dissenting.
The majority concludes that, even assuming the state court unreasonably applied the first prong of Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), we must defer to their determination that no prejudice occurred under AEDPA. Because I believe the state court’s prejudice determination was based on both an unreasonable application of clearly established federal law and an unreasonable determination of the facts, I dissent.
I disagree with the majority’s articulation of the AEDPA standard to evaluate the question of prejudice related to counsel’s failure to investigate and present mitigating evidence about Williams’s troubled background. The majority concludes that it would have been reasonable for a state court to determine that the mitigating evidence was outweighed by the aggravating factors of Williams’s crime. However, that is not what the state court determined. The state court decided that the mitigating evidence would have been inconsistent with the theory counsel actually presented at sentencing. The state court thus concluded that Williams was not prejudiced by counsel’s failure to present that evidence because the evidence was inconsistent with counsel’s penalty-phase strategy, even though the strategy was not adopted after a constitutionally sufficient investigation into Williams’s background. The state court’s prejudice determination is, in this respect, both an unreasonable application of Strickland and based on an unreasonable determination of the facts. The state court’s judgment cannot survive § 2254(d) analysis simply because, as the majority concludes, “a reasonable state court” might have determined for a different reason that Williams was not prejudiced. Furthermore, the cases cited by the majority do not support such a broad proposition.
I describe this disagreement in more detail below. However, because my conclusion requires a finding that Williams established the first Strickland prong, and because the majority merely assumed that the first prong was met for the sake of analysis, I first set forth further procedural history relevant to Williams’s claim.
I. Background
On May 30, 2003, Williams filed for post-conviction relief pursuant to Supreme Court of Missouri Rule 29.15, alleging, inter alia, several instances of ineffective assistance of counsel. The only issue di*840rectly relevant to the current appeal was Williams’s claim that counsel failed to investigate and present evidence that he was the victim of an abusive childhood and that he suffered from mental deficiencies. In support of this argument, Williams offered the affidavit of Dr. Cross, who described how family members and a church deacon sexually abused Williams as a child, how Williams grew up in a violent household with repeated physical abuse, how he suffered from borderline mental retardation before dropping out of school, and how he grew up and dealt with issues of mental illness, such as post-traumatic shock, depression, and drug addiction. Attorney Joseph Green, who handled the penalty phase of Williams’s trial, later stated that counsel was not aware of the details of abuse contained in Cross’s affidavit.
The state postconviction motion court granted an evidentiary hearing for Williams’s postconviction appeal, but limited the testimony to Williams’s claim that counsel was ineffective for failing to allow him to testify. The court did not grant Williams’s motion for a hearing specific to the new mitigation evidence. Despite this limitation, some of the testimony happened to be directly relevant to his failure-to-investigate claim, though the evidence on that issue was not fully developed on the record. During the hearing, both Green and co-counsel Chris McGraugh testified about the strategy used during trial and sentencing. In addition, Williams submitted a deposition that indicated what he would have testified to at sentencing. Green explained that counsel “had several theories at penalty phase. We had some that were developed and some that were underdeveloped.” Post Conviction Relief Hr’g Tr. at 98:18-20, Mar. 12, 2004. Green described the central theory as “residual doubt,” though they also tried to demonstrate Williams’s positive impact on his family. Green indicated that he had wanted to demonstrate Williams’s ability to adjust to incarceration, but never fully developed that theory.
At the hearing, Green was not asked about Dr. Cross’s affidavit. Green was asked about Williams’s deposition, which included statements about an abusive childhood, but Green had not read Williams’s deposition before the hearing. Green testified generally about the importance of using social and mental history as mitigating evidence during sentencing — in Green’s experience, he found that this evidence was important to humanize a defendant — but Green never demonstrated any knowledge about Williams’s abusive childhood.
On cross-examination, the state asked Green whether evidence of an abusive childhood might have conflicted with the residual doubt strategy used during sentencing. Green’s answers indicate that he did not see a direct conflict.
Q: Now, with residual doubt as a theory in the second phase, aren’t you then going to try and portray the Defendant in the most favorable light to the jury as you can? In other words, you don’t want to portray him as being a mean, spiteful, vengeful angry-type person, do you, having been abused as a child?
A: Okay. In my mind, those are two different questions.
Id. at 123:5-12. Rather than describe evidence of abuse as something that would create problems for his theory at sentencing, Green suggested that “[tjhere’s a number of different ways [evidence of abuse] can be used in mitigation.” Id. at 138:24-25. Williams’s postconviction counsel asked Green generally about why information about abuse would be important during sentencing:
*841Q. And the fact that Marcellus testified about abuse in his deposition, I know you haven’t read that, would that be something that you would want the jury to hear, that the abuse may have had something to do with his later behavior if you had an expert to testify to that?
A. Yes, if I could show a nexus, yes.
Q. And why would that be important?
A. Well, if it — why it would be important to show the nexus?
Q. Yes, for the jury to hear this.
A. Oh, sure. It gives an explanation as to how he became to be the person that they see before them, you know, and that to some extent to give the jury an understanding that had maybe society or the government or somebody interceded before at that point, we may not be here today.
Or another way is to give the jurors an explanation that if it is an abuse problem, now that it’s been identified as a problem — this to this individual, a doctor may testify that now we have been able to identify the cause of it we can start treating it and alleviate this problem. There’s a number of different ways that can be used in mitigation.
Q. But did you explore those ways in Marcellus’ case, or did you want to?
A. I — well, I definitely wanted to explore them, and to what extent we explored them, I think we made an attempt to do that; but I don’t think we finished our attempt, though.
Id. at 138:2-139:6. On re-cross, the state again tried to question whether describing a defendant as a victim of abuse as a child would conflict with a residual doubt strategy:
Q. If you went into the abuse to explain to the jury, well, maybe if it wasn’t for all the abuse as a child maybe we wouldn’t be here today, what you’re doing is — aren’t you conceding that he committed the murder? As a result of this abusive background you allege?
A. If you take that position with it, right. If you’re trying to explain who the person is that’s before them today, I think is the wording I used, and it doesn’t necessarily mean that you’re conceding the murder. If you’re just saying this is why — because you have all this other criminal history that’s before the jury.
So if you’re just to say why is this a guy who constantly finds himself incarcerated and charged with crimes? Well, because he has an addiction problem, or because he’s been abused and has a frontal lobe problem or, you know, whatever the reason may be. It doesn’t necessarily mean you’re conceding the act of guilt for the murder; it just explains the person who sits before you.
Id. at 139:20-140:13. All of the questions put to Green about childhood abuse, and all of Green’s answers, appear to be hypothetical. Green does not discuss Williams’s abusive childhood in particular, and none of Green’s answers suggest he knew of any details of abuse.
During the hearing, Green also testified that he had hired a mitigation expert, Jeff Eno, to obtain child development and family history. Green described Eno as the man who helped set up his interviews with Williams’s family members. Green also testified that Eno helped him “get some child development and family history,” id. at 93:14-15, but there is no indication that either Eno or Green learned anything about the abusive childhood or mental retardation described in Cross’s affidavit. Green also said that he hired a psycholo*842gist, Mark Cunningham, to diagnose Williams’s future dangerousness. There is no indication, however, that Cunningham investigated or knew about Williams’s abusive childhood. See id. at 100. It does appear that Green knew something about Williams’s history of trouble in school, but he did not offer a reason for not pursuing that possible mitigating evidence.
Q. As you were aware that Marcellus had problems in school as early as kindergarden [sic], would it have been helpful to have expert testimony to explain Marcellus’s behavior before going to prison and as a child growing up and what was involved in those?
A. Yes, that always can be a hedge.
Q. Did you have a strategic reason for not seeking that type of evidence?
A. As I sit here now, I can’t think of any.
Id. at 100:20-101:3.
After the hearing, the state motion court denied Williams’s motion to vacate the judgment and sentence. The motion court appeared to accept the truth of Dr. Cross’s affidavit, but did not discuss the affidavit’s particulars. The court considered the issue of childhood abuse to be mooted by Green’s chosen strategy during sentencing. The motion court found that evidence of abuse “would have defied trial counsel’s reasonable trial strategy of presenting Movant as a family man, who is innocent of such a violent murder.” State’s Ex. 33 at 802 (hereinafter “Postconviction Order”). The court stated that Williams was not permitted to argue a different theory would have worked better and that “there is no reasonable probability that the omitted evidence would have changed the re-suit of Movant’s sentencing.” Id. at 802-03.
Williams then filed a motion for reconsideration and offered a sworn affidavit from Green about what Green would have testified to had Williams been allowed to address the failure-to-investigate claim. Green stated that he “ran out of time” investigating Williams’s case because of his work on another capital trial at the same time. Green Aff. at ¶ 35. Acting as sole counsel in the other capital case, Green “had to prepare for trial during the same period I was preparing for Mareellus’[s] trial,” and because a motion to continue Williams’s trial was denied, Green admitted that he “did not have sufficient time to thoroughly prepare for Marcellus’[s] trial.” Id. at ¶ 23-24. Green stated that he had reviewed Dr. Cross’s affidavit and that, had he obtained Dr. Cross’s diagnosis during Williams’s penalty-phase case, he would have put it before the jury. Green stated that such evidence would have been important because “it would have provided explanations for his prior criminal history” and would have presented him in “a more sympathetic picture ... than the facts surrounding the crime.” Id. at 31-32. The motion court struck Green’s affidavit, finding it to be an improper attempt to expand the record.5
The Supreme Court of Missouri affirmed the motion court’s denial of relief. The court predicated its decision on the rule that “[t]he selection of witnesses and evidence are matters of trial strategy, virtually unchallengeable in an ineffective assistance claim.” Williams v. State, 168 S.W.3d 433, 443 (Mo.2005). The Supreme Court of Missouri then approved the lower court’s reasoning, and stated: “The motion *843court found that an abusive childhood defense would have been inconsistent with the penalty phase strategy and would not have changed the jury’s sentence. The motion court did not clearly err in denying this claim without an evidentiary hearing.” Id. The Supreme Court of Missouri did not address whether counsel had or had not met its duty to investigate.
II. Ineffective Assistance
As the majority noted, the state motion court concluded that Williams’s counsel had employed a reasonable penalty-phase strategy and that Williams could not establish ineffective assistance of counsel by arguing that a different strategy would have worked better. The Supreme Court of Missouri similarly concluded that the presentation of evidence was “virtually unchallengeable” as a matter of trial strategy. Id. at 443. However, for a state court to conclude that counsel reasonably employed a strategy without assessing whether that strategy was supported by thorough investigation — or a reasonable and informed decision to cut short a thorough investigation — is an unreasonable application of Supreme Court precedent. A finding in this case that there was either a thorough investigation or an informed decision to cut short such an investigation would be an unreasonable determination of the facts. The Supreme Court of Missouri’s decision that counsel’s performance was not deficient thus fails either prong of § 2254(d).
A. Reasonable trial strategy must be based on thorough investigation
It is unassailable that under the prevailing professional norms at the time of Williams’s trial, counsel had an “obligation to conduct a thorough investigation of the defendant’s background.” Williams v. Taylor, 529 U.S. 362, 396, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (hereinafter “Terry Williams”)', Strickland, 466 U.S. at 691, 104 S.Ct. 2052 (“[Cjounsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.”). While it may be proper for counsel to choose not to investigate all claims that a defendant urges him to advance if counsel believes such claims to be meritless, the Supreme Court has explained that “strategic choices made after less than complete investigation are reasonable [only] to the extent that reasonable professional judgments support the limitations on investigation.” Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052. “[I]f limiting the investigation was not reasonable, then neither was the subsequent strategic choice. ‘[Sjtrategy resulting from lack of diligence in preparation and investigation is not protected by the presumption in favor of counsel.’” Antwine v. Delo, 54 F.3d 1357, 1367 (8th Cir.1995) (quoting Kenley v. Armontrout, 937 F.2d 1298, 1304 (8th Cir.1991)). This duty was especially important in Williams’s case, where the prosecution had indicated from the start that they would pursue the death penalty. See Antwine, 54 F.3d at 1367 (“ ‘Given the severity of the potential sentence and the reality that the life of [the defendant] was at stake, we believe that it was [counsel’s] duty ... to collect as much information as possible about [the defendant] for use at the penalty phase of his state court trial.’” (quoting Hill v. Lockhart, 28 F.3d 832, 845 (8th Cir.1994))).
Neither the Supreme Court of Missouri, nor the motion court that it affirmed, discussed the clearly established federal law about counsel’s duty to investigate. This is not, in itself, a problem. The Supreme Court has stated that § 2254(d) does “not require citation of our cases — indeed, it does not even require awareness of our cases, so long as neither the reasoning nor the result of the state-court decision con*844tradicts them.” Early v. Packer, 537 U.S. 3, 8, 123 S.Ct. 362, 154 L.Ed.2d 263 (2002) (per curiam). However, both state courts clearly applied a standard concerning the introduction of mitigating evidence at trial — a strategic decision — rather than the investigation of that evidence in the first place — a constitutional duty. This is an unreasonable application of Strickland.
Strickland did say that strategic decisions were “virtually unchallengeable.” What the Supreme Court actually said in full, however, was that “strategic choices made after thorough investigation of law and facts relevant to plausible options are virtually unchallengeable; and strategic choices made after less than complete investigation are reasonable precisely to the extent that reasonable professional judgments support the limitations on investigation.” 466 U.S. at 690-91, 104 S.Ct. 2052 (emphasis added). Strickland makes it clear that the reasonableness of a strategic decision is dependent on the reasonableness of counsel’s investigation into possible mitigating evidence. Furthermore, “a particular decision not to investigate must be directly assessed for reasonableness in all the circumstances.” Id. at 691, 104 S.Ct. 2052.
The Court further explained the analysis of strategic decisions and reasonable investigations in Wiggins v. Smith, 539 U.S. 510, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003). There, the Court applied AEDPA deference to a similar state-court rejection of relief where “the Maryland Court of Appeals appears to have assumed that because counsel had some information with respect to petitioner’s background ... they were in a position to make a tactical choice not to present a mitigation defense.” Id. at 527, 123 S.Ct. 2527. The Court explained, however, that such a choice could not be based on assumptions about counsel’s investigations; a court must first assess the reasonableness of an attorney’s investigation by considering,
not only the quantum of evidence already known to counsel, but also whether the known evidence would lead a reasonable attorney to investigate further .... Strickland does not establish that a cursory investigation automatically justifies a tactical decision with respect to sentencing strategy. Rather, a reviewing court must consider the reasonableness of the investigation said to support that strategy.
Id. Wiggins makes clear that a state court’s failure to assess the reasonableness of counsel’s investigation of possible mitigating factors meets the AEDPA “unreasonable application” prong. Id. at 528, 123 S.Ct. 2527 (“The Court of Appeals’ assumption that the investigation was adequate thus reflected an unreasonable application of Strickland. As a result, the court’s subsequent deference to counsel’s strategic decision not ‘to present every conceivable mitigation defense,’ despite the fact that counsel based this alleged choice on what we have made clear was an unreasonable investigation, was also objectively unreasonable.” (internal citations omitted)). By failing to analyze the reasonableness of counsel’s investigation, both the Supreme Court of Missouri and the state motion court failed to apply clearly established federal law.
B. The record does not reflect a constitutionally sufficient investigation
Even if the Supreme Court of Missouri had not unreasonably assumed that counsel met its constitutional duty to investigate in Williams’s case and had specifically found that a sufficient investigation supported a decision not to introduce evidence about Williams’s childhood, such a finding would be an unreasonable determination of the facts under § 2254(d)(2). There is no evidence in the record that counsel chose *845to avoid evidence of an abusive childhood or mental deficiencies at sentencing. Green testified about his intent to conduct investigation about Williams’s history, but he admits that he was not able to complete this investigation — in his words, he “ran out of time.” Even when confronted with the state’s view that evidence of Williams’s abusive childhood might conflict with other evidence Green offered at sentencing, Green testified that he thought evidence of abuse still could have been useful, had he been aware of it at sentencing. Green’s testimony at the hearing and his later affidavit both support the district court’s conclusion that counsel’s family-man/residual-doubt strategy became a strategy by default and was not adopted after a reasonable decision to avoid evidence of Williams’s childhood.
The state argues on appeal that counsel did investigate Williams’s social history, and therefore counsel’s decision not to present it at sentencing was part of a reasonable trial strategy. To support this argument, the state points to Green’s hearing testimony, where he talked about Jeff Eno, the mitigation specialist he had hired, and Mark Cunningham, a psychologist. But the record does not demonstrate that either Eno or Cunningham presented Green with any evidence of Williams’s abusive childhood. Even if Green did learn something about an abusive childhood, it would have been objectively unreasonable not to pursue such leads. In Wiggins, the Court reviewed the performance of counsel who had some information about his Ghent’s social history; for example, a presentencing report and a report from the Department of Social Services that included information about the defendant’s troubled childhood. The Court still found that it was “unreasonable in light of what counsel actually discovered” to not pursue those leads and expand the investigation of his social history. 539 U.S. at 525, 123 S.Ct. 2527.
When presented with some evidence of an abusive childhood,
any reasonably competent attorney would have realized that pursuing these leads was necessary to making an informed choice among possible defenses.... Indeed, counsel uncovered no evidence in their investigation to suggest that a mitigation case, in its own right, would have been counterproductive, or that further investigation would have been fruitless; this case is therefore distinguishable from our precedents in which we have found limited investigations into mitigating evidence to be reasonable.
Id. (citing, inter alia, Strickland, 466 U.S. at 699, 104 S.Ct. 2052); see also Rompilla v. Beard, 545 U.S. 374, 389, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005) (ineffective assistance where counsel failed to investigate mitigating information that would have been apparent from documents any reasonable attorney would have obtained). Even if the state is right that counsel knew something about Williams’s abusive childhood, it would have been unreasonable for the state court not to hold an evidentiary hearing to directly assess the reasonableness of counsel’s decision not to investigate further.6
*846Although it might be reasonable, after a thorough investigation of Williams’s background, to decide to present Williams “as a family man, who is innocent of such a violent murder,” instead of presenting evidence of his troubled childhood, Green never admitted to making such a decision. After denying Williams an evidentiary hearing on this issue, which could have revealed more details about counsel’s actual decisions in preparing trial strategy, the state court merely assumed that such a decision had been made. “When viewed in this light, the ‘strategic decision’ the state courts and respondents all invoke to justify counsel’s limited pursuit of mitigating evidence resembles more a post hoc rationalization of counsel’s conduct than an accurate description of their deliberations prior to sentencing.” Wiggins, 539 U.S. at 526-27, 123 S.Ct. 2527; see also Harris v. Reed, 894 F.2d 871, 878 (7th Cir.1990) (“Just as a reviewing court should not second guess the strategic decisions of counsel with the benefit of hindsight, it should also not construct strategic defenses which counsel does not offer.”).
III. Prejudice
The majority concludes that, even if the Supreme Court of Missouri unreasonably applied federal law, we must defer to its decision that Williams was not prejudiced. I strongly disagree. Both state courts explicitly misapplied the first Strickland prong. Their prejudice determination was premised on this misapplication because both state courts failed to assess how a constitutionally insufficient investigation— rather than a conscious and informed decision not to introduce evidence — might prejudice a defendant. Furthermore, the state court erroneously concluded that the mitigating evidence could only have been introduced in conjunction with counsel’s decision to present Williams as a family man who was actually innocent of the crime, even though Green directly contradicted this contention in an evidentiary hearing. Because the state court both unreasonably applied clearly established federal law and unreasonably determined the facts in light of the evidence presented in state proceedings, § 2254(d) does not bar habeas relief for Williams. Rather, Williams has established a reasonable probability that, but for his counsel’s ineffective assistance, his sentence would have been different.
A. The majority misapplies the AJEDPA standard
Instead of reviewing whether the prejudice decision actually offered by the Supreme Court of Missouri was either an unreasonable application of clearly established law or rested upon an unreasonable *847determination of the facts, the majority-looks at whether a reasonable state court could have concluded that Williams was not prejudiced by the deficiencies of counsel. Although such a standard of review is appropriate in some AEDPA cases, it is not appropriate in a case like this one, where a state court has clearly explained its decision on the merits; this is a subtle, but vitally important distinction.
The majority correctly states that “the proper question is whether there is ‘any reasonable argument’ that the state court’s judgment is consistent with Strickland.” Supra at 831-82. However, in determining what “judgment” should be considered, the majority distinguishes “decisions” from a state court’s explanation of its reasons. This distinction comes from a line of cases that deal with how the AEDPA standard applies to state courts’ summary, unexplained, or indeterminate denials, and does not justify ignoring reasons given by a state court to support its decision. Unlike in the cases cited by the majority, the present case does not involve a state court’s failure to offer an explanation for its rejection of post-conviction relief, see Harrington v. Richter, — U.S.—, 131 S.Ct. 770, 788, 178 L.Ed.2d 624 (2011); Wright v. Sec’y for Dep’t of Corr., 278 F.3d 1245, 1254 (11th Cir.2002), nor is it a case where a state court offered only “sparse or otherwise unsatisfactory” explanation for its decision, see Wright, 278 F.3d at 1254 n. 2 (describing Neal v. Puckett, 239 F.3d 683, 696 (5th Cir.2001) and Hennon v. Cooper, 109 F.3d 330, 335 (7th Cir.1997)). It is true that a petitioner challenging a summary or unexplained denial “can satisfy the ‘unreasonable application’ prong of § 2254(d)(1) only by showing that ‘there was no reasonable basis’ for” the state court decision. Cullen v. Pinholster, — U.S. —, 131 S.Ct. 1388, 1402, 179 L.Ed.2d 557 (2011) (quoting Harrington, 131 S.Ct. at 784). However, unlike in the summary denial cases, we are here presented with an explicit decision by the state court that is an unreasonable application of clearly established federal law.
The majority cites no authority that would require — or for that matter, allow— us to cure this defect in the state court’s reasoning by looking for some other reasonable basis upon which the state court could have denied Williams’s claim. While AEDPA may not require state courts to explain their decisions, it does not allow federal courts to ignore a state court’s explanation that fails to meet the AEDPA standard. See Harrington, 131 S.Ct. at 786 (“Under § 2254(d), a habeas court must determine what arguments or theories supported or, as here [i.e., where the state court provided no rationale], could have supported, the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of [the Supreme] Court.”)7; Packer, 537 U.S. at 8, 123 S.Ct. 362 (noting that § 2254(d) does “not require citation of [Supreme Court] cases — indeed, it does not even require awareness of [Supreme Court] cases, so long as neither the reasoning nor the result of the state-court decision contradicts them.” (emphases added)); Wright, 278 F.3d at 1256 n. 3 (“The failure of a state court to set out its reasoning is not *848equivalent to the conspicuous misapplication of Supreme Court precedent.”); Johnson v. Sec’y, DOC, 643 F.3d 907, 930 n. 9 (11th Cir.2011) (noting that “[t]he Court’s instruction from Harrington ” that a petitioner must show there was no reasonable basis for a summary denials “does not apply here because the Florida Supreme Court did provide an explanation of its decision”)8. To defer to a hypothetical conclusion rather than the conclusion actually offered by the state court, as the majority now does, is to create an even more heightened standard than has been previously acknowledged in AEDPA cases.
The majority asserts that Premo v. Moore, — U.S. —, 131 S.Ct. 733, 178 L.Ed.2d 649 (2011), supports its position that deference to unstated hypothetical conclusions is required even in the face of clearly articulated but unreasonable applications of Supreme Court precedent. The United States Supreme Court in Premo, however, expressly characterized the state-court judgment under review as being indeterminate in its rationale. Id. at 740 (“The state court did not specify whether this [rejection of a Strickland claim] was because there was no deficient performance under Strickland or because Moore suffered no Strickland prejudice, or both.”). The Supreme Court did not suggest that it was disregarding a state court’s articulation of an unreasonable application of clearly established Supreme Court precedent. In fact, nothing in the lower state court opinion in the Premo case as quoted by today’s majority seems to suggest an unreasonable application of any Supreme Court precedent. In any event, I find it unnecessary to look behind the Supreme Court’s own clear characterization of the state court record in Premo.
Notwithstanding the extensive discussion of Premo, then, the fact remains that the majority today expands the scope of AEDPA deference beyond the limits previously imposed by statute or by Supreme Court precedent. Even if the majority is correct that this newly articulated and greater degree of AEDPA deference may someday become the next step in the Supreme Court’s habeas jurisprudence, I do not share the majority’s willingness to pioneer a path for this type of broader sanctioning of unreasonable applications of existing law. I believe instead that the Supreme Court itself should be the source of such new statements of law and that lower courts and intermediate courts like ours should-be not only transparent in our actions when we step ahead of the Court’s rulings, but hesitant to take such steps.
Here, Williams’s claim for post-conviction relief was clearly adjudicated on the *849merits by the Supreme Court of Missouri. The state court clearly concluded that Williams was not prejudiced by his counsel’s performance. The reason offered for this conclusion is equally clear: “The motion court found that an abusive childhood defense would have been inconsistent with the penalty phase strategy and would not have changed the jury’s sentence. The motion court did not clearly err in denying this claim.... ” 168 S.W.3d at 443; see Postconviction Order at 802 (“This evidence would have defied trial counsel’s reasonable trial strategy of presenting Movant as a family man, who is innocent of such a violent murder. Movant cannot now plead ineffective assistance alleging that a different strategy would have worked better.”). As explained above, this reasoning about “reasonable trial strategy” is based on an unreasonable application of Strickland.
B. The Supreme Court of Missouri failed to consider how a deficient investigation could prejudice a defendant
The fact that the state courts correctly recited Strickland’s prejudice standard, as identified by the majority, supra at 831, does not satisfy our review of whether the state court reasonably applied that standard. See Sears v. Upton, — U.S.—, 130 S.Ct. 3259, 3264, 177 L.Ed.2d 1025 (2010) (“Although the court appears to have stated the proper prejudice standard, it did not correctly conceptualize how that standard applies to the circumstances of this case.”). Here, the state court’s prejudice determination reflects a misunderstanding about how counsel’s failure to investigate affected the penalty-phase strategy to begin with.
Wiggins makes clear that it is objectively unreasonable for a state court to defer to a “strategic decision” unsupported by reasonable investigation. 539 U.S. at 528, 123 S.Ct. 2527.9 Furthermore, the reasonableness of a penalty-phase strategy is irrelevant to whether the investigation supporting it was reasonable. Sears, 130 S.Ct. at 3265 n. 10 (“[T]he reasonableness of the theory is not relevant when evaluating the impact of evidence that would have been available and likely introduced, had counsel completed a constitutionally adequate investigation before settling on a particular mitigation theory.”) Yet the prejudice determination of both state courts deferred to what they considered a reasonable mitigation strategy, despite not considering the adequacy of the investigation supporting that strategy.
Both courts viewed the theory that was actually presented at sentencing as a reasonable one and concluded that Williams was not permitted to “plead ineffective assistance alleging that a different strate*850gy would have worked better.” Postconviction Order at 802; 168 S.W.3d at 443 (declining to review counsel’s trial strategy because it was “virtually unchallengeable in an ineffective assistance claim”). Thus, instead of weighing the mitigating evidence against the aggravating factors, as the majority now does, both state courts considered only whether the new mitigating evidence could have fit within the family-man/residual-doubt theory that was actually offered. The state motion court concluded that evidence of Williams’s troubled childhood would have “defied trial counsel’s reasonable trial strategy” because it would have “been tantamount to a concession of guilt” and portrayed him as “angry, aggressive, and violent.” Postconviction Order at 802. The Supreme Court of Missouri agreed with the motion court that such a portrayal “would have been inconsistent with the penalty phase strategy and would not have changed the jury’s sentence.” 168 S.W.3d at 443.
This is not only an unreasonable deference to an unreasonable strategy based on unreasonable investigation, but it is also contradicted by counsel’s own testimony. The motion court concluded that offering the new mitigation evidence “would have been tantamount to a concession of guilt,” and thus would have “defied” counsel’s strategy; however, Green testified before that court that introducing evidence of Williams’s troubled background wouldn’t “necessarily mean you’re conceding the act of guilt for the murder; it just explains the person who sits before you.” Hr’g Tr. at 140:11 — 13.10 This is also contrary to Supreme Court precedent. See Wiggins, 539 U.S. at 535, 123 S.Ct. 2527 (“While it may well have been strategically defensible upon a reasonably thorough investigation to focus on [defendant’s] direct responsibility for the murder, the two sentencing strategies are not necessarily mutually exclusive.”).
Furthermore, the assumption that the mitigation evidence would “defy” or be “inconsistent with” the strategy employed by counsel at sentencing unreasonably assumes that this evidence could only have been introduced in concert with the family-man/residual-doubt theory, even though a thorough investigation into Williams’s background might have persuaded counsel to offer the new mitigation evidence instead of the theory that was actually presented. See Wiggins, 539 U.S. at 535, 123 S.Ct. 2527 (noting that “a reasonable attorney might well have chosen to prioritize the mitigation case over” the doubt strategy actually employed at sentencing). The fact that neither state court considered this possibility is further evidence of a failure to recognize the importance of a thorough and competent investigation as required by Strickland. This is why the state court determination that no prejudice occurred is both an unreasonable application of Strickland, as well as an unreasonable determination of the facts.
C. We must weigh de novo the mitigating evidence against the aggravating factors
The majority may be correct that it would have been reasonable for the state court to conclude that there was not a reasonable probability that the new mitigating evidence would have changed the outcome of Williams’s trial. The case against Williams was strong. In light of the state court’s unreasonable application *851of Strickland, however, we do not defer to what a reasonable state court might have decided. Instead, we must now conduct a de novo review of whether Williams suffered prejudice. See Johnson, 643 F.3d at 930 n. 9 (de novo review of prejudice is required where there has been an unreasonable application of Supreme Court precedent); see also Frantz v. Hazey, 533 F.3d 724, 739 (9th Cir.2008) (en banc) (“To identify a § 2254(d)(1) ‘contrary to’ error, we analyze the court’s actual reasoning. ... Identification of such an error is not the end of a federal habeas court’s analysis____ Instead, pursuant to § 2254(a) and pre-AEDPA standards of review, we must also evaluate de novo the petitioner’s constitutional claims, without limiting ourselves to the reasoning of the state court.”); Young v. Sirmons, 486 F.3d 655, 674 (10th Cir.2007) (“Because the [state court] applied a legal principle that is contrary to clearly established federal law in evaluating [Petitioner’s] ineffective assistance claim — as the state concedes— no deference is given to the [state court’s] decision under 28 U.S.C. § 2254(d). Thus, the district court properly reviewed this claim de novo.”).
In assessing whether Williams was prejudiced by his counsel’s deficient performance, we must ask “whether there is a reasonable probability that, absent the errors, the sentencer ... would have concluded that the balance of aggravating and mitigating circumstances did not warrant death.” Strickland, 466 U.S. at 695, 104 S.Ct. 2052. This review “do[es] not require a defendant to show ‘that counsel’s deficient conduct more likely than not altered the outcome’ of his penalty proceeding, but rather that he establish ‘a probability sufficient to undermine confidence in [that] outcome.’ ” Porter v. McCollum, 558 U.S. 30, 130 S.Ct. 447, 455-56, 175 L.Ed.2d 398 (2009) (quoting Strickland, 466 U.S. at 693-94, 104 S.Ct. 2052). “In assessing prejudice, we reweigh the evidence in aggravation against the totality of available mitigating evidence.” Wiggins, 539 U.S. at 534,123 S.Ct. 2527.
In his post-conviction appeal, Williams submitted evidence that he was sexually abused as a child, that he grew up in a violent household with constant physical abuse, that he suffered from borderline mental retardation before dropping out of school, and that he grew up and dealt with issues of mental illness, such as post-traumatic shock, depression, and drug addiction. Had the jury been presented with mitigating evidence of Williams’s abusive childhood and mental impairments instead of the family-man/residual-doubt strategy that was actually offered at sentencing, it would have been considering evidence of a completely different character than that which it actually considered. This is not a case where the newly discovered mitigation evidence was merely duplicative of evidence that was presented, see Paul v. United States, 534 F.3d 832, 842 (8th Cir.2008), or where the post-conviction court disbelieved the evidence favorable to the defendant, see Roll v. Bowersox, 177 F.3d 697, 700-01 (8th Cir.1999). This is not a case where the new evidence “would barely have altered the sentencing profile presented to the sentencing judge.” Strickland, 466 U.S. at 700, 104 S.Ct. 2052. The court does not have to determine whether the evidence would have more likely than not altered the outcome; it need only establish a probability sufficient to undermine confidence in the outcome. Porter, 130 S.Ct. at 455-56. I believe Williams has demonstrated sufficient evidence to undermine that confidence.
The state now argues that the new mitigating evidence is of only negative value. The state believes that the evidence of his abusive and violent childhood would portray Williams as a bad man, an irredeema*852ble monster. It is possible that a jury might adopt the state’s interpretation of the evidence; however, that interpretation conflicts with longstanding Supreme Court views on the use of such evidence. See Terry Williams, 529 U.S. at 398, 120 S.Ct. 1495 (noting that a “graphic description of [defendant’s] childhood, filled with abuse and privation, or the reality that he was ‘borderline mentally retarded,’ might well have influenced the jury’s appraisal of his moral culpability”); Penry v. Lynaugh, 492 U.S. 302, 319, 109 S.Ct. 2934, 106 L.Ed.2d 256 (1989) (“[E]vidence about the defendant’s background and character is relevant because of the belief, long held by this society, that defendants who commit criminal acts that are attributable to a disadvantaged background ... may be less culpable than defendants who have no such excuse.”), abrogated on other grounds by Atkins v. Virginia, 536 U.S. 304, 122 S.Ct. 2242, 153 L.Ed.2d 335 (2002).
I agree with the district court that, “[i]n light of the overwhelming amount of aggravating evidence that was proffered by the state, the jury would not think worse of Williams based on evidence of his dysfunctional youth and mental issues.” Order at 32; see also Simmons v. Luebbers, 299 F.3d 929, 938-39 (8th Cir.2002) (“By the time the state was finished with its case, the jury’s perception of [Defendant] could not have been more unpleasant. Mitigating evidence was essential to provide some sort of explanation for [Defendant’s] abhorrent behavior.”). Indeed, even if some of the mitigating evidence might have proved harmful to Williams in the end, that does not prevent us from finding there is a reasonable probability that the evidence would have changed the outcome of his sentence. See Terry Williams, 529 U.S. at 396, 120 S.Ct. 1495 (finding prejudice even though “not all of the additional evidence was favorable to” the defendant); see also Porter, 130 S.Ct. at 455 (rejecting as unreasonable a state court’s determination that negative aspects of petitioner’s military history would undermine positive aspects of mitigating evidence).
The majority is correct that the state presented strong aggravating evidence against Williams. I agree with the majority that the mitigating factors in Williams’s case may not be as strong as those in Porter, nor the aggravating factors as weak. However, because the other cases compared by the majority applied a deferential “any reasonable basis” standard that is not appropriate for this case, those cases are not useful guides for determining whether the evidence in this case is sufficient to undermine our confidence in the outcome. For instance, this case is wholly unlike Woodford v. Visciotti, 537 U.S. 19, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002), where the Supreme Court deferred to a state-court decision that the petitioner suffered no prejudice from counsel’s inadequacy in the face of the aggravating factors the state court explicitly described as “overwhelming” and “devastating.” Id. at 26, 123 S.Ct. 357.11 Although the majority now weighs the mitigating evidence against the aggravating factors in Williams’s case, it does this in the first instance.
Nor is this case like Cullen v. Pinholster, where the state-court decision was nothing more than a summary denial. Because the state court had not explained its decision in that case, the petitioner could *853“satisfy the ‘unreasonable application’ prong of § 2254(d)(1) only by showing that ‘there was no reasonable basis’ for the California Supreme Court’s decision.” 131 S.Ct. at 1402 (quoting Harrington, 131 S.Ct. at 784). Unlike in Cullen or Harrington, the state court in Williams’s case did not issue a summary denial or fail to explain its reasoning for finding no prejudice.
I believe the more relevant cases for our prejudice review are Wiggins, Rompilla, and Simmons, which the majority dismisses as irrelevant because “the prejudice analysis in those cases was not governed by the deferential review that applies under AEDPA.” Supra at 838 n. 4. While it is true that those cases conducted de novo review of Strickland prejudice, they are more useful guides to us in the face of an objectively unreasonable state-court prejudice determination than are cases like Woodford or Cullen, where a petitioner failed to meet the standard set forth in § 2254(d). Because we need not defer to the state court’s unreasonable application of Strickland, we still must decide in the first instance whether the ineffectiveness of Williams’s counsel undermines our confidence that, but for counsel’s deficient performance, the jury would have concluded that the balance of aggravating and mitigating circumstances did not warrant death. I believe Williams, like the petitioners in Wiggins, Rompilla, and Simmons, has demonstrated sufficient evidence to undermine that confidence.
IV. Conclusion
In his post-conviction appeal, Williams submitted evidence that he was sexually abused as a child, that he grew up in a violent household with constant physical abuse, that he suffered from borderline mental retardation before dropping out of school, that he grew up and dealt with issues of mental illness, such as post-traumatic shock, depression, and drug addiction. His appeal alleged that his trial counsel was ineffective for not investigating or presenting any of this possible mitigating information at sentencing. His counsel has stated that he did not know this information, that he would have used this information had he known it, and that he had no strategic reason for not using it. Based on clearly established federal law, counsel’s performance was constitutionally ineffective.
The state-court determination that Williams was not prejudiced by counsel’s failure to present this information is also objectively unreasonable. The Supreme Court of Missouri did not assess how a failure to investigate can undermine counsel’s “strategic decisions” during the penalty phase. To find that Williams was not prejudiced because the new evidence would have “defied” or been “inconsistent with” the otherwise adopted strategy is an unreasonable conclusion based on an unreasonable premise.
The majority may offer a sensible explanation of why a hypothetical state court could have been reasonable if it had found the mitigating evidence offered by Williams might not have changed the outcome of his sentence. However, that explanation was not offered by either state court in this case. A reasonable hypothetical decision does not cure the objectively unreasonable explanation that the Supreme Court of Missouri did put forward, even under AEDPA’s strict standard. Under a proper application of Strickland, I believe Williams has presented evidence that establishes a probability sufficient to undermine confidence in the outcome of his sentence. I would therefore affirm the district court’s decision to grant the habeas petition.
. The state argued below that, because it was struck by the state motion court, this affidavit was not properly part of the record. At oral argument, however, the state conceded that the affidavit appeared to be part of the record considered by the Supreme Court of Missouri and is thus properly part of the record before us.
. The state argues that the record is underdeveloped and does not sufficiently demonstrate that Green had not investigated into or known about Williams’s abusive past before adopting a his penalty-phase theoiy. The state believes that "absent definitive evidence in the record of what trial counsel knew and when he knew it about Williams' actual childhood background ... it is reasonable to conclude that the presumption that counsel acted within the wide range of professional competence has not been overcome.” Reply Br. at 13. While I believe there is sufficient evidence in the record to affirmatively demon*846strate that counsel's investigation was lacking, the state's argument can be rejected as contrary to Supreme Court precedent. Strickland established that, although “a heavy measure of deference to counsel’s judgments” is due, any "particular decision not to investigate must be directly assessed for reasonableness in all the circumstances.” 466 U.S. at 691, 104 S.Ct. 2052. The state court denied Williams's request for an evidentiary hearing on his failure-to-investigate claim. Any- absence of information in the record about counsel’s decisions is not due to a lack of diligence on Williams's part. Deference to counsel's decisions would not justify the denial of an evidentiary hearing when the sufficiency of counsel’s investigation is in question. See Sinisterra v. United States, 600 F.3d 900, 907 (8th Cir.2010) (remanding a federal habeas claim for an evidentiary hearing because "counsel had an obligation to conduct a thorough background investigation and to exercise reasonable, professional judgment in determining the mitigation evidence to present during the penalty phase of [petitioner's] trial. [Petitioner] argues that counsel failed to meet those obligations, and the record does not affirmatively refute the factual assertions upon which [petitioner's] claim is based.” (internal citations omitted)).
. The state argues that Harrington requires us to ask whether fairminded jurists could disagree that there was no prejudice in this case. This argument makes the same mistake as that in the majority’s holding. Instead, Harrington requires us to ask whether fairminded jurists could disagree that the arguments or theories supporting the state court’s decision were inconsistent with clearly established federal law, as decided by the Supreme Court. See 131 S.Ct. at 786.
. In fact, in Johnson, the Eleventh Circuit determined that de novo review of prejudice was required given the Florida Supreme Court’s unreasonable application of the deficient-performance prong of Strickland. Johnson stated in full:
The Supreme Court's recent decision in Harnngton, where the state supreme court had issued a summary order denying relief, tells us that "[w]here a state court's decision is unaccompanied by an explanation, the habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief.” The Court’s instruction from Harrington does not apply here because the Florida Supreme Court did provide an explanation of its decision which makes clear that it ruled on the deficiency prong but did not rule on the prejudice prong, and it is also clear that the trial court's ruling on the prejudice prong did not address counsel's investigation and presentation of non-statutory mitigating circumstances evidence. Johnson II, No. CR 80-101 at 3-4. As a result, we are still required to follow the Court’s instructions from Rompilla and Wiggins and conduct a de novo review.
643 F.3d at 930 n. 9 (internal citation omitted).
. The majority does not believe that Wiggins is relevant to our review of the state court prejudice determination because the review of Strickland's second prong in Wiggins was not governed by AEDPA deference. Supra at 838 n. 4. While I would agree that any case offering de novo prejudice review would provide poor guidance under a "no reasonable basis” standard, see, e.g., Cullen, 131 S.Ct. at 1410-11 (discounting the usefulness of de novo prejudice reviews in determining reasonableness of a summary habeas denial), that is not the proper standard for a case where a state court affirmatively misapplies clearly established federal law. Wiggins is important to our review because it demonstrates how a failure to investigate affects a penalty-phase strategy. Had the state courts not affirmatively misapplied Supreme Court precedent and still concluded that no prejudice occurred, I would agree that § 2254(d) would prevent us from granting habeas relief. Because both state courts based their prejudice determination on a misunderstanding of the first Strickland prong that Wiggins declared objectively unreasonable under deferential AEDPA review, see 539 U.S. at 527, 123 S.Ct. 2527, Wiggins is relevant to our prejudice analysis.
. It is not a reasonable determination of the facts for the court to concoct a decision between two purportedly-mutually-exclusive options when counsel never testified to making such a decision. See Wiggins, 539 U.S. at 526-27, 123 S.Ct. 2527; Harris, 894 F.2d at 878.
. Similarly, Link v. Luebbers, 469 F.3d 1197 (8th Cir.2006), also cited by the majority, dealt with a state court that explicitly weighed the new mitigating evidence against the aggravating factors and concluded that there was no reasonable probability that a jury would have come to a different result had the mitigating evidence been introduced. Id. at 1204. Link is thus distinguishable for the same reason as Woodford.