Dissent by Judge Callahan
CALLAHAN, Circuit Judge,dissenting in part:
The Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”) “circumscribe[s]” a federal court’s role in reviewing a habeas claim that was “adjudicated on the merits in State court proceedings.” Johnson v. Williams, 568 U.S. 289, 298, 133 S.Ct. 1088, 185 L.Ed.2d 105 (2013) (quoting 28 U.S.C. § 2254(d)); Wiggins v. Smith, 539 U.S. 510, 520, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003). The Supreme Court has made clear, time and again, that our task is limited to deciding whether the state court was “objectively unreasonable” in its application of federal law, as determined by the United States Supreme Court, or in its determination of the facts that were before the trial court. See Lockyer v. Andrade, 538 U.S. 63, 75, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003); see also Johnson, 133 S.Ct. at 1094.
Today’s majority gives short shrift to the Supreme Court’s admonitions. Along the way, it misapplies Supreme Court case law, embarks on its own fact-finding mission, and overrules the jury’s credibility determinations. A meaningful application of our deferential standard of review under AEDPA, by contrast, compels the conclusion that the Nevada Supreme Court was not objectively unreasonable in rejecting Browning’s Ineffective Assistance of Counsel (“LAC”) claim, as well as his claims under Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963) and Napue v. Illinois, 360 U.S. 264, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959). I respectfully dis*477sent and would deny Browning’s habeas petition.1
I. Background
Sometime between 4:00 p.m. and 4:30 p.m. on November 8, 1985, Hugo Elsen was stabbed to death inside his jewelry store. In the course of the murder, the assailant smashed a glass display case and stole some 72 pieces of jewelry.
The police immediately identified Browning as a suspect based on eyewitness accounts and circumstantial evidence. Three individuals identified Browning at or near the jewelry store around the time of the murder. Hugo’s wife, Josy, actually witnessed the murder. She saw a “black man with a blue cap” raise a knife over Hugo. While she got only a side view of the attacker, she noted that his hair “was a little bit puffed out on the bottom” of his cap. That description was consistent with Browning’s Afro-like hairstyle. Police later recovered a blue cap with the word “Hollywood” on it in a dumpster near Browning’s motel room. At trial, Josy identified that cap as the one worn by the killer. Presented with a photographic lineup approximately a month after the murder, which included a photo of Browning, Josy did not choose Browning’s photo. However, when confronted with Browning in person at trial, Josy gave a positive identification of Browning as the assailant.
After Josy witnessed the murder, she ran out the back of the shop to the business next door and asked someone to call the police. Debra Coe was in the neighboring office, located just south of the Elsens’ store. When Josy told her that “there was a man standing over Hugo with a knife,” Coe went to the front of her office and saw a black man running “south.” Coe identified the man as wearing a blue cap, a jacket, and Levi’s, with hair sticking out about an inch from underneath the cap. At trial, Coe identified the blue “Hollywood” cap—presented as a state trial exhibit—as the one worn by the person she saw run by her office.
Later the same day, the police presented Coe with two men. She stated that the first man was “definitely not” the person she saw. The police then presented Browning, who was wearing no cap, shirt, or jacket. Coe said she “thought” Browning was the person she saw run by her office, but that she would have been more certain had he been wearing the cap and jacket. She noted, however, that his hair was “pressed down” as though he had been wearing a cap. When asked if she could be “more sure in [her] mind,” Coe said “[n]o, I wouldn’t think so, no.... they all look the same and that’s just what I think when I see a black person, that they all look the same.” Coe retracted this statement at trial, testifying that she did not really think that all black people looked the same. At trial, Coe positively identified Browning as the man she saw run by her office.
Charles Woods owned a jewelry store three ¡doors south of the Elsens’ shop. At around 4:30 p.m., he was standing outside the front of his store when he saw a black man jogging toward him. The man was wearing a dark-colored, “beret”-style cap, a light-colored shirt, and dark pants. Later that day, the police presented Woods with Browning, as they did with Coe, and Woods positively identified Browning as the man he saw. He also positively identified Browning at trial, though, unlike Josy and Coe, he testified that the hat found *478near Browning was not the same one worn by Browning.
Bradley Hoffman owned á store two doors from the Elsens’. At trial, he testified that he saw a man walk by his shop and approach the Elsens’ store about 20 minutes before the robbery-murder. He described the man as a Cuban, with a slight build, wearing Levi’s jeans and a blue baseball cap! Like Coe and Woods, Hoffman was presented with Browning by the police later that day. He stated that Browning was not the man he saw. He also testified that the blue “Hollywood” cap recovered by the police, and positively identified by Josy and Coe, was not the cap worn by the man he saw.
Hugo also identified his assailant. Officer David Radcliffe was. one of the first officers on the scene. He found Hugo lying in a pool of blood and in an “extremely serious” condition. Hugo identified his attacker as “a black man wearing a blue baseball cap.” Officer Gregory Branon also arrived early on the scene, and he, too, received a description of the attacker from Hugo. Hugo stated that the suspect was a “black male” who.was “wearing a blue baseball cap, blue windbreaker-type jacket, blue Levi’s” and who had “shoulder length”'hair. At trial, Branon testified that the description he received2 included a description of the attacker’s hair as a “J[h]eri-type curl.”
Besides the - eyewitness identifications, two witnesses—Randy Wolfe and his wife, Vanessa—testified that Browning confessed to them to committing the crime. At the time, the Wolfes, as well as Browning and his girlfriend, Marcia Gaylord, resided at the Normandy Motel. The two couples were acquainted. According to Randy, shortly after the robbery-murder, Randy found Browning in the Wolfes’ room, wearing a blue .“Hollywood” cap and surrounded by some of the stolen jewelry. Browning admitted to Randy to stealing the jewelry and killing Hugo. Randy then left to get the police, at which point Vanessa entered the room. According to Vanessa, she found Browning with a knife, and saw the “Hollywood” cap on the floor. Like Josy and Coe, Vanessa identified the state’s trial exhibit as the cap she saw near Browning.
According to Vanessa, Browning asked her to help him dispose of the evidence. Vanessa threw his shirt and' cap in a dumpster and hid the knife in a small closét under the stairway of the motel. The police later recovered the items. At trial, expert testimony established that the knife was “consistent” with Hugo’s wounds. Browning was arrested in the Wolfes’ motel room approximately half an hour after the murder. Several pieces of the stolen jewelry were in the room with him, as well as a tan jacket.
The prosecution also presented physical evidence. The tan jacket had blood on it, which was later identified as Type B— Hugo’s blood type. Browning’s fingerprints were also found at the crime, scene. Identification Specialist David.Horn testified at trial that several of the showcase counters had been “disturbed,” and that a glass door on the vendor’s side of one of the counters was broken. Browning’s prints were found on top of one of the counters and also on a fragment from the vendor-side glass door, which is the employee area.
Several pieces of exculpatory evidence were presented at trial. Horn testified to finding bloody shoeprints leading from *479Hugo’s body to the front door of the El-sens’ store. Those prints were consistent with a tennis shoe and did not match the shoes Browning was wearing at the time of his arrest. Browning’s trial counsel, Randall Pike, also called a hairstyle expert to testify to the difference between a Jheri curl and an Afro. Branon had testified that the description he received was of a person with “shoulder-length,” “J[h]eri curl” hair, whereas Browning had an Afro-style haircut. Pike presented the hairstylist with the same photographic lineup that the police showed Josy. She stated that four of the photos depicted individuals with Jheri curls—none of them was Browning.
Finally, the jury was presented with substantial evidence relevant to the Wolfes’ credibility. The jury knew that the Wolfes were habitual heroin and cocaine users, that Vanessa was a prostitute, that Randy had several prior convictions, that Randy was awaiting sentencing in another cage, that Vanessa used to “bilk people out of their money,” that Randy had kept some of the stolen jewelry, and that Randy had lied under oath about doing so at a preliminary hearing.
Ultimately, the exculpatory evidence was not enough to create reasonable doubt in any of the jurors’ minds. A Nevada jury found Browning guilty of four counts related to a robbery and murder at a Las Vegas jewelry store and sentenced him to death. Nee Browning v. State, 104 Nev. 269, 757 P.2d 351 (1988).
II. Procedural History
In 2004, the Nevada Supreme Court affirmed Browning’s conviction in a state habeas proceeding. Browning v. State, 120 Nev. 347, 91 P.3d 39 (2004). Browning argued, as is relevant to the instant appeal, that the prosecution withheld Brady material—i.e.,- exculpatory evidence—and that his trial counsel, Randall Pike, provided ineffective assistance for failing to adequately investigate his- case. See id. at 45, 54.
As 4s relevant here, Browning identified three pieces of allegedly exculpatory evidence in the state court post-conviction proceedings. . First,, he argued that the prosecution should have disclosed Officer Branon’s observation that the bloody shoe-prints at the crime scene predated the arrival of police and other first responders. Id. at 55. Second, he argued that the prosecution should _ have disclosed that the term “Jheri curl” came from, a black police officer (Branon), rather than the white victim. Id. Third, he claimed that the prosecution withheld information regarding a benefit given to Randy Wolfe.3 Id. at 54-55..
Browning also argued that Pike was ineffective for, among other things, failing to interview Branon. Id. at 46. Browning reasoned that had Pike done so, he would have discovered that the bloody shoeprints could not have been left by first responders, and that Hugo’s description of his attacker’s hair did not include the term “Jheri curl.’’ Id. Moreover, Browning argued that Pike should have interviewed the Wolfes, as théy were the prosecution’s key witnesses. Id.
The. Nevada Supreme Court upheld *480Browning’s guilty verdict.4 As to Browning’s IAC claim, the court ruled that it was a reasonable trial strategy to not interview Branon to discover his knowledge of the bloody shoeprints. Id. at 46. Pike’s investigation was sufficient to determine that the prints did not match Browning. Id. The court held that it was a “reasonable, tactical decision to leave the source of the prints uncertain.” Id. That way, Pike could argue that the real killer was the source. Id. Had he investigated further, he may have discovered that first responders were responsible for the prints, thereby neutralizing this defense. Id.
As for Pike’s failure to interview the Wolfes, the court held that it was a reasonable tactic for Pike to delegate witness interviews to his investigator, lest he interview witnesses personally and risk becoming a percipient witness himself. Id. Moreover, Pike’s investigator had “gathered enough information to permit [Pike] to adequately cross-examine the Wolfes on their version of events, their drug usage, their informer status, their lying, and their convictions and arrests.” Id.
Finally, while the court concluded that Pike was deficient for not discovering that “Jheri curl” was Branon’s term and not Hugo’s, it held that this was not prejudicial because the “issue of Browning’s hairstyle was extensively explored at trial.” Id.
As to Browning’s Brady claims, the Nevada Supreme Court held that the prosecution should have disclosed the benefit to Randy Wolfe. Id. at 54-55. The court reasoned that, “[u]nder Brady, even if the State and a witness have not made an explicit agreement, the State is required to disclose to the defense any evidence implying agreement or an understanding.” Id. at 55. Even so, the court ruled that there was no “reasonable probability of a different result” had the information been disclosed because Randy’s “credibility was extensively challenged at trial.” Id. The court also rejected Browning’s-bloody shoeprint Brady claim, noting that it had already deemed that information to be immaterial in the IAC context, and that the information was available to the defense had Pike interviewed Branon. Id. The court similarly rejected Browning’s Brady claim’ regarding the hairstyle evidence, noting that this information was, like the shoeprint evidence, available to the defense. Id.
Finally, the Nevada Supreme Court considered whether any errors, considered cumulatively, were prejudicial. Id. at 56. As is relevant here, the court considered Pike’s failure to discover Hugo’s true description of the killer’s hair, and the prosecution’s failure to turn over impeachment evidence regarding Randy Wolfe. Id. It determined that there was no “reasonable probability” that Browning would not have been convicted but for the cumulative effect of the errors. Id. The court reasoned that the “evidence of Browning’s guilt remains overwhelming.” Id.
Browning filed a petition for a writ of habeas corpus in the United States District Court for the District of Nevada. On November 28, 2011, Browning filed his ’ Fifth Amended Petition, which is the petition before us. On August 1, 2014, the district court denied Browning’s petition. The district court granted Certificates of Appealability (“COA”) on the issues of (i) whether the prosecution committed a Brady or Napue violation by failing to turn over information regarding the bloody shoeprints, (ii) whether the prosecution *481committed a Brady violation by failing to turn over evidence of a benefit to Randy Wolfe, and (iii) whether Pike provided ineffective assistance by failing to investigate the source of the bloody shoeprints, Hugo’s description of his attacker’s hair, and the Wolfes’ credibility. Browning appealed.
III. Standard of Review
We review the district court’s decision de novo, while applying AEDPA’s “highly deferential standards” to the last reasoned state court decision—here, the Nevada Supreme Court’s 2004 denial of post-conviction relief. Davis v. Ayala, — U.S. -, 135 S.Ct. 2187, 2198, 192 L.Ed.2d 323 (2015); see Reis-Campos v. Biter, 832 F.3d 968, 973 (9th Cir. 2016), cert. denied, — U.S. -, 137 S.Ct. 1447, 197 L.Ed.2d 658 (2017). The state court’s decision receives binding deference unless it is “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States” at the time of the state court’s decision, or if it was “based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” 28 U.S.C. § 2254(d)(1), (2); Wiggins, 539 U.S. at 520, 123 S.Ct. 2527.
Surmounting AEDPA deference is “daunting.” Taylor v. Maddox, 366 F.3d 992, 1000 (9th Cir. 2004). This is by design out of respect for state court proceedings, and is “satisfied in relatively few cases.” IcL; see also Williams, 133 S.Ct. at 1094. “[AEDPA] reflects the view that habeas corpus is a ‘guard against extreme malfunctions in the state criminal justice systems,’ not a substitute for ordinary error correction through appeal.” Harrington v. Richter, 562 U.S. 86, 102-03, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011) (quoting Jackson v. Virginia, 443 U.S. 307, 332 n.5, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979) (Stevens, J., concurring)). It is not enough that a federal court determine, in its “independent judgment,” that the “state-court decision applied clearly established federal law erroneously or incorrectly.... Rather, that application must be objectively unreasonable.” Lockyer, 538 U.S. at 75-76 (emphasis added and internal citation omitted). Where “it is possible to read the state court’s decision in a way that comports with clearly established federal law ... we must do so.” Mann v. Ryan, 828 F.3d 1143, 1157-58 (9th Cir. 2016). In other words, we must uphold a state court determination even if we would have concluded, on de novo review, that the state court committed legal error, so long as a fair-minded jurist could decide otherwise. See Lockyer, 538 U.S. at 75-76, 123 S.Ct. 1166. To overrule a state court’s decision requires that its ultimate conclusion be so unreasonable that it there is no “possibility for fairminded disagreement.” Ayala, 135 S.Ct. at 2199 (internal quotation marks omitted). Finally, a federal court may not review the facts of a case de novo; we must begin with the “presumption” that the state court’s factual determinations are correct. Taylor, 366 F.3d at 999.
IV. Browning’s Claims for Relief
Browning’s petition comes down to three pieces of evidence: the description of the assailant provided by the victim, Hugo El-sen; bloody shoeprints leading from Hugo’s body; and the benefit received by Randy Wolfe for his testimony. None of this evidence compels a finding that the Nevada Supreme Court was objectively unreasonable in rejecting Browning’s petition for post-conviction relief.
First, Hugo’s description of the killer was presented to the jury, and the jury knew that the description conflicted with other eyewitness testimony. That the jury *482did not know the term “Jheri curl” was the testifying officer’s and not the victim’s did not appreciably diminish this conflict. Second, Pike’s ignorance of the fact that the bloody shoeprints predated the arrival of first responders did not undercut his defense that someone other than Browning committed the murder because he did know that the prints did not match Browning. And third, the jury was presented with a eavalcade of impeachment evidence against Randy and Vanessa Wolfe. That the jury did not know that the prosecution would later help Randy secure a benefit for his testimony against Browning—a fact that apparently Randy did not even know—makes no difference because it was not reasonably probative of his credibility. And even if it was, the information was, at most, cumulative.
The weakness of the alleged exculpatory evidence is enough to reject Browning’s habeas petition on its own. But it positively blanches when set against the substantial evidence inculpating Browning: his fingerprints were found on the vendor’s side of a glass display case, which is off-limits to customers; he was found by police with some of the stolen jewelry; numerous eyewitnesses identified him; and he confessed to the Wolfes.
a. Browning Fails to Establish a Brady Violation Because the Purported Brady Evidence Is Either Not Exculpatory or Not Material
The Nevada Supreme Court decided Browning’s Brady claim on the merits. We therefore apply AEDPA’s deferential standard of review, and may only find a Brady violation if the state court’s decision was an objectively .unreasonable application of Supreme Court law or an unreasonable determination of the facts: 28 U.S.C. § 2254(d)(1), (2).
Under Brady v. Maryland, the prosecution must disclose “evidence that is both favorable to the accused and material either to guilt or to punishment.” United States v. Bagley, 473 U.S. 667, 674, 105 S,Ct. 3375, 87 L.Ed.2d 481 (1985) (internal quotation marks omitted). “‘Evidence is material within the meaning of Brady when there is a reasonable probability that, had the evidence been disclosed, the result of the . proceeding would have been different.’ ” Turner v. United States, — U.S. -, 137 S.Ct. 1885, 1893, 198 L.Ed.2d 443 (2017) (quoting Cone v. Bell, 556 U.S. 449, 469-70, 129 S.Ct. 1769, 173 L.Ed.2d 701 (2009)). “ ‘A reasonable probability of a. different result is one in which the suppressed evidence undermines, confidence in the outcome of the trial.’” Id. (quoting Kyles v. Whitley, 514 U.S. 419, 434, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995)).
Thus, to establish a Brady violation, a defendant must prove: 1) the evidence at issue is favorable to the accused, either because it is exculpatory, or because it is impeaching, 2) the evidence was suppressed ... either willfully or inadvertently, and 3) prejudice resulted, meaning there is a reasonable probability that disclosing the evidence to the defense would have changed the result.
Andrews v. Davis, 798 F.3d 759, 793 (9th Cir. 2015) (internal quotation marks and adjustments omitted) (citing Strickler v. Greene, 527 U.S. 263, 281-82, 119 S.Ct. 1936, 144 L.Ed.2d 286 (1999); Bagley, 473 U.S. at 682, 105 S.Ct. 3375).
The Nevada Supreme Court reasonably concluded that no Brady violation occurred because the hairstyle and impeachment evidence is not exculpatory, and, while the shoeprint evidence is exculpatory, a fair-minded jurist could deem it immaterial when viewed collectively with the abundant evidence of Browning’s guilt.
*483i. Whether the Purported Brady Material Is Exculpatory
1. The Hairstyle Evidence
The jury did not know that Officer Bra-non received a description of the assailant from the victim himself. Instead, Branon recounted the description that Hugo gave him, without identifying its source. He told the jury that
[t]he description we received was black male, adult in his late twenties, wearing a blue baseball cap, blue windbreaker-type jacket, blue Levi’s, He was medium ■complexioned, bore a mustache and what was described as a shoulder length J[h]eri-type curl.
This was a mostly accurate reporting of Hugo’s dying declaration, except that the term “J[h]eri-type curl” was Branon’s, not Hugo’s. Browning argues that the prosecution’s failure to turn over the fact that Branon’s description came from the victim himself, and that the term “J[h]eri-type curl” was Branon’s, not Hugo’s, was materially exculpatory.
The Nevada Supreme Court held that no Brady violation occurred because the exculpatory information was “reasonably available to the defense.” Browning, 91 P.3d at 55. And, at any rate, the information did not give rise to a “reasonable probability of a different result.” Browning, 91 P,3d at 46. This was because the “issue of Browning’s hairstyle was extensively explored at trial.” Id.
The Nevada Supreme Court’s conclusion was not objectively unreasonable because a fair-minded jurist could conclude that the hairstyle evidence was not exculpatory, let alone materially so. See Ayala, 135 S.Ct. at 2199. The majority concludes otherwise only by asserting a novel view of Brady that extends the state’s obligations into the murky zone of interpretations of otherwise neutral- facts. That “Jheri curl” was Bra-non’s term and not Hugo’s does not help Browning. The only basis for deeming this information potentially exculpatory is that Prosecutor Daniel Seaton leveraged its purported source—Hugo, a white male—to argue that the speaker probably confused the terms Afro and Jheri curl. Seaton reasoned that the declarant was just “some white person” who didn’t “really know[] the true definition of J[h]eri-curl.” ■ But while it is true that the source of the term was a black male, the evidence itself is not exculpatory, and the jury was free to disregard Seaton’s unsubstantiated speculation as just that.
2. The Benefit to Randy Wolfe for His Testimony
It is undisputed that the prosecution intervened on Randy Wolfe’s behalf in a separate trial in which Randy was the defendant after Randy testified in Browning’s trial. Yet the Nevada Supreme Court determined that the prosecution should have disclosed the benefit .to the defense. Browning, 91 P.3d at 54-55. The court reasoned that “the State is required to disclose to the defense any evidence implying an .agreement or an .understanding.” Id.
It appears that the Nevada Supreme Court’s determination was not based on any facts in the record. Both Randy and Prosecutor Seaton stated that there was no plea bargaining with Randy Wolfe regarding Browning’s ease, and Browning points to nothing in the record showing that Randy expected a benefit for his testimony. The Nevada Supreme Court therefore erred in concluding that the evidence was exculpatory. See Amado v. Gonzalez, 758 F.3d 1119, 1136 (9th Cir. 2014) (quoting 28 U.S.C. § 2254(d)(2)) (state court made' an “unreasonable determination of the facts” where “[tjhere was nothing in the record that could support” its finding).
*484The majority correctly notes that the evidence could only be exculpatory if Randy actually expected a benefit for his testimony that was not already disclosed. But the Nevada Supreme Court never made such a finding. Under AEDPA, we are not entitled to weave facts out of whole cloth just to make sense of a state court’s determination. See id. Because nothing in the record shows that any deal was made— expressly or otherwise—between Seaton and Randy at the time of trial, nothing was suppressed and no Brady violation occurred.5
3. The Bloody Shoeprint Evidence
The jury knew that the bloody shoe-prints leading from Hugo’s body to the front of the Elsens’ store did not match the loafers Browning wore at the time of his arrest. But post-conviction testimony also established that Branon knew that the shoeprints were present before first responders arrived at the scene. Under Brady, that knowledge is imputed to the prosecution as a whole. Kyles, 514 U.S. at 437-38, 115 S.Ct. 1555 (the “individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the case, including the police”). Yet Specialist Horn testified that the prints were consistent with tennis shoes, and noted that first responders often wear tennis shoes at crime scenes. The jury was therefore left with the impression that first responders may have left the shoeprints. That was false. Moreover, the prosecution offered no evidence to suggest that someone besides the true killer could have been the source of the prints. Evidence that the shoeprints were present before the first responders arrived was therefore exculpatory.
ii. Whether the Purported Brady Evidence is Material
The majority correctly notes that the three Brady claims must be considered “collectively” to determine whether they are material. Kyles, 514 U.S. at 436, 115 5.Ct. 1555; Turner, 137 S.Ct. at 1895 (considering the “cumulative effect of the withheld evidence”). Materiality is a determination of whether disclosure of all pieces of exculpatory evidence, taken together, gives rise to a “reasonable probability” of a different outcome. See Cullen v. Pinholster, 563 U.S. 170, 189, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011). “A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Id.
The majority concludes that the cumulative effect of three pieces of evidence— Hugo’s description of his killer’s hairstyle, the source of the bloody shoeprints, and the benefit for Randy Wolfe’s testimony— tips the scales in favor of a finding of materiality.6 The majority’s conclusion is *485not compelled either by the evidence or binding Supreme Court law.
1. The majority errs by assuming that the benefit to Randy Wolfe and the hairstyle evidence is Brady material. As discussed (see Part IV.a.l, 2), both pieces of evidence are reasonably viewed as not exculpatory and so withholding the information has no material impact on the case. But even if the hairstyle evidence could be deemed exculpatory because its disclosure would have foreclosed Seaton’s argument that “some white person” confused the terms “Jheri curl” and “Afro,” it was not unreasonable for the Nevada Supreme Court to deem it immaterial. The majority, concedes that all of the words Branon used to describe Browning’s hair are inconsistent with an Afro—“Jheri curl,” “shoulder length,” “loosely curled,” and “wet.” In other words, Branon’s use of the term “Jheri curl” did nothing to diminish the conflict in the eyewitness testimony that was squarely before the jury.7 On the one hand, Hugo described Browning as having wet, shoulder-length hair. On the other, Josy and Coe described the person they saw as having short hair that “puffed” or stuck out the back of a blue cap. That the jury was also exposed to the term “Jheri curl” does not somehow reconcile these inconsistent accounts.
The term “Jheri curl” also lost its salience over the course of the trial. Browning’s counsel presented a hairstyle expert, Annie Yates, who testified to the difference between a Jheri curl and an Afro. Pike then showed her a 12-person photographic array and asked her to identify which persons had Jheri curls. The array included a picture of Browning. Yates stated that four of the persons had Jheri curls, none of whom was Browning. The jury therefore knew that Browning’s Afro hairstyle was inconsistent with a Jheri curl. Accordingly, it was not objectively unreasonable for the Nevada Supreme Court to reject the hairstyle Brady claim because the hairstyle evidence was not materially exculpatory.
Undeterred, the majority insists jhat the information was critical to the defense because, it reasons, “Jheri curl” is the only term “susceptible to the argument that the speaker could have seen an Afro and used the wrong term because he was unfamiliar with African American hairstyles.” But the fact that Seaton leveraged the term’s reía- - five obscurity to spin speculation does not detract from the fact that a fair-minded juror could reasonably dismiss such conjecture as unfounded. See Ayala, 135 S.Ct. at 2199. Moreover, Seaton’s argument makes no sense, because the description Branon received—and which was before the jury—included the fact that the assailant’s hair was “shoulder length.” Thus, even if the jury entertained the far-fetched notion that the speaker said “Jheri-curl” when he meant “Afro,” it was still faced with a clear conflict in the evidence: was the assailant’s hair shoulder-length or in an Afro?
2. As to the bloody shoeprint evidence, the majority chides the Nevada Supreme Court for engaging in “pure speculation” *486for suggesting that the prints were probar bly left by Josy or Coe. To be sure, the evidence did not support such a conclusion. But neither did it suggest otherwise. Under AEDPA, our review is limited to the original trial record, Pinholster, 563 U.S. at 181, 131 S.Ct. 1388, and nothing in that record indicates that the shoeprints were not those of Josy or Coe. The majority buries this fundamental rule in a footnote, while simultaneously embarking- on its own fact-finding mission, concluding that, based on its review of the images of the shoe-prints, “the shoeprints appear to us larger than those of a typical woman’s shoe.” Tellingly, the majority cites no expert testimony from the trial record to support its observation—because there is none. Reading with its chin, the majority commits double error under AEDPA: not only does it draw its own evidentiary conclusions, see Taylor, 366 F.3d at 999, but it engages in the very “speculation” for which it roundly criticizes the Nevada Supreme Court.
Limiting our review to the trial record, as we must, the shoeprints’ provenance is unknown. What was known at the time of trial, and what was presented to the jury, was the fact that the shoeprints did npt match Browning. In- other words, they pointed to someone .else, which is consistent with the defense’s theory that a black Cuban associate of the Wolfes committed the crime. Had the defense known that the prints predated the arrival of the first responders, its theory would have been the same. And to the extent the information had some exculpatory value by eliminating one innocuous explanation for the prints (i.e., that they were left by first responders), the record does not compel a conclusion that the information was material. Indeed, even had the jury known that first responders were not the source, it could have reasonably inferred that the prints may have been left by a person who was not the killer—e.g., Josy, Coe, or someone else entirely. Either way, while the jury was left speculating -about the prints’ origin, it knew that they did not incriminate Browning.
3. Finally, even if the shoeprint evidence, viewed in a vacuum, was significant, it is reasonably deemed immaterial when considered collectively with the evidence inculpating Browning. We asséss the combined effect' of both undisclosed exculpatory evidence and the evidence that was before the jury as a whole in determining whether there is a “reasonable probability” that disclosure of Brady material would have changed the outcome. See Kyles, 514 U.S. at 437, 115 S.Ct. 1555 (materiality of Brady evidence is judged according to the “net effect” of the evidence). As the Nevada Supreme Court found, substantial evidence inculpated Browning:
• Browning’s fingerprints in the El-sens’ store. Browning’s fingerprints were found on the disturbed jewelry counter—both on the top side of the glass and on a fragment .from the broken sliding-glass door on the vendor’s side of the display case. The majority minimizes this fact, which it coneedés is “probably the strongest evidence against Browning,” by surmising various innocent explanations: the fingerprints could have predated the murder, Browning may have made the prints' during the commission of the crime but someone else stabbed • Hugo, or the other unidentified prints at the scene could belong to the true killer. But the jury could have easily drawn a contrary conclusion: the fact that Browning’s prints were found on a piece of glass broken in the commission of the. robbery-murder, and on the vendor’s side of the glass case, points to his guilt. Combined with the other inculpatory evidence (discussed below), this would not have been an *487unreasonable inference. Cf. Mikes v. Borg, 947 F.2d 353, 361 (9th Cir. 1991) (fingerprints alone—absent evidence that they were made “during the commission of the crime”—are insufficient evidence of guilt where the government’s case is “supported solely by evidence that the defendant’s fingerprints were found, on th[e] object” (emphasis added)).
• The eyewitness identifications. Three eyewitnesses—Josy, Coe, and Woods—identified Browning as the person they saw in or near the Elsens’ store around the time of the murder. The majority insists the identifications were “flawed” because Josy and Coe were at first equivocal, and the police used a suggestive “show up” identification procedure at the crime scene. But the majority ignores Josy’s testimony that the person she saw raise a knife over her husband wore a blue “Hollywood” cap with a “puff[ ]” of hair protruding out the bottom—an account that fits with Coe’s description of the alleged assailant, matches Browning’s hairstyle at the time of his arrest, and is consistent with the fact that the same blue cap was found in a dumpster outside the Wolfes’ motel room.8 Moreover, Woods’ contemporaneous identification of Browning was unequivocal. A fairminded jurist could therefore reasonably conclude that the identifications were strong evidence of Browning’s guilt. ,
• The jewelry in the motel room. Police officers discovered Browning with some of the stolen jewelry in the Wolfes’ motel room shortly after the murder. The majority notes, that Browning’s fingerprints were not found on the stolen jewelry, and the rest of the loot was later turned over by the Wolfes. But the jury knew these facts and was entitled to infer guilt from this evidence when considered with the other evidence inculpating Browning.
• Browning’s Confession to Bandy arid Vanessa Wolfe. The Wolfes’ testimony was devastating to the defense. According to the Wolfes, Browning confessed to them to committing the robbery-murder, and then asked .Vanessa to help cover it up.
To be sure, the Wolfes’ testimony was susceptible to substantial impeachment. They were habitual drug users with prior convictions and a penchant for lying. But all of this was presented to the jury. The jury learned that Randy had a’ history of illegal drug use,-used heroin four days before testifying, stole property, used his wife’s prostitution to support his drug use, lied under oath in Browning’s case about keeping some of. the stolen jewelry, had three prior felony .convictions, and was awaiting, sentencing in a pending casé. .The jury was entitled to credit the Wolfes’ testimony notwithstanding that the' Wolfes were, by all accounts, thoroughly unscrupulous characters.
Tellingly, the majority provides no support for the .suggestion that the jury could not have reasonably believed the' Wolfes. Instead, it spee-ulates that perhaps one more piece of exculpatory evidence—e.g., the source of the bloody shoeprints—would have tipped the scales for at least one ju*488ror. But our role is not to reweigh the evidence and make fresh credibility-determinations. See Williams v. Ryan, 623 F.3d 1258, 1266 (9th Cir. 2010). Because the jury was entitled to credit the Wolfes’ testimony, and because that testimony directly implicated Browning as the murderer, a fair-minded jurist could have concluded that Browning’s confession to the Wolfes was strong evidence of his guilt.
• The Knife. Vanessa Wolfe testified that she found Browning with a knife in her motel room, and that he asked ■ her to help him dispose of it. If the jury believed her, this was compelling evidence inculpating Browning. Moreover, at trial, a prosecution expert testified that Hugo’s injuries were “consistent” with wounds made by the knife, though he did not know whether the particular knife recovered by the police was the actual murder weapon.9
In sum, the jury had before it substantial evidence inculpating Browning. The majority concludes otherwise only by reweighing the evidence: by deciding that the Wolfes were not credible, dismissing the eyewitness identifications of Browning as “flawed,” and minimizing the highly in-culpatory fingerprint evidence.10 In so doing, the majority ignores the presumption owed to the Nevada Supreme Court’s factual determinations and decides for itself the strength of the case against Browning. See Taylor, 366 F.3d at 1000. That is error. Considering the limited exculpatory value of the shoeprint evidence, combined with the substantial evidence pointing to Browning’s guilt, the jury could have convicted Browning even if it was presented with the purported Brady material. The Nevada Supreme Court was therefore not objectively unreasonable in rejecting Browning’s Brady claim.
b. Browning’s Ineffective Assistance of Counsel Claim Fails Because the Nevada Supreme Court Was Not Objectively Unreasonable in Concluding That Pike Acted According to a Reasonable Trial Strategy
To prevail on his IAC claim, Browning must show (i) that his trial counsel’s assistance fell below an objective standard of reasonableness, and (ii) that but for his counsel’s deficient performance, there is a “reasonable probability that ... the result of the proceeding would have been different.” Strickland v. Washington, 466 U.S. 668, 691-94, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984); accord Hinton v. Alabama, — U.S. -, 134 S.Ct. 1081, 1087-88, 188 L.Ed.2d 1 (2014); Wiggins, 539 U.S. at 521, 123 S.Ct. 2527. “A reasonable probability is a probability sufficient to undermine *489confidence in the outcome.” Pennsylvania v. Ritchie, 480 U.S. 39, 57, 107 S.Ct. 989, 94 L.Ed.2d 40 (1987) (internal quotation marks omitted). This is a rigorous standard. The defendant must show both -that “counsel made errors so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment,” and that the “deficient performance prejudiced the defense.” Strickland, 466 U.S. at 687, 104 S.Ct. 2052.
The Nevada Supreme Court decided Browning’s IAC claim on the merits, and therefore our review is governed by AED-PA’s deferential review standard. See Johnson, 133 S.Ct. at 1094. Because we must also afford counsel’s performance a presumption of reasonableness, Strickland, 466 U.S. at 690, 104 S.Ct. 2052, claims of IAC under AEDPA are “doubly deferential.” Yarborough v. Gentry, 540 U.S. 1, 6, 124 S.Ct. 1, 157 L.Ed.2d 1 (2003). When applying Strickland in the AEDPA context, the question is “whether there is a reasonable argument that counsel satisfied Strickland’s deferential standard, such that the state court’s rejection of the IAC claim was not an unreasonable application of Strickland.” Murray v. Schriro, 746 F.3d 418, 465-66 (9th Cir. 2014) (internal quotation marks and citation omitted).
Browning contends that his trial counsel, Randall Pike, provided ineffective assistance because he (i) did not interview Officer Branon and learn about Hugo’s description of his assailant’s hairstyle, (ii) did not investigate the source of the bloody shoeprints; and (iii) did not interview the Wolfes. The Nevada Supreme Court agreed on (i), but not on (ii) or (iii). As to (i), the Nevada Supreme Court held that the deficiency was not prejudicial because the hairstyle evidence was “extensively explored at trial.”11 Browning, 91 P.3d at 46.
i. The Hairstyle Evidence
The Nevada Supreme Court noted that, while Pike’s performance in not discovering Hugo’s true description of his assailant’s hair was deficient, the evidentiary conflict was squarely before the jury. See id. It concluded that, in light of the conflicting testimony, combined with the “strong evidence of Browning’s guilt,” there was “no reasonable probability of a different result if counsel had discovered and presented the evidence that ‘j[h]eri-curl’ was the officer’s term, not the victim’s.” Id.
Under AEDPA, our task is to decide whether the Nevada Supreme Court’s conclusion that Pike’s deficient performance did not prejudice Browning’s defense was objectively unreasonable. The majority concludes that it' is, noting that “Hugo’s description of his assailant’s hair is powerful evidence of Browning’s innocence.” But, as described in the context of Browning’s Brady claim (see Part IV.a, supra), the jury did hear Hugo’s description of his killer’s hair, and so the conflict between his account and the accounts of Josy and Coe was squarely presented. The only question is whether there exists a reasonable probability that the jury would have acquitted Browning had it known that “Jheri curl” was Branon’s term and not Hugo’s. For the reasons already discussed, the Nevada Supreme Court’s determination was not objectively unreasonable because the hairstyle evidence could reasonably be viewed as not exculpatory.
*490ii. The Bloody Shoeprint Evidence
The Nevada Supreme Court held that it was not deficient performance for Pike to forgo investigating the actual source of the bloody shoeprints. Id. at 46. In the post-conviction proceeding, Pike explained that his decision was part of a strategy to “ov-ereaste ] a shadow of a doubt,” by pointing to someone else—a black Cuban associate of the Wolfes—as the assailant. If he had investigated the time source of the shoe-prints, he may have discovered that someone other than the killer—e.g., a first responder—was the true source, which would have undercut this theory. The Nevada Supreme Court reasoned that “[a]s long as the source of the prints was unknown, [Pike] could argue to the jury that the actual murderer had left them.” Browning, 91 P.3d at 46.
The Nevada Supreme Court was not objectively unreasonable in concluding that Pike executed a reasonable trial strategy based on his investigation of the evidence. An attorney’s decision must be “evaluate[dj ... from counsel’s perspective at the time” of the decision, thereby “elimi-nat[ing] the distorting effects' of hindsight.” Strickland, 466 U.S. at 689, 104 S.Ct. 2052. At the time of trial, Pike had gathered enough information to know that the shoeprints did not match his client. It was an open question whether first responders, Josy, Coe, or someone else was the source. Interviewing first responders about their shoe's could have resulted in the discovery that a first responder was the source. This is therefore not, as the majority insists, an “extreme instance of strategy determining investigation.” To the contrary, it is an example of trial counsel making a “strategic ehoice[]” after “less than complete investigation” that is rooted in a sound theory of the defense. See Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052.
, The majority relies on Weeden v. Johnson, 854 F.3d 1063 (9th Cir. 2017), but that case is not on point. There, a split panel of this court considered counsel’s failure to obtain a psychological evaluation of the defendant. Id. at 1067-68. The defendant had been convicted of attempted robbery and first degree felony-murder after a botched robbery attempt resulted" in the victim’s death. Id. at 1067. Because the defendant was not present at the crijne scene, the prosecution pressed the theory that she planned and facilitated the crime. Id. at 1066, 1070. Thus, the defendant’s “mental condition was an essential factor in deciding whether she actually , had the required mental states for the crime.” Id. at 1070. Yet defense counsel did not pursue psychological evidence that could have shown that the defendant lacked the requisite mental capacity to plan the robbery. Id. at 1068-69. He argued that it was a reasonable tactic to remain ignorant, because a psychological profile could have revealed that.his.client was easily manipulated. Id. That may have given the prosecution an opening to argue that even if the defendant “did not understand the magnitude of the robbery, she nonetheless went along with it.” Id. (internal quotation marks omitted).
The court held that defense counsel had unreasonably “put[] the cart before the horse” by allowing trial strategy to dictate the scope of the investigation. Id. at 1070. While I stand by my dissent in Weeden, even on the Weeden majority’s own terms that case is distinguishable in relevant part. There, counsel conducted no investigation on an issue that was central to the prosecution’s burden of proof. See id. Here, by contrast, Pike collected enough information to know that the shoeprints did not match his client, and this discovery supported his trial strategy of arguing that someone else committed the murder.
*491Moreover, unlike the defendant’s psychological competency in Weeden, the importance of discovering the source of the shoeprints was not evident until Branon revealed—some 15 years later—that the prints predated his arrival at the scene. At the time of trial, Pike knew that the prints were exculpatory because they did not match his client, and there was no reason to believe that someone other than the true killer was the source. Indeed, Pike may have reasonably assumed that first responders would have exercised care to preserve the crime scene. The fact that first responders did not make the prints only became relevant after Horn’s testimony suggested that they may have been the source. Whether Pike performed adequately is a measure of his own actions in preparing for trial, not a function of misleading testimony introduced by the prosecution. Had Horn not testified about first responders’ shoe. preferences, Browning would have no claim of deficient performance based on the shoeprint evidence. To the contrary, Pike made clear to the jury the shoeprints’ exculpatory value.
The majority’s reasoning also creates a tension between the prosecution’s Brady obligations and defense counsel’s performance responsibilities. A Brady violation occurs where the prosecution fails to turn over evidence requested by the defense, or where it fails to “volunteer exculpatory evidence never requested.” Kyles, 514 U.S. at 433, 115 S.Ct. 1555 (emphasis added); see Strickler, 527 U.S. at 280, 119 S.Ct. 1936. The underlying premise is that some evidence is discoverable by diligent inquiry, while other evidence- is not. The shoe-print evidence falls into the latter category because, until Horn’s testimony, Pike reasonably did not think to ask whether the prints were left by someone, other than the killer. Boiled.down, Browping’s grievance reduces to a Brady, not an IÁC, claim.
Finally, in Weeden, even if counsel had obtained a psychological report that was unfavorable to his theory of the case, he was not required to disclose it to the prosecution. 864 F.3d at 1070, Nor would an adverse psychological' finding have precluded him from arguing, as he did, that the defendant, a 14-year-old girl, could be “easily manipulated by older people” because of her age; See id. at ’1067 (internal quotation marks omitted). Thus, he arguably had nothing to lose but potentially much to gain by investigating his client’s psychological profile. Here, by contrast, had Pike discovered that first responders were actually responsible for the shoe-prints, this would have undercut his argument that a black Cuban was the true killer.12
The majority’s reliance on Harrington v. Richter, a case in which the Supreme Court rejected an IAC claim, is equally puzzling. There, the Court—reversing an en banc decision of this court—upheld a state court’s ruling that defense counsel’s failure to test blood evidence was a reasonable trial strategy, Richter, 562 U.S. at 107-08, 131 S.Ct. 770, Had defense counsel tested the blood, he would have discovered—as post-conviction results revealed— that the evidence supported the defendant’s version of events. See id. But with*492out the benefit of hindsight, defense counsel faced two possible outcomes from a blood test: a result that corroborated his client’s account and one that undermined it. See id. at 108, 131 S.Ct. 770. Faced with the “serious risk[]” of an adverse test result, the Court held that the attorney was not required to “pursue an investigation that ... might be harmful to the defense.” Id.
Pike faced a similar dilemma here. Investigating the bloody shoeprints could have bolstered his theory—and Browning’s account—that a black Cuban murdered Hugo, or it could have undermined it.13 Only in the “harsh light of hindsight” does Pike’s strategy appear unreasonable. As in Richter, “[i]t was at least arguable that a reasonable attorney could decide to forgo inquiry into the [shoeprint] evidence in the circumstances here.” See Richter, 562 U.S. at 106, 131 S.Ct. 770.
The majority distinguishes Richter on the ground that defense counsel there did not completely trust his client’s version of events. Id. at 108,131 S.Ct. 770. According to the majority, this distinction makes all the difference because, here, “Pike had no reason to disbelieve Browning’s assertions that he had been framed by the Wolfes.”14 But Richter does not teeter on so thin a reed. That counsel there “had reason to question the truth of his client’s account” was only one factor considered by the Court. See id. at 108, 131 S.Ct. 770 (noting that “[ejven apart from th[e] danger” that the defendant was lying, testing the blood could have “shift[ed] attention to esoteric matters of forensic science, distract the jury from whether Johnson was telling the truth, or transform the case into a battle of the experts” (emphasis added)). Ultimately, the Court rejected the IAC claim because defense counsel’s tactic was consistent with a strategy of “try[ing] to cast pervasive suspicion of doubt [rather] than *493to strive to prove a certainty that exonerates.” Id. at 109, 131 S.Ct. 770. Same here. Pike, in his words, executed a strategy of “overcasting a shadow of doubt, as opposed to proving.” The Nevada Supreme Court’s determination that Pike’s performance was not deficient for failing to discover the shoeprint evidence was therefore not objectively unreasonable. See Browning, 91 P.3d at 46.
iii. The Wolfes
The Nevada Supreme Court concluded that Pike acted reasonably in not interviewing the Wolfes. Browning, 91 P.3d at 46. The court noted Pike’s policy of delegating the responsibility of interviewing witnesses to his investigator rather than conducting interviews himself, which could have made Pike a percipient witness. Id. The court concluded that this was a “reasonable tactic.” Id. But Pike’s investigator never interviewed the Wolfes. "While he sought permission to do so, Pike denied his requests. The Nevada Supreme Court’s conclusion that Browning “has failed to show that counsel was ineffective” because it was “reasonable” to delegate interview responsibility to an investigator was therefore an unreasonable determination of the facts. See id.-, 28 U.S.C. § 2264(d)(2).
Even so, the Nevada Supreme Court’s ultimate conclusion was not unreasonable because Browning fails to show prejudice. See Richter, 562 U.S. at 88, 131 S.Ct. 770. Assuming it was deficient performance to not interview the Wolfes, it is unclear how any additional information that Pike may have uncovered would have likely changed the outcome of the trial. As the majority correctly notes, the jury was presented with a “mountain of evidence providing potential reasons to doubt the Wolfes’ credibility.”15 The jury knew that the Wolfes had a history of lying, stealing, drug use, and prior convictions. Piling on one more bad act would have simply added to the already formidable “mountain.” See Lewis v. Cardwell, 609 F.2d 926, 928 (9th Cir. 1979) (counsel’s failure to discover impeachment evidence that was merely cumulative did not prejudice the defendant).
Because Browning fails to show how interviewing the Wolfes would have resulted in a “reasonable probability” of a different outcome, the Nevada Supreme Court was not objectively unreasonable in rejecting Browning’s IAC claim on this ground.
Y. Conclusion
The role of the federal judiciary in reviewing habeas petitions from state courts is limited, and for good reason. AEDPA, when properly applied, prevents federal courts from unnecessarily intruding on states’ broad authority to administer their own criminal justice systems. That is why we are tasked with considering not whether we would decide a case differently, but whether the state court’s determination is beyond fair-minded debate. Today’s majority repeatedly loses sight of this standard. In light of the substantial evidence inculpating Paul Browning in Hugo Elseris murder, the limited exculpatory value of the alleged Brady material, and the fact that Pike’s representation reasonably did not prejudice Browning’s defense, I would affirm the district court’s denial of Browning’s petition for habeas relief. I respectfully dissent.
. Branon did not identify Hugo as the source of the description until post-conviction proceedings some 15 years later.
. Pretrial, the prosecution stipulated that Randy was not promised anything for his testimony, a point that Randy confirmed on the stand. See Browning, 91 P.3d at 55; Trial Tr. at 4 (Dec. 8, 1986). During post-conviction proceedings, Prosecutor Seaton admitted that after Browning’s trial, he told the judge in a case in which Randy was the defendant that Randy had helped with Browning’s trial. Id. Randy ended up receiving probation for attempted possession of stolen property—a conviction that could have resulted in a 5-year sentence. Seaton also helped Randy secure a job. Id.
. The Nevada Supreme Court reversed the district court’s denial of Browning’s challenge to his sentence of death, Browning, 91 P.3d at 56, but the jury subsequently reinstated the death sentence on remand and the Nevada Supreme Court affirmed in Browning v. State, 124 Nev. 517, 188 P.3d 60 (2008).
. The majority goes to great lengths in an attempt to shore up the Nevada Supreme Court’s determination. Besides imputing to that court a finding that it did not make, the majority argues that it doesn’t matter anyway because federal courts are powerless to review a state court's findings that favor the habeas petitioner. But the federal habeas statutes are only a one-way ratchet with respect to who may seek federal court review. See 28 U.S.C. § 2254(a). While only a prisoner may invoke a federal court’s jurisdiction to challenge his detention, nothing in the habeas statutes prevents a federal court from reviewing the entire record—including facts that the state court construed as favorable to the petitioner—once it has properly asserted jurisdiction. The majority cites no authority for the proposition that a state court’s factual (or legal) errors are impervious to challenge in such a situation.
. Tellingly, the majority arrives at this conclusion only after sua sponte expanding the certificate of appealability to include an alleged Brady violation not even raised by Browning on appeal: Hugo's description of his assailant’s hairstyle. Considering our obligation to *485consider Brady evidence cumulatively to determine its prejudicial effect, it is unclear whether the majority could have found a Brady violation if the certificate was limited to the two claims that Browning actually raised before this court.
. The majority suggests otherwise, asserting that the "precise words Hugo used to describe his assailant’s hair” is material under Brady because Hugo’s description of his assailant as having shoulder-length hair "was markedly different from Browning’s hair on the day of the murder—an Afro.” But the defense did know that Hugo’s description included the term “shoulder-length hair,” and that information was presented to the jury.
. The majority also does not acknowledge the fact that Coe's identification is consistent with Browning's own admission that he was walking south near the store. However, because Browning testified to this fact only in post-conviction proceedings, I do not consider it in the Brady analysis. See Pinholster, 563 U.S. at 181, 131 S.Ct. 1388.
. In post-trial proceedings, the defense submitted an affidavit by Dr. William Chisum in which Dr. Chisum determined that Hugo’s wounds did not “coherently coincide” with those of the recovered knife. But that testimony, at most, created a conflict in the evidence. And, at any rate, we are limited to considering the evidence that was before the trial court. See Pinholster, 563 U.S. at 181, 131 S.Ct. 1388.
. The jury likely also considered the tan jacket found with Browning in the Wolfes' motel room, and which the prosecution identified as the jacket worn by Browning in a photograph shown to the jury. The jacket had Type B blood on it, which is the same blood type as Hugo. Prosecutor Seaton argued in his rebuttal closing argument that this “proves [Browning’s] guilt probably as much as anything.” Post-conviction forensic testing revealed, however, that the blood was not Hugo's. Seaton's statement was therefore unfairly prejudicial. Even so, the jury was not left with the unassailable impression that the blood was, in fact, Hugo’s. Indeed, Seaton conceded that “[t]here are other people in this world with B blood.”
. I agree with the majority that we must review Pike’s performance as a whole, and that the district court erred in granting COAs only on specific aspects of Pike’s alleged defi-dent performance. See 28 U.S.C. § 2253(c)(2); see also Wong v. Belmontes, 558 U.S. 15, 17, 130 S.Ct. 383, 175 L.Ed.2d 328 (2009).
. To be sure, it appears that Pike never connected the shoeprints directly to the enigmatic black Cuban. This renders his post-trial explanation for not investigating the prints somewhat suspect. Even so, our task is not to review Pike’s performance as if on a direct appeal or to second-guess his intent at the time of trial. We may only reject the state court’s determination if Pike's decision cannot be construed by a fairminded jurist as a "sound trial strategy.” See Strickland, 466 U.S. at 689, 104 S.Ct. 2052; see also Richter, 562 U.S. at 110, 131 S.Ct. 770 ("Strickland ,.. calls for an inquiry into the objective reasonableness of counsel's performance, not counsel’s subjective state of mind”).
. The majority speculates that Pike had nothing to lose by interviewing Branon because no matter what Branon said, that "still would have inflicted no harm on [Pike’s] theory” that the black Cuban committed the murder. Not necessarily. In response to a question about whether first responders entered the store before Branon's arrival, Pike could not' have ensured that Branon would answer with a simple “yes” and nothing more, as the majority assumes. Branon may very well have elaborated, saying something like: "Yes, and they’re the reason why there were bloody shoeprints all over the place.” This observation also disposes of Browning’s argument that Pike should have asked Branon about the prints after Horn’s testimony indicated that they may have been left by first responders. Browning suggests that, at that point, at worst Branon could have confirmed Horn’s testimony. But Horn did not testify that the first responders had made the prints. He only observed that first responders wear tennis shoes to crime scenes. Thus, even after Horn’s testimony, Pike could plausibly argue his black-Cuban-did-it theory. Had Pike asked Branon about what he saw when he arrived at the scene, and had Branon told him that the first responders were the source of the prints, Pike would have had significantly less latitude to press this defense. All this is to say that a fair-minded jurist could conclude that Pike was reasonable in seeking to avoid obtaining the shoeprint information by not interviewing Branon.
. The majority also makes the bald allegation that Pike "thought Browning was guilty.” The record does not appear to support this statement, yet it is key to the majority's ominous warning that condoning Pike's strategy would result in blanket cover for attorneys who shirk their investigatory obligations. The majority reasons that defense counsel need merely cite a belief that their clients are untrustworthy to justify conducting little or no investigation. This is a red herring: The issue is not whether Pike believed Browning was guilty; it is whether Pike’s trial strategy made sense. Had Pike’s decision to forgo further investigation of the shoeprints' provenance been untethered to any potential benefit to his client, that decision may very well have constituted inadequate performance. But, as explained, that is not the case.
. The record indicates that Pike was told, prior to trial, that the Wolfes had, on another occasion, falsely accused someone of committing crimes against Vanessa Wolfe.