OPINION
THOMAS, Chief Judge:Eric Owen Mann, who was convicted and sentenced to death in Arizona state court for the murders of two men, appeals the district court’s denial of his habeas corpus petition. We have jurisdiction under 28 U.S.C. § 1291 and 28 U.S.C. § 2253. We affirm in part and reverse in part.
I
A
This case involves another tragic tale of drugs and violence. On the evening of November 23, 1989, Eric Mann and his then-girlfriend, Karen Miller, returned to the house they rented in Tucson after having Thanksgiving dinner with Mann’s mother. Two weeks before, Mann had arranged to sell about a kilogram of cocaine to his friend, Richard Alberts, for about $20,000, and the transaction was to take place at the house that evening. However, according to Miller, Mann never actually planned to provide cocaine to Al-berts. Rather, he planned to “rip off’ Alberts by taking the money and giving Alberts a shoebox filled with paper instead of cocaine. He told Miller that he planned to “whack” Alberts because he knew he would not be able to get away with the theft otherwise.
About fifteen minutes after Mann and Miller returned to their house, Alberts arrived with another man, Ramon Bazurto. Mann was initially upset when he saw that Alberts had unexpectedly brought along a second person, but after some reflection he told Miller that he had “to do it.” Mann let both men into the house, and after *1207about ten minutes of idle talk, they followed Mann to the master bedroom in the back of the house. Miller followed the men and stood in the doorway, behind and to the left of Bazurto. Both Mann and Miller were armed.
In the master bedroom, Mann laid his .45 caliber pistol on the bed, picked up the paper-filled shoebox, and handed it to Al-berts, who in turn handed Mann a black bag containing the money. Mann placed the bag on the bed. Alberts then opened the shoebox and immediately realized that it was filled with paper. At that moment, Mann grabbed his gun and immediately shot Alberts and then Bazurto.
The bullet that struck Alberts traveled through his heart, killing him almost instantly. The bullet that struck Bazurto traveled through his lung and severed his aorta, but he remained alive for a short while. He fell through-the bedroom doorway and landed on 1¿he floor by Miller’s feet. As Bazurto lied there on his back, he made feeble attempts to reach for the gun that was tucked into the front of his pants. Miller testified that Bazurto’s sporadic movements lasted approximately three to five minutes before he died.
Miller testified that she and Mann did not make a plan beforehand to deal with the aftermath of the killings, so after Mann shot Alberts and Bazurto, he left to find somebody to help move the bodies. He picked up a friend, Carlos Alejandro, who agreed to help. Mann and Alejandro loaded the bodies into Alberts’s car, drove them to a remote location on a dirt road near Fort Grant prison, and left them there.
The next day, Mann and Miller thoroughly cleaned the house and concealed any indication of what happened the night before. They recovered the bullets, patched the bullet holes in the walls, repainted, and scrubbed the bedroom with ammonia. Mann dismantled his gun and Miller’s gun, hammered down the pieces, and threw them, along with the recovered bullets, in a lake. Mann then gave Al-berts’s car to another man to pay off a debt.
Police searched Mann’s and Miller’s house on November 28, 1989 pursuant to a search warrant, and found evidence of repairs to the two rear bedrooms. Mann told the police that Alberts and Bazurto had been to the house for a drug deal on the evening of November 23, but left after they failed to agree on a price. The police made no arrests at that time, and the case did not develop further until 1994.
Miller ended her relationship with Mann in October 1993. She testified that their relationship deteriorated as Mann became increasingly violent and abusive toward her, and that in late October he threatened to kill her. Miller took their young daughter and left Tucson to live with Miller’s father in Washington state. In January 1994, Miller contacted the Pima County Sheriffs Department to report the 1989 murders, and her information led investigators to Alejandro. Mann was subsequently arrested and charged with two counts of first-degree murder.
B
Mann’s trial in the Pima County Superi- or Court began on October 25, 1994, and lasted five days. The prosecution’s case relied primarily on the testimony of Karen Miller and Carlos Alejandro, who were both granted immunity from prosecution for their roles in the murders and coverup. Mann’s court-appointed attorney did not call any witnesses for the defense, but rather sought to hold the prosecution to its burden of proof and establish through cross-examination that Mann shot Alberts *1208and Bazurto out of self-defense, not premeditation.
On November 1, 1994, the jury found Mann guilty of both counts of first-degree murder. Following Mann’s conviction, defense counsel realized that he needed more time to investigate Mann’s background— including his education, criminal history, and medical records — so he requested a continuance of the sentencing hearing. He also requested that the court appoint a psychologist from the superior court clinic to evaluate Mann’s mental health. The court granted both requests.
C
During the sentencing phase of the trial, defense counsel argued in support of ten non-statutory mitigating factors: Mann’s positive, nonviolent relationship with his two daughters; his positive influence on his mother; his unstable and abusive family background; his poor educational experience; his history of substance abuse; his remorse; his cooperation with the authorities; his previously nonviolent history; his good conduct while incarcerated; and the disparity of his treatment compared to that of Miller and Alejandro. In addition to his sentencing memorandum, Mann’s attorney presented Mann’s handwritten autobiography, which included vivid details about how, when Mann was growing up, his father abused his brother and beat his mother so badly that she could not leave their house. The autobiography also mentioned a 1985 traffic accident in which Mann sustained a head injury and his two passengers were killed.
At the sentencing hearing, Mann’s counsel presented the testimony of four witnesses. .Mann’s former employer testified that Mann was dependable and responsible, and Mann’s former co-worker testified that Mann was a hard worker. Mann’s oldest daughter testified that she loved her father and wanted to continue to have a relationship with him, and Mann’s mother confirmed Mann’s account of his father abusing his family and introducing Mann to crime at an early age.
The sentencing judge also had before him the report of the court-appointed psychologist, Dr. Todd C. Flynn. In that report, Dr. Flynn diagnosed Mann with alcohol abuse, polysubstance abuse, and antisocial personality disorder. He hypothesized that Mann’s antisocial personality disorder was the result of extreme immaturity, which itself was due to substance abuse at an early age. He also concluded that Mann probably fit the designation of “Psychopath.”
The sentencing judge found three statutory aggravating factors: offenses committed for pecuniary gain, Ariz.Rev.Stat. § 13-703(F)(5); multiple homicides, § 13-703(F)(8); and, with respect to Bazurto, murder committed in an especially cruel, heinous, and depraved manner, § 13703(F)(6). Of the ten mitigating factors that Mann’s counsel sought — all of which were nom-statutory — the judge found six. He rejected cooperation with authorities, nonviolent history, disparity of treatment, and remorse. The judge specifically and vehemently denied that Mann had proven the latter factor. The judge found that the only evidence of Mann’s remorse for the murders was a letter he had written to the court before the sentencing hearing. The judge also pointed out that in Mann’s handwritten autobiography, Mann indicated no remorse for the. deaths of the two passengers in his 1985 traffic accident. This omission, combined with Dr. Flynn’s report diagnosing Mann as a psychopath with antisocial personality disorder, led the judge to conclude that “the Defendant is incapable of remorse” and “has no conscience.” The judge sen*1209tenced Mann to death for each of the murders.
D
On direct appeal, the Arizona Supreme Court affirmed Mann’s convictions. State v. Mann, 188 Ariz. 220, 934 P.2d 784, 795 (1997). The state supreme court also affirmed the trial court’s findings concerning the aggravating and mitigating factors and, after independently reweighing those factors, concluded that the mitigating factors were insufficient to justify leniency. Id. at 790-91, 794-795. The United States Supreme Court denied Mann’s petition for writ of certiorari. Mann v. Arizona, 522 U.S. 895, 118 S.Ct. 238, 139 L.Ed.2d 169 (1997).
Mann timely petitioned the state trial court for post-conviction relief pursuant to Arizona Rule of Criminal Procedure 32. Among other claims, he asserted violations of his constitutional right to the effective assistance of counsel under Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Mann claimed ineffective assistance during the guilt phase due to counsel’s decision to not call him to testify, and he claimed ineffective assistance during the sentencing phase due to counsel’s failure to investigate and present reasonably available mitigating evidence— specifically, evidence pertaining to the effects of a serious traffic accident in 1985. Mann also claimed ineffective assistance due to counsel’s failure to retain an independent mental health expert.
Judge John F. Kelly, the same judge who presided over Mann’s trial and sentenced him to death, presided over Mann’s post-conviction proceedings. During several evidentiary hearings, the judge heard testimony from Mann’s trial counsel, Karen Miller, and Dr. James Comer, a clinical neuropsychologist who conducted a battery of tests on Mann to detect evidence of organic brain injury, The judge also had before him Dr. Comer’s report, a psychological evaluation of Mann conducted by Dr. Richard Hinton, and Mann’s medical records from his 1985 traffic accident.
At the close of the Rule 32 proceedings, the state court denied Mann’s petition for post-conviction relief. With regard to Mann’s claim of ineffective assistance for counsel’s failure to call him to testify as a guilt-phase witness, the state court concluded that counsel’s performance did not constitute deficient performance. With regard to Mann’s claim of ineffective assistance for his trial counsel’s failure to investigate and present reasonably available mitigating evidence, the state court found that Mann failed to show that his counsel’s conduct prejudiced him. The Arizona Supreme Court denied Mann’s petition for review without citation or comment.
Mann filed a timely petition for writ of habeas corpus under 28 U.S.C. § 2254, and subsequently filed an amended petition. The district court denied Mann’s request for an evidentiary hearing, and on August 10, 2009, the district court issued an order denying Mann’s petition' on the merits. The district court certified three claims for appeal, and Mann timely appealed two of them: his claim of ineffective assistance of counsel at the guilt phase of his trial, and his claim of ineffective assistance of counsel at the sentencing phase of his trial.
We review de novo a district court’s denial of a petition for writ of habeas corpus. Estrada v. Scribner, 512 F.3d 1227, 1235 (9th Cir.2008). We review the district court’s factual findings for clear error. Hurles v. Ryan, 706 F.3d 1021, 1029 (9th Cir.2013).
Because Mann filed his habeas petition after April 24, 1996, we apply the Antiter-rorism and Effective Death Penalty Act of 1996 (“AEDPA”), Pub.L. No. 104-132, 110 *1210Stat. 1214. Lindh v. Murphy, 521 U.S. 320, 336, 117 S.Ct. 2059, 138 L.Ed.2d 481 (1997). Under AEDPA, a state prisoner may not obtain federal habeas relief for any claim that was adjudicated on the merits by a state court unless the state court’s decision was (1) “contrary to” clearly established'federal law as determined by the Supreme Court, (2) “involved an unreasonable application of’ such clearly established law, or (3) “was based on an unreasonable determination of the facts” in light of the record before the state court. 28 U.S.C. § 2254(d); Harrington v. Richter, 562 U.S. 86, 131 S.Ct. 770, 785, 178 L.Ed.2d 624 (2011).
For the purpose of determining whether AEDPA bars federal habeas relief, we review the last reasoned state court decision. Gill v. Ayers, 342 F.3d 911, 917 n. 5 (9th Cir.2003). Because the Arizona Supreme Court denied Mann’s petition for review without citation or comment, we “look through” that decision and review the state trial court’s decision denying post-conviction relief. See id.
II
Mann is not entitled to habeas relief on his guilt phase claim for ineffective assistance of counsel because, under either version of the facts, counsel’s decision not to call Mann as a witness was strategic, and therefore fell within the wide range of reasonable professional assistance deemed constitutionally adequate under Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).
A
To prevail on a claim that his counsel was constitutionally ineffective, a petitioner must first establish that his counsel’s performance was constitutionally deficient. Strickland, 466 U.S. at 687, 104 S.Ct. 2052. In order to do so, he must show that the representation “fell below an objective standard of reasonableness” under all the circumstances. Id. at 688, 104 S.Ct. 2052. There are no rigid rules for judging attorney performance, but the American Bar Association (“ABA”) standards serve as guides for determining what is reasonable under prevailing professional norms. Id. “Because of the difficulties inherent in making the evaluation, a court must indulge a strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance; that is, the defendant must overcome the presumption that, under the circumstances, the challenged action might be considered sound trial strategy.” Id. at 689, 104 S.Ct. 2052 (internal quotation marks and citation omitted).
B
1
This appeal involves two competing versions of why Mann’s counsel did not call him to the witness stand: one advanced by Mann and one by his counsel. During the post-conviction evidentiary hearings, Mann’s trial counsel testified that he did not allow Mann to take the stand because he had an obligation not to suborn perjury. According to counsel, at one of their first attorney-client meetings Mann confessed to premeditating the murders, so if Mann had testified that he acted in self-defense, he would have committed perjury. In addition counsel would have been required to withdraw, as putting Mann on the stand to testify in a manner that he knew to be fake would have violated his ethical duties. Counsel claimed that he presented Mann with two options: either counsel would continue to represent him, in which case Mann would have to decline to testify; or Mann could obtain new counsel who might put him on the stand. Counsel claimed *1211that, given this choice, Mann elected not to change counsel, to pursue a self-defense strategy, and not to take the stand.1
Mann recalled a starkly different version of the events before and during trial. He testified that he never confessed to counsel that he premeditated the murders. Rather, he claimed he told counsel that he acted in self-defense against what he perceived to be an immediate threat from Bazurto. Mann testified that he and his counsel discussed the possibility of Mann testifying at trial, but counsel was “adamant” that Mann not testify because he was concerned about the admission of Mann’s prior convictions. Nevertheless, Mann felt that he should testify, and he asked his counsel a number of times to let him do so. Despite Mann’s requests, counsel decided not to call Mann as a witness.
Under either version of the facts, counsel’s performance was not deficient under Strickland. Taking counsel’s version of the facts as true, he did not provide deficient performance by not calling Mann as a witness so that he would not commit perjury. See Nix v. Whiteside, 475 U.S. 157, 171, 106 S.Ct. 988, 89 L.Ed.2d 123 (1986). Counsel gave Mann the choice to either keep him as his lawyer and not testify, or obtain new counsel. This was consistent with counsel’s professional responsibilities. Counsel’s duty in such a situation is first to “attempt to dissuade the client from the unlawful course of conduct” and, if the client continues to “threaten to commit perjury,” the Model Rules on Professional Conduct provide for “withdrawal from representation as an appropriate response.” Id. at 169-70, 106 S.Ct. 988 (citations omitted). If counsel’s version of the facts is correct, he had no choice but to comply with his ethical obligation to not suborn perjury. See id. at 166, 106 S.Ct. 988 (“[Cjounsel is precluded from taking steps or in any way assisting the client in presenting false evidence or otherwise violating the law.”). Moreover, we do not need to address this claim in further detail because Mann conceded that if his counsel’s account is credible then he has not established a claim for ineffective assistance of counsel at the guilt phase of his trial.
Taking Mann’s version of the facts as true, the district court correctly determined that he failed to establish a claim for ineffective assistance of counsel under Strickland. The district court properly determined that defense counsel’s decision not to call Mann to testify was constitutionally adequate because it was based on a strategic decision. As we have noted, *1212such strategic decisions are presumptively within the wide range of professional assistance, and “virtually unehallengable” under a Sixth Amendment analysis. Id. at 690, 104 S.Ct. 2052.
Here, taking the stand would have opened the door to potentially damaging cross-examination, including the admission of one of Mann’s prior felony convictions and, perhaps, evidence of Mann’s past abuse of Miller and another ex-girlfriend. Therefore, defense counsel’s decision to keep Mann off the stand was “within the wide range of reasonable professional assistance.” Id. at 689, 104 S.Ct. 2052; see Bell v. Cone, 535 U.S. 685, 700, 122 S.Ct. 1843, 152 L.Ed.2d 914 (2002) (finding that defense counsel “had sound reasons” for deciding not to call the defendant and certain other individuals as guilt-phase witnesses where defendant’s testimony could have “alienated him in the eyes of the jury” and where the other witnesses’ testimony could have opened the door to the defendant’s criminal history); see also Gulbrandson v. Ryan, 711 F.3d 1026, 1038-39 (9th Cir.2013) (holding that it was a reasonable application of Strickland for a state court to conclude that defense counsel’s decision not to call the defendant as a guilt-phase witness was “within the wide range of reasonable professional assistance” where the defendant’s testimony could have harmed the defense).
Mann argues that the potential damage from his prior conviction was marginal because the trial court limited the prosecution to showing only its existence, not its nature. Even if Mann may be right about the impeachment value of his prior eonviction, given the “strong presumption that, under the circumstances, the challenged action might be considered sound trial strategy,” we decline to “second-guess” counsel’s assessment of the risks associated with putting Mann on the witness stand. Strickland, 466 U.S. at 689, 104 S.Ct. 2052.
2
Mánn also claims that defense counsel’s decision to keep him off the witness stand was unreasonable because counsel’s opening statement to the jury implicitly promised that Mann would testify.2 The district- court did not err in rejecting this claim.
Counsel may render constitutionally ineffective assistance when counsel breaks a promise to the jury that the defendant will testify. See, e.g. United States ex rel. Hampton v. Leibach, 347 F.3d 219, 258 (7th Cir.2003) (counsel’s performance objectively unreasonable where counsel told the jury in his opening statement that the defendant would testify about the circumstances of the alleged offense, but then declined to call the defendant to the witness stand for reasons that were apparent at the time he made his opening statement); Ouber v. Guarino, 293 F.3d 19, 22, 30 (1st Cir.2002) (defense counsel’s trial performance deficient where counsel failed to put the defendant on the stand after telling the jury four times in his opening statement that the defendant would testify); McAleese v. Mazurkiewicz, 1 F.3d 159, 166-67 (3d Cir.1993); Harris v. Reed, 894 F.2d 871, 879 (7th Cir.1990).
*1213In those circumstances, defense counsel made a specific promise to the jury to present important evidence, such ás the defendant’s testimony or other evidence central to the defense. Ouber, 293 F.3d at 22; Hampton, 347 F.3d at 257; Harris, 894 F.2d at 873. And in each case, defense counsel made an about-face by declining to present the promised evidence in the absence of unforeseen circumstances that would have justified the change in strategy. Ouber, 293 F.3d at 28, 29; Hampton, 347 F.3d at 258; Harris, 894 F.2d at 877-78.
Here, Mann’s counsel made no promise to the jury that Mann would testify or that the defense would present specific evidence. In his opening statement, counsel told the jury that “the facts will be” that Mann planned a real drug deal with Alberts on the night of the shooting, and that “just when the transaction took place, Ray Basurto [sic] did something that made Eric think he was about to be the victim of a drug rip-off himself.” Counsel continued by telling the jury that “in a spontaneous act, [Mann] fired at both people and killed them both because he felt like he was forced to do so.” None of these statements specifically-promised the jury that Mann would testify; they were nothing more than counsel’s spin on the evidence that would later be produced. See McAleese, 1 F.3d at 167 (holding that defense counsel made no promises to the jury where “[h]e merely summarized evidence that was later produced from which a jury could be left with a reasonable doubt about McAleese’s identity as his ex-wife’s killer”).
Moreover, to the extent Mann’s counsel somehow implicitly promised the jurors they would hear evidence from which they could conclude that Mann planned a real drug transaction and spontaneously reacted to a perceived threat from Bazurto, we cannot conclude that any such implicit promise went unfulfilled. To suggest that Bazurto may have done something to make Mann feel threatened, defense counsel elicited testimony from the medical examiner that, given Miller’s description of Bazurto’s arm movements as he was dying, Bazurto could have been reaching for his gun when Mann shot him. To show that Mann acted spontaneously and never planned to rip-off or kill the men, defense counsel first elicited testimony from Carlos Alejandro that Mann felt he had no choice but to shoot the men. Counsel then elicited Miller’s admission that she chose to stand in the line of fire behind Bazurto, which undermined her earlier testimony that she knew Mann was planning to shoot Bazurto and Alberts. Finally, defense counsel elicited Miller’s admission that she and Mann had no plan beforehand to dispose of the bodies or clean up the house. To the extent this testimony undermined the allegation that Mann premeditated the murders, it also undermined the allegation that Mann planned to rip-off Alberts and Bazurto; according to Miller, the only reason Mann intended to kill the men was so he could get away with the ploy to rip them off.
In sum, the record, when read closely, shows that defense counsel did not make, and therefore did not break, a specific promise to the jury that Mann would testify. Accordingly, the district court was correct to conclude that Mann failed to establish that he was denied effective assistance of counsel due to a broken promise to the jury.3
*1214Because Mann offers no other basis for claiming a Strickland violation, we conclude that counsel did not provide ineffective assistance at the guilt phase.
Ill
The district court erred in determining that AEDPA bars habeas relief on Mann’s claim that his counsel was constitutionally ineffective for failing to investigate and present reasonably available mitigating evidence at sentencing. The district court also erred in determining that, even if AEDPA does not bar relief, Mann failed to establish that counsel’s performance prejudiced him under Strickland.
A
AEDPA does not bar habeas relief on this claim. The state post-conviction court’s conclusion that Mann was not prejudiced by counsel’s performance at sentencing was contrary to clearly established federal law because it applied the incorrect standard for prejudice.
The Supreme Court has clearly established that the governing legal standard for assessing the prejudice from counsel’s errors during the sentencing phase of a capital ease is “whether there is a reasonable probability that, absent the errors, the sentencer — including an appellate court, to the extent it independently reweighs the evidence — would have concluded that the balance of aggravating and mitigating circumstances did not warrant death.” Strickland, 466 U.S. at 695, 104 S.Ct. 2052. Under the Strickland standard, “a defendant need not show that counsel’s deficient conduct more likely than not altered the outcome in the case.” Id. at 693, 104 S.Ct. 2052; see also Richter, 131 S.Ct. at 791 (“In assessing prejudice under Strickland, the question is not whether a court can be certain counsel’s performance had no effect on the outcome. ...”); Rompilla v. Beard, 545 U.S. 374, 393, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005)“ (“[Ajlthough we suppose it is possible that a jury could have heard [all of the mitigating evidence counsel failed to present] and still have decided on the death penalty, that is not the test.”). Indeed, the Strickland Court expressly rejected the more-likely-than-not standard for determining whether newly discovered evidence warrants a new trial. 466 U.S. at 694, 104 S.Ct. 2052. Where a state court “reject[s] a prisoner’s claim of ineffective assistance of counsel on the grounds that the prisoner had not established by a preponderance of the evidence that the result of his criminal proceeding would have been different,” the state court’s decision is “contrary to” clearly established federal law and “a federal court will be unconstrained by § 2254(d)(1).” Williams v. Taylor, 529 U.S. 362, 405-06, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000).
In concluding that Mann failed to establish prejudice from counsel’s performance, the state post-conviction court imported the reasoning from an earlier section of its decision where it had rejected Mann’s separate claim under Arizona Rule of Criminal Procedure 32.1(e) that newly discovered evidence concerning the effects of his 1985 traffic accident required a new sentencing hearing:
Here, Defendant has failed to show prejudice .... Additional evidence that pertains to the 1985 accident and its effects is discussed under “Issues Three and Four” above, where this Court found that it would not have changed the sentence imposed. For that reason, Defendant was not prejudiced by counsel’s performance and the claim is denied.
In that earlier section, titled “Issues Three and Four,” the state court denied Mann a new sentencing hearing because it found *1215that “nothing presented would have changed the verdict or the sentence imposed,” and because it found that Mann failed to prove a causal connection between the accident and the murders.
At the time of the post-conviction court’s decision, a state court could grant a prisoner a new sentencing hearing only if “[n]ewly discovered material facts probably exist[ed] and such facts probably would have changed the verdict or sentence.” Ariz. R.Crim. P. 32.1(e) (2000). Under Rule 32.1(e), “probably” meant “more likely than not.” See State v. Orantez, 183 Ariz. 218, 902 P.2d 824, 829 (1995) (in banc) (holding that a new trial was warranted where new evidence “would have likely resulted in a different verdict”); State v. Pac, 175 Ariz. 189, 854 P.2d 1175, 1178 (1993) (explaining that to justify relief, “the evidence ‘must be such that it would likely have altered the verdict, finding or sentence if known at the time of trial’ ” (quoting State v. Cooper, 166 Ariz. 126, 800 P.2d 992, 995 (1990))). Thus, when the state post-conviction court concluded that Mann was not prejudiced for the same reason it found that Mann was not entitled to a new sentencing hearing under Rule 32.1(e), it held Mann to the more-likely-than-not standard, which the Supreme Court expressly rejected in Strickland, 466 U.S. at 694, 104 S.Ct. 2052. And despite the dissent’s insistence to the contrary, this error in reasoning is actionable. The state court’s application of the wrong standard renders its decision “contrary to” clearly established federal law and removes AEDPA as a bar to relief. See Lafler v. Cooper, — U.S. -, 132 S.Ct. 1376, 1390, 182 L.Ed.2d 398 (2012) (holding that when a state court fails to apply Strickland to assess an ineffective assistance of counsel claim, its rationale is subject to de novo review by a federal court reviewing a habeas petition); see also Williams, 529 U.S. at 405-06, 120 S.Ct. 1495; Amado v. Gonzalez, 758 F.3d 1119, 1137-38 (9th Cir.2014).4
The district court concluded that the state post-conviction court did not hold Mann to the wrong standard because “by determining that there was no probability that a different sentence would have resulted if [counsel] had presented the omitted information concerning Petitioner’s social history and the effects of the 1985 traffic accident,” the state court “necessarily found that Petitioner failed to satisfy the Strickland ‘reasonable probability’ standard for prejudice.” However, the. district court misconstrues what the state court decided. To say that the omitted mitigating evidence “would not have changed the sentence imposed” is not at all the same as saying that “there was no probability that a different sentence would have resulted.” Even if the mitigating evidence did not, in fact, change Judge Kelly’s mind, there could still have been a reasonable probability (i.e., a less than fifty percent chance) that it would have changed his mind. See Rompilla, 545 U.S. at 393, 125 S.Ct. 2456 (holding that there was a reasonable probability the omitted *1216mitigating evidence would have swayed the jury’s verdict, even though “it is possible that a jury could have heard it all and still have decided on the death penalty”). In addition, under Arizona law, the Arizona Supreme Court is required to re-weigh the aggravating and mitigating factors. Thus, prejudice cannot be assessed by its impact on the sentencing court alone. Correll v. Ryan, 539 F.3d 938, 956 (9th Cir.2008). Requiring Mann to prove anything more than a reasonable probability was contrary to clearly established Supreme Court precedent. Williams, 529 U.S. at 405-06, 120 S.Ct. 1495.
Relying on Woodford v. Visciotti, 537 U.S. 19, 24, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002) (per curiam), the State argues that we should give the post-conviction court the benefit of the doubt that it applied the correct legal standard. However, in Vis-ciotti the state court “painstakingly describe[d] the Strickland standard” and 'only failed to use the modifier “reasonably” before the term “probable” in three instances. Id. at 23-34. By contrast, the state court here never even mentioned the Strickland standard. Even if the state court had recited the Strickland standard word for word, it would make no difference because, as already explained, the record demonstrates that the court actually held Mann to a different standard. See Sears v. Upton, 561 U.S. 945, 130 S.Ct. 3259, 3264, 177 L.Ed.2d 1025 (2010) (per curiam) (“Although the court appears to have stated the proper prejudice standard, it did not correctly conceptualize how that standard applies to the circumstances of this case.”) (footnote omitted).
In sum, the state post-conviction court wrongly held Mann to the more-likely-than-not standard when it imported the reasoning from its decision denying Mann a new sentencing hearing based on newly discovered evidence. Strickland, 466 U.S. at 694, 104 S.Ct. 2052. Accordingly, AED-PA does not constrain this Court from finding that Mann was prejudiced by his counsel’s performance. Williams, 529 U.S. at 406, 120 S.Ct. 1495.
B
Because the state post-conviction court did not reach the deficiency prong of the Strickland analysis, our review of this prong is not circumscribed by AEDPA.5 See Wiggins, 539 U.S. at 534, 123 S.Ct. 2527. Reviewing this prong de novo, we conclude that counsel’s performance at sentencing was constitutionally deficient.
The Supreme Court has “declined to articulate specific guidelines for appropriate attorney conduct, and instead [has] emphasized that ‘the proper measure of attorney performance remains simply reasonableness under prevailing professional norms.’ ” Id. at 521, 123 S.Ct. 2527 (quoting Strickland, 466 U.S. at 688, 104 S.Ct. 2052) (internal brackets omitted). However, the Court has recognized that counsel has an obligation at the penalty phase “to conduct a thorough investigation of the defendant’s background.” Williams, 529 U.S. at 396, 120 S.Ct. 1495 (citing ABA Standards for Criminal Justice 4-41, commentary, p. 4-55 (2d ed.1980)).
The ABA standards have long provided guidance for judging the reasonableness of counsel’s conduct, including counsel’s investigation into the defendant’s background and potential mitigating evidence. Wiggins, 539 U.S. at 524, 123 S.Ct. 2527. According to the relevant standards in effect at the time of Mann’s sentencing, counsel’s investigation “should comprise efforts to discover all reasonably available mitigating evidence and evidence .to rebut *1217any aggravating evidence that may be introduced by the prosecutor.” ABA Guidelines for the Appointment and Performance of Counsel in Death Penalty Cases 11.4.1(C) (1989). The investigation into potential mitigating evidence was to “begin immediately upon counsel’s entry into the case” and was “to be pursued expeditiously.” Id. 11.4.1(A). Counsel was advised to interview the client within twenty-four hours — or otherwise as soon as possible— after counsel’s entry into the case to “explore the existence of ... any mitigating factors.” Id. 11.4.1(D)(2)(B). Counsel was also advised to consider interviewing potential “witnesses familiar with aspects of the client’s life history that might affect ... mitigating evidence to show why the client should not be sentenced to death.” Id. 11.4.1(D)(3)(B).
As measured against prevailing professional norms and the recognized obligations of capital defense counsel, there can be little doubt that counsel’s performance fell below an objective standard of reasonableness. In his post-conviction hearing testimony, counsel admitted he was “not focused” on mitigation prior to Mann’s conviction. Counsel testified that neither he nor the investigator he employed interviewed Mann prior to trial about potential sources of mitigating evidence. Counsel never interviewed Miller, and never even investigated the possibility that she would testify regarding mitigating evidence, even though she lived with Mann for ten years and was familiar with Mann’s life history. And he made no attempt to obtain Mann’s school, prison, or medical records, even though this Court has found that such records are “fundamental to preparing for virtually every capital sentencing proceeding.” Robinson v. Schriro, 595 F.3d 1086, 1108-09 (9th Cir.2010). These basic failures left counsel totally unprepared to proceed to the sentencing phase after conviction, as evidenced by counsel’s request for a continuance to conduct further investigation. Clearly, counsel did not conduct his investigation “immediately” and “expeditiously” according to prevailing professional norms.
The investigation that counsel did conduct was unreasonable “in light of what counsel actually discovered.” Wiggins, 539 U.S. at 525, 123 S.Ct. 2527. From Mann’s handwritten autobiography, counsel learned that Mann had sustained a concussion in his 1985 traffic accident, yet counsel made no effort to obtain medical records or otherwise investigate whether Mann had ever suffered from organic brain damage.
In declining to pursue that lead, counsel ignored a death penalty expert at the Phoenix Capital Representation Project who advised him to seek neuropsychological testing to detect the existence of organic brain damage, and he disregarded the fact that Arizona courts at the time placed “significant weight” on brain injuries as mitigating evidence. Correll v. Ryan, 539 F.3d 938, 950 n. 3 (9th Cir.2008). Counsel’s failure to pursue these leads further renders his performance deficient. See Wiggins, 539 U.S. at 525, 123 S.Ct. 2527 (“[A]ny reasonably competent attorney would have realized that pursuing these leads was necessary to making an informed choice among possible defenses.... ”); Lambright v. Schriro, 490 F.3d 1103, 1117 (9th Cir.2007) (“[Wjhen tantalizing indications in the record suggest that certain mitigating evidence may be available, those leads must be pursued.”) (internal quotation marks and citation omitted).
This is not a case where counsel made a reasoned strategic decision to not pursue a mitigation strategy or not present certain mitigating evidence. A decision to not pursue or present mitigating evidence cannot be considered a reason*1218able strategic decision unless counsel supports that decision with investigation. Wiggins, 539 U.S. at 521, 123 S.Ct. 2527 (explaining that “the deference owed such strategic judgments” is defined “in terms of the adequacy of the investigations supporting those judgments”); Strickland, 466 U.S. at 691, 104 S.Ct. 2052 (“[CJounsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.”). In other words, “[a]n uninformed strategy is not a reasoned strategy.” Correll, 539 F.3d at 949. “It is, in fact, no strategy at all.” Id.
Here, counsel could not have made a reasoned strategic decision to not present mitigating evidence regarding Mann’s organic brain damage or other life history because he did not even attempt to explore that evidence. Rather, the record “suggest[s] that [his] failure to investigate thoroughly resulted from inattention, not reasoned strategic judgment.” Wiggins, 539 U.S. at 526, 123 S.Ct. 2527. Counsel was not focused on investigating mitigating evidence until after Mann’s conviction, even though he was concerned he was not going to be successful in convincing the jury that Mann acted in self-defense. Indeed, he did not even direct his investigator compile a biographical sketch of Mann until after conviction. When counsel realized that he needed more time to conduct an adequate investigation, he requested and obtained a continuance of the sentencing hearing. But despite gaining the additional time, and without any explanation, counsel never attempted to obtain Mann’s school, prison, or medical records, he never looked further into the circumstances or consequences of Mann’s 1985 traffic accident, and he made no effort to investigate whether Mann had ever sustained organic brain damage.6
Our conclusion that counsel unreasonably failed to thoroughly investigate Mann’s medical and social history is buttressed by the fact that counsel had no indication that such an investigation would be counterproductive, or that presenting additional mitigating evidence would be harmful to Mann’s case. See id. at 525, 123 S.Ct. 2527 (finding that counsel’s failure to pursue additional leads was unreasonable where “counsel uncovered no evidence in their investigation to suggest that a mitigation case, in its own light, would have been counterproductive, or that further investigation would have been fruitless”). Nothing suggested to counsel that interviewing Miller or exploring the possibility that Mann sustained organic brain damage would be fruitless, nor did anything suggest that presenting evidence about Mann’s life history or medical history would be harmful to his case. The absence of such double-edged mitigating evidence, combined with the lack of even a preliminary investigation, sets this case apart from those where counsel’s decision to limit further investigation was reasonable. Cf. Pinholster, 131 S.Ct. at 1406 (holding that counsel could have reasonably decided to not present additional evidence about the defendant’s background where the record showed that counsel spent time investigating the evidence and the defendant’s history of psychotic behavior would likely have engendered a nega*1219tive reaction in the jury); Burger v. Kemp, 483 U.S. 776, 794-95, 107 S.Ct. 3114, 97 L.Ed.2d 638 (1987) (concluding that counsel’s limited investigation was reasonable where he interviewed all potential witnesses and discovered evidence that could have affected the jury adversely and contradicted the defense strategy); Strickland, 466 U.S. at 699, 104 S.Ct. 2052 (concluding that counsel made a reasonable strategic choice to not present evidence on the defendant’s character and psychological history because such evidence would have been of little help and may have been harmful).
Under these circumstances, we cannot conclude that counsel made a reasoned decision to not pursue or present additional mitigating evidence. Counsel acted contrary to prevailing professional norms by waiting until after conviction to begin his investigation into Mann’s background. Then, for no apparent reason, counsel never completed that investigation, despite obtaining additional time. Therefore, we conclude that counsel provided constitutionally deficient representation during the sentencing phase of Mann’s trial.
C
Because the state post-conviction court’s determination that Mann was not prejudiced by counsel’s deficient performance was contrary to clearly established federal law, see supra Section III.A, we review the prejudice prong of the Strickland analysis de novo. See Lafler, 132 S.Ct at 1390; see also Williams, 529 U.S. at 406, 120 S.Ct. 1495; Amado, 758 F.3d at 1137-38. We conclude that the district court erred in determining that even if AEDPA is not an obstacle to relief, defense counsel’s deficient performance did not prejudice Mann’s case.
To establish prejudice from counsel’s errors during the sentencing phase of a capital case, the petitioner must show “a reasonable probability that, absent the errors, the sentencer — including an appellate court, to the extent it independently reweighs the evidence — would have concluded that the balance of aggravating and mitigating circumstances did not warrant death.” Strickland, 466 U.S. at 695, 104 S.Ct. 2052. “In making this determination, a court hearing an ineffectiveness claim must consider the totality of the evidence before the judge or jury.” Id.
Here, the mitigating evidence that defense counsel failed to uncover and present would have significantly altered the sentencing profile presented to the sentencing judge. Cf. id. at 699-700, 104 S.Ct. 2052 (finding no prejudice because the omitted mitigating evidence “would barely have altered the sentencing profile”). None of the mitigating evidence counsel actually presented to the sentencing judge mentioned the likelihood that Mann sustained organic brain damage in his 1985 traffic accident, and nothing hinted at Mann’s capacity for remorse or empathy. If counsel had adequately investigated Mann’s background, he could have presented the sentencing judge with a wealth of non-cumulative mitigating evidence from Karen Miller, Mann’s medical records, and the results of neuropsychological testing.
According to Miller’s post-conviction hearing testimony, had counsel interviewed her he would have discovered that she was opposed to sentencing Mann to death and that she was willing to testify in support of a mitigation case at the sentencing hearing. In particular, had she testified at the sentencing hearing she would have recounted the horrific details of Mann’s 1985 traffic accident. According to Miller, Mann was driving his friend’s ex-wife and fourteen-year-old daughter as a *1220favor when his Jeep rolled over and tumbled into a canyon. Mann remained unconscious at the bottom of the canyon for several hours before help arrived and he was airlifted to a hospital.' Meanwhile, his two passengers died.
Miller would have testified that Mann received treatment at the Maricopa County Medical Center in Phoenix, where the medical staff were very concerned about injuries he sustained from blows to his head: She would have recalled that when Mann learned of the deaths of his passengers — especially the report that the girl had been decapitated — he immediately fell into a deep depression. In the subsequent months Mann expressed regret and remorse for the deaths of his passengers, and he told Miller many times that he wished he had died with them.
Miller would have testified that following the accident, she noticed a dramatic and long-term change in Mann’s personality. Before the accident Mann was energetic and supportive, but afterward he was very depressed and worried about the financial pressures his medical treatment had brought on his family. He started dealing cocaine to bring in money, and he became abusive and aggressive. According to Miller, Mann became different “in every way.”
Had counsel sought and presented Mann’s hospital records from his treatment at the Maricopa County Medical Center, the sentencing judge would have seen that Mann had no memory of the accident and had sustained an injury over his right eye. He would also have seen notations that Mann was “suffering from grief reaction.”
Counsel’s failure to seek and present the results of neuropsychological testing was especially damaging to Mann’s defense at sentencing. In May 2001, Dr. Comer, a clinical neuropsychologist, conducted a battery of neuropsychological tests on Mann that had never been performed on him before.' His report concluded that Mann “appears to have experienced a TBI [traumatic brain injury] leading to subtle though lasting cognitive defects and other symptoms of post-concussional syndrome.” Mann’s performance on many of the tests was within the normal range, but Dr. Comer found that “[t]he exceptions to this pattern ... represent the type of subtle cognitive impairment that has been identified in persons who have suffered traumatic brain injury (TBI).” For example, Dr. Comer found that Mann was “seriously impaired” (i.e., scored in the 0.6 percentile rank) on a test that is a highly sensitive indicator for previous traumatic brain injury in individuals who have otherwise recovered well cognitively. He noted that the reported changes in Mann’s personality and behavior following his 1985 accident are consistent with a history of traumatic brain injury. He also substantially ruled out depression and past cocaine use as an explanation for Mann’s test results.
During post-conviction proceedings, Dr. Comer testified that he would have expected the same tests to show signs of traumatic brain injury if they had been performed on Mann at the time of his trial and sentencing. He also testified that the behavioral effects of the brain injury— aggression, irritability, and egocentricity— would have been more pronounced closer in time to the original trauma.
Counsel’s failure to fully investigate the circumstances and effects of Mann’s 1985 traffic accident likely affected the psychological evaluation presented to the sentencing judge. Dr. Flynn diagnosed Mann with antisocial personality disorder and concluded that he was likely a psychopath, but Dr. Comer testified that a diagnosis of antisocial personality disorder requires ruling out organic brain injury because the *1221symptoms of the former tend to mimic the symptoms of the latter. Dr. Comer testified that Dr. Flynn’s method of evaluation — a clinical interview — could not 'have detected Mann’s organic brain injury, which suggests that Dr. Flynn’s report presented the sentencing judge with an incomplete and misleading picture of Mann’s mental and cognitive health. '
Had counsel provided Dr. Flynn with the results of an adequate investigation into Mann’s past, including Miller’s description of the effects of the 1985 traffic accident, the sentencing judge likely would have seen a different psychological profile. On December 1, 2000, Dr. Richard Hinton, a clinical psychologist, conducted a psychological evaluation of Mann. Unlike Dr. Flynn, Dr. Hinton was aware of the details of the traffic accident, and he was aware that Mann experienced lingering physical and emotional effects, including frequent headaches, a sense of guilt, and an increase in his use of drugs and alcohol. Dr. Hinton found that Mann’s “functioning changed dramatically following the automobile accident,” that “he was tormented by a sense of guilt that he was somewhat responsible for the death of the two passengers in his car,” and that following the accident “he began to behave much more aggressively and to use cocaine much more regularly.” Unlike Dr. Flynn, Dr. Hinton recognized that this change in functioning occurred independently of Mann’s difficult childhood, substance abuse, and past criminal history. Consequently, although Dr. Hinton suggested that a personality disorder diagnosis was possible, his evaluation did not contain any suggestion that Mann was a psychopath.
The district court denied that Mann’s sentencing profile would have been different had counsel properly uncovered and presented mitigating evidence. Specifically, the district court concluded that “Mann cannot show prejudice because Miller’s testimony at the [post-conviction] hearing about the circumstances of Bazurto’s death was not substantively different from her trial testimony.” The district court’s conclusion is only halfway correct. While it is true that Miller’s testimony would not have altered Mann’s sentencing profile with regard to the aggravating factors, the district court overlooked the importance of Miller’s testimony as a source of mitigat- ‘ ing evidence. Miller’s testimony would have clearly shown Mann’s capacity for profound remorse, evidence of which was utterly lacking from the profile counsel presented, as the sentencing judge clearly pointed out.
In a similar vein, the State denies prejudice by questioning the value of Dr. Comer’s neuropsychological report as mitigating evidence. First, the State attempts to discredit Dr. Comer’s report by pointing out that he did not review Mann’s hospital records from 1985. But this attack falls short because Dr. Comer’s report states that if Mann’s medical records showed signs of unconsciousness and amnesia— which they do — then the diagnosis of traumatic brain injury wquld be strengthened. Second, the State attempts to discredit Dr. Comer’s report by pointing out that he did not review Dr. Flynn’s report diagnosing Mann with antisocial personality disorder. But, as already noted, Dr. Comer’s testimony made clear that Dr. Flynn’s report was irrelevant to the neuropsychological evaluation because a diagnosis of antisocial personality disorder requires ruling out organic brain injury, not the other way around. Third, the State questions the value of Dr. Comer’s report because the report found that Mann had “recovered rather well” and that the effects of his injury were “subtle.” ■ But the fact that Mann had “recovered rather well” by 2001 does nothing to refute the conclusion that he suffered a significant brain injury in *12221985. Furthermore, Dr. Comer expressly rejected any suggestion that his use of the word “subtle” meant “trivial”; by using that word, he meant to indicate that after more than a decade, the lasting effects of Mann’s 1985 brain injury could only be detected with sophisticated tests.
Finally, the State echoes the district court’s conclusion that the results of the neuropsychological testing were of little mitigating value because they were “equivocal at best.” However, Dr. Comer’s report and testimony are anything but equivocal — he was clear that based on his evaluation there is a high probability that Mann suffered traumatic brain injury.
Apart from questioning the mitigating value of Miller’s testimony and Dr. Comer’s report, the district court found no prejudice because the sentencing judge declared that the proffered evidence “would not have changed the sentence imposed.” However, as already explained, the standard for prejudice under Strickland is not whether the omitted mitigating evidence would necessarily have changed the outcome. More importantly, under Strickland the sentencing judge is not the only sentencer who matters; it also matters whether there is a reasonable probability that “an appellate court, to the extent it independently reweighs the evidence [,] would have concluded that the balance of aggravating and mitigating circumstances did not warrant death.” Strickland, 466 U.S. at 695, 104 S.Ct. 2052 (emphasis added). In this case, as in every capital case, the Arizona Supreme Court reviewed Mann’s death sentence and independently reweighed the aggravating and mitigating factors. Mann, 934 P.2d at 790 (citing State v. Brewer, 170 Ariz. 486, 826 P.2d 783, 797 (1992)). Thus, we must also look to the effect of the omitted mitigating evidence on the Arizona Supreme Court in determining prejudice.
Comparing the sentencing profile that was actually presented to the sentencing judge and the Arizona Supreme Court with the sentencing profile that would have been presented absent counsel’s deficient performance, there is a reasonable probability that Mann would have received a sentence other than death had counsel performed adequately. In rejecting the mitigating factor of remorse, the sentencing judge paid special attention to the fact that Mann’s handwritten autobiography indicated no remorse for the deaths of his two passengers in the 1985 accident. The sentencing judge also focused on Dr. Flynn’s psychological report diagnosing Mann as a psychopath with anti-social personality disorder. Both of these observations led the sentencing judge to conclude that Mann “is incapable of remorse” and “has no conscience,” a conclusion which the Arizona Supreme Court left undisturbed. Mann, 934 P.2d at 794.
Had Mann’s counsel performed according to professional norms, he would have presented the sentencing judge and the Arizona Supreme Court with evidence that Mann was in fact capable of profound remorse. Mann’s counsel also would have presented evidence that Mann had suffered a serious brain injury that altered his personality, which could very well have altered the balance of aggravating and mitigating factors because Arizona courts at the time gave significant weight to organic brain injuries as mitigating evidence. Correll, 539 F.3d at 950 n. 3. Given that the aggravating and mitigating factors were already closely -balanced, the failure of counsel to perform adequately “undermine[s] confidence in the outcome” of Mann’s sentencing. Strickland, 466 U.S. at 694, 104 S.Ct. 2052; see also Sears, 130 S.Ct. at 3266 (finding prejudice even where counsel had proven some mitigating factors). Therefore, we conclude that Mann’s *1223counsel’s deficient performance prejudiced his penalty-phase defense.
IV
Defense counsel’s decision to not call Mann to testify at the guilt phase of his trial was within the wide range of reasonable professional assistance. Counsel did not promise the jury that Mann would testify, and to the extent he made any implicit promises regarding what the evidence would show, he did not break them.
By contrast, defense counsel’s performance during the penalty phase of Mann’s trial fell below an objective standard of reasonableness. Counsel failed to expeditiously conduct a reasonable investigation of Mann’s background and potential sources of mitigation. When counsel gained additional time to conduct an adequate investigation, he never followed through. There was no explanation for this failure, and nothing indicated that additional investigation would be fruitless or that the omitted mitigating evidence would be harmful.
Counsel’s failure to uncover and present reasonably available mitigating evidence left the sentencing judge and the Arizona Supreme Court with an incomplete and inaccurate picture of Mann. Had counsel performed adequately, there is a reasonable probability that the sentencers would have concluded that the balance of aggravating and mitigating factors did not warrant death.
Therefore, we conclude that Mann was denied his Sixth Amendment right to the effective assistance of counsel. We reverse the judgment of the district court, and remand with instructions to grant the writ of habeas corpus conditional on the state conducting a new sentencing.
Because we grant relief on Mann’s claim of ineffective assistance for failure to investigate and present reasonably available mitigating evidence, we do not address his claim of ineffective assistance for counsel’s failure to retain an independent mental health expert.
Each side should bear its own costs.
AFFIRMED in part; REVERSED in part.
. A few facts in the record appear to be inconsistent with counsel's version of the events. Before trial, counsel wrote a letter to Mann, stating: “The only remaining decision for .you to make is whether you will testify.’’ He explained that the benefit of testifying would be that Mann could bolster his claim of self-defense by telling the jury that Bazurto "went for his gun” but the "huge risk” in testifying was that he would "be cross-examined regarding [] previous criminal activity, and ... abuse toward [his ex-girlfriends].” Moreover, prior to trial, counsel told the judge that Mann would be the defense’s only witness. Counsel also filed motions to limit the impeachment of Mann's testimony, and he submitted a jury instruction on the competence of defendants to testify as witnesses. Moreover, if counsel intended to withdraw during trial to allow Mann to testify, then he was virtually guaranteeing a mistrial.
Even if we were to conclude from these inconsistencies that counsel was not credible and that the state post-conviction court was unreasonable in concluding otherwise, the result would be that we would be compelled to accept Mann’s version of the facts as true. Mann’s version is plausible, and if counsel’s version is discredited, there is no basis for rejecting it. As explained infra, however, even under his version of the facts, Mann is not entitled to relief under Strickland.
. Mann never presented the substance of this claim to the state post-conviction court or the Arizona Supreme Court, so it is unexhausted. Picard v. Connor, 404 U.S. 270, 275, 92 S.Ct. 509, 30 L.Ed.2d 438 (1971). We may nevertheless consider it because the state expressly waived the exhaustion requirement when, in its answer to Mann’s amended habeas petition, it stated that Mann exhausted this claim and it responded at length to Mann’s arguments that defense counsel was ineffective for making unfulfilled promises to the jury. 28 U.S.C. § 2254(b)(3); see Sharrieff v. Cathel, 574 F.3d 225, 229-30 (3d Cir.2009).
. Because we conclude that the district court correctly rejected Mann's guilt-phase Strickland claim, we need not and do not determine' whether habeas relief on this claim is barred* under 28 U.S.C. § 2254(d).
. The dissent claims "the state court's reasoning here was not unambiguously contrary to federal law” because nothing in the opinion speaks to “the probability that the new evidence would change the sentence.” In its analysis of Mann’s state law claim for a new hearing based on newly discovered evidence, the state court expressly stated that "the evidence must satisfy the requirement that it probably would have changed the verdict or sentence.” When it referred to and incorporated the foregoing analysis in its discussion of Mann’s Strickland claim, the state court did not reformulate the standard of review. This reliance on a more-likely-than-not standard is not consonant with the reasonable probability standard demanded by precedent of the United States Supreme Court. See 466 U.S. at 695, 104 S.Ct. 2052.
. The district court did not reach this prong of the Strickland analysis either.
. The dissent claims that the absence of any reference to brain injury in the court-appointed psychologist’s report excuses counsel’s deficient performance. Dis. Op. 1227. This is not a case where an expert affirmatively "told counsel that [the defendant] did not appear to suffer from brain damage[.]” Cullen v. Pinholster, -U.S. -, 131 S.Ct. 1388, 1405, 179 L.Ed.2d 557 (2011). Rather, the expert report makes no reference to the possibility of brain injury based on Mann’s 1985 traffic accident. This does not excuse counsel's failure to investigate the possibility that Mann suffered brain damage as a result.