitioner-Appellee, Cross-Appellant, v. STEPHANIE DORETHY, Warden, Respondent-Appellant, Cross-Appellee. ____________________ Appeals from the United States District Court for the Northern District of Illinois, Eastern Division. No. 1:10-cv-04878 — Sharon Johnson Coleman, Judge. ____________________ ARGUED JANUARY 7, 2020 — DECIDED APRIL 13, 2020 ____________________ Before BRENNAN, SCUDDER, and ST. EVE, Circuit Judges. ST. EVE, Circuit Judge. Gerald Winfield confessed to police that he shot Jarlon Garrett. Based on that confession, a judge on the Circuit Court of Cook County convicted Winfield of attempted murder. Winfield was also accused of killing Dominick Stovall in the same shooting, but the trial judge ac- quitted him of that charge because no credible witness had placed Winfield at the scene of the crime and his confession did not mention Stovall. The judge rejected Winfield’s 2 Nos. 19-1441 & 19-1547 argument that his confession had been coerced, as well as his half-hearted alibi defense, and sentenced him to thirty years’ imprisonment. In his direct appeal, Winfield’s new counsel raised one unsuccessful argument—that the judge had abused his discretion at sentencing. These appeals require us to consider the performance of Winfield’s trial and appellate counsel. The Illinois state courts, on post-conviction review, concluded that trial counsel’s presentation of Winfield’s alibi was not so deficient that it vi- olated the Constitution, but they did not address the perfor- mance of appellate counsel to any meaningful degree. The district court, therefore, applied the stringent and deferential standard of the Antiterrorism and Effective Death Penalty Act (AEDPA), 28 U.S.C. § 2254(d), to Winfield’s claim that he re- ceived ineffective assistance of trial counsel and denied that part of his petition for writ of habeas corpus. On the matter of appellate counsel, the district court concluded that AEDPA did not apply because the claim had not been “adjudicated on the merits in State court,” id., but had instead been over- looked. It considered the claim without any deference to the state courts’ denial of relief. Through that lens, and although it believed it to be a close case, the court found appellate coun- sel had rendered ineffective assistance by omitting an argu- ment that there was insufficient evidence to convict because Winfield’s confession was uncorroborated. Both parties have appealed. The state argues that the dis- trict court erred in granting relief on the appellate counsel claim; Winfield contends that the court erred in denying relief on the trial counsel claim. We affirm the judgment in part and reverse it in part, as we conclude that Winfield is not entitled to habeas corpus relief under either theory. Nos. 19-1441 & 19-1547 3 I. Background A. Trial and Direct Appeal On the afternoon of October 17, 1998, Garrett and Stovall were walking on West Huron Street, in Chicago, when some- one shot at them. Stovall was killed and Garrett, injured, though who shot them—or even how many people shot them—was not immediately clear. Police arrested Winfield as the primary suspect. He ultimately was indicted for the mur- der of Stovall, the attempted murder of Garrett, and lesser in- cluded offenses. He proceeded to a bench trial. The state’s first eyewitness—and “[t]he only real credible witness” according to the trial judge—was Lonnie Hartman. Hartman had been sitting in his car, facing north, when he saw a black SUV traveling southbound on North Central Park Avenue. The SUV stopped at the corner of Huron, and a man exited the vehicle’s passenger side. The man shot at Stovall and Garrett before reentering the car and driving off. Hart- man could not identify the shooter, though he described him as at least six feet tall and about 25 years old. Winfield was only 5ʹ7ʺ and 19 at the time of the shooting. Consistent with that difference, Hartman denied that Winfield was the shooter he had seen, though he declined to rule out the possibility that there were other shooters he could not see. In contrast to his assessment of Hartman, the trial judge described the testimony of the state’s second witness, Lorenzo Curry, as “worthless.” Curry said he knew Winfield, Garrett, and Stovall from playing dice with the three of them about a month before the shooting. At that dice game, Winfield’s brother, Terrance, had bumped into Stovall’s arm, causing him to drop the dice into a bad roll and leading to a fight. On 4 Nos. 19-1441 & 19-1547 the day of the shooting, Curry was sitting on a porch near the intersection of Huron and Central Park. He, too, testified that he saw a black SUV stop at the corner, but he saw at least two shooters exit: Winfield and another man. Both men returned to the car after the shooting. Few details of this testimony matched what Curry had told police or the grand jury, so the trial judge noted his “extensive impeachment,” among other credibility problems. The trial judge described the testimony of the last witness, Garrett, the surviving victim, as “not much better” than Curry’s. Garrett spoke about the dice game in terms like Curry’s, and said he also saw a black SUV pull up at the cor- ner of Huron and Central Park before Winfield emerged, be- gan shooting, and hit Garrett twice, in the hip and arm. The trial judge did not credit Garrett’s identification largely be- cause he had told police after the shooting that he did not know the shooter, who he said had returned to the car. Weeks after the crime, Garrett and Curry both identified Winfield through a photo array and line-up conducted by a detective. The detective testified to these identifications, as well as a custodial statement that Winfield gave. An Assistant State’s Attorney had written down the statement, Winfield had signed it, and at trial, the detective read it into the record. In the statement, Winfield explained that Terrance had bumped into Stovall at the dice game, leading to a fight. Win- field thought he and Stovall had mended their relationship until a few weeks later, when members of Garrett and Stovall’s gang beat Terrance. By the day of the shooting, he had started carrying a gun because Garrett had shot at him recently. When members of his gang offered him a ride in a black SUV, Winfield joined them as they drove southbound Nos. 19-1441 & 19-1547 5 on Central Park until the car stopped at the corner of Huron. There, Winfield spotted Garrett and Stovall. He said that he then exited the vehicle and started shooting at Garrett (with- out mentioning Stovall) while another person fired in the same direction. Contrary to all witnesses’ testimony, though, he stated that the car drove off without him, so he had to flee on foot. Winfield’s statement also asserted he had voluntarily given it based on no promises, and that the detective and ASA had treated him well. He testified otherwise at trial and said he had been abused and the ASA had promised that they would release him if he signed some papers (which he did without reading them). The detective and ASA denied these assertions. When his trial counsel asked him where he was on the day of the shooting, Winfield responded, “I probably say I was at home,” and denied being involved in a shooting. Counsel did not follow up on this testimony and no other evidence was introduced to support it. Winfield otherwise confirmed through his testimony that he and Terrance had argued with Stovall at the dice game. As the trial judge summed the evidence up, “no credible witness” placed Winfield at the scene. Given this gap and the open question of how many shooters there were, the judge had doubts that Winfield was responsible for Stovall’s mur- der, either as the killer or on an accountability theory. So, he acquitted Winfield of all charges relating to Stovall. Garrett’s attempted murder was different, however, and the difference was Winfield’s confession. As the judge put it, the lack of a credible witness was not a problem because he 6 Nos. 19-1441 & 19-1547 had Winfield’s “assurance he was present.” The judge de- clined to find that law enforcement had coerced his confes- sion, reasoning that Winfield was too intelligent to believe that police would let him go just for signing papers. The judge framed the remaining question as whether there was enough evidence to convict based on the statement, “or stated differently, whether the State has proven the corpus de- licti of any crimes that this statement confesses.” The corpus delicti of an offense is simply proof apart from a confession that a crime occurred. See United States v. Kerley, 838 F.2d 932, 939 (7th Cir. 1988). Under common law, proof of the corpus delicti was required for conviction based on a con- fession, but that rule no longer holds weight in federal courts or those of many states. Id. at 940. In Illinois, however, the common law rule maintains its vitality. See People v. Lara, 983 N.E.2d 959, 964 (Ill. 2012), as modified on denial of rehʹg (2013). The trial judge understood the corpus delicti rule to re- quire that the elements of attempted first-degree murder— that Winfield “intended unjustifiably to take a life”—be proved with evidence “outside of the confession of the de- fendant or aliunde the defendant’s confession.” The judge saw this proof in the mere fact that Garrett had been shot twice. The judge continued, however, and found the neces- sary intent also in the statement itself, which described Win- field’s conflict with Garrett that led him to carry a gun. From these admissions, the judge said, “no other conclusion can be drawn but that he intended to kill him.” Accordingly, the judge convicted Winfield of attempted murder. At sentencing a few months later, the trial judge opined at length about sentencing theories and described rehabilitation Nos. 19-1441 & 19-1547 7 as “almost a joke.” Still, he emphasized, he was “fully aware” that he had “to take into consideration the rehabilitative po- tential of the defendant.” The judge, however, thought this potential rested “somewhere between nil and zero” and sen- tenced Winfield to 30 years’ imprisonment. Winfield obtained new appellate counsel, who challenged only the judge’s sentencing decision. He contended that the judge had not weighed Winfield’s potential for rehabilitation. The Illinois Appellate Court rejected this argument, pointing to the judge’s express statements that he had considered Win- field’s rehabilitative potential and simply found it lacking. The Illinois Supreme Court denied leave to appeal. B. State Post-Conviction Review Winfield soon filed a petition for post-conviction review, alleging he had been deprived of the effective assistance of appellate counsel in violation of the Sixth and Fourteenth Amendments. In particular, he criticized counsel for not rais- ing an argument based on the sufficiency of the evidence or the corpus delicti. After some procedural hiccups (including a mistaken dismissal), Winfield filed an amended petition raising three new claims relating to both his trial and appel- late counsel. He alleged trial counsel was ineffective for not moving to quash his arrest warrant and suppress his confes- sion, and for failing to interview, investigate, or call alibi wit- nesses—his family. Regarding appellate counsel, his amended petition claimed counsel had overlooked “obvious and significant issues … such as, sufficiency of evidence.” Winfield attached to this amended petition affidavits from his family members, attesting that on the date of the shooting Winfield was sleeping at their shared home. He also attached 8 Nos. 19-1441 & 19-1547 his own affidavits, though neither related to his alibi. One at- tested to the truth of the contents of his first petition (which, again, addressed only appellate counsel), and the other said that he had told trial counsel his confession had been coerced. The Circuit Court of Cook County held an evidentiary hearing solely on the claim that trial counsel failed to investi- gate alibi witnesses. Winfield himself did not testify at the hearing, but his family members did. His aunt recalled the day of the shooting because it was Sweetest Day. She had spent the day preparing her daughter for a homecoming dance and saw Winfield sleeping in his room until she left for the dance that evening. She told the court that she had given Winfield’s trial counsel this information (though she could not recall when), but counsel had not asked her to testify. Winfield’s mother offered a similar story. She was helping her niece get ready for the dance and saw Winfield asleep at the time of the shooting. She testified that she told this to trial counsel shortly after retaining him, but counsel “didn’t re- spond well with it” and told her “he was going to do it the way he wanted to.” The court next heard from Winfield’s trial counsel. As this hearing took place about seven years after the trial, counsel could not remember many details about his preparation for the case. When the state asked counsel if Winfield had ever told him of an alibi, counsel responded that Winfield had al- ways denied he did the shooting: “the problem was he did mention an alibi”—a different one. Winfield had thought he might have been at a store at the time of the shooting. Counsel investigated that story but found it wanting. Winfield’s family had produced a receipt with a printed time that did not line up with the shooting. Counsel could not recall if he spoke Nos. 19-1441 & 19-1547 9 with Winfield’s aunt, but he “specifically” did not remember speaking with Winfield’s mother. Had he been made aware of a verifiable alibi, he explained, he would have pursued it, even post-trial. The Circuit Court denied Winfield’s petition for post-con- viction relief. Applying the two-pronged analysis of Strickland v. Washington, 466 U.S. 668 (1984), the court concluded that Winfield had not demonstrated deficient performance of trial counsel and did not reach the second prong, whether he was prejudiced. The court found that neither Winfield nor his fam- ily had told counsel of his home alibi, and so counsel could not have been ineffective for failing to present witnesses to support that story. Even assuming counsel had been aware of Winfield’s alibi and the witnesses, the court continued, it nonetheless would have found that the decision not to call im- peachable family members was reasonable. Moving on to appellate counsel, the Circuit Court mistak- enly framed Winfield’s claim to be “that appellate counsel was ineffective for not raising the issue of trial counsel’s own incompetence.” The omission of this argument was neither deficient nor prejudicial, the court said, because “the under- lying claims of ineffectiveness lack support.” The court did not mention sufficiency of the evidence or any other appellate argument independent of trial counsel. Winfield appealed the post-conviction decision raising what he identified as a single issue: whether the Circuit Court had erred in finding trial counsel was not ineffective. Never- theless, he maintained that appellate counsel had overlooked “significant issues, such as, sufficiency of evidence.” 10 Nos. 19-1441 & 19-1547 The Illinois Appellate Court upheld the Circuit Court’s findings that Winfield had simply never informed trial coun- sel he was at home at the time of the shooting, and that, even if he had, it was reasonable not to call family members as wit- nesses. It further found no prejudice in light of “the over- whelming evidence” of Winfield’s guilt, namely “his inculpa- tory statement, and the two eyewitness identifications of him as the shooter.” On the issue of appellate counsel, the Appel- late Court treated Winfield’s claim as wholly derivative of the claim regarding trial counsel, just as the Circuit Court had. Thus, both arguments failed together. The Illinois Supreme Court again denied leave to appeal. C. District Court and Prior Appeal At this point, Winfield proceeded to federal court seeking a writ of habeas corpus. His petition continued his two real claims: first, trial counsel was ineffective for failing to present alibi witnesses, and second, appellate counsel was ineffective for failing to challenge the sufficiency of the evidence. The state answered the petition and asserted that, alt- hough deference was owed to the state court’s resolution of the first claim under AEDPA, the second had neither been ad- judicated on the merits in state court, nor procedurally de- faulted, and so the district court was tasked with reviewing that claim de novo. Accepting this concession, the court per- mitted discovery, including a deposition of appellate counsel. Appellate counsel, like trial counsel before him, had lim- ited memory of his case preparation, which, at that point, was nearly a decade and a half ago. He based most of his recollec- tion on what he had written in the brief itself, as he thought it accurately reflected what was in his mind at the time. He Nos. 19-1441 & 19-1547 11 suspected that an argument based on the sufficiency of the evidence would have failed considering the trial judge’s find- ing that the confession was voluntary. Moreover, he noted that there was “physical evidence supporting the confession because Mr. Garrett … was, in fact, shot.” And though the wit- nesses disagreed on the details, counsel thought there was some corroborating statements in that they all testified to see- ing the crime occur. Altogether, he suspected he had believed at the time that there existed only a remote possibility of over- turning the conviction based on sufficiency of the evidence. Winfield thought otherwise and pressed his claim that ap- pellate counsel was ineffective for not challenging the suffi- ciency of the evidence. The thrust of his argument was that the Illinois courts were searching in their review of convic- tions and would have acquitted him on these facts, even if a federal court would not have. Winfield further contended that a reasonable attorney would have challenged the state’s proof of the corpus delicti. The record, in his view, did not contain evidence that anyone had intended to kill Garrett (as opposed to his merely being hit by two stray bullets). Moreover, he posited that his confession was uncorroborated, because no one had asked Garrett about the time he shot at Winfield, sup- posedly prompting Winfield to carry a gun. The district court concluded that appellate counsel was in- effective for not raising a corpus delicti argument, though not the same one Winfield pressed. It summarized Illinois’s cor- pus delicti rule as requiring the state to both establish that a crime occurred and corroborate the confession. The court ac- cepted the trial judge’s logic, contrary to Winfield’s argument, that the mere fact that Garrett was shot twice was proof that an attempted murder occurred. But, the court concluded, the 12 Nos. 19-1441 & 19-1547 evidence did not corroborate the confession and connect Win- field to that attempt. Hartman—the sole credible witness— had testified that there was one shooter, not two; this shooter was not Winfield; and the shooter left the scene by car (not on foot, as Winfield said he had). Though the court accepted that “tangential components” of Winfield’s statement had been corroborated, it concluded there was a reasonable probability that this was not enough under Illinois law and granted the petition. The state moved to amend the judgment, arguing that the district court had misconstrued Illinois’s corpus delicti rule, which it said requires proof only that a crime occurred before the state can use a confession to convict. The court acknowl- edged it was a close case, with no clear rule and authorities going each way, but was convinced its result was still right. The state also tried to walk back its concession that AEDPA deference did not apply to the appellate counsel claim, but the district court deemed the argument waived. On the state’s appeal, we reversed that determination and re- manded after concluding that the standard of review owed under AEDPA is not waivable. Winfield v. Dorethy (Winfield I), 871 F.3d 555, 560 (7th Cir. 2017). On remand, Winfield maintained that AEDPA did not ap- ply to his appellate counsel claim. Regardless of the state’s earlier concession, he argued that both state courts had in fact overlooked his arguments about appellate counsel and not adjudicated them on the merits. He also revived his trial coun- sel claim, contending he was entitled to relief despite AEDPA. The district court denied the petition insofar as it sought a new trial based on ineffective assistance of trial counsel. The Nos. 19-1441 & 19-1547 13 court deferred to the Illinois courts’ finding that Winfield had never told counsel that he was home at the time of the shoot- ing and that one factual finding defeated Winfield’s claim. Moving back to appellate counsel, the district court agreed that the state courts had not adjudicated the claim on the mer- its. It found that this was the rare case where a petitioner had rebutted the strong presumption that state courts adjudicate all claims on their merits. Both state courts had “inadvertently overlooked” Winfield’s appellate counsel claim, and so the district court concluded it would still review the claim de novo. It reached the same result it had before and ordered the state to reopen Winfield’s appeal. The state appealed the grant of relief. The district court granted Winfield a certificate of appealability on the denial of his trial counsel claim, and he cross-appealed. II. Standard of Review We start by laying down some groundwork applicable to both appeals before diving into the specifics of each one. We review de novo the district court’s decisions on a petition for writ of habeas corpus. Schmidt v. Foster, 911 F.3d 469, 476 (7th Cir. 2018) (en banc), cert. denied, 140 S. Ct. 96 (2019). Under AEDPA, a federal court can grant a petition for writ of habeas corpus after a state-court adjudication of the merits only if that adjudication “(1) resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Su- preme Court of the United States; or (2) resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceed- ing.” 28 U.S.C. § 2254(d)(1)–(2); Schmidt, 911 F.3d at 476. This 14 Nos. 19-1441 & 19-1547 is a high standard. Unreasonable means more than incorrect. Rather, the state court’s ruling must be “so lacking in justifi- cation that there was an error … beyond any possibility of fairminded disagreement.” Burt v. Titlow, 571 U.S. 12, 19–20 (2013) (alteration in original) (quoting Harrington v. Richter, 562 U.S. 86, 103 (2011)). To be subject to that deference, though, the claims must have been “adjudicated on the mer- its” in state court. 28 U.S.C. § 2254(d); Cone v. Bell, 556 U.S. 449, 472 (2009). Strickland v. Washington, 466 U.S. 668 (1984), provides the clearly established federal law for both of Winfield’s claims. A petitioner raising a Strickland claim is required to demon- strate two things. First, he must show that counsel provided constitutionally deficient performance, meaning counsel made errors so serious he “was not functioning as the ‘coun- sel’ guaranteed the defendant by the Sixth Amendment.” Id. at 687. In assessing counsel’s performance, courts are ex- pected to “indulge a strong presumption that counsel’s con- duct falls within the wide range of reasonable professional as- sistance.” Id. at 689. Under AEDPA, a federal court must be “doubly deferential” and give a deferential look at counsel’s performance through the “deferential lens” of AEDPA. Cullen v. Pinholster, 563 U.S. 170, 190 (2011) (quoting Knowles v. Mir- zayance, 556 U.S. 111, 121 n.2 (2009)). Second, the petitioner must show that this deficient performance prejudiced his de- fense—meaning there is a “reasonable probability that, but for counsel’s unprofessional errors, the result of the proceed- ing would have been different.” Strickland, 466 U.S. at 694. Failing to prove either element defeats a petitioner’s claim. Id. at 697. Nos. 19-1441 & 19-1547 15 III. Winfield’s Appeal – Trial Counsel We now turn to Winfield’s appeal. Winfield concedes that AEDPA applies to his claim of ineffective assistance of trial counsel. We conclude that he cannot overcome the double layer of deference we owe to the state court’s finding that counsel performed reasonably. Winfield fails to identify any basis to reject the state court’s presumptively valid factual finding that Winfield never told counsel that he was at home at the time of the shooting. Winfield asserts that the state court’s decision was “based on an unreasonable determination of the facts.” 28 U.S.C. § 2254(d)(2). We start with the presumption that a state court’s determination of a fact is correct. 28 U.S.C. § 2254(e)(1). To re- but that presumption, a petitioner must show that the finding was unreasonable by clear and convincing evidence. Id.; Ja- nusiak v. Cooper, 937 F.3d 880, 888 (7th Cir. 2019). A finding cannot be unreasonable if “reasonable minds reviewing the record might disagree about the finding in question.” Brum- field v. Cain, 135 S. Ct. 2269, 2277 (2015). The evidence before the Circuit Court presented a classic credibility dispute, on whose resolution reasonable minds could differ and which we have “no license” to disturb. Mar- shall v. Lonberger, 459 U.S. 422, 434 (1983). Winfield’s aunt and mother said they told trial counsel they could testify regard- ing Winfield’s whereabouts at the time of the shooting. Coun- sel, however, “specifically” denied any recollection of talking to Winfield’s mother and could not remember if he spoke with Winfield’s aunt. Counsel also said that he would have presented a verifiable alibi, had anyone made him aware of one. The Circuit Court evidently believed counsel over 16 Nos. 19-1441 & 19-1547 Winfield’s family and inferred from the fact that counsel did not present the home alibi that he was not aware of it. Winfield has pointed to no clear and convincing evidence that it was unreasonable to credit counsel’s testimony. When questioned whether Winfield told him of an alibi, counsel an- swered that “the problem was he did mention an alibi”—“the alibi” was that he was at a store. Counsel investigated this story with the help of Winfield’s family, and found it wanting. Counsel was further asked to confirm only that he had spoken to Winfield about his trial testimony before he gave it. From this, Winfield infers that counsel must have known of the alibi that he testified to at trial. Even assuming Winfield told coun- sel exactly what he was going to say at trial, what Winfield said during his trial testimony was indefinite: “I probably say I was at home.” Winfield himself never offered anything more than this equivocal statement; he neither testified at the post- conviction hearing nor submitted an affidavit about his dis- cussions with counsel. It was not unreasonable for the state courts to infer that, when counsel spoke of “the alibi” he in- vestigated and “an alibi” that Winfield mentioned, he had in- vestigated the only verifiable alibi Winfield had offered. Winfield emphasizes that trial counsel admitted to not re- membering his trial preparation and “never claimed that Winfield definitively ruled out being at home.” True, but that is far from enough to rebut AEDPA deference. At best, Win- field has identified a gap in the record. The absence of evi- dence, however, cannot overcome either of the layers of def- erence we owe to the state court’s adjudication of counsel’s performance under AEDPA and Strickland. Titlow, 571 U.S. at 22–23. Without clear and convincing evidence otherwise, it was not unreasonable for the state courts to find, as a matter Nos. 19-1441 & 19-1547 17 of fact, that Winfield never told counsel he was home at the time of the shooting and that counsel was unaware he had witnesses who could testify so. From that finding, the state courts reasonably concluded that counsel’s failure to investigate or present Winfield’s alibi was not constitutionally deficient performance. The Supreme Court recognized in Strickland itself that the scope of an attor- ney’s duty to investigate defenses is informed by the “infor- mation supplied by the defendant.” 466 U.S. at 691. If Win- field told counsel only that he was at a store, and counsel checked that alibi and found it wanting, then the Illinois Ap- pellate Court could reasonably conclude that the Constitution mandates nothing more. Cf. Morris v. Bartow, 832 F.3d 705, 711 (7th Cir. 2016) (concluding Strickland was not applied unrea- sonably when state courts found defendant had not claimed his plea was coerced). Winfield does not argue otherwise. We need not address the state court’s other findings—that counsel properly refused to call familial witnesses (assuming he knew of them) and that Winfield was not prejudiced. We affirm the denial of relief on the trial counsel claim. IV. The State’s Appeal That brings us to the state’s appeal. The state contends that the district court erred in issuing a writ of habeas corpus on Winfield’s claim that he received ineffective assistance of ap- pellate counsel. It principally argues that the district court erred in reviewing the claim de novo and not with the defer- ence provided by AEDPA. Alternatively, it contends that even with plenary review, Winfield has not established the ele- ments of a Strickland claim. 18 Nos. 19-1441 & 19-1547 We agree with this latter argument. Even if we owe no def- erence to the state court’s determination of the claim, we would reverse the district court’s judgment granting habeas relief to Winfield. We start by briefly explaining why we are assuming without deciding that the state court is not owed AEDPA deference before addressing the claim on its merits. A. AEDPA As in our previous appeal in this case, the primary dispute between the parties is whether the deferential view of AEDPA applies to this claim. See Winfield I, 871 F.3d at 560. That ques- tion depends, as we have noted, on whether the claim was “adjudicated on the merits.” 28 U.S.C. § 2254(d). The Supreme Court has instructed that “[w]hen a federal claim has been presented to a state court and the state court has denied relief, it may be presumed that the state court ad- judicated the claim on the merits in the absence of any indica- tion or state-law procedural principles to the contrary.” Rich- ter, 562 U.S. at 99. In Richter, the state court gave no reasons for denying relief, but the Court has extended this presump- tion to state courts that have addressed some but not all claims. See Johnson v. Williams, 568 U.S. 289, 300–01 (2013). The presumption is rebuttable, if a petitioner provides “reason to think some other explanation for the state court’s decision is more likely.” Richter, 562 U.S. at 99–100. One such situation is when “the evidence leads very clearly to the con- clusion that a federal claim was inadvertently overlooked in state court.” Williams, 568 U.S. at 303; Sarfraz v. Smith, 885 F.3d 1029, 1036 (7th Cir. 2018). The district court determined that this is precisely what happened here. Though Winfield consistently asserted that Nos. 19-1441 & 19-1547 19 appellate counsel had failed to raise “significant issues, such as, sufficiency of evidence,” the Illinois Appellate Court never mentioned those issues or considered appellate counsel’s per- formance in any way separate from trial counsel’s. The state raises several reasons why the district court’s analysis was inconsistent with the presumption that state courts adjudicate claims on their merits. For our limited pur- poses here, we need address only its strongest. The Appellate Court found that overwhelming evidence eliminated any prejudice from trial counsel’s failure to present alibi wit- nesses. Giving the “benefit of the doubt” to the state courts, as we are required, Brady v. Pfister, 711 F.3d 818, 826 (7th Cir. 2013), we agree that this finding implicitly means the Illinois Appellate Court thought a sufficiency of the evidence argu- ment would have been frivolous for appellate counsel to raise.1 The logic is simple. If one believes it improbable that a fact- finder would acquit with new, favorable evidence, then, nec- essarily, one could not also believe it possible that the same factfinder was obligated to acquit without that evidence. Moreover, counsel could neither have performed deficiently nor prejudiced Winfield by failing to raise a doomed argu- ment. See McNary v. Lemke, 708 F.3d 905, 921 (7th Cir. 2013); see also Howard v. Gramley, 225 F.3d 784, 790 (7th Cir. 2000) 1 Winfield asserts that the state forfeited this argument in this court by waiting until its reply brief to raise it (although it had raised it in the pre- vious appeal and in the district court). We explained in the last appeal, however, that “the actions of the state courts essentially speak for them- selves, rather than depend upon characterizations by the stateʹs prosecut- ing arm.” Winfield I, 871 F.3d at 561. We defer to the state court’s adjudica- tion, not to the state’s interpretation of that adjudication. 20 Nos. 19-1441 & 19-1547 (recognizing overlap between the two Strickland prongs in claims of ineffective appellate counsel). Thus, the Illinois Appellate Court’s finding that there was “overwhelming evidence” resolved both prongs of Winfield’s appellate counsel claim. Having found the state court’s “spe- cific reasons” for denying relief, the next question is whether that explanation was reasonable thereby requiring our defer- ence. Wilson v. Sellers, 138 S. Ct. 1188, 1192 (2018). The Appellate Court pointed to Winfield’s “inculpatory statement and the two eyewitness identifications of him as the shooter.” The trial judge, however discredited both of those eyewitness identifications and found that no witness placed Winfield at the scene of the crime. We are hesitant to say, in this case, that Winfield’s statement alone can reasonably be deemed “overwhelming”—as opposed to merely sufficient— evidence of guilt. Cf. Harris v. Thompson, 698 F.3d 609, 630–31 (7th Cir. 2012) (rejecting state’s argument that confession is overwhelming evidence of guilt for harmless-error purposes). The state insists that the judge did not fully discredit Gar- rett. We question that, considering the judge’s explicit com- ment that no evidence placed Winfield at the scene, but even the state agrees that Curry’s testimony is out. Reducing the eyewitness identifications from two to one might be enough to make the Appellate Court’s decision unreasonable. See Wig- gins v. Smith, 539 U.S. 510, 528 (2003) (recognizing that partial reliance on an erroneous factual finding can be unreasonable); Ben-Yisrayl v. Buss, 540 F.3d 542, 550 (7th Cir. 2008) (same). Although we could remand again for the district court to decide these issues (and other criticisms Winfield has levied against the “overwhelming evidence” finding), we see little Nos. 19-1441 & 19-1547 21 point in doing so. Even if Winfield can get around the defer- ence of 28 U.S.C. § 2254(d), he still must prove that he is “in custody in violation of the Constitution or laws or treaties of the United States” under § 2254(a). See Mosley v. Atchison, 689 F.3d 838, 853 (7th Cir. 2012); Ben-Yisrayl, 540 F.3d at 550; see also Adorno v. Melvin, 876 F.3d 917, 919 (7th Cir. 2017) (denying habeas relief on de novo review, assuming petitioner could rebut Richter presumption on remand).2 We therefore accept, for the sake of argument, that the dis- trict court properly reviewed Winfield’s appellate counsel claim de novo. The court nevertheless erred in granting relief. B. Appellate Counsel To prove that he is entitled to a writ of habeas corpus, Win- field must show that his appellate counsel overlooked an ar- gument that was both “’obvious and clearly stronger’ than is- sues that appellate counsel did raise.” Walker v. Griffin, 835 F.3d 705, 709 (7th Cir. 2016) (quoting Sanders v. Cotton, 398 F.3d 572, 585 (7th Cir. 2005)); see also Smith v. Robbins, 528 U.S. 259, 285, 288 (2000) (describing this standard as an application of Strickland). The issue that counsel overlooked can be one purely of state law and we may still resolve the federal con- stitutional question whether counsel’s omission was both de- ficient and prejudicial under Strickland. See Shaw v. Wilson, 721 F.3d 908, 914–15 (7th Cir. 2013). 2 We have said that a petitioner is not entitled to de novo review “simply because the state court’s rationale is unsound.” Whatley v. Zatecky, 833 F.3d 762, 775 (7th Cir. 2016) (citing Brady, 711 F.3d at 827). Instead, we defer to the state court’s judgment (notwithstanding its reasons). Id. We need not decide if or how this standard might apply here. Winfield is not entitled to relief even under de novo review. 22 Nos. 19-1441 & 19-1547 Winfield identifies two arguments that, he says, meet this hurdle. First, he raises a “general” sufficiency argument— there was simply not enough evidence to convict him of at- tempted murder. Second, he asserts the corpus delicti argu- ment—the evidence did not corroborate his confession. (We call the first argument general because a corpus delicti argu- ment is a type of sufficiency argument. See, e.g., Lara, 983 N.E.2d at 969; People v. Harris, 776 N.E.2d 743, 752–54 (Ill. App. Ct. 2002) (explaining how the two arguments interact).) 1. General Sufficiency of the Evidence Winfield continues to pursue the general sufficiency argu- ment, though that is not the basis on which he obtained relief. Under federal law and the Due Process Clause of the Four- teenth Amendment, there is sufficient evidence to convict when “after viewing the evidence in the light most favorable to the prosecution, any rational trier of fact could have found the essential elements of the crime beyond a reasonable doubt.” Jackson v. Virginia, 443 U.S. 307, 319 (1979). We have described this Jackson standard as a “nearly insurmountable hurdle.” United States v. Torres-Chavez, 744 F.3d 988, 993 (7th Cir. 2014). Winfield recognizes that we would almost cer- tainly uphold his conviction on direct appeal. See, e.g., United States v. Curtis, 324 F.3d 501, 507 (7th Cir. 2003) (affirming murder conviction based on defendant’s admissions and the victim’s “bullet-riddled body”). He insists, however, that Illinois courts do not apply the same standards as federal courts, but instead a more generous one that would lead to acquittal here. In support of this argu- ment, he relies on two decisions in which the Illinois Appel- late Courts vacated convictions despite the defendants’ Nos. 19-1441 & 19-1547 23 admissions: People v. Rivera, 962 N.E.2d 53 (Ill. App. Ct. 2011), and People v. Rodriguez, 728 N.E.2d 695 (Ill. App. Ct. 2000). The first and most obvious problem with this argument is that both cases cited Jackson as the governing standard. Rivera, 962 N.E.2d at 60; Rodriguez, 728 N.E.2d at 707. That makes sense, as the Illinois Supreme Court adopted Jackson in People v. Young, 538 N.E.2d 461, 473 (Ill. 1989), and continues to fol- low it, e.g., People v. Drake, 131 N.E.3d 555, 561 (Ill. 2019). We have not found a single case in which an Illinois court has said its own review is more searching or defendant-friendly than Jackson requires. All Winfield directs us to is language in Rivera, Rodriguez, and a great many Illinois decisions, that “[a] conviction will be reversed where the evidence is so unreasonable, improba- ble, or unsatisfactory that there remains a reasonable doubt of defendantʹs guilt.” People v. Washington, 969 N.E.2d 349, 355 (Ill. 2012). Though the United States Supreme Court does not use this precise formulation, we do not see how it is anything more than a rephrasing of the holding in Jackson that “a properly instructed jury may occasionally convict even when it can be said that no rational trier of fact could find guilt be- yond a reasonable doubt.” 443 U.S. at 317. That seems to be how the Illinois Supreme Court understands it—even when it relies on that language to vacate a conviction, it still cites Jack- son as the governing standard. See People v. Smith, 708 N.E.2d 365, 369–70 (Ill. 1999). We have even said that this formulation was neither “contrary to” Jackson nor “meaningfully differ- ent” from it. Cabrera v. Hinsley, 324 F.3d 527, 533 (7th Cir. 2003). Courts sometimes vacate convictions despite the high hurdle of Jackson. See, e.g., United States v. Garcia, 919 F.3d 489, 24 Nos. 19-1441 & 19-1547 496 (7th Cir. 2019). That they do so is no reason to believe that they were, in fact, applying a different, less onerous standard. Factually, Rivera and Rodriguez do nothing to demonstrate that a general sufficiency argument would succeed for Win- field. Rivera involved a unique situation in which a series of acquittals and conclusive DNA evidence left the state with only a convoluted and improbable theory on which to sustain its conviction. 962 N.E.2d at 63. There is nothing so improba- ble about the state’s theory here. And although Rodriguez in- volved admissions, they were of facts that the state posited only the murderer would know, not a confession to the crime, like Winfield’s. 728 N.E.2d at 709. The district court was right to skip over the general suffi- ciency argument. Winfield all but concedes there was not a reasonable probability the argument would succeed under the Jackson standard, and that is the standard the Illinois courts apply. It follows that counsel neither performed defi- ciently nor prejudiced Winfield by omitting this argument. 2. Corpus Delicti We next turn to the argument on which the district court granted relief: corpus delicti. As noted above, a corpus delicti argument is a type of sufficiency argument. E.g., Lara, 983 N.E.2d at 969. The key difference is that Illinois courts strin- gently (and intentionally) enforce the corpus delicti rule to as- sess confession evidence more carefully than Jackson alone de- mands. See id. at 982 (Thomas, J., concurring) (recognizing that the federal courts and several state courts have aban- doned corpus delicti rule); People v. Sargent, 940 N.E.2d 1045, 1057 (Ill. 2010) (rejecting request to abandon rule). Nos. 19-1441 & 19-1547 25 The corpus delicti is, traditionally, nothing more than the commission of a crime. See Kerley, 838 F.2d at 939; Lara, 983 N.E.2d at 964. It encompasses both the existence of an injury and the fact that the injury had a criminal cause. People v. Furby, 563 N.E.2d 421, 425 (Ill. 1990); 1 Kenneth S. Broun et al., McCormick on Evidence § 146 (8th ed. 2020). Under some for- mulations of the common law, a conviction based on a con- fession could not stand unless the corpus delicti had been proved through independent evidence. See David A. Moran, In Defense of the Corpus Delicti Rule, 64 Ohio St. L.J. 817, 817 (2003). This rule evolved out of a concern that the state might obtain a conviction based on a confession to a crime that never occurred. In perhaps the most dramatic example of this error, three people were hanged for murder even though the victim was still alive. Perrys’ Case, 14 How. St. Tr. 1312 (Eng. 1661). In this traditional sense, it is indisputable that there was evidence of the corpus delicti here, and Winfield conceded as much at argument. Someone attempted to murder Garrett; he was shot at between two and fifteen times. (He was hit twice, and police found fifteen shell casings.) Multiple shots permit a factfinder to infer an intent to kill, as the trial judge did here. E.g., People v. Howery, 687 N.E.2d 836, 856 (Ill. 1997). The district court thought that the corpus delicti rule also extends to whether Winfield was the shooter and requires the state to corroborate the confession. The court candidly acknowledged, however, that some authorities supported this reading of the rule and others did not, so it was a close case with no definitive answer and no case directly on point. We need not decide the scope of the corpus delicti rule or whether the state met the requirements of it. See Crockett v. Butler, 807 F.3d 160, 168 (7th Cir. 2015) (explaining that habeas 26 Nos. 19-1441 & 19-1547 review cannot correct perceived errors of state law). As the district court recognized, the issue before the federal courts is whether the argument that the state had not met its burden was both obvious and clearly stronger than the argument that appellate counsel raised on appeal. Only if both conditions are met can we say that appellate counsel’s performance fell below the constitutional minimum. See Makiel v. Butler, 782 F.3d 882, 898 (7th Cir. 2015). That the underlying corpus delicti argument is close and novel cuts against Winfield on both prongs. First, it implies that the argument is not all that strong, even if it could ulti- mately have prevailed. “[T]he comparative strength of two claims is usually debatable.” Shaw, 721 F.3d at 915. Reasona- ble minds can differ on the relative strength of an argument that ekes out a win compared to one that falls just short. Strick- land instructs that we must be “highly deferential” to coun- sel’s choices to avoid the temptation of hindsight. Morris, 832 F.3d at 710. Counsel is entitled to—expected to—“select[] the most promising issues for review” and “focus[] on one central issue if possible.” Jones v. Barnes, 463 U.S. 745, 751–52 (1983). Winfield cannot succeed by proving only that the corpus de- licti argument is not frivolous. Indeed, the Constitution does not obligate appellate counsel to raise every nonfrivolous ar- gument. Id. at 751. Although the state court rejected the sentencing argument raised on direct appeal, it was not quite so weak as to be “the equivalent of filing no brief at all,” or “effectively a substitute for a [no-merit] brief,” like the sufficiency argument in Shaw, 721 F.3d at 915. The Illinois Appellate Court has reversed when judges have failed to consider a defendant’s rehabilita- tive potential. E.g., People v. Jeter, 616 N.E.2d 1256, 1264 (Ill. Nos. 19-1441 & 19-1547 27 App. Ct. 1993). The problem was that the trial judge did not ignore Winfield’s rehabilitative potential, he just found that he had none. Second, a close case makes the argument less obvious. “Strickland does not guarantee perfect representation, only a ‘reasonably competent attorney.’” Richter, 562 U.S. at 110. The more precedents one must distinguish (and the more care- fully one must do so) to reach a favorable result, the less un- reasonable it likely is for an appellate lawyer not to spot the issue and brief it. This is especially true where, as here, the legal issue is still unclear twenty years later. See Loden v. McCarty, 778 F.3d 484, 501 (5th Cir. 2015) (finding no deficient performance when law was unclear at time of appeal); cf. Shaw, 721 F.3d at 916–17 (discussing circumstances in which counsel performs deficiently by not predicting change in the law). Counsel had Jeter to support his sentencing argument, even if it turned out not to be enough. Winfield, however, has still not identified any case that directly supports his corpus delicti argument—only general principles that could favor him if the court were to interpret them a certain way. These ambiguities in the law are not enough to carry his burden to prove that the corpus delicti argument was obvious such that any reasonable attorney would have raised it. Win- field faced two questions before he could prevail on a corpus delicti argument. One, was the state obligated to corroborate the details of his confession? Two, did the state fail to corrob- orate them? His petition could potentially succeed only if a reasonable attorney would find a basis to argue that both an- swers are yes. Although we, like the district court, see no de- finitive answer to these questions in the decisions of the Illi- nois Supreme Court, we think the implied answer to both is 28 Nos. 19-1441 & 19-1547 no. The state was likely not required to corroborate Winfield’s confession beyond proving the traditional corpus delicti, and even if it were, it likely succeeded. We, thus, cannot say that counsel was not acting as counsel when he failed to argue oth- erwise. In assessing the obviousness of this argument, it is best to start with what the Illinois Supreme Court considers “the most precise explanation” of the corpus delicti rule. Lara, 983 N.E.2d at 967 (citing People v. Willingham, 432 N.E.2d 861, 864 (Ill. 1982)). This version was set out in People v. Perfecto: The rule is that an uncorroborated confes- sion is insufficient to convict, but the corpus de- licti is not required to be proved beyond a rea- sonable doubt exclusively by evidence aliunde the confession or admissions of the accused, nor is it necessary that it be established by evidence other than that which tends to connect the de- fendant with the crime. “The true rule is that if there is evidence of corroborating circumstances which tend to prove the corpus delicti and correspond with the circumstances related in the confession, both the circumstances and the confession may be consid- ered in determining whether the corpus delicti is suf- ficiently proved in a given case.” 186 N.E.2d 258, 258–59 (1962) (citations omitted) (emphasis added) (quoting People v. Gavurnik, 117 N.E.2d 782, 785 (Ill. 1954)). Winfield and the district court read the emphasized sen- tence to mean that the evidence must always both “tend to prove the corpus delicti” and “correspond with the Nos. 19-1441 & 19-1547 29 circumstances related in the confession,” i.e., corroborate the confession, to sustain a conviction. The district court found that the state had met the first prong but failed the second. The state contends the evidence must meet only one prong because the rule is actually disjunctive. Under its reading of Perfecto, corroboration is necessary only before the confession “may be considered in determining whether the corpus delicti is sufficiently proved.” Though it is not “required” to prove the traditional corpus delicti exclusively with evidence inde- pendent of the confession, it may do so. And if it does (as we explained is true here), then the confession need not be cor- roborated more. Another way to phrase this reasoning is that the independent proof of the corpus delicti is itself sufficient corroboration. Cf. United States v. Fleming, 504 F.2d 1045, 1049 (7th Cir. 1974) (Stevens, J.) (concluding that proof of tradi- tional corpus delicti is sufficient but not necessary to corrobo- rate confession under federal law). The Illinois Supreme Court’s frequent emphasis that the corpus delicti and the identity of the offender are separate concepts supports the state’s argument. See, e.g., Lara, 983 N.E.2d at 964; People v. Cloutier, 622 N.E.2d 774, 784 (Ill. 1993).3 In People v. Holmes, 367 N.E.2d 663 (Ill. 1977), the court made explicit the disjunctive test the state advocates: “It is enough if the other evidence either tends to show that a crime did in fact occur or to corroborate the confession.” Id. at 665 (emphasis added) (citations omitted) (quoting People v. 3 The ordinary rule is that we resolve the performance prong based on the law at the time of the appeal (here 2001). See Shaw, 721 F.3d at 915. Nevertheless, we cite to Lara, a 2012 case, because both parties rely on it, neither contends that it represents a change in the law, and it cogently synthesizes preexisting caselaw. See Lara, 983 N.E.2d at 966–71. 30 Nos. 19-1441 & 19-1547 Norcutt, 255 N.E.2d 442, 446 (Ill. 1970)). That the victim was shot and died was enough to prove the corpus delicti of mur- der. Id. More recently, the Illinois Appellate Court (relying on Holmes) has said that “it is not a requirement of corpus delicti that evidence apart from the confession tend to connect de- fendant to the crime charged.” People v. Valladares, 994 N.E.2d 938, 960 (Ill. App. Ct. 2013). It was a failure to connect Winfield to the crime—to prove identity—that led to relief here. We think the state’s reading likely reflects the proper in- terpretation of the corpus delicti rule. The state could comply with the traditional rule by proving through independent ev- idence that a crime occurred. Alternatively, if there were in- sufficient independent evidence that the crime occurred, the state could follow Perfecto and corroborate the confession with evidence corresponding to the circumstances of the confes- sion. Either independent evidence or the confession plus the corroborating evidence together could be used to establish that a crime occurred—the ultimate requirement of the corpus delicti rule. Willingham, 432 N.E.2d at 865. Once the state passed this legal hurdle, all that was left was to prove the other element, identity, as a factual matter, and it could do that with the confession alone. See People v. Taylor, 317 N.E.2d 97, 102 (Ill. 1974).4 4 It is notable that this interpretation of the corpus delicti rule appears to be the one under which everyone was operating until the district court’s order. The state trial judge, though bolstering his conclusion with the con- fession, initially found the corpus delicti from the fact that Garrett was shot twice. Appellate counsel likewise pointed to this simple fact to ex- plain why he did not bring a sufficiency argument. Indeed, even Winfield pressed in the district court a corpus delicti theory that there was no Nos. 19-1441 & 19-1547 31 Although we doubt the state was required, as a matter of law, to corroborate the confession to any degree more than to prove that someone attempted to murder Garrett, we accept the district court’s point that there might be room to argue other- wise. Some cases use “and” instead of Holmes’s “or”; others discuss corroboration, despite independent proof of the cor- pus delicti; many, in that discussion, further emphasize that this corroborating evidence proved a connection between the crime and the accused. (Willingham, 432 N.E.2d at 864–66, does all three.) Even under this interpretation, though, the Il- linois Appellate Court would not vacate the conviction unless the proof fell short of corroborating the confession. We think it far from obvious that it did. The Illinois Supreme Court has expressed just how low the bar is for corroboration. It requires “far less independent evi- dence to corroborate a defendant’s confession under the cor- pus delicti rule than to show guilt beyond a reasonable doubt.” Lara, 983 N.E.2d at 970; accord Furby, 563 N.E.2d at 426; Willingham, 432 N.E.2d at 864. Moreover, “there is no require- ment that the independent evidence and the details of the con- fession correspond in every particular.” Furby, 563 N.E.2d at 428. The law requires “only some ‘consistency’ ‘tending to confirm and strengthen the confession.’” Lara, 983 N.E.2d at 970 (quoting Furby, 563 N.E.2d at 421). The district court identified two inconsistencies between Hartman’s testimony and Winfield’s statement that it thought made it obvious that an Illinois Court might deem the evidence anyone had attempted to murder Garrett—i.e., that the tradi- tional rule had not been met (and thus corroboration was necessary). 32 Nos. 19-1441 & 19-1547 confession uncorroborated. The first was the number of shoot- ers, the second, Winfield’s exit.5 The first of these discrepancies—that Hartman saw only one shooter who was not Winfield—is immaterial. It is not even an inconsistency. Hartman declined to rule out the pos- sibility that there might have been a second shooter he could not see. Even if he had not been so candid, the Illinois Su- preme Court has held that eyewitness testimony that there were two robbers “did not necessarily create inconsistencies” with a confession that the defendant was the third robber. Willingham, 432 N.E.2d at 866. The second difference is more substantial: all the witnesses (including Curry and Garrett) said the shooter(s) returned to the car and no one saw anyone fleeing on foot, as Winfield said he did. This is an inconsistency, but again, Illinois law does not require perfect parity between the evidence and the confession. Furby, 563 N.E.2d at 428. All that is needed is a “loose[] ‘correspondence’”; “corroboration of only some of the circumstances related in a defendant’s confession is suffi- cient.” Lara, 983 N.E.2d at 970–71. The independent evidence corresponded with Winfield’s statement in many critical respects. Hartman confirmed that a black SUV traveled south on Central Park and stopped at the corner of Huron when at least one shooter exited the 5 We, like the district court, accept that the trial judge discredited both Garrett and Curry’s testimony. As we noted above, the state argues that the court partially credited Garrett’s testimony. We doubt that is true, but a state court might reasonably disagree. This possibility is just one more barrier in the way of a finding that a corpus delicti argument is both obvi- ous and clearly stronger than counsel’s sentencing argument. Nos. 19-1441 & 19-1547 33 vehicle and began firing. Further, Winfield’s own testimony was independent evidence that he and his brother had an ar- gument with Stovall at a dice game weeks before the shooting, just as he stated to the detective. See Willingham, 432 N.E.2d at 866 (relying on defendant’s trial testimony about events be- fore crime as corroborating evidence). An Illinois court would likely deem this enough to ensure that Winfield’s confession was “reasonably reliable,” allowing the trial judge to consider it in his “role in deciding credibility issues, weighing the evi- dence and drawing reasonable inferences, and resolving evi- dentiary conflicts.” Lara, 983 N.E.2d at 971–72. Certainly, it was not unreasonable for Winfield’s counsel to conclude that these inconsistencies were for the trial judge, not the Appel- late Court, to resolve and so a sufficiency argument had only a remote possibility of success, if that. The district court thought the Illinois courts might dismiss this independent evidence as confirming only tangential com- ponents of the statement, as in People v. Lambert, 472 N.E.2d 427 (Ill. 1984) (per curiam). One crucial difference between this case and Lambert is that in the latter there was no inde- pendent evidence that a crime occurred at all. Id at 429. The state was, thus, undoubtedly obligated to corroborate the de- fendant’s confession to sexual assault of a child under any reading of the rule. The independent evidence, however, showed only that the alleged child victim spent a single night in the defendant’s ordinary sleeping quarters (a basement) and that a few weeks later, the child’s rectum appeared “pink- ish and swollen.” Id. There was no evidence of a cause for this inflammation (criminal or otherwise) and the child never complained of being assaulted. All the evidence confirmed was the opportunity to commit a crime, which is to say that the defendant slept where his bed was. 34 Nos. 19-1441 & 19-1547 Here, the independent evidence did not line up only with confessed facts of Winfield’s daily life but with details of how the crime occurred: the car’s color, path of travel, and his un- derlying motive for the shooting. The corpus delicti rule does not require that the evidence and confession “precisely align … on each element of the charged offense, or indeed to any particular element of the charged offense.” Lara, 983 N.E.2d at 972. It is difficult to see why a reasonable attorney would ex- pect an Illinois court to deem these key facts tangential rela- tive to Winfield’s purported escape method. The district court carved a careful path for Winfield to raise a nonfrivolous corpus delicti argument. Even if that path exists, though, it is not an obvious one. It is narrow and winds carefully around many barriers, both factual and legal. It is so complex that we cannot say that all reasonably competent at- torneys would spot it. Nor can we be sure the path had a fa- vorable end. For our purposes, we need not be certain where it leads; it suffices to say that counsel’s decision to take a dif- ferent way was not constitutionally deficient. We therefore va- cate the district court’s grant of habeas corpus relief. V. Conclusion Winfield argued that two of his lawyers mishandled his case. He said his trial counsel should have presented an alibi defense, but the state court reasonably found he was not con- stitutionally obligated to present a defense of which he was simply not aware. Likewise, we conclude that the Constitu- tion did not obligate Winfield’s appellate counsel to discover and present a complex and novel legal argument that may or may not have succeeded. We affirm the denial of relief and reverse the grant of relief, so that Winfield’s petition for writ of habeas corpus is denied in full.