dissenting:
I respectfully dissent from the court’s conclusion that Correll has met the “highly demanding and heavy burden of establishing actual prejudice” in the pursuit of his claim of ineffective assistance of counsel during the penalty phase of the trial. Allen v. Woodford, 395 F.3d 979, 1000 (9th Cir.2005) (quoting Williams v. Taylor, 529 U.S. 362, 394, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000)) (internal quotation marks omitted). The majority ignores the mountain of precedent which requires us, in assessing prejudice, to consider not only the likely benefits of the mitigating evidence Correll’s counsel failed to present, but also its likely drawbacks. In addition, the majority substitutes its independent analysis of the record for that of the district court, relying on its own view of the evidence rather than considering, as we must, the effect the evidence would have had on an Arizona sentencing judge 23 years ago. Because I do not believe that Correll has met his burden “affirmatively [to] prove prejudice,” I would affirm the judgment of the district court denying the petition for writ of habeas corpus. See Strickland v. Washington, 466 U.S. 668, 693, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).
I
The facts of Correll’s brutal crimes are disturbing, but must be recounted to illustrate the unlikelihood that Correll’s new evidence would have convinced the sentencing judge not to impose the death penalty.1
A
On the night of April 11, 1984, as Guy Snelling and his girlfriend Debra Rosen were getting ready to go to sleep, a knock came at the door. Snelling answered the door and found John Nabors, his co-worker, and Correll, whom he had not met.
After Snelling let the two men into his home, Nabors pulled a gun and demanded money. Correll secured Snelling and Ro-sen with duct tape. When Robin Cady and Shawn D’Brito, two friends of Snell-ing, unwittingly arrived at the house, Cor-rell secured them with duct tape as well. *957Then Correll and Nabors escorted Snelling throughout his home to search for money and valuables.
After raiding the house for approximately 45 minutes, Nabors and Correll exited with Cady, D’Brito, and Snelling, whom they forced into Cady’s car. Nabors briefly went back inside to secure Rosen. While holding the gun on the three victims, Correll drove to a deserted area where Nabors’s truck was parked. Na-bors took his truck and followed Correll, who was still driving Cady’s car with the three victims, to a desert area north of Phoenix. There, they forced the three victims out of the car and made them lie face down on the ground. Correll shot Snelling in the back of the head. Nabors then shot and killed D’Brito, and then tried to shoot Cady. The gun misfired a couple of times and Correll said “hurry up, hurry up, ... okay, it’s cool, no cars coming, get a shell chambered.” After reloading the gun, Nabors was finally successful in shooting and killing Cady. After Correll and Nabors left, Snelling, who miraculously did not die, reported the crime. Rosen, whom Nabors and Correll had left in the house when they drove the other three victims into the desert, was later found in the house, killed by strangulation.
B
At trial, Correll’s sole defense was mis-identification — namely, that Snelling, who was under the influence of drugs and alcohol when the crimes occurred, had wrongly identified Correll as one of his assailants, and that it was reasonably likely that Cor-rell’s brother Terry, who resembled Cor-rell, had committed the crimes instead. Unpersuaded by this defense, a jury convicted Correll of three counts of first degree murder, one count of attempted first degree murder, one count of armed robbery, one count of first degree burglary, and four counts of kidnaping.
At sentencing, the government urged the court to impose the death penalty. The government asserted that five statutory aggravating factors were present: (1) a previous violent felony conviction;2 (2) grave risk of death to others in addition to the persons murdered;3 (3) commission of the murders in anticipation of pecuniary gain;4 (4) commission of the murders in an especially heinous, cruel or depraved manner;5 and (5) convictions for multiple murders during the offense.6
In response, Correll’s attorney argued that the prosecution had failed to prove, as required by Enmund v. Florida, 458 U.S. 782, 102 S.Ct. 3368, 73 L.Ed.2d 1140 (1982), that Correll intended to kill Rosen, Cady, and D’Brito. Although the sentencing court did not accept this argument, Correll’s attorney preserved it for appeal and the Arizona Supreme Court later modified one of Correll’s death sentences to life imprisonment on this ground. See State v. Correll, 148 Ariz. 468, 715 P.2d 721, 730-31 (1986). Correll’s attorney also countered each of the government’s proffered aggravating factors.7 He argued— *958and the sentencing court agreed — that the “grave risk of death to others” aggravating factor did not apply. He also argued that the multiple murder aggravating factor could not be considered. Although the sentencing court did not accept this argument, Correll’s attorney preserved it for appeal and the Arizona Supreme Court later invalidated this aggravating factor. See id. at 734-35. Correll’s attorney further argued, unsuccessfully, that the evidence did not support the remaining aggravating factors.
In addition to challenging the government’s aggravating factors, Correll’s attorney also presented substantial mitigating evidence.8 First, counsel emphasized that it was John Nabors, not Correll, who actually shot the three victims who died. Second, counsel endeavored to present Nabors as the “leader” and “planner” of the criminal endeavor, accentuating the facts that “Mr. Nabors was the one that knew Guy Snelling was a drug dealer,” and that “Snelling would have money and drugs [when] the robbery occurred.” In addition, counsel drew the court’s attention to Snelling’s statement that “it appeared to him that John Nabors was the leader, was the one calling the shots, so to speak.” Finally, counsel pointed out that, prior to the robbery, it was impossible for Correll reasonably to have anticipated that anyone would be present in the home other than Snelling, and that, consequently, Correll could not have planned the three deaths.9
Third, Correll’s attorney also argued for mitigation, both in his sentencing memorandum and at oral argument, on the grounds that Correll was under the influence of drugs and alcohol at the time of the murders.10 Counsel specifically drew the sentencing judge’s attention to Snell-ing’s statement to the police that he smelled alcohol on Correll’s breath during the crimes.
Fourth, counsel presented Correll’s troubled family history, explaining that “the reason that Mike [Correll] has had problems is the fact that when he was 14 years old, that both of his parents abandoned him and what can be expected when someone is abandoned by their parents at such an early age?” Finally, Correll’s attorney also argued that Correll’s age— 24 — was mitigating.11
Although Correll’s attorney knew that Correll had received psychological counseling, he declined to develop psychological evidence because he believed, based on his conversations with Correll, that the only possible diagnosis was antisocial personality disorder. As counsel explained at the evidentiary hearing, he believed such a diagnosis would carry little, if any, mitigat*959ing weight with the sentencing judge and would, in fact, make it easier for the judge to sentence Correll to death because it would cause him to view Correll as permanently psychologically damaged.12
C
The sentencing judge ultimately found four statutory aggravating circumstances.13 Determining that the mitigating evidence did not outweigh these factors, the judge sentenced Correll to death on each of the murder counts. The Arizona Supreme Court affirmed Correll’s convictions, with the modifications previously mentioned. It then re-weighed the aggravating and mitigating factors and determined that the death penalty was appropriate. Correll, 715 P.2d at 736.
In his state petition for postconviction relief, Correll alleged that his counsel rendered ineffective assistance at sentencing. He contended that during the month that elapsed between the jury verdict and the sentencing hearing, his attorney met with him for just five minutes. He also contended that his attorney failed to investigate and to develop available evidence relating to his psychiatric history and condition at the time of the crimes. The state trial court summarily dismissed Correll’s petition, concluding that Correll raised “no colorable issues” relating to ineffective assistance of counsel. The court went on to explain that “the Court specifically recalls that the trial work of defense counsel was precise, careful, and competent, and manifested strategic and tactical judgments of the same high quality.” The Arizona Supreme Court denied review without comment.
Correll later filed a federal petition for writ of habeas corpus and the district court entered summary judgment against him. On appeal (“Correll I ”), we held that Cornell's ineffective assistance allegations, which had not been fully explored in state court, entitled him to an evidentiary hearing. We held that Correll had established (1) that the state court trier of fact had not conducted a full and fair hearing to find the relevant facts, and (2) that his allegations, if proven, might constitute a color-able ineffective assistance claim. Correll v. Stewart, 137 F.3d 1404, 1411-12 (9th Cir.1998).
D
Pursuant to our instructions on remand, the district court conducted an evidentiary hearing on Correll’s ineffective assistance of counsel claim. The evidentiary hearing lasted nine days. The district court heard testimony from 17 witnesses, 14 called by Correll (who waived his appearance), and three called by the government. In addition, the district court reviewed reams of documents, including Correll’s attorney’s notes, which were nearly a quarter-century *960old, and Correll’s childhood medical records, which were two decades older.
After outlining all the evidence in a detailed 109-page disposition, the district court made several findings. First, as to the sufficiency of counsel’s consultation with Correll, the court rejected Correll’s allegation that counsel spent only five minutes with him between conviction and sentence. Instead, the district court found that “[pjrior to sentencing, [counsel] had multiple face-to-face meetings and phone calls with Petitioner” in which he “discussed] with Petitioner the overall mitigation case and the specific reasons he would present to the court in favor of a life sentence rather than the death penalty.”
Second, as to the sufficiency of counsel’s investigation of possible mitigating evidence, the district court found that counsel spoke to between 40 and 50 witnesses, including every member of Correll’s family who would cooperate. The district court further found that, unfortunately, “[t]he witnesses were not able to provide relevant useful mitigation information.” In fact, “in many instances, the witnesses only provided inculpatory and non-mitigating information.”
The district court did find that counsel’s performance was constitutionally deficient in two respects: (1) counsel’s failure to obtain medical treatment records relating to the head injury Correll suffered at seven years old and (2) counsel’s failure to thoroughly review Correll’s mental health records. The court determined that a reasonable attorney would have investigated these matters for possible mitigating evidence rather than relying on his own impression, based on his interaction with the defendant, that the defendant had no intellectual or psychological deficits that could serve as mitigating evidence.
Nevertheless, the district court found that Correll was not prejudiced by these errors. After Correll’s postconviction counsel developed all the evidence relating to Correll’s head injury and mental health history, the district court still found Cor-rell to be a “highly functioning adult” who never suffered from brain damage or a major psychological disorder. Thus, the district court found that Correll’s medical and mental health records provided no substantial evidence of mitigation. Furthermore, the district court found that much of the new evidence Correll offered would have been counterproductive if put before the sentencing judge because it would have “opened the door for the prosecution to come forward with strong damaging rebuttal information to counter its mitigating effect.”
II
In reversing the district court’s judgment, the majority concludes that the district court committed “clear error” in finding that counsel’s investigation and presentation of possible mitigating defenses was constitutionally sufficient. Under the clearly erroneous standard of review, our scrutiny of a district court’s factual findings must be “significantly deferential, in that we must accept the district court’s factual findings absent a definite and firm conviction that a mistake has been committed.” Hovey v. Ayers, 458 F.3d 892, 900 (9th Cir.2006) (quoting Silva v. Woodford, 279 F.3d 825, 835 (9th Cir.2002)) (internal quotation marks omitted).14 In other words, as long as the district court’s account of the evidence “ ‘is plausible in *961light of the record viewed in its entirety, the court of appeals may not reverse it even though convinced that had it been sitting as the trier of fact, it would have weighed the evidence differently.’ ” Phoenix Engineering and Supply Inc. v. Universal Elec. Co., Inc., 104 F.3d 1137, 1141 (9th Cir.1997) (quoting Anderson v. Bessemer City, 470 U.S. 564, 573-74, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985)).
Unfortunately, the majority ignores these instructions and embarks on its own independent examination of the facts presented to the district court. By contrast, as indicated by its exhaustive 109-page disposition, the district court’s findings were well-supported by the facts and reached only after a thorough review of all available evidence. I simply cannot agree that the district court’s findings were erroneous at all, let alone clearly erroneous.
More alarming than its reconstruction of the record, however, the majority jumps with startling speed from its new factual determination that Correll received ineffective assistance of counsel to its ultimate conclusion that his habeas petition must be granted. In so doing, the majority ignores Strickland’s second requirement, that even if Correll proves ineffective assistance of counsel, he must also prove that the result was actually prejudicial. Wiggins, 539 U.S. at 534, 123 S.Ct. 2527 (citing Strickland, 466 U.S. at 692, 104 S.Ct. 2052). As the Supreme Court has made clear, we do not presume prejudice from counsel’s ineffective assistance. Strickland, 466 U.S. at 693, 104 S.Ct. 2052. Rather, even if counsel’s performance was deficient, Correll still bears the “highly demanding and heavy burden of establishing actual prejudice.” Allen, 395 F.3d at 1000 (internal quotation marks omitted) (emphasis added). This burden “affirmatively[to] prove prejudice” requires Correll to show more than the mere possibility that counsel’s performance prejudiced the outcome. Strickland, 466 U.S. at 693, 104 S.Ct. 2052. Instead, Correll must demonstrate “a reasonable probability” that, but for counsel’s constitutionally deficient performance, he would have received a lesser sentence. Id. at 695, 104 S.Ct. 2052.
In assessing prejudice in this case, it is important to remember that “we are not asked to imagine what the effect of certain testimony would have been upon us personally,” Smith v. Stewart, 140 F.3d 1263, 1270 (9th Cir.1998), or even to imagine the effect of such testimony on an abstract juror. Instead, we must determine what the effect of Correll’s new evidence would have been upon the Arizona sentencing judge at the time of Correll’s sentencing hearing 23 years ago. Id. As discussed below, none of the evidence unearthed during the district court’s evidentiary hearing creates a “reasonable probability” that Correll would have received a lesser sentence had it been presented at the sentencing phase of his trial. Accordingly, I cannot conclude that Correll has met the heavy burden required to establish prejudice.
A
First, Correll’s attorney’s failure to obtain the medical records relating to Cor-rell’s childhood head injury was not prejudicial because these records did not, in fact, demonstrate any brain damage. After receiving testimony from neuropsycholo-gists, the district court found that Correll “did not suffer any brain injury from the block wall that fell on him when he was 7 years old.” Quite to the contrary, the district court credited a neuropsychologist’s testimony that “of all the capital defendants he has tested, Petitioner is one of the highest functioning.”
The medical records from the incident support this assessment. After his childhood injury, Correll was diagnosed with a *962subgaleal hematoma, which is a bruise or collection of blood under the scalp, but above the skull. The hematoma cleared in five days, at which time a doctor described the seven-year-old Correll as alert and well. I accordingly cannot agree with the majority’s conclusion that Correll has carried his burden to establish a reasonable probability that he would have received a lesser sentence if the records relating to his childhood head injury had been before the sentencing judge. Indeed, as we have suggested before, counsel’s failure to present “mitigating evidence may be irrelevant when no substantial mitigating evidence is available.” Smith v. Stewart, 189 F.3d 1004, 1013 n. 4 (9th Cir.1999) (citing Ger laugh v. Stewart, 129 F.3d 1027, 1042 (9th Cir.1997)). The medical records, which led the district court to conclude that Correll was a “highly functioning adult,” presented no opportunity for mitigation.
B
In assessing the remainder of Correll’s mitigating evidence, it is important to emphasize that the majority’s conclusion that Correll has met the heavy burden of demonstrating actual prejudice ignores a mountain of precedent which requires us to consider not only the benefits of the ostensibly mitigating evidence counsel failed to present, but also its potential drawbacks. In Darden v. Wainwright, 477 U.S. 168, 186, 106 S.Ct. 2464, 91 L.Ed.2d 144 (1986), trial counsel’s failure to present any mitigating evidence did not constitute deficient performance because the presentation of such evidence would have opened the door to damaging rebuttal evidence. Similarly, in Burger v. Kemp, 483 U.S. 776, 107 S.Ct. 3114, 97 L.Ed.2d 638 (1987), trial counsel’s failure to present psychological records did not amount to ineffective assistance because the records were “by no means uniformly helpful to petitioner,” as they suggested “violent tendencies” that would have undermined counsel’s strategy of portraying petitioner’s actions as the result of another person’s “strong influence upon his will.” Id. at 793, 107 S.Ct. 3114. Based on these cases, we have held that an attorney who failed to present psychological testimony relating to the defendant’s antisocial personality disorder was not ineffective because such testimony “would have allowed the prosecution during cross-examination and rebuttal to rehash the horrific details of [the] crimes.” Bonin v. Calderon, 59 F.3d 815, 836 (9th Cir.1995). Most recently, the Supreme Court, in Wiggins v. Smith, 539 U.S. 510, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003), repeatedly emphasized that the Darden-Burger line of cases remains in effect. In Wiggins, the Court held that the petitioner had met his heavy burden of proving actual prejudice because “Wiggins d[id] not have a record of violent conduct” that the State could have introduced to offset the mitigating evidence his attorney failed to offer. Id. at 537, 123 S.Ct. 2527. The majority fails to realize that unlike Wiggins, much of the new mitigating evidence Correll offers would have enabled the prosecution to present very damaging evidence in rebuttal. Indeed, Correll’s mitigating evidence presents precisely the type of “double edge” the Supreme Court found lacking in Wiggins’s case. Id. at 535, 123 S.Ct. 2527 (distinguishing the mitigating evidence presented by Wiggins from the double-edged evidence presented in Burger, 483 U.S. 776, 107 S.Ct. 3114, and Darden, 477 U.S. 168, 106 S.Ct. 2464).
1
I begin with Correll’s new psychiatric evidence, which would not have significantly helped his case. The district court found that “there is insufficient evidence to support that Petitioner has ever suffered from any major mental illness, whether PTSD [post traumatic stress disorder], a major depressive disorder, or a bipolar *963disorder.” The district court reached this factual finding after two psychological experts testified that there was no evidence Correll has ever suffered from these disorders. The sole witness who speculated that Correll might have suffered from post traumatic stress disorder acknowledged that such a diagnosis was “only a possibility.” The district court found Correll’s self-reporting of bipolar disorder and severe depression incredible in light of Cor-rell’s obvious motive to fabricate and in light of the fact that these diagnoses do not appear in his records and Correll indicated that he was never given medication to treat them.
The district court also found that the evidence did not support Correll’s contention that he was given anti-psychotic medications while in custody. In reaching this factual finding, the district court noted that the mental health experts for both parties scrutinized Correll’s medical records from the California Department of Corrections (“CDC”) and reported the absence of any indication that anti-psychotic medication was ever prescribed. Although it appears that Correll was given Mellaril for a period of time as a juvenile, the government’s mental health expert, Dr. John Scialli, M.D., testified without opposition that the dosage — 25 milligrams— would have served as a mild tranquilizer and was far lower than the dosage that would be utilized to counteract psychosis (approximately 625 milligrams).
Accordingly, had Correll’s attorney thoroughly reviewed Correll’s mental health records, he would have only had credible evidence for the diagnosis he already suspected: antisocial personality disorder accompanied by mild depression. As we have repeatedly acknowledged, a diagnosis of antisocial personality disorder may be “potentially more harmful to [a] petitioner than [helpful].” Gerlaugh, 129 F.3d at 1035. We have explained that, because of its “obvious countervailing tactical dangers,” such evidence “[i]n its best possible light, it is a basket of cobras.” Id. Accordingly, in a prior case, “we c[ould] identify no prejudice flowing from counsel’s failure to develop” psychiatric testimony relating to a defendant’s anti-social personality disorder. Id.; see also Darden, 477 U.S. at 186-87, 106 S.Ct. 2464 (counsel’s decision not to present mitigating character or mental-state evidence was sound trial strategy because it would have opened the door to damaging rebuttal evidence, including a psychiatric opinion that the defendant had a sociopathic personality); Daniels v. Woodford, 428 F.3d 1181, 1204, 1210 (9th Cir.2005) (indicating that testimony suggesting that a capital defendant is a “sociopath” is aggravating rather than mitigating); Beardslee v. Woodford, 358 F.3d 560, 583 (9th Cir.2004) (acknowledging that an antisocial personality diagnosis can be damaging to a capital defendant); Caro v. Woodford, 280 F.3d 1247, 1257 (9th Cir.2002) (concluding that a psychologist’s testimony did not help the defendant’s mitigation case because it tended “to paint him as a violent psychopath”); Clabourne v. Lewis, 64 F.3d 1373, 1384 (9th Cir.1995) (noting that mental health records omitted from the sentencing hearing “hardly turned out to be helpful” because they indicated that the defendant had “an antisocial personality”); Williams v. Calderon, 52 F.3d 1465, 1472 (9th Cir. 1995) (“We have no doubt that ... statements[suggesting that the defendant is so-ciopathic] did nothing to advance Williams’s cause.”).
Furthermore, had counsel presented Correll’s mental health records at sentencing, he would have opened the door for the prosecution to present extremely damaging rebuttal evidence that would have likely eviscerated the minimal mitigating impact these records carried. The district court “credited counsel’s] testimony that the prosecutor, Sidney Davis, had a repu-
*964tation for excellent preparation and that she would have left no stone unturned in her opportunity to rebut any mitigation evidence presented.” Accordingly, the district court found that, had Correll’s attorney presented mental health evidence, this “highly skilled” prosecutor would have presented a great deal of aggravating evidence that was not already before the sentencing judge, specifically: (1) Cornell's rape of a female psychotic patient while being treated for his antisocial personality disorder and mild depression; (2) Cornell's numerous escapes from mental health treatment facilities and numerous rejections of institutional efforts to provide him with mental health treatment; (3) an incident where Correll took hostages in an armed attempt to escape from a mental health treatment facility; (4) the underlying factual basis of Cornell's prior convictions for armed robbery; (5) the conclusion of a social evaluation at age 18 that Correll was a danger to the community and was not a candidate for probation; (6) additional information showing the efforts of Correll’s parents to deal with his drug abuse problem and to obtain psychological treatment for him following his armed threat against a teacher at school; (7) Correll’s statement that he had no desire to work but only wished to enjoy himself; and (8) Cornell's statement that when he committed the 1978 armed robberies that it gave him a strong sense of power and excitement.
Finally, presentation of Cornell's antisocial personality disorder at sentencing would have severely undermined counsel’s strategy of arguing that Correll was merely following Nabors’s lead during the commission of the crimes. Had counsel introduced evidence of Cornell's antisocial personality disorder diagnosis, the prosecution would have almost certainly responded by pointing out that Correll, at age 18, instigated the armed robbery of three convenience stores at gunpoint, an effort in which he enlisted the assistance of his 13-year-old brother and 15-year old girlfriend. We have previously held that counsel’s failure to present psychological evidence is not prejudicial where it would have distracted the fact-finder from counsel’s main mitigation theory and other mitigation evidence. See Bonin, 59 F.3d at 836 (finding that counsel’s failure to present expert psychological testimony was not prejudicial because it “would have distracted jurors ..., reduced [the defendant’s] credibility with the jury, and opened the door to powerful cross-examination and rebuttal”); see also Burger, 483 U.S. at 793, 107 S.Ct. 3114 (holding that a petitioner failed to prove ineffective assistance where the affidavits detailing the defendant’s behavioral history his attorney failed to present “are by no means uniformly helpful to petitioner because they suggest violent tendencies that are at odds with the defense’s strategy of portraying petitioner’s actions on the night of the murder as the result of [another person’s] strong influence upon his will”).
In sum, the psychological evidence, if presented, would have demonstrated only that Correll has an antisocial personality with mild depression. Such evidence has tremendous potential to be more harmful than helpful. Further, such evidence would have opened the door for the prosecution to introduce a laundry list of extremely damaging information not already before the sentencing judge and would have crippled Correll’s chances of convincing the sentencing judge that he was merely following Nabors’s lead during the crimes.15 Accordingly, contrary to the ma*965jority’s conclusion, Correll cannot prove a reasonable probability that he would have received a lesser sentence if the available psychological evidence had been before the sentencing judge.
2
Given the lack of substantial mitigation found in Correll’s medical and psychiatric records, Correll cannot claim to have been prejudiced by counsel’s failure to offer further evidence of Correll’s drug use beyond what he already presented to the sentencing judge. The district court found that there was no evidence — other than Cor-rell’s self-serving statements — that Correll was significantly impaired at the time of the crimes. Arizona law at the time provided that “[a] defendant’s intoxication or alcoholism at the time of the offense is a mitigating circumstance if the evidence shows that it significantly impaired the defendant’s capacity to appreciate the wrongfulness of his conduct or to conform his conduct to the requirements of the law.” State v. Zaragoza, 135 Ariz. 63, 659 P.2d 22, 30 (1983) (emphasis added). The district court specifically found that Cor-rell’s behavior during the murders indicated he was not intoxicated:
[I]t was Petitioner who remained calm when the gun misfired as Nabors was trying to kill Robin Cady. It was Petitioner who encouraged Nabors to remain calm as there were no cars coming, to get a shell chambered and shoot Cady. Such behavior at the time of the crime does not demonstrate intoxication and, in fact, undercuts an assertion of intoxication.
See Williams v. Woodford, 384 F.3d 567, 624 (9th Cir.2004) (reasoning that there is little basis for believing that drugs materially affected the defendant’s behavior at the time of the crimes when the facts of the crimes reflect deliberate and methodical action).
Furthermore, no witnesses could have established that Correll was intoxicated on the date of the crimes. The best evidence Correll can point to would have come from his sister, who could have testified that Correll used methamphetamine in the morning on the day before the crimes. Correll was not prejudiced by counsel’s decision not to present his sister’s testimony, however, because cross-examination would have eviscerated any remaining residual doubt in the sentencing judge’s mind as to Correll’s guilt. Correll maintained his innocence throughout the sentencing proceedings. However, Correll’s sister knew he was with Nabors at the time of the crimes and that they had sought a ride out of the state very soon after the murders occurred. Accordingly, as the district court found, her testimony would have “totally eliminated any mitigating weight” and residual doubt from Cor-rell’s assertion at the guilt phase of his trial that it was his brother, not he, who had committed the murders. See Allen, 395 F.3d at 1004 (explaining that “mitigation witnesses proffered by [the defendant] would not have proved helpful given their own involvement in [the defendant’s criminal enterprise.”); Williams v. Woodford, 384 F.3d 567, 624 (9th Cir.2004) (“[T]he best thing a capital defendant can do to *966improve his chances of receiving a life sentence has nothing to do with mitigating evidence strictly speaking. The best thing he can do, all else being equal, is to raise doubt about his guilt.”).
The only other witness Correll’s post-conviction counsel presented relating to drug use was Dawn Day, who testified that she used methamphetamine with Correll during a four month period from November 1982 until February 1983. We cannot consider Day’s testimony, however, because Correll failed to establish that Day was available to testify at his sentencing hearing. See Douglas v. Woodford, 316 F.3d 1079, at 1086 n. 2 (9th Cir.2003) (explaining that testimony presented at a district court evidentiary hearing that was not available to counsel at the sentencing hearing may not be considered for prejudice purposes). Furthermore, even if Cor-rell had established that Day would have been available, Day’s testimony that Cor-rell used methamphetamine more than one year before the crime would have provided little support for Correll’s argument that, at the time of the crime, he was so impaired that he was unable “to appreciate the wrongfulness of his conduct or to conform his conduct to the requirements of the law.” Zaragoza, 659 P.2d at 30.16
The majority attempts to minimize counsel’s complete inability to present any corroborating evidence that Correll either used methamphetamine on the date of the crimes or appeared intoxicated to anyone he encountered that day by instead pointing to expert testimony that “gross methamphetamine intoxication, unlike gross alcohol intoxication, is not necessarily apparent to outside observers.” Maj. Op. at 954. Consequently, the majority appears content to rely exclusively on Correll’s self-serving statement that he was intoxicated at the time of the crimes to reach the conclusion that counsel’s failure to present further evidence of his drug use was prejudicial. I, on the other hand, would prefer to rely on the credibility findings made by the district court. Those findings bear repeating in full:
The Court does not credit [Correll’s] unsubstantiated self-report that he abused methamphetamine every day before the crimes were committed. Petitioner chose not to testify at the eviden-tiary hearing; Petitioner chose not to fully cooperate with [the government’s drug abuse expert’s] examination of him regarding the issue of drug abuse. Because of the obvious motive to fabricate, Petitioner’s self-serving statements about his drug usage prior to the crimes is [sic] unreliable and subject to searching skepticism. See, e.g., [State v.] Medrano, [185 Ariz. 192] 914 P.2d [225,] 227 [(Ariz.1996) (“the defendant provided most of the information concerning his use of cocaine in the past and on the night of the murder, as well as the drug’s effect on him. Because of the obvious motive to fabricate, such self-serving testimony is subject to skepticism and may be deemed insufficient to establish mitigation.”)]; see also Bernard Smith [v. Stewart ], 140 F.3d [1263,] 1270 [1998] (evaluating evidence based on impartial sentencing judge applying Arizona law); see generally, Strickland, 466 U.S. at 695, 104 S.Ct. 2052 (“The assessment of prejudice should proceed on the assumption that the decision maker is reasonably, conscientiously, and impartially applying the standards that govern the decision.”). The Court’s searching skepticism toward *967Petitioner’s self report is corroborated by Respondent’s drug abuse expert, Dr. Matthews, who opined as follows: “Antisocial personality disorder is characterized by malingering and deceit; instances of [Petitioner’s] lifelong pattern of deceptiveness abound throughout his penal and other records. He has been deceitful about a great many matters, including his history of substance abuse. Because of [Petitioner’s] history of deceit, it is a major clinical error to accept [Petitioner’s] self-serving view of his condition at the time of the offense as accurate.”
Because there was no other evidence to establish that Correll was intoxicated at the time of the crimes, I cannot agree with the majority that Correll was prejudiced by counsel’s failure to present expert testimony regarding the effects of methamphetamine addiction at the sentencing hearing. Conversely, I agree with the district court that counsel’s decision reflects a reasonable strategic choice. First, the district court found that “any expert would have to take into account the underlying facts of the crimes, which show that [Cor-rell] was involved in deliberative acts, such as planning, conspiring, avoiding detection, ... awareness of wrongdoing,” and the fact that he was “generally orientated [as] to time, place, and reality.” This would have materially undermined counsel’s strategy of portraying Correll as merely following Nabor’s lead. Second, the district court acknowledged that any expert “would have been forced to utilize hypothetical supposition regarding [Correll’s] conduct at the time of the crimes,” and that “[s]uch hypothetical supposition would have opened the door for contrary rebuttal argument and reiteration by the prosecution regarding the lack of factual support and incredulity of [Correll’s] alleged intoxicated condition at the time of the crimes.”
Yet it is solely on the strength of such “hypothetical supposition” that the majority now declares that the district court was “clearly wrong” to conclude that there was no evidence to support Correll’s assertion that, on the night of the crimes, he was “grossly intoxicated — to the point of being unable to appreciate the wrongfulness of his conduct,” as required for mitigation. Maj. Op. at 955. The majority points to the testimony of two drug abuse experts presented by Correll’s post-conviction counsel at the evidentiary hearing. The district court, however, reasonably declined to credit these experts’ opinions because they were not based on an examination of Correll but instead were based on a hypothetical set of facts provided by Cor-rell’s postconviction counsel. As the district court explained:
Dr. Sullivan did not examine Petitioner nor did he look at Petitioner’s Arizona Department of Corrections or CDOC records. Rather, Dr. Sullivan was asked to assume [a set of] hypothetical facts[that] do not accurately or reliably portray Petitioner’s alleged drug abuse.... [H]is opinion was based on unsubstantiated and unreliable assumptions.
In addition, Correll’s other expert witness on drug addiction, Dr. Shaw, whom the majority quotes for the proposition that Correll “may have been experiencing drug-induced paranoia” at the time of the murders, Maj. Op. at 953, was “thoroughly impeached” at the evidentiary hearing. As the district court explained, “Dr. Shaw admitted that he only minimally considered the facts of the crime before reaching his conclusion.” The district court found Dr. Shaw’s opinion “entirely not credible and wholly speculative” because, like Dr. Sullivan’s opinion, it was “based upon hypothetical drug usage at the time of the crimes that was not established.”
*968In stark contract to the hypothetical assumptions on which Drs. Sullivan and Shaw based their opinions, the district court found that, except for 229 days, Cor-rell was incarcerated throughout the nine-year period between October 1975 (when he was first incarcerated, at age 14) and March 1984 (one month before the murders) and that Correll “was not a methamphetamine addict or a long-term abuser of methamphetamine during the time he was incarcerated.”
Consequently, I agree with the district court that counsel’s failure to present further evidence of Correll’s drug use was not prejudicial. Counsel had already stated that Correll had been using alcohol and drugs and presented Snelling’s statement that he smelled alcohol on his captor’s breath. I agree with the district court that if Correll’s attorney had called an expert to testify, “it is highly likely any lay witness basis for the expert’s opinion could have been cross-examined at sentencing and impeached by virtue of the fact that no lay witness could testify that Petitioner was intoxicated at the time of the crimes.” I also credit the district court’s observation that “if an expert had testified based solely on Petitioner’s self-reporting ... it is very likely that the expert’s opinion would have been severely undermined by undisputed evidence that Petitioner had spent almost 9 of the last 10 years incarcerated with little or no access to drugs.”
Nevertheless, the majority’s independent review of the expert testimony leads it to conclude that the evidence “dearly established that methamphetamine use, in the quantities that Correll indisputably used the drug on a regular basis, would significantly impair judgment and consciousness without causing perceptible symptoms of intoxication.” Maj. Op. at 955 (emphasis added). I do not quarrel with the notion that severe use of methamphetamine may significantly impair judgment and consciousness. Whether the symptoms of methamphetamine intoxication are perceptible or not, however, it is quite disputable that Correll used the drug “on a regular basis” and it is entirely unproven that Correll used the drug on the date of the crime.
Accordingly, I share the district court’s inability to find that Correll was prejudiced by counsel’s decision not to present additional evidence of drug use beyond what he already had.17
3
Finally, Correll has presented no credible evidence about his childhood that his attorney could have placed before the sentencing judge other than the evidence the sentencing judge already had before him. The district court, who is in the best position to determine credibility, found Cor-rell’s uncorroborated allegation that his mother banged his head against a kitchen table incredible. In regard to the head injury Correll suffered at age seven when a cinder block wall fell on him, the district court expressly found that Correll’s parents were not negligent in securing medical care. After reviewing the medical records presented at the evidentiary hearing, *969the district court found that Correll’s parents took him to the family doctor the same day the accident occurred and “acted reasonably in caring for Petitioner, which included two visits to their family doctor, one emergency room visit and a follow-up visit for additional specialized testing.”
The majority cites evidence of Correll’s family history provided by Reverend Curry, whom the district court found “was not an available witness” for counsel at the time of the sentencing hearing. The district court found “that if [Reverend Curry] had been contacted by [counsel] prior to sentencing, he would have informed him that he would not discuss information about Petitioner or appear at sentencing because it was against California law for him to discuss former residents of the CYA.”18 Accordingly, Reverend Curry’s testimony cannot factor into the prejudice analysis. See Douglas, 316 F.3d at 1086 (explaining that testimony presented at a district court evidentiary hearing that was not available to counsel at the sentencing hearing may not be considered for prejudice purposes).
The majority suggests that counsel should have presented evidence of Cor-rell’s parents’ religious fanaticism as Jehovah’s Witnesses, specifically their decision, “[a]fter Correll was shot in the arm at age 14,” to “cut off all communication with their son and considerf ] him dead, as required by their church’s teachings.” Maj. Op. at 953. Of course, the majority’s analysis ignores the district court’s finding that, by age 14, Correll had “already been arrested several times,” that his parents had responded by providing him with “extensive psychological treatment,” and that only after “another arrest” did his parents allow him to become a ward of the state.
Similarly, the majority suggests that counsel should have presented evidence that Correll’s parents used corporal punishment “in response to his obvious substance abuse problems.” Maj. Op. at 952. Once again, the majority ignores the district court’s finding that, had counsel emphasized such evidence, the prosecution would have countered with evidence that Correll’s parents took him to a private psychologist and participated in a six-month treatment program with him after Correll was expelled from the eighth grade for threatening a teacher with a knife.
Finally, the majority indicates that evidence of incest in the family could have served as mitigating evidence. Maj. Op. at 952. At the evidentiary hearing, Correll’s sister Patty testified that their father had been arrested and convicted of child molestation. Correll’s sister Robin testified that she suffered “repeated and continual” sexual molestation at the hands of her father and her brothers, specifically Correll himself. Because the prosecution almost certainly would have presented such evidence in rebuttal, Correll cannot claim that he was prejudiced by counsel’s failure to present evidence of incest as a mitigating fact.
Accordingly, on balance, presentation of family history evidence would have been counterproductive. I cannot agree with the majority’s conclusion that Correll has *970met his burden to prove that, had counsel presented more detailed evidence about his childhood, he would have received a lesser sentence.
Ill
The sum of the majority’s analysis in this case simply eviscerates the requirement that a habeas petitioner demonstrate actual prejudice in order to prevail on a claim for ineffective assistance of counsel. See Wiggins, 539 U.S. 510, 123 S.Ct. 2527. Not satisfied with merely reconstructing the facts, the majority also reinvents Supreme Court authority, asserting that Cor-rell presented evidence sufficient to establish a presumption of prejudice under Wiggins, and that this “classic mitigation evidence ... certainly had the potential to persuade at least one fact-finder that Correll was, at the time of the crimes, incapable of appreciating the wrongfulness of his conduct.” Maj. Op. at 954. These statements, of course, are patently absurd, as even a cursory review of the facts in Wiggins reveals that Correll fell drastically short of carrying the demanding burden of proving actual prejudice the Supreme Court found sufficient in that case.
In holding that Wiggins had met his burden to prove actual prejudice, the Supreme Court explained that Wiggins “experienced severe privation and abuse in the first six years of his life while in the custody of his alcoholic, absentee mother,” that he suffered “physical torment, sexual molestation, and repeated rape” during his subsequent years in foster care, and that he spent time homeless. Id. at 512, 123 S.Ct. 2527. Perhaps most critically, Wiggins was mentally retarded. Id.
In stark contrast, CorrelPs history, which reveals that he was “a highly functioning adult” at the time of his crimes, comes nowhere close to the “powerful mitigating narrative” present in Wiggins. Id. at 513, 123 S.Ct. 2527. Furthermore, the Supreme Court noted that Wiggins lacked a “record of violent conduct,” id. at 537, 123 S.Ct. 2527, and found no evidence “suggesting] that a mitigation ease, in its own right, would have been counterproductive.” Id. at 525, 123 S.Ct. 2527. CorrelPs history, on the other hand, littered with numerous examples of his violent and destructive lifestyle, stands at the very opposite end of the spectrum. Indeed, the district judge, who was in the best position to evaluate all the evidence, concluded that, after considering both the positive and negative repercussions of CorrelPs new evidence, the balance of aggravation and mitigation had “barely been altered.”
Viewed against the standard set forth by Strickland and rearticulated in Wiggins, the majority’s conclusion that the insubstantial mitigating evidence Correll now offers was sufficient to meet the “highly demanding and heavy burden of establishing actual prejudice” not only misapplies the test these cases impose, it essentially writes the prejudice requirement out of our circuit jurisprudence altogether.
Accordingly, I must respectfully dissent.
CALLAHAN, Circuit Judge, with whom KOZINSKI, Chief Judge, and O’SCANNLAIN, KLEINFELD, TALLMAN, and BEA, Circuit Judges, join, in dissenting from the denial of rehearing en banc:I respectfully dissent from our denial of rehearing en banc because the panel majority fails to give deference to the district court’s factual findings as required by Hovey v. Ayers, 458 F.3d 892, 900 (9th Cir. 2006), and improperly interprets the test for ineffective assistance of counsel set forth in Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), so as to create an almost irrebuta-ble presumption of prejudice.
*971Over twenty years ago, Michael Correll was convicted of three counts of first-degree murder, with four aggravating circumstances, and sentenced to death. State v. Correll, 148 Ariz. 468, 471, 478-81, 715 P.2d 721 (1986). The Arizona Supreme Court affirmed the convictions, three of the aggravating circumstances, and the sentence. Id. at 485, 715 P.2d 721. In his federal habeas petition Correll contends that he was denied the effective assistance of counsel at trial as guaranteed by the Sixth Amendment. On remand from this court, the district court conducted a nine-day evidentiary hearing. The district court concluded that although the performance of Correll’s attorney at sentencing was deficient, Correll was not prejudiced.
The panel majority’s opinion reweighs the evidence before the district court and reverses its conclusion by ignoring the district court’s factual findings as well as the second prong of the Strickland test for ineffective assistance of counsel. The majority opinion collapses the two Strickland prongs into one prong. The opinion implies that if counsel makes a strategic decision not to investigate or present what it calls “classic mitigating circumstances” that would nonetheless open the door to more damaging aggravating evidence, prejudice will be presumed. It compounds this mistake by failing to appreciate that in this case even if a presumption of prejudice arises, the presumption was, as the district court found, rebutted. Moreover, if the facts in this case do not rebut the majority’s presumption of prejudice, the presumption in effect becomes irrebutable. For these reasons, I dissent from our decision not to rehear this matter en banc.
I
Although it is not clear from the panel majority opinion, the district court in its 109-page opinion found that trial counsel’s performance had been deficient on only two matters. First, the district court held:
Notwithstanding Strickland’s recognition that defense counsel’s duty to investigate, develop and present mitigating evidence can be reasonably based on a judge’s sentencing tendencies, the Court reluctantly and narrowly concludes that [counsel’s] performance was deficient because he failed to review Petitioner’s mental health records ... before making sentencing strategy decisions.
Second, the district court again narrowly concluded that given the overwhelming aggravating circumstances that Correll faced, counsel “should have obtained the medical treatment records” concerning Correll’s head injury when a wall fell on him when he was seven years old.
The district court, however, rejected a number of other challenges to the attorney’s performance. For example, the district court noted:
The Court specifically finds that [counsel] did maintain regular contact with Petitioner prior to sentencing and rejects Petitioner’s allegation that [counsel] only spent five minutes with him between conviction and sentence.... Petitioner did provide names of persons for [counsel] to contact prior to sentencing, including Susan Curry. [Counsel] followed-up and interviewed or tried to interview the persons Petitioner suggested .... The witnesses were not able to provide relevant useful mitigation information. In fact, in many instances, the witnesses only provided inculpato-ry and non-mitigating information.
(Emphasis in original.) The district court rejected the contention that counsel had improperly failed to present mitigating evidence concerning drug use. It also found that counsel’s performance was not deficient in failing to present expert testimony on methamphetamine intoxication at the *972time of the crime because “there was no lay witness testimony to support Petitioner’s intoxication at the time of the crimes.” In addition, the district court rejected challenges to counsel’s limited investigation of Correll’s family background. It concluded that information concerning incest in the family was not available to counsel because neither Correll nor any of the family members that counsel interviewed provided him with any information. The district court also concluded that counsel “was not deficient in failing to present evidence corroborating the child abuse allegations because such corroborating testimony was not reasonably available to, and thus could not have been obtained by [counsel] at sentencing.” As to the charge that Correll suffered from his mother’s religious fanaticism, the district court concluded that counsel had the available information regarding Petitioner’s mother being a Jehovah’s Witness, but reasonably chose to present such evidence as an abandonment issue.
Thus, as indicated by Judge O’Scann-lain’s dissent, a review of the district court’s 109-page memorandum of decision and order, although confirming that counsel provided deficient representation when he failed to seek documents relating to Correll’s mental health and medical conditions, also shows that counsel’s efforts on behalf of his client were considerably more nuanced than implied by the panel majority.
II
The panel majority alleges that defense counsel basically abandoned his client at sentencing because Arizona law, as it then existed, mandated the death penalty when a defendant had a qualifying prior conviction, and there was no mitigating evidence. Indeed, this would be a much easier case if this assertion were true. However, it misstates the law in a critical manner, and implies the existence of clearly mitigating evidence where no clearly mitigating evidence exists.
The Arizona Supreme Court affirmed Correll’s conviction, the presence of three aggravating circumstances, and his sentence.1 The Arizona Supreme Court did not find that counsel had abandoned Cor-rell. Rather it held:
The trial court found no mitigating circumstances which called for leniency. Defendant offered five mitigating circumstances: upbringing, cooperation in preparation of the pre-sentence report, psychological problems such that he did not understand the wrongfulness of his conduct, his minor participation in the murders, and age. A.R.S. § 13-703(G) provides that any relevant mitigating circumstance proffered must be considered in determining whether to impose the death penalty. We find that none of these factors, alone or in combination, are sufficiently substantial to call for leniency.
State v. Correll, 148 Ariz. at 482, 715 P.2d 721. Later in its opinion, after affirming the existence of three aggravating factors, the Arizona Supreme Court noted that it had “also considered the mitigating circumstances offered by defendant, and[] conclude[d] that even in combination the mitigating circumstances are not sufficiently substantial to call for leniency.” Id. at 483, 715 P.2d 721. Thus, it appears that Arizona law did not mandate the death penalty, but required that the courts determine whether there were factors that were “sufficient to call for leniency.” Furthermore, the Arizona Supreme Court’s opinion rebuts the panel majority’s sugges*973tion that counsel had abandoned Correll at the sentencing hearing.
In addition, much of the majority’s criticism of counsel’s performance fails to recognize the critical difference between raising reasonable doubt as to the prosecutor’s case for the death penalty and presenting affirmative evidence of mitigating circumstances. In Williams v. Woodford, 384 F.3d 567 (9th Cir.2004), we recognized the validity of the lingering-doubt defense at the penalty phase, particularly as it “did not require introduction of mitigating evidence that would open the door to damaging rebuttal evidence.” Id. at 624. We wrote:
based upon our review of the reasons underlying [counsel’s] penalty-phase strategy, we cannot fault [counsel’s] sound tactical decision to present a lingering-doubt defense in lieu of a defense based upon mitigating evidence of Williams’s family and life history, drug use, or mental state. We note in this regard that the defense of “‘residual doubt has been recognized as an extremely effective argument for defendants in capital cases.’ ” Lockhart v. McCree, 476 U.S. 162, 181, 106 S.Ct. 1758, 90 L.Ed.2d 137 (1986) (quoting Grigsby v. Mabry, 758 F.2d 226, 248 (8th Cir.1985) (en banc) (Gibson, J., dissenting)). A comprehensive study on the opinions of jurors in capital cases concluded:
‘Residual doubt’ over the defendant’s guilt is the most powerful “mitigating fact.” ... [T]he best thing a capital defendant can do to improve his chances of receiving a life sentence has nothing to do with mitigating evidence strictly speaking. The best thing he can do, all else being equal, is •to raise doubt about his guilt.
Stephen P. Garvey, Aggravation and Mitigation in Capital Cases: What Do Jurors Think?, 98 Colum. L.Rev. 1538, 1563 (1998) (footnote omitted); accord William S. Geimer & Jonathan Amsterdam, Why Jurors Vote Life or Death: Operative Factors in Ten Florida Death Penalty Cases, 15 Am. J.Crim. L. 1, 28 (1988) (“The existence of some degree of doubt about the guilt of the accused was the most often recurring explanatory factor in the life recommendation cases studied.”).
Williams, 384 F.3d at 624.
Here counsel represented a client who insisted on his innocence.2 Under such circumstances, a lingering-doubt defense was not only reasonable, but in light of then existing Arizona law and the double-edged nature of the so-called “classic mitigating evidence,” perhaps the only reasonable approach available.
Ill
It follows that because Arizona law did not mandate the entry of the death penalty and counsel did not abandon Correll, the second prong of the Strickland test for ineffective assistance of counsel cannot be presumed. Indeed, the purposes of the district court’s nine-day evidentiary hearing was to determine whether the evidence that counsel failed to discover could possibly have been “sufficiently substantive to call for leniency.” What the district court found, and the majority does not really dispute, is that there was no evidence that might humanize Correll or portray Correll as sympathetic. Rather, the evidence con*974cerning Correll’s sociopathic or antisocial personality disorder, drug use, and troubled family was double-edged. Although the evidence might offer some explanation for Correll’s criminal acts, the evidence would also shed light on his prior criminal acts, violent tendencies, and unremorseful attitude. The Supreme Court has held that evidence of a troubled upbringing and mental issues can be mitigating. See Rompilla v. Beard, 545 U.S. 374, 392-93, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005); Wiggins v. Smith, 539 U.S. 510, 534-35, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003). However, such evidence is also recognized to be double-edged, and it can be a reasonable strategic choice not to present such evidence. Williams, 384 F.3d at 619-20.3 This case requires a determination of whether the mitigating evidence could have been sufficiently substantial to call for leniency. In other words, whether there is “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052.
This question must be asked in the context of Arizona law as it then existed. When Correll was tried, the death penalty was not the province of a jury, but the responsibility of the trial judge and the Arizona Supreme Court. Accordingly, the inquiry is not whether a juror might possibly have been moved to alter his or her view of the case, but whether there is a reasonable possibility that any of the mitigating evidence would have changed the trial judge’s or the Arizona Supreme Court’s positions.
IV
By not asking this question, the panel majority fails to appreciate that Correll has not, and cannot, meet the second prong of the Strickland test for ineffective assistance of counsel. Instead the majority (a) improperly substitutes its view of the evidence for the district court’s findings, (b) fails to appreciate Arizona’s death penalty provisions that were in existence when Correll was tried, and (c) ignores the fact that presentation of the alleged “classic mitigating evidence” would have opened the door to overwhelmingly damaging rebuttal evidence.
A. The panel majority’s improper substitution of its independent analysis of the record is well presented in Judge O’Scann-lain’s dissent; however, its treatment of two issues — Correll’s alleged brain injury and the allegation that he was under the influence of drugs at the time of the crimes — illustrates the extent to which the majority’s conclusions differ from the evidence in the record and the district court’s perspective.
Addressing Correll’s alleged brain injury, the panel stated:
When Correll was seven, a brick wall collapsed on his head. Although he was *975unconscious for some time after the accident, his parents did not seek medical treatment until several days later when he was still not back to normal. Several experts testified that this type of accident and the symptoms Correll exhibited then and now indicate a high likelihood of brain impairment.
The district court, however, made the following findings:
Petitioner received a head injury on April 8, 1967, when he was 7, at which time his parents took him to see their family doctor. Four days later, Petitioner was vomiting and again taken to the family doctor where an X-ray was taken and an EEG scheduled. On April 14, 1967, an EEG was done. On April 15, in response to more vomiting, Petitioner’s parents took him to the emergency room at Children’s Hospital of Los Angeles. At the hospital, he was seen by a treating physician, who diagnosed a subgaleal hematoma, which is a bruise or collection of blood under the scalp, but above the skull. The treating physician recommended a neurosurgery consultation, which was done. The doctor in the neurosurgery clinic also diagnosed Petitioner with a subgaleal hema-toma. On May 3, 1967, Petitioner was brought back to the neurosurgery clinic for a follow-up visit. The follow-up visit noted that Petitioner’s hematoma cleared in 5 days and that Petitioner was alert and well.
In addition, the record indicates that when Correll was examined by experts around 2000, his brain functioned at a high level, although there was some evidence of impairment in the prefrontal lobe.4 Also, there was little in the record to connect the alleged impairment in 2000 to the 1967 accident. The district court, having had the benefit of hearing from both parties’ neuropsychologists, concluded that Correll did not suffer any brain injury from the 1967 accident.
The majority, however, chastises the district court for making such a finding asserting that it should “have decided only whether there existed a ‘reasonable probability’ that ‘an objective fact-finder’ in a state sentencing hearing would have concluded that Correll had a brain injury that impaired his judgment at the time of the crimes.” But the majority’s correction of the standard cannot change the fact that the district court determined that there was no reasonable probability that a fact-finder would find .that Correll had a brain injury that impaired his judgment, and that the majority, rather than defer to this reasonable perspective, improperly substitutes its view of the record for that of the district court.
The majority’s myopic view of the record also allows it to conclude that “the evidence of Correll’s methamphetamine use on the night of the crimes, had it been fully presented,' could have risen to the level of a statutory mitigator.” The majority states that there was undisputed evidence induced that Correll was addicted to methamphetamine and had used it on the day of the crime, and further relies on expert testimony “that gross methamphetamine intoxication, unlike gross alcohol intoxication, is not necessarily apparent to outside observers.”
The majority’s conclusion cannot be reconciled with Arizona law at the time of the crimes and the evidence in the record. As noted by Judge O’Scannlain, Arizona law provided “[a] defendant’s intoxication or alcoholism at the time of the offense is a mitigating circumstance if the evidence *976shows that it significantly impaired the defendant’s capacity to appreciate the wrongfulness of his conduct or to conform his conduct to the requirements of the law.” State v. Zaragoza, 135 Ariz. 63, 659 P.2d 22, 30 (1983) (emphasis added).
The district court, after holding a nine-day evidentiary hearing, found:
* due to Correll’s continued maintenance of his innocence, Correll did not discuss his mental state with counsel and did not attempt to help counsel prove that he was intoxicated when he committed the crimes;
* counsel interviewed the persons Cor-rell indicated he was with prior to when the crimes were committed and they did not indicate that he was grossly intoxicated by alcohol abuse or drug abuse on the day of the crimes, but did indicate that Correll had been using methamphetamine prior to the day the crimes were committed;
* “none of the witnesses could have testified that they observed Petitioner injecting methamphetamine in close proximity to the time the crimes occurred”;5
* the first thing that Correll and Nabors asked Snelling when they gained entry into his trailer home was whether he had any speed;
* Correll had spent the vast majority of his life incarcerated in prison where his access to drugs was limited;6
* Correll’s self-reporting of drug use was severely limited due to the lack of corroboration;7
* Correll’s expert’s testimony was “thoroughly impeached” because he admitted that he only minimally considered the facts of the crimes before reaching his conclusion, he admitted that the facts did not necessary establish that Correll had prominent hallucinations or delusions, and the facts of the crimes show that Correll “was involved in deliberate acts, planning, conspiring, avoiding detection, awareness of wrongdoing, and that he was oriented to time, place and reality.”
* the determination by Arizona’s expert in addiction that Correll was not in a substance abuse psychosis was credible because “he utilized the facts of the case to support his opinion and tied his opinion to the facts of the case;”
* “the evidence shows that it was Petitioner who remained calm when the gun *977misfired as Nabors was trying to kill Robin Cady. It was Petitioner who encouraged Nabors to remain calm as there were no cars coming, to get a shell chambered and shoot Cady. Such behavior at the time of the crime does not demonstrate intoxication and, in fact, undercuts an assertion of intoxication.”
Accordingly, the majority’s assertions that there is a reasonable probability that a fact-finder could have found that Correll had a brain injury or that he was intoxicated by drugs at the time of the crimes are not supported by the record. More importantly, the majority’s reweighing of the evidence violates our established law of deferring to the district court’s findings. Hovey, 458 F.3d at 900 (“[F’Jactual findings made by the district court are reviewed under the ‘significantly deferential’ clearly erroneous standard, in which we accept the district court’s findings of fact absent a ‘definite and firm conviction that a mistake has been committed.’ ”) (citations omitted).8
Here, the district court held a nine-day evidentiary hearing and made detailed findings of fact on remand from this court. We abuse our role as an appellate court when we cavalierly ignore the findings that a district court makes on remand. Again, the question is not whether reasonable minds might differ, but whether a review of the record creates “a definite and firm conviction that a mistake has been committed.” 9 The majority, in pursuing “some mitigating evidence[that] could have spared Correll’s life,” fads to appreciate— as it is required to — that the district court’s contrary position is reasonable and entitled to deference.
B. Although counsel’s decision not to investigate Correll’s medical and psychiatric records fell below an acceptable level of competence, his underlying reasons may be relevant to a determination of whether the failure to investigate was prejudicial. The district court explained defense counsel’s perspective as follows:
[Counsel] estimated that Petitioner had one chance in twenty for a life sentence if he did not present the psychological diagnosis but only one chance in fifty if he presented Petitioner’s psychological diagnosis to Judge Howe [the trial judge].... Rather than argue Petitioner’s personality disorder to Judge Howe, [counsel] decided that Petitioner had a better chance to avoid the death penalty if he portrayed that Petitioner was involved in a drug, ripoff which had gone terribly wrong, that Petitioner had only been a follower in the matter, that he had not been the trigger-man as to the three people who died, that Gary Snell-ing had reported to police that Petitioner was under the influence of drugs and/or alcohol at the time of the crimes, *978and that he should be shown sympathy because his family abandoned him at the age of 14.... [Counsel] concluded that any psychological diagnosis of Petitioner could not be presented and argued without producing witnesses and other evidence that might destroy any residual doubt arising from the presentation of his misidentification defense at trial and the mitigation theory [counsel] intended to present regarding Petitioner being only a follower in the drug mishap and not the trigger-man.
The district court further explained:
[Counsel] acknowledged that Judge Howe would consider and give effect to constitutionally relevant mental health evidence, but believed that once Judge Howe knew Petitioner’s diagnosis, he would find it easier not to show sympathy and sentence him to death.... Likewise, [counsel] did not believe that Judge Howe would give substantial mitigating weight to antisocial personality disorder evidence. ... On the other hand, [counsel] did believe that there was some possibility that Judge Howe might give Petitioner a break for not being the trigger-man in the murders.
Although the panel majority opines that defense counsel “was afraid of the sentencing judge,” a review of the record indicates that counsel’s evaluation of Judge Howe’s outlook was probably accurate and definitely reasonable. Furthermore, the panel majority’s criticism of counsel’s focus on Judge Howe ignores that under existing Arizona law, it was Judge Howe who was responsible for sentencing Correll. If Judge Howe imposed the death sentence, the Arizona Supreme Court would review that decision. However, if Judge Howe had declined to impose the death sentence, it is doubtful that Arizona or the Arizona Supreme Court could have, or would have, sought to change that decision.10
C. Perhaps the most critical factor, which is not really denied by the majority, is that the introduction of the “classic mitigating circumstances” would open the door to the admission of overwhelming negative evidence. The district court explained:
Although the Court has attempted to recount the mitigation evidence presented at Petitioner’s evidentiary hearing in detail, the bottom line is clear: a strategy of presenting Petitioner’s emotional and mental problems, his condition at the time of the crime, or as a victim of an abusive and tragic upbringing, would have required the following additional disclosures of facts, none of which are “mitigating:” (i) Petitioner’s rape of a female psychotic patient while he was undergoing mental health treatment for his antisocial personality disorder and mild depression ...; (ii) Petitioner’s numerous escapes from mental health treatment facilities and rejections of institutional efforts to provide him with mental health treatment ...; (iii) Petitioner’s hostage taking and armed aggression against mental health workers in an escape attempt from a mental *979health treatment facility ...; (iv) the underlying factual basis of Petitioner’s prior conviction for armed robbery ...; (v) the revelation that, shortly after the murders were committed, Petitioner and Nabors woke up Robin Correll and informed her that they needed a ride out-of-state right away ...; (vi) Petitioner’s lack of effort to seek any type of treatment for his substance abuse problem ...; (vii) Petitioner’s acts of regularly molesting his sister Robin ...; (viii) the conclusion of the social evaluation at age 18 that Petitioner was not a candidate for probation and was a danger to the community ...; (ix) additional information showing the efforts of Petitioner’s parents to deal with his drug abuse problem and obtain psychological treatment for him following his armed threat against a teacher at school ...; (x) that Petitioner had no desire to work but only wished to enjoy himself ...; and (xi) Petitioner’s statement that when he committed the 1978 armed robberies that it gave him a strong sense of power and excitement....
The district court also agreed with counsel that the prosecutor “had a reputation of excellent preparation and that she would have left no stone unturned in her opportunity to rebut any mitigation evidence presented.”
The panel majority seeks to minimize the negative impact of this evidence by suggesting that a “significant portion” of the “damaging rebuttal evidence was already available through the pre-sentence report.” This is a misleading overstatement. Such facts as Correll’s molestation of his sister, his use of minors to facilitate armed robberies, and his rape of a psychotic female patient were not set forth in the pre-sentence report. Moreover, as has already been noted, there is a world of difference between raising questions as to the sufficiency of the State’s presentation and introducing “mitigating evidence that would open the door to damaging rebuttal evidence.” Williams, 384 F.3d at 624.
Regardless of how much of the negative evidence was already before the trial court, the inquiry remains whether further presentation of the evidence would have been substantial enough to mitigate a death sentence.11 The majority does not really take issue with the district court’s determination that Correll, at most, had an antisocial personality disorder, and that there was “insufficient evidence to support that [Correll] has ever suffered from any major mental illness, whether PTSD, a major depressive disorder, or a bipolar disorder.” Instead, the panel majority opines that the damaging rebuttal evidence “could, in the hands of a competent attorney, have been used to support Correll’s claims of dysfunctional upbringing and continuing mental disorder.” This may be true in the abstract, but in light of the horrific nature of the crimes, Correll’s defense of innocence, the extant standard for the imposition of the death penalty, and the damaging rebuttal evidence, the majority engages in wishful thinking.
Here we are concerned with the possible impact of damaging rebuttal evidence on Judge Howe and the Arizona Supreme Court. The district court properly noted that the standard is high:
Based on both the horrific facts surrounding these murders and [Correll’s] prior criminal history, this is the type of case that demands powerful mitigation before it may be said that confidence in the outcome at sentencing has been undermined. See Bonin v. Calderon, 59 *980F.3d 815, 836 (9th Cir.1995); see also Gerlaugh v. Stewart, 129 F.3d 1027, 1042-43 (9th Cir.1997) (horrific crime facts require substantial mitigation before it may be said that the balance of aggravating factors did not warrant death); Campbell [v. Kincheloe], 829 F.2d [1453,] 1464 [ (9th Cir.1987) ] (overwhelming aggravating factors and the heinous nature of the crime required more than insubstantial mitigation to establish prejudice).
Also, Arizona argues that under Arizona law:
Correll’s personality disorder, his alleged drug addiction, his past psychological and medical history, and his dysfunctional family, as non-statutory mitigators, are not entitled to any significant mitigating weight, because Cor-rell failed to demonstrate any causal nexus between these mitigators and the crimes he committed. [State] v. Murdaugh, 209 Ariz. 19, 35 [97 P.3d 844] ... (2004) (drug impairment, personality disorder, and paranoia not entitled to significant mitigating weight because there was no proven causal nexus between them and the defendant’s crimes); [State ] v. Hoskins, 199 Ariz. 127, 151-53 [14 P.3d 997] ... (2000) (antisocial or borderline personality disorder, and dysfunctional family, not mitigating in absence of causal link to crime.).
Furthermore, even the panel majority does not question that the state proved three aggravating factors. In light of the horrific nature of the murders, the reasonably perceived nature of the trial judge’s jurisprudence, and the incredibly damaging nature of some of the rebuttal evidence, it is not reasonable to conclude that the admission of the evidence would have had any mitigating impact on either Judge Howe or the Arizona Supreme Court.
y
This case presents an instance in which counsel’s instinct that an investigation into Correll’s medical and mental history would not yield any positive evidence, although an unacceptable reason for not conducting an investigation, turns out after 17 years, a full investigation, and a 9-day evidentiary hearing, to have been correct. The panel majority does not really deny that there is no positive evidence, but argues that evidence concerning Correll’s alleged brain damage, sociopathic or antisocial personality disorder, drug use, and troubled family, constitute “classic mitigating circumstances.” The second prong of the Strickland test, however, does not call for an abstract analysis of what might be mitigating evidence, but a determination of whether there is “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland, 466 U.S. at 694,104 S.Ct. 2052.
The Supreme Court reiterated this standard in Woodford v. Visciotti, 537 U.S. 19, 22, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002), when it reversed this court for failing to defer to the California Supreme Court’s determination that trial counsel’s inadequacy was not prejudicial.12 Moreover, in *981Allen v. Woodford, 395 F.3d 979 (9th Cir. 2005), we first found that “counsel’s failure to prepare for the sentencing phase until a week before that phase began, and his resulting failure to thoroughly investigate and present Allen’s mitigation case, was constitutionally deficient,” id. at 1002, but denied relief because we could not “conclude that there is a reasonable probability, had trial counsel presented the potential mitigation evidence developed during habeas, that the jury would have weighed the evidence in favor of a life sentence.” Id. at 1005. Although Allen presented a very different factual situation, it does require that we look beyond the fact of counsel’s deficient performance to determine whether it had any effect on the result.
The record in this case clearly shows that the presentation of evidence of Cor-rell’s alleged brain damage, sociopathic or antisocial personality disorder, drug use, and troubled family would not have made any difference to the trial judge or the Arizona Supreme Court. This conclusion is solidly based on the horrific nature of the murders, the applicable constitutional and state law as it existed when Correll was tried, the perceived nature of the trial judge’s jurisprudence, and the incredibly damaging nature of the rebuttal evidence. It is one thing to cast about for alternate theories after the imposition of the death penalty, and an entirely different thing to argue that a defendant who has been convicted by a jury of first-degree murder should not receive the death penalty because he is a sociopath who cannot control himself. Correll’s counsel thought that there was a one in twenty chance that Judge Howe would not impose the death penalty if the murders were presented as the result of a “routine robbery, drug rip off that went bad.” The presentation of further evidence of Correll’s mental and medical records, antisocial behavior, and prior crimes — far from eliciting sympathy — would have rendered a death sentence a certainty rather than a probability.
The panel majority’s opinion is not only factually wrong, but more importantly for Ninth Circuit law, fails to follow the standard for ineffective assistance of .counsel mandated by the Supreme Court and followed in our prior cases. First, instead of fairly asking separately whether counsel’s performance was deficient and whether the deficient performance was prejudicial, the opinion collapses the two standards into a single inquiry of whether counsel’s performance was deficient. In other words, it in effect allows a finding of a deficient strategic decision to carry an ir-rebutable presumption of prejudice. Second, it fails to recognize that even assuming that a presumption of prejudice may arise from a determination of deficient performance, here the presumption was rebutted. It does this by insisting that there were “classic mitigating circumstances,” without acknowledging the seriously damaging nature of this evidence.13
The panel majority’s opinion makes it almost impossible for defense counsel in a death penalty to render effective assistance of counsel. Where, as here, defense counsel recognizes that what might argu*982ably be mitigating evidence is also damaging, he or she faces an impossible decision. If counsel decides to forego presentation of the evidence, counsel’s performance may be subsequently determined to be ineffective because the choice was prejudicial (the client received the death penalty). If defense counsel proffers the controversial evidence, and the client gets the death penalty, counsel will be chastised for introducing evidence that was prejudicial to the client. Moreover, every criminal defendant who persuades a court that his or her counsel was ineffective will argue that if the deficiency was prejudicial in this case, it must be prejudicial in his or her case. Because the panel majority misconceives the Strickland standard for ineffective assistance of counsel and then applies it in such a way as to suggest an irrebutable presumption of prejudice from a counsel’s deficient strategic decision, I dissent from our decision not to rehear this matter en banc.
. Although it is normally not necessary to restate the facts and procedural history in a dissenting opinion, the reader will understand that this exercise is necessary due to the sharp divergence between the majority's presentation of the facts and the district court's factual findings.
. Ariz.Rev.Stat. § 13 — 703(F)(2).
. Id. § 13 — 703(F)(3).
. Id. § 13-703(F)(5).
. Id. § 13 — 703(F)(6).
. Id. § 13-703(F)(8).
. The majority unduly discounts defense counsel’s attack of the government’s asserted aggravating factors. See Maj. Op. at 947-48. Both the Arizona Supreme Court and the state trial court disagreed with the majority's assessment of counsel’s performance with respect to the "grave risk of death to others” and the "multiple murders” aggravating factors, agreeing with counsel’s assertion that the first factor was unsupported and the second was unconstitutional in this case. The Arizona Supreme Court also found persuasive defense counsel’s argument that the govern*958ment failed to prove beyond a reasonable doubt that Correll intended to kill one victim and therefore the death penalty could not be imposed on that count. Furthermore, counsel made compelling substantive legal and factual arguments with respect to the other aggravating factors.
.The majority quotes the state trial court record, which reads that "Defendant waive[d] presentation of mitigating evidence.” Maj. Op. at 946. This excerpt, however, was merely the conclusion of the court clerk. In the district court proceedings, defense counsel adamantly maintained that "[w]e didn’t waive” the presentation of mitigating evidence.
. At the time of sentencing, Arizona state law explicitly provided that such inability reasonably to foresee that one’s conduct would cause death to another person was a statutory mitigating factor. Ariz.Rev.Stat. § 13-703(G)(4).
. A defendant’s inability to appreciate the wrongfulness of his conduct or to conform his conduct to the requirements of law is a statutory mitigating factor. See Ariz.Rev.Stat. § 13-703(G)(1).
. A defendant's age is a statutory mitigating factor. See Ariz.Rev.Stat. § 13-703(G)(5).
. As the district court summarized the evidence presented at the sentencing hearing: Rather than argue Petitioner’s personality disorder to Judge Howe, [counsel] decided that Petitioner had a better chance to avoid the death penalty if he portrayed that Petitioner was involved in a drug ripoff which had gone terribly wrong, that Petitioner had only been a follower in the matter, that he had not been the trigger-man as to the three people who died, that Guy Snelling had reported to police that Petitioner was under the influence of drugs and/or alcohol at the time of the crimes, and that he should be shown sympathy because his family abandoned him at the age of 14.
. The Supreme Court has since held that Arizona's practice of judges finding aggravating factors violates the Sixth Amendment right to a jury. See Ring v. Arizona, 536 U.S. 584, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002). Ring does not apply, however to cases such as this one that were already final on direct review. See Schriro v. Summerlin, 542 U.S. 348, 358, 124 S.Ct. 2519, 159 L.Ed.2d 442 (2004).
. Because Correll’s petition for a writ of habeas corpus was filed before the effective date of the Antiterrorism and Effective Death Penalty Act of 1996 (''AEDPA”), Pub.L. No. 104-132, 110 Stat. 1214, pre-AEDPA law governs our review. Lindh v. Murphy, 521 U.S. 320, 327, 117 S.Ct. 2059, 138 L.Ed.2d 481 (1997).
. While the majority concludes that "a significant portion of that damaging rebuttal evidence was already available through the pre-sentence report,” Maj. Op. at 955, it fails to acknowledge what Correll’s counsel realized, *965that the introduction of some potentially mitigating evidence would open the door to a parade of horribles. For example, while the presentence report summarily discloses Cor-rell’s conviction of three counts of armed robbery in 1978, Correll's attorney understandably wanted to preclude damning rebuttal evidence revealing that Correll enlisted his 13 year-old younger brother and his 15-year-old girlfriend in these crimes. Furthermore, the presentence report is silent regarding other extremely damaging information that the prosecutor would have surely brought to light in rebutting certain potentially mitigating evidence.
. Indeed, there was evidence in the record that suggested that Correll was not intoxicated at the time of the crimes. Specifically, when Correll and Nabors entered Snelling's trailer home, the first thing they asked Snell-ing was "whether he had any speed.”
. I would further note that a drug defense likely would have evoked less sympathy from an Arizona sentencing judge 22 years ago than it does from the court today. See May-field v. Woodford, 270 F.3d 915, 931 (9th Cir.2001) (crediting testimony that there were “no death penalty cases tried in San Bernar-dino County prior to 1983 where a drug defense had been successful in gaining either an acquittal or in reducing the sentence from death to life without parole.”). The sentencing judge very likely would have taken note of the fact that Correll never sought treatment for his substance abuse problem and repeatedly secured his removal from the mental health programs in which he was placed either by escaping or by violently assaulting the staff.
. While the majority quotes Reverend Curry’s testimony that he "would have unhesitatingly come to help" Correll, see Maj. Op. at 946 n. 2, I credit the district court's finding that at the time of the sentencing hearing he was unavailable to help. Reverend Curry testified that he "cannot offer testimony or assertions regarding people who have been in California Youth Authority [because] [i]t is forbidden by law.” Reverend Curry testified that while others could contact him, he "could not make contact with” counsel and when he "talked with [his] supervisors about it, ... they said no.” Furthermore, defense counsel testified that when he contacted Reverend Curry’s wife, she informed him that the Reverend "didn’t really want to be involved.”
. It should be noted that the majority does not suggest that any investigation or effort by counsel could have prevented the Arizona courts’ finding three aggravating factors.
. The district court noted:
Based on Petitioner's continuing claim of innocence and Petitioner’s failure to prove otherwise, the Court concludes that Petitioner did not discuss or attempt to help [counsel] prove that he was intoxicated when he committed the crime or that his condition at that time of the crimes would mitigate his sentence.
. In Williams, we stated:
The Supreme Court and this court have consistently held that counsel’s performance is not deficient for the failure to present evidence in mitigation at the penalty phase when counsel’s decision is based upon a reasonable tactical determination that the mitigating evidence would allow for the introduction of rebuttal evidence 'that might be literally fatal.’ Burger [v. Kemp ], 483 U.S. [776,] at 791-94, 107 S.Ct. 3114 [97 L.Ed.2d 638] [(1987)] (counsel’s failure to present any mitigating evidence, including the defendant's own testimony or the testimony of the defendant’s mother that he had an exceptionally unhappy and physically abusive childhood, or the expert testimony of a psychologist, was reasonable professional judgment because the testimony would risk bringing before the jury evidence of the defendant’s unremorseful attitude, violent tendencies, and prior criminal acts).
384 F.3d at 619-20.
. The district court noted that Dr. Martell, a neuropsychologist, testified that this mild impairment did not "have a lot of import for [Correll’s] everyday behavior” and that "of all the capital defendants he has tested, [Correll] is one of the highest functioning defendants.”
.The district court noted that at the eviden-tiary hearing, Robin Correll, petitioner’s sister, testified that on the morning before the crimes she witnessed Correll inject some amount of methamphetamine. However, Robin did not testify at the trial. The district court explained:
Had Robin testified about Petitioner’s use of methamphetamine the day prior to the crimes, she' could have been cross-examined about her knowledge of Petitioner’s whereabouts at the time of the crimes. Had Robin testified concerning Petitioner’s conversations about his need to leave town quickly, she would have disclosed that Petitioner was with John Nabors and had wanted an immediate ride out-of-state very soon after the murders occurred. Such testimony would have totally eliminated any mitigating weight from Petitioner’s claim of innocence and residual doubt (i.e., the guilt phase misidentification defense).
. The district court noted that except for 229 days, Correll had been incarcerated for the 9-year period between October 1975 and March 1984, a month before the crimes.
. The district court concluded:
The court does not credit Petitioner's unsubstantiated self-report that he abused methamphetamine every day before the crimes were committed. Petitioner chose not to testify at the evidentiary hearing; Petitioner chose not to fully cooperate with Dr. Matthews’s examination regarding the issue of drug abuse. Because of the obvious motive to fabricate, Petitioner's self-serving statements about his drug usage prior to the crimes is unreliable and subject to searching skepticism.
. Our deference to the district court's factual findings did not arise in Hovey, 458 F.3d at 900, but is of a longstanding principle of habeas review. See Silva v. Woodford, 279 F.3d 825, 835 (9th Cir.2002) (‘‘[0]ur review for clear error is 'significantly deferential,’ in that we must accept the district court's factual findings absent a 'definite and firm conviction that a mistake has been committed.’ ”); United States v. Syrax, 235 F.3d 422, 427 (9th Cir.2000) (“Clearly erroneous review is 'significantly deferential,' requiring that the appellate court accept the district court’s findings absent a 'definite and firm conviction that a mistake has been committed.' ”); McMillan v. United States, 112 F.3d 1040, 1044 (9th Cir. 1997) (quoting Concrete Pipe & Prods, of Cal., Inc. v. Construction Laborers Pension Trust, 508 U.S. 602, 113 S.Ct. 2264, 124 L.Ed.2d 539 (1993).).
. In Hayes v. Woodford, 301 F.3d 1054, 1067 n. 8 (9th Cir.2002), we noted that “[t]o be clearly erroneous, a decision must strike us as more than just maybe or probably wrong; it must, as one member of this court recently stated during oral argument, strike us as wrong with the force of a five-week-old, unrefrigerated dead fish.” (internal quotation omitted).
. The majority dismisses counsel’s perspective concerning Judge Howe by labeling it a presumption that "the judge would not follow the law.” The district court, however, specifically rejected the suggestion that Judge Howe would not follow the law. Rather, counsel thought that if Judge Howe found out that Correll was a sociopath or psychopath, he was more likely to sentence him to death. The district court found counsel’s perspective to be credible. Moreover, Judge Howe’s alleged perspective does not appear to be either irrational or illegal. In Strickland, the Supreme Court recognized that the trial judge’s known views could be considered by counsel. 468 U.S. at 699, 104 S.Ct. 3262. The trial judge’s views were particularly important in Correll's case as three aggravating circumstances were established beyond dispute and a life sentence depended on Judge Howe finding sufficient mitigating circumstances to warrant leniency.
. Ironically, the panel majority’s suggestion that the evidence was in the pre-sentence report would appear to weigh against its con-elusion that the presentation of the evidence would have made a difference.
. The Supreme Court’s description of the California Supreme Court's determination— that it held we had improperly failed to accept-bears an uncanny resemblance to the district court’s determination in this case. The Supreme Court noted:
The California Supreme Court concluded that despite the failure to present evidence of respondent’s "troubled family background,” which included his being "berated,” being "markedly lacking in self-esteem and depressed,” having been "born with club feet,” having "feelings of inadequacy, incompetence, inferiority,” and the like, moving "20 times” while he was growing up, and possibly suffering a "seizure disorder," the aggravating factors were overwhelming. In the state court’s judgment, the circumstances of the crime (a coldblooded execution-style killing of one victim *981and attempted execution-style killing of another, both during the course of a preplanned armed robbery) coupled with the aggravating evidence of prior offenses (the knifing of one man, and the stabbing of a pregnant woman as she lay in bed trying to protect her unborn baby) was devastating. The California Supreme Court found these aggravating factors to be so severe that it concluded respondent suffered no prejudice from trial counsel's (assumed) inadequacy.
Woodford, 537 U.S. at 26, 123 S.Ct. 357 (citations omitted).
. As previously noted, we have held that the double-edged nature of arguably mitigating evidence may justify a decision not to present such evidence. Williams, 384 F.3d at 619-20.