ORDER
Judges Wardlaw and Gould AMEND their majority opinion in the above captioned case filed September 20, 2017 as follows:
The paragraph on page 55 of the slip opinion that begins with the sentence < Finally, Browning lists in a footnote of his brief a litany of other asserted deficiencies in Pike’s representation. > shall be deleted in its entirety and replaced with the following language:
< Finally, in arguing for an expansion of the COA, Browning lists a number of other alleged deficiencies in Pike’s representation. Because we find that Browning’s ineffective assistance of counsel claim succeeds on other grounds, we do not here assess these other alleged deficiencies^
Existing footnote 19 shall be inserted in its entirety after <Pike’s representations in the above-inserted text.
Judge Callahan objects to any basis for expanding the COA, does not concur in amending the majority opinion, and stands by her dissent.
Judges Wardlaw, Gould, and Callahan vote to deny the Petition for Panel Rehearing.
The Petition for Panel Rehearing is DENIED. No further petitions for panel rehearing or rehearing en banc will be accepted.
IT IS SO ORDERED.
Dissent by Judge Callahan
OPINION
GOULD, Circuit Judge:Nevada state prisoner Paul Browning appeals the district court’s denial of his petition for a writ of habeas corpus under 28 U.S.C. § 2254. In 1986, a Nevada jury found Browning guilty of four crimes involving the robbery and murder of Hugo Elsen in a Las Vegas jewelry store. The jury sentenced Browning to death.
In his habeas corpus petition, Browning challenges his convictions. He asserts that he is entitled to habeas relief on two grounds: prosecutorial misconduct and ineffective assistance of trial counsel (“IAC”). Browning contends that the prosecutor in his case withheld material evidence favorable to the defense and presented false and misleading evidence at trial. He also contends that his trial counsel’s pretrial investigation and preparation were constitutionally inadequate. The Supreme Court of Nevada previously rejected these claims.
Under this procedural posture, a federal court’s role is limited. Our role is *450only “to guard against extreme malfunctions in the state criminal justice systems.” Davis v. Ayala, — U.S. —, 135 S.Ct. 2187, 2202, 192 L.Ed.2d 323 (2015) (internal quotation marks omitted). In Paul Browning’s case, a mixture of disturbing prosecutorial misconduct and woefully inadequate assistance of counsel produced just that. Because the Supreme Court of Nevada unreasonably applied clearly established Supreme Court precedent in denying some of Browning’s claims, we reverse the district court’s denial of habeas relief and remand for further proceedings.
I
We start with the factual background: Between 4:00 p.m. and. 4:30 p.m. on November 8, 1985, Hugo Elsen was stabbed to death during a robbery of the jewelry store he operated with his wife, Josy El-sen. Soon after this brutal murder, police officers arrested Paul Browning as the primary suspect; Browning was staying at the Normandy Motel, located a few blocks from the Elsens’ store. The state charged Browning with (1) burglary, (2) robbery with the use of a deadly weapon, (3) murder with the use of a deadly weapon, and (4) escape.' Because the public defenders’ office was representing a potential witness in Browning’s case, the court appointed former Clark County prosecutor Randall Pike to represent Browning. At the time of his appointment, Pike had been practicing as a defense attorney for less than a year. He represented in a state- habeas proceeding that Browning may have been his first capital defendant.1
Browning pleaded not guilty. The court scheduled trial for March 3, 1986. A week before that date, the prosecution requested a continuance, explaining that it was not prepared to begin trial because someone in its office had written the wrong trial date in the case file. Over the defense’s objection, the court granted the continuance. Because of the delay, Browning sought dismissal of his case from the Supreme Court of Nevada and federal court. He was unsuccessful. In the meantime, Pike lost contact with Browning’s girlfriend, Marsha Gaylord—an essential witness for Browning’s trial defense, according to Pike. Trial commenced on December 9, 1986, with Gaylord still unreachable.
A
The prosecution’s first witness was Josy Elsen, the spouse of the 'victim. Josy testified that in the late afternoon of November 8,1985, she was napping in a back room of the jewelry store when she heard commotion in the showroom. She awoke, entered the showroom, and saw a black man with a blue cap holding a knife and kneeling over Hugo. Hugo and the assailant were in the opposite corner of the room, and a showcase stood between them and Josy. All Josy could see was the side of the assailant’s head and hair that “puffed” out of the back of his cap. Josy at once ran through the back door of the jewelry store, knocked on the window' of an office next door, and asked the occupants to call the police. Debra Cóe, an employee in that -office, then accompanied Josy back into the jewelry store through the back entrance; victim Hugo was lying in the. same corner in- a pool of blood, but the assailant was gone. Later that night, police brought spouse Josy to a station, where she positively identified many pieces of jewelry as coming from her store. At trial, Josy identified a picture of a blue hat with the word *451“Hollywood” written on the front as the one she saw the assailant wearing in the showroom.
Josy testified that in December 1985, a month after Hugo’s murder, police called her back to the station and presented her with a photographic lineup of twelve black men. The officers placed Browning’s picture—taken in November 1985—in the “#5” position. According to an officer’s report, Josy “immediately” explained to the officers that she thought she would not be able to identify the assailant because “she only saw him for a very slight moment from the side.” Nonetheless, Josy examined the photos and stated that the men in photos #1, #6, and #11 had hair “somewhat like” the assailant’s. She did not then indicate any recognition of Browning’s photo. Yet at trial, when Josy was asked to identify the man who had killed Hugo,’She said that, although» she had a limited view of the assailant, she was certain that it was Browning.
The prosecution also called a business neighbor and witness, Debra Coe. Coe testified that when Josy Elsen frantically arrived at Coe’s office, Coe ran to the front window to see if she could see anyone leaving the Elsens’ store. Coe saw a man run by her office from the direction of the jewelry shop, but later that day told the officers that the man had not come out of the Elsens’ store and instead “must have run past it.” She told the officers that it was “hard for [her] to see how he could’ve come out of the door and was running at the angles he was at.” She initially told the officers that the man she saw was white, but in a later interview the same day said he was “definitely black.” In the interview, Coe stated, “when I see a black person, that they all look the same.” At trial, Coe described the man as black, about six feet tall and 27 years old, with a mustache and hair sticking out about an inch beyond a blue cap. She also said he was wearing Levi’s and a dark blue jacket. When asked at trial if she truly believed that all black people “look the same,” Coe said she did not. Coe admitted, however, that she did not “really know any black persons personally.”
Coe testified that later on the evening of November 8, an officer asked her to accompany him around the corner to “identify the man that they had picked up.” Coe obliged,, and a minute or two later, an officer pulled up in a police vehicle. The police first showed her someone whom Coe stated was “definitely not” the man she had seen. The officers then presented Browning, who was shirtless and in handcuffs. Browning had a large Afro-style haircut. Coe indicated to the officers that she “thought” Browning was the person she had seen running by her office, but she “would have been able to identify him better, if he had the hat on and the jacket.” According to Coe, during the showup, Browning’s hair" was “pressed down” as if he had been wearing a hat.
At trial,. Coe testified that she was now “sure” Browning was the man she had seen running by her office, on.November 8, 1985. When Pike asked her how, a year after her equivocal identification on the night of the crime, she was so sure that it was Browning that she had seen, Coe stated that she had “had time to think .about it.” Coe also identified the blue “Hollywood” hat as the one worn by the man who ran by her office.
The prosecution also called Charles Woods, who in 1985 operated a jewelry store three doors down from the Elsens’ store. Woods was standing outside his store with a friend around 4:30 p.m. when he saw a man jogging towards him. The man was not holding anything, and had no blood on him. The man passed Woods within touching distance. Woods told police *452that the man he saw was about six-feet tall, slim, muscular, about 180 pounds, and was wearing dark pants, a light-colored shirt, and a “darker color” hat. When shown a picture of the Hollywood hat at trial, Woods said that it was not the hat he saw the man wearing, which was more of a “beret sort of thing.”
Woods testified that the officers at the scene asked him to join Coe at the nearby corner to “stick around and identify” a suspect. When police presented Woods with a shirtless and handcuffed Browning, Woods identified Browning as 'the man who ran by Woods earlier that day.
The prosecution then called Randy Wolfe. Randy and his wife, Vanessa Wolfe, lived in the same motel where Browning was staying. Randy testified that between 4:00 p.m. and 4:30 p.m. on November 8, 1985, he was working on his landlady’s car when Browning yelled his name from the motel’s upper level. Randy went upstairs and found Browning in the Wolfes’ apartment sitting on the bed and wearing a tan windbreaker and the Hollywood hat. Jewelry had been dumped on the bed. According to Randy Wolfe, Browning told Randy that he had just robbed a jewelry store and thought he had killed someone. Browning also told Randy that he planned to use the jewelry to get Gaylord out of jail. Randy told Browning that he wanted nothing to do with the murder, and was going to go finish working on his landlady’s car and then get some heroin. Browning asked if Randy would get some heroin for him too. On his way down the stairs, Randy encountered Vanessa, whom he instructed to keep Browning “cool” while Randy found the police. Randy then located the crime scene and led several officers back to his apartment. Randy testified that when the officers got there, they promptly arrested Browning, and that after they removed Browning from the apartment, Randy found a cup under his sink filled with additional jewelry.
Randy Wolfe made several admissions during his testimony that bore on his credibility. He admitted that he and Vanessa were addicted to heroin and cocaine, and that he would break into cars to support their habits. He had prior convictions for selling a controlled substance and prison escape. He also admitted that he had kept some of the jewelry he claimed he found under his sink, and lied during Browning’s preliminary hearing by stating that he did not keep any of it.
Before the time of Browning’s trial, Randy had been charged with possession of stolen property, a charge unrelated to the Elsens’ stolen jewelry. Prior to Randy’s testimony at trial in Browning’s case, the state permitted Randy to plead to a lesser charge of attempted possession of stolen property, for which he faced one to five years in prison. Despite Randy Wolfe’s failure to appear in court almost thirty times in the past, he was released on his own recognizance before Browning’s trial. Randy Wolfe testified, however, that he had not received “anything” for his testimony against Browning, and that no one had promised that his sentence might be “diminished” if he testified. The prosecutor in Browning’s case, Daniel Seaton, was not the prosecutor in Randy’s case.
The prosecution next called Vanessa Wolfe. Around 1977, Vanessa and Gaylord, Browning’s girlfriend at the time of the murder, had lived in southern California, where they ran con games and “bilk[ed] people out of their money.” At the time of Browning’s trial, Vanessa worked as a prostitute in Las Vegas. On the afternoon of November 8, 1985, Vanessa was bringing a client to the Wolfes’ apartment when she encountered Randy on the stairs. Randy told Vanessa about what Browning had done, and that Browning was going to take *453Vanessa hostage if Randy called the police. Vanessa then entered the apartment. Browning was inside, shaking water off of a knife. The Hollywood hat and tan windbreaker were on the floor. Browning instructed Vanessa to throw the knife and hat on the roof, but instead she put the knife in a pizza box under the stairs and threw the hat in a dumpster. Police arrived soon after.
Officer David Radcliffe also testified for the prosecution. He said that he arrived at the Elsens’ store to find Hugo Elsen conscious, but in an “extremely serious” condition. Hugo told Radcliffe that a black man wearing a blue baseball cap had stabbed him. Radcliffe then joined some officers standing outside of the storefront, and soon after, Randy Wolfe approached him. Radcliffe had known Randy for several years because Randy was a regular narcotics user. Randy led Radcliffe and several other officers to Randy’s apartment, which the officers forcibly entered. Inside, they found Browning sitting on the corner of the bed. Pieces of jewelry were scattered along the floor on the opposite side of the room from the bed.
The prosecution also called several forensics specialists to testify about evidence at the crime scene. Identification specialist David Horn testified that three of the showcase counters in the store had been “disturbed,” and that the merchant-side sliding glass of one of the showcases had been broken. Horn lifted “approximately twenty some odd” fingerprints from the scene. Two were most relevant: one from the top glass of one of the counters, and another from a fragment of the counter’s broken sliding-glass door. A different fingerprint examiner concluded that these two prints matched Browning’s.
Horn also testified that he observed bloody tennis shoe-style shoeprints leading away from the comer where Hugo was lying and towards the store’s front door. After leaving the scene, Horn compared the shoeprints to the loafers Browning was wearing when he was arrested; they did not match. Horn testified that paramedics and off-duty officers often wear tennis shoes at crime scenes, so he did not think any further investigation into the source of the shoeprints was necessary.
State criminalist Minoru Aoki testified that Hugo had Type B blood. Blood had been found on the tan jacket that was lying on the floor in the Wolfes’ apartment, and it too was type B. Aoki did not test Randy Wolfe or Vanessa Wolfe’s blood type.
Pathologist Giles Green testified that the knife recovered from the pizza box under the stairs at the Normandy Motel was “consistent” with the wound configurations in Hugo’s body, but “nothing about that knife t[old him] that the knife made those wounds.”
Browning’s trial counsel, Randall Pike, called three witnesses. First was Bradley Hoffman, who operated a store two doors down from the Elsens’ store. Hoffman testified that around 4:00 p.m. on the day of the crime, he saw a man walking down the street towards the Elsens’ store. The man was Cuban, “probably five seven, slight build,” and wearing Levi’s jeans, the shirt that Hoffman vaguely recalled as plaid, and a blue baseball cap. Later that night, officers brought Hoffman to the showup with Coe and Woods, and presented him with a shirtless Browning. Hoffman stated that Browning’s hair, a “medium sized Afro,” did not appear as though Browning had recently been wearing a hat. He also stated at trial that the Hollywood hat was not the one worn by the man he saw walking towards the Elsens’ store.
Pike also called Officer Gregory Branon, who testified that he was “one of the first *454two officers” to arrive at the scene, Branon receivéd a description of the suspect: a “black male, adult in his late twenties, wearing a blue baseball cap, blue windbreaker-type jacket, "blue Levi’s[,] ... medium complexioned, bore a mustache and what was described as a shoulder length J[h]eri-type curl.” Pike did not ask Branon who gave him that description.
Last, Pike called Annie Yates—a hair stylist—who testified to the difference between a Jheri Curl (the assailant’s hairstyle as described to Officer Branon), and an Afro (Browning’s hairstyle on November 8,1985). Yates stated that a Jheri Curl requires the use of chemicals, whereas an Afro does not. Pike presented Yates with the twelve-person photographic array pre-. viously shown to Josy Elsen, which had Browning at position #5. Yates stated that pictures #1, #2, #4, and #10 had Jheri Curls.
In his closing, Seaton laid out his theory: Browning robbed the jewelry store tjo bail Gaylord out of jail because he relied on her prostitution income to feed his heroin addition. Seaton’s closing argument was incendiary, but Pike rarely objected. Seaton began by characterizing the presumption of innocence as follows:
Now we are talking about that wonderful constitutional element called the presumption of innocence, we are now talking about piercing that veil, dropping that facade because, in fact, as a person sits in a courtroom he may not be innocent. He may be guilty.
[Browning] has the presumption of innocence. And, of course, it is one when his guilt is shown that the farce of that presumption is known and it’s been done in this case.
Seaton gave the following description of Browning’s murder of Hugo Elsen:
[Browning, t]his man whose girlfriend prostituted for him so he could get drugs, money to get drugs, this man who took heroin, he wanted Randy Wolfe to get him to cop some heroin for him after the murder. He.shot the life of Hugo Elsen right up his arm.. That’s what he was doing that day. That’s what we have here,
Seaton also described Josy Elsen’s identification of Browning at trial as “as good as you can ask for.” Anticipating -that Pike would argue that Browning’s hair on the night of the murder (an-Afro) did not match the assailant’s hairstyle as de-1 scribed to Officer Branon (a Jheri- Curl), Seaton explained that Officer Branon received the description from “some white person” who did not understand “the true definition” of a Jheri Curl. Seaton concluded by saying that it was the jury’s “duty to go out, decide that and come back in here and tell [Browning] just exactly that, that he’s the one that has to pay for these crimes.”
Pike’s closing argument set forth a theory that the Wolfes’ friend, a Cuban man, committed the robbery-murder, and the Wolfes were now framing Browning.
The jury found Browning guilty on all four counts2 and sentenced him to death. Browning directly appealed to the Supreme Court of Nevada, which affirmed. Browning v. State, 104 Nev. 269, 757 P.2d 351 (1988).
*455B
Browning filed a petition for a writ of habeas corpús in Nevada state district court, arguing in relevant part that Seaton had withheld exculpatory evidence from the defense and that Pike was ineffective by failing to perforin an adequate investigation before trial. The petition included three new pieces of evidence. First,' the state stipulated that post-conviction DNA analysis had proven that the blood on the tan windbreaker found in the Wolfes’ apartment did not belong to Hugo Elsen. Second, a forensics report indicated that Hugo Elsen’s wounds did not “coherently coincide” with the knife found in the pizza box under the stairs at the Normandy Motel. Third, a forensics report suggested that the bloody shoeprints were too large to belong to Josy Elsen or Debra Coe.
The Nevada district court held an evi-dentiary hearing, at which attorneys Jason Isaacs and Daniel Lamb represented Browning. Robert Shomer, a forensic psychologist, testified that Josy Elsen’s identification of Browning as the assailant was questionable because (1) her in-court identification of Browning was 14 months after the incident; (2) the stress of the moment might have made her memory more vivid, but no more accurate; (3) the in-court identification was extremely suggestive because Browning was the sole available “choice”; (4) cross-racial identifications are unreliable (Josy is white, Browning is black); and (5) Josy’s observation of Browning’s picture during the photo array in December 1985 may have implanted Browning’s' face in Josy’s memory and prompted a false in-court identification. Shomer also criticized Coe’s identification, explaining that (1) the 14-month gap between Coe’s observations and the trial likely distorted her memory; (2) Coe initially reported to the police that the man she saw did not appear to come from the El-sens’ store, and Shomer suggested that Coe probably did not focus intently on his characteristics; (3) Coe had reported to the police that the man she saw was white, but one of the officers’ statements to Coe that the suspect was black might have impacted Coe’s memory; and (4) given the highly suggestive procedures of the in-person showup on the night of November-8,1985, Coe’s identification was, at best, equivocal. Finally, Shomer criticized Woods’s identification because (1) Woods stated that he saw nothing notable about- the man who ran towards him on the afternoon of November 8,1985, suggesting that Woods did not pay close attention to the man’s appearance; and (2) the showup was particularly suggestive in light of the officers’ telling Woods that they had a suspect whom they wanted Woods to identify.
Browning’s counsel then called Michael Sweedo, a fingerprint examiner and crime scene analyst, to give his opinion on the officers’ forensic investigation. Sweedo testified that Browning’s fingerprints' on the showcase glass could have 'been the result of Browning leaning over the case. Sweedo noted that it was unusual that there were no other identifications on the remainiiig twenty-some latent prints lifted from the crime scene. Sweedo also said that it was abnormal that the officers did not investigate the source of the bloody shoeprints.
Browning’s attorneys then called Pike, Browning’s trial counsel. Pike told the court that Marsha Gaylord would have testified to two crucial pieces of evidence that never came out at trial: (1) Gaylord and Browning had been in the Elsens’ store prior to November 8, 1985, which could have explained the presence of Browning’s fingerprints in. the .store; and (2) the Wolfes had a friend that was of Cuban descent. Pike could not call Gaylord as a witness, however, because she “disappeared” after the initial trial continuance. *456Pike asserted that he tried to locate Gay-lord by using Martin Schopp, his investigator. Other than Browning himself, Gaylord was the only person who knew that Browning had been in the Elsens’ store before November 8,1985.
Pike then described his investigation of Browning’s case. He explained that although he visited the Elsens’ store after it was cleaned and reopened to the public, he never went when it was a crime scene. Pike explained that, to avoid becoming a witness himself, he had Schopp conduct all witness interviews. Pike did not have Schopp interview the Wolfes before trial, despite Pike’s knowing that the Wolfes were “long time informants.” Pike suggested that any inquiry into whether the Wolfes were receiving a benefit for their testimonies would have been futile because, in 1986, plea bargaining was “informal,” and “basically, there were a lot of things that were done just with passive agreement.” At some point prior to trial, Pike, was told that the Wolfes had falsely accused a man named Jerold Morell of assaulting Vanessa with a knife, but Pike could not recall making any attempt to locate Morell.
Pike did not retain a fingerprint expert because, as a former prosecutor, he “knew” all of the state’s forensics witnesses and relied on informal conversations with them. Pike said that he could trust the state’s main fingerprints expert to be “straight” with him.
Pike did not conduct any investigation into the source of the bloody shoeprints, and never authorized any interviews to determine when the responding officers observed the shoeprints. He explained that if he investigated the shoeprints’ source and determined that they belonged to one of the paramedics, he would not be able to argue that the shoeprints exculpated Browning as the murderer. Pike characterized his overall trial strategy as “overcasting a shadow of doubt, as opposed to proving” Browning’s innocence.
Before trial, Pike was told that a man named Thomas Stamps had information suggesting that Randy Wolfe and another man, Mike Hines, were attempting to sell some of the jewelry stolen from the El-sens’ store. Pike could not recall why he did not have Schopp interview Stamps. Pike also could not recall why he did not instruct Schopp to interview Martha Hay-gard (the Wolfes’ landlady), who had seen the Wolfes with a Cuban individual. Nor could Pike recall why he had not followed up on Coe’s initial statement to the police that the man whom she saw running by her office was white, not black.
Investigator Martin Schopp also testified at the state habeas hearing. Schopp performed “substantially all the investigative work” for Browning’s defense. Schopp, however, did not have autonomy—Pike directed all of his inquiries. While the court appointed Schopp soon after Browning’s arrest, Schopp was not contacted to perform any work until five months later, when Browning himself reached out. Schopp testified that such a significant delay was unusual and likely allowed evidence to get cold. Pike did not give Schopp a discovery file until August 1986—ten months after Browning’s arrest—and the file included only police reports and a voluntary statement by Randy Wolfe. Pike gave Schopp no other information to create a foundation for his investigation. Schopp performed a total of 12 hours of investigative services for Pike—a number Schopp thought was low under the circumstances. Pike limited Schopp’s investigation by denying Schopp’s requests for additional investigation funds. Schopp and Pike spoke no more than five times, and each time only briefly.
*457Schopp explained that after preparing a preliminary report—which included a statement from Haygard that she saw the Wolfes wearing “big gold wedding bands” after the robbery—he felt that there were various other leads to follow. He requested that Pike permit him to interview the Wolfes and Thomas Stamps. He also wanted to interview Jerold Morell, who had told Pike that the Wolfes falsely accused him of sexually assaulting Vanessa with a knife. (A jury acquitted Morell of those accusations.) Pike denied each of these investigation requests. Pike also never asked Schopp to interview any police officers or detectives. In Schopp’s opinion, the investigation into Browning’s case was never “completed.”
Browning’s state habeas counsel also called prosecutor Daniel Seaton. Seaton testified that after Browning was convicted of the robbery-murder, Seaton gave Randy Wolfe two concrete benefits. First, he helped Randy get a drywalling job. Second, and more importantly, before Randy was sentenced on his conviction for attempted possession of stolen property, Seaton spoke to the sentencing judge on Randy’s behalf. At Randy’s sentencing hearing, the judge explained that Seaton had told him that Randy was “a witness in a recent trial,” and in light of Randy’s being “somewhat helpful” to the prosecution on “several occasions,” Seaton felt Randy “deserve[d] something positive for doing that.” The judge later noted that Seaton told him that Randy “more than fulfilled his obligation” in Browning’s trial “and as a matter of fact put himself in some jeopardy and deserves something for it.” In light of Seaton’s statements to the sentencing judge, the prosecution in Randy’s case withdrew its recommendation of five years imprisonment. The judge imposed only probation.
At the state habeas hearing, Seaton testified that he never promised Randy any benefits in exchange for his testimony against Browning, and decided to speak to Randy’s sentencing judge only after Browning was convicted. Seaton admitted, however, that he had engaged in off-the-record plea bargaining in the past in at least one other case. Seaton also admitted that after he learned that the Wolfes kept some of the Elsens’ stolen jewelry, Seaton did not impound the jewelry or instruct anyone else to do so. Nor did the state prosecute the Wolfes with any crime relating to the jewelry they kept.
Browning’s state habeas counsel also called Officer Branon. Branon testified that he and another officer were the first to arrive at the crime scene—before paramedics or other officers entered the Elsens’ store. Upon arrival, Branon immediately noticed bloody shoeprints on the floor. Branon encountered Hugo Elsen lying in the corner of the store. Hugo was scared, but lucid. In Branon’s original report, he wrote that the assailant had a Jheri Curl-type hairstyle. As noted above, Branon never explained at Browning’s trial who gave him that description, and the speaker was not revealed. During the state habeas hearing, however, Branon explained that it was victim Hugo Elsen himself who had given the description. Branon also testified that Hugo did not use the term “Jheri Curl” when describing the assailant; rather, Hugo described the assailant’s hair as shoulder length, loosely curled, and wet. It was Branon—who is black—who first used the term Jheri Curl to describe the assailant’s hair. When shown pictures and a video of Browning from the night of November 8, 1985, Bra-non stated that Browning’s hair was a “four inch Afro with braids on top of it,” and could not be described as a Jheri Curl or shoulder length, loosely curled, and wet. The revealed testimony about the victim’s *458description of the murderer's hair raised a critical identification issue.
Finally, Browning himself testified at his state habeas hearing. He told the court that on November 8, 1985, around 4:00 p.m., he was walking down the street when he saw Randy Wolfe driving a yellow Dát-sun. Browning asked Wolfe for a ride downtown, and as he approached, a Cuban man—whom Browning knew as Randy’s friend—pushed Browning out of the way and entered the car. The Cuban man was wearing both the Hollywood hat and tan jacket found in the Wolfes’ apartment when Browning was arrested. Randy told Browning to meet him back at the Wolfes’ motel room, and then drove off.
C
The state ‘ district court denied Browning’s habeas petition on December 7, 2001, and filed its findings of fact and conclusions of law on October 24, 2002. Browning appealed to the Supreme Court of Nevada, which on June 10, 2004 affirmed the denial of Browning’s challenge to his convictions, but reversed the district court’s denial of Browning’s challenge to his sentence. Browning v. State, 120 Nev. 347, 91 P.3d 39 (2004). On remand, a jury again sentenced Browning to death. Browning appealed to the Supreme Court of Nevada, which affirmed. Browning v. State, 124 Nev. 517, 188 P.3d 60 (2008). The United States Supreme .Court denied a subsequent petition for a writ of certiorari, Browning v. Nevada, 556 U.S. 1134, 129 S.Ct. 1625, 173 L.Ed.2d 1006 (2009).
While he was being resentenced, 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, the operative version before our court now. After Browning abandoned several unexhausted claims, the district court denied Browning’s petition in full on August 1, 2014. The district court granted Certificates of Appealability (“COA”) on the following issues: (1) whether the prosecution's failure to produce evidence relating to the bloody shoeprints constituted a violation of Browning’s rights as described in Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), and/or Napue v. Illinois, 360 U.S. 264, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959); (2) whether evidence impeaching Randy Wolfe’s credibility was withheld in violation of Browning’s rights under Brady,- and (3) whether Pike was ineffective in light of his failure to investigate the source of the bloody shoeprints, Hugo Elsen’s description of the assailant, and the credibility of Browning’s accusers. Browning timely appealed.3
II
We have jurisdiction under 28 U.S.C. §§ 1291 and 2253(a), and “review de novo the district court’s dismissal of a habeas petition.” Runningeagle v, Ryan, 825 F.3d 970, 978 (9th Cir. 2016). Under the Antiterrorism and Effective Death Penalty Act (“AEDPA”), if a state court adjudicates a petitioner’s federal law claim on the merits, a federal court may grant *459habeas relief only if the state court’s adjudication “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1).4
“[A]n unreasonable application of federal law is different from an incorrect application of federal. law.” Williams v. Taylor, 529 U.S. 362, 410, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). “A state court’s decision can involve an unreasonable application of Federal law if it either [(1)] correctly identifies the governing rule but then applies it to a new set of facts in a way that is objectively unreasonable, or [ (2) ] extends or fails to extend a clearly established legal principle to a new context in a way that is objectively unreasonable.” Hernandez v. Small, 282 F.3d 1132, 1142 (9th Cir. 2002) (internal quotation marks omitted).
Browning asks that wé review' some of his claims de novo rather than with deference to the Supreme Court of Nevada. He contends that the Supreme Court of Nevada’s rulings were not on the merits, and that its reasoning was based on standards contrary to federal law. See 28 U.S.C. § 2254(d). Because we hold that Browning is entitled to relief based on an unreasonable application of United States Supreme Court precedent, we need not, and do not, address whether the Supreme Court of Nevada’s decisions were on the merits or contrary to federal law.
Ill
Under Brady, prosecutors are responsible for disclosing “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). The failure to turn over such evidence violates due process. Wearry v. Cain, — U.S. -, 136 S.Ct. 1002, 1006, 194 L.Ed.2d 78 (2016) (per curiam). The prosecutor’s duty to disclose material evidence favorable to the defense “is applicable even though there has been no request by the accused, and ... encompasses impeachment evidence as well as exculpatory evidence.” Strickler v. Greene, 527 U.S. 263, 280, 119 S.Ct. 1936, 144 L.Ed.2d 286 (1999) (citation omitted).
Under Napue, convictions obtained through the use of false testimony also violate due process. 360 U.S. at 269, 79 S.Ct. 1173. A violation occurs whether the prosecutor solicits false statements or merely allows false testimony to go uncorrected. Id. The constitutional prohibition applies even when the testimony is relevant only to a witness’s credibility, id., and where the testimony- misrepresents the truth, see Miller v. Pate, 386 U.S. 1, 6, 87 S.Ct. 785, 17 L.Ed.2d 690 (1967) (prosecu*460tor “deliberately misrepresented the truth” by presenting testimony that shorts with large reddish-brown stains tested positive for blood, while leaving out that the stains were made by paint).
For claims under Brady, the prosecutor’s personal knowledge does not define the limits of constitutional liability. Brady imposes a duty on prosecutors to learn of material exculpatory and impeachment evidence in the possession of state agents, such as police officers. See Youngblood v. West Virginia, 547 U.S. 867, 869-70, 126 S.Ct. 2188, 165 L.Ed.2d 269 (2006) (“Brady suppression occurs when the government fails to turn over even evidence that is ‘known only to police investigators and not to the prosecutor.’” (quoting Kyles v. Whitley, 514 U.S. 419, 438, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995))).
In the Ninth Circuit, the same is true for claims under Napue. First, in Giglio v. United States, the Supreme Court held that it would impute to an entire prosecution office one prosecutor’s knowledge that a government witness’s testimony was false, even though the prosecutor with knowledge of the false testimony was not the trial attorney on the case. 405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972). Then, in Jackson v. Brown, we applied the same principle to police officers with knowledge that trial testimony offered by the government was false, holding that “Napue and Giglio make perfectly clear that the constitutional prohibition on the ‘knowing1 use of perjured testimony applies when any of the State’s representatives would know the testimony was false.” 513 F.3d 1057,1075 (9th Cir. 2008).
However, the dispositive question on the Napue claim here is what “clearly established Federal law, as determined by the Supreme Court of the United States,” says on the issue. See 28 U.S.C. § 2254(d)(1). We recently answered that question. Despite our holding in Jackson, we held in Reis-Campos v. Biter that “it is not clearly established that a police officer’s knowledge of false testimony may be attributed to the prosecution under Napue.” 832 F.3d 968, 977 (9th Cir. 2016).
As the habeas petition in Jackson was filed before AEDPA’s effective date, Jackson did not directly address whether there was clearly established Supreme Court precedent as required by 28 U.S.C. § 2254(d)(1). As such, Reis-Campos—a case decided under the AEDPA standard—is controlling on that question. See 832 F.3d at 973.
The district court granted a COA as to whether the prosecution violated (1) Brady or Napue with respect to Officer Branon’s undisclosed observation of the bloody shoeprints, and (2) Brady with respect to evidence of an undisclosed benefit for Randy Wolfe’s testimony. We expand the COA to include a third claim: whether the prosecutor violated Brady by not disclosing the actual description that Hugo Elsen gave to Officer Branon of the assailant’s hair.5 We first address whether *461each piece of evidence was exculpatory, triggering a potential duty to disclose under Brady, and for the shoeprint evidence, whether it involved the prosecution’s knowing presentation of misleading testimony in violation of Napue. We then turn to materiality.
A
The bloody shoeprints. At trial, Browning argued that the bloody shoe-prints—which did not match the shoes Browning was wearing when he was arrested—demonstrated that someone else committed the murder. The prosecution responded with Officer Horn’s testimony that responding paramedics and off-duty detectives often wear tennis shoes at crime scenes, misleadingly suggesting that the shoeprints came from them. But during the state habeas hearing, Branon testified that he and Officer Robertson were the first responders at the store, before the paramedics or other officers, and that the shoeprints were there when he arrived. Branon’s observation of the shoeprints was directly contrary to Horn’s suggestion that paramedics or other officers left the prints. Had Branon’s observation been disclosed, Browning could have used that evidence to bolster his contention that the shoeprints were left by the real killer. This makes Branon’s observation exculpatory under Brady. See Kyles, 514 U.S. at 441, 115 S.Ct. 1555 (undisclosed witness observation did not match defendant, and so was exculpatory). And, under Brady, Branon’s knowledge of the shoeprints is imputed to the government as a whole. See Youngblood, 547 U.S. at 869-70, 126 S.Ct. 2188.
Browning contends that the prosecution’s handling of the shoeprint evidence similarly implicates Napue. He asserts that Branon’s observation, which was written in Branon’s original report, made Horn’s testimony that paramedics or off-duty detectives often wear tennis shoes misleading, because it suggested a source of the shoeprints that could not have been true. See Miller, 386 U.S. at 6-7, 87 S.Ct. 785. But there is no evidence suggesting that the prosecution knew that Horn misrepresented the truth. And, as we held in Reis-Campos, it is not clearly established under Supreme Court precedent (and was not clearly established under Supreme Court*precedent on June 10, 2004, the date of the Supreme Court of Nevada’s decision rejecting Browning’s Napue claim) that the prosecution had a duty to learn from Branon about his observation. See 832 F.3d at 977. Browning contends that the evidence suggests Horn knew that his testimony was misleading. But this theory runs into the same obstacle: it is not, and was not on June 10, 2004, clearly established that Horn’s knowledge would be imputed to the prosecution. The record before the Supreme Court of Nevada does not suggest that the prosecution knew that Horn’s testimony was false or misleading. As a result, Browning has not shown that the Supreme Court of Nevada unreasonably applied clearly established Supreme Court precedent in denying his Napue claim.
Benefit for Randy Wolfe’s Testimony. When Pike learned that Randy had been allowed to plead guilty in an unrelated case to a lesser charge of attempted possession of stolen property, Pike moved for a continuance in Browning’s case to investigate whether Randy and Seaton had made a deal. Seaton responded in court: “I can tell the court categorically ... there *462has never been any plea bargaining with Randy Wolfe regarding this case.” At Browning’s trial, Randy similarly testified that he had not been promised anything for his testimony, including any promise of a more lenient sentence on his recent conviction. But after Browning’s trial, Seaton spoke with Randy’s sentencing judge on Randy’s behalf. This led Randy’s prosecutor to withdraw his recommendation <jf five years, and the. judge to sentence Randy to only probation. The Supreme Court of Nevada held that this constituted withholding of impeachment evidence favorable to Browning at his trial,6 Browning, 91 P.3d at 54-55, and the state does not dispute that conclusion.
While the Supreme Court of Nevada explicitly concluded that Seaton improperly withheld evidence in this context, it never specified precisely what evidence the prosecution should have disclosed. It stated:
[T]he prosecutor withheld information regarding benefits given to an important witness for the State, Randy Wolfe. ... [A]t th[e] time [of trial], Wolfe was the defendant in a separate criminal prosecution, and the prosecutor' admitted at the post-conviction evidentiary hearing that after Browning’s trial he told the district judge assigned to Wolfe’s case that Wolfe had helped in prosecuting Browning.... Though the prosecutor maintained that he acted unilaterally and never made any deal with Wolfe, this information still should have been disclosed to the defense. Under 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 an agreement or an understanding. .
Id. (citing Jimenez v. State, 112 Nev. 610, 918 P.2d 687, 694-95 (1996)). The-only way this information could be “evidence implying an agreement or an understanding” would be if Randy knew that Seaton was contemplating speaking to Randy’s sentencing judge. If Randy did not know, then Seaton’s intentions would have, had no impact on Randy’s motivations' to tell the truth, or not, at trial. We therefore read the Supreme Court of Nevada’s decision as concluding that Randy knew that Seaton might help reduce his sentence if he testified against Browning.7- It is that piece .of evidence—Randy’s expectation of a poten*463tial benefit in exchange for his testimony— that constituted impeachment evidence that should have been disclosed to Pike. See, e.g., Arizona v. Fulminante, 499 U.S. 279, 300, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991) (recognizing that benefits conferred by authorities may motivate a witness to lie).
Hugo Elsen’s Description of the Killer’s Hair. Browning’s hairstyle at the time of the robbery was an Afro. At trial, Officer Branon testified that he received a description of the suspect at the scene as sporting a “shoulder length J[h]eri-type curl.” At closing, the prosecution argued that whoever gave this description to Bra-non did not know the difference between a Jheri Curl and an Afro. But during the state habeas hearing, Branon testified that the description he was given did not actually include the words “Jheri Curl.” Rather, Hugo told him that the assailant’s hair was “shoulder length,” “loosely curled,” and “wet.” Branon, who is African American, then interpreted those words to mean a Jheri Curl, and used that term in his original report.
Neither “Jheri Curl” nor “shoulder length,” “loosely curled,” and “wet” are descriptions of an Afro. But only “Jheri Curl” is 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. Had the prosecution disclosed before trial that victim Hugo Elsen’s description of his assailant’s hair was not a “shoulder length J[h]eri-type curl,” but “shoulder length,” “loosely curled,” and “wet,” Browning could have easily refuted the prosecution’s argument. This makes the exact words Hugo used to describe his assailant evidence favorable to the defense under Brady.8
We hold that Officer Branon’s shoeprint observation, Randy’s understanding that Seaton was considering speaking with Randy’s sentencing judge in exchange for Randy’s testimony against Browning, and the precise hair description Branon received from Hugo Elsén were all favorable to Browning under Brady. We also hold that Browning’s Napue claim fails because it was not clearly established at the time of the Supreme Court of Nevada’s decision that a police officer’s knowledge of false or misleading testimony would be imputed to the prosecution.
For the Brady evidence, except for Randy’s expectation of a benefit for his testimony, the Supreme Court of Nevada did not explicitly address whether this evidence was favorable- to Browning. But in light of our above analysis, we hold that had the Supreme Court of Nevada not viewed the evidence as favorable to the *464defense, it would have been an unreasonable application of Supreme Court precedent.
B
We turn now to materiality as an element of the Brady claims. Under Brady, evidence is material “if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different. A • ‘reasonable probability is a probability sufficient to undermine confidence in the outcome.” Bagley, 473 U.S. at 682, 105 S.Ct. 3375. When there are multiple Brady claims, the Supreme Court instructs that we consider materiality “collectively.” Kyles, 514 U.S. at 436, 115 S.Ct. 1555. We must imagine that every piece of suppressed evidence had been disclosed, and then ask whether, assuming those disclosures, there is a reasonable probability that the jury would have reached a different result. See, e.g., Turner v. United States, — U.S. —, 137 S.Ct. 1885, 1893, 198 L.Ed.2d 443 (2017); Cone v. Bell, 556 U.S. 449, 473-74, 129 S.Ct. 1769, 173 L.Ed.2d 701 (2009).
Applying this procedure to the facts before us, and incorporating AEDPA deference, we address the following question: Imagine that the prosecution had disclosed (1) that Officer Branon observed that the shoeprints existed before paramedics or other officers arrived; (2) that Randy expected a benefit for his testimony; and (3) that Hugo Elsen described the assailant as having shoulder length, loosely curled, and wet hair, rather than a Jheri Curl. Was it objectively unreasonable for the Supreme Court of Nevada to conclude that there was not a reasonable probability that the jury would have reached a different result?
We conclude that the answer is yes. Officer Branon’s undisclosed shoeprint observation disproves the prosecution’s primary rebuttal against Browning’s strongest piece of evidence that someone else killed Hugo. The undisclosed evidence of a benefit for Randy’s testimony adds a powerful reason to disbelieve him and his wife, the prosecution’s most critical witnesses. And the undisclosed evidence of Hugo’s exact dying words defeats the prosecution’s central argument against its proba-tiveness.
Also, the prosecution’s trial evidence was remarkably weak. Its case relied on flawed identifications and the Wolfes’ unreliable testimony. And the physical evidence was just as consistent with Browning having been framed as with him being the killer.
We conclude that it was an objectively unreasonable application of Supreme Court precedent to hold that the Brady materiality standard was not met here. Below, we discuss materiality in more detail, analyzing the relevant evidence at trial piece-by-piece, with an aim to showing the probable ultimate effect on the jury’s decision.
The Bloody Shoeprints. Browning’s trial theory was that someone else killed Hugo Elsen. The shoeprints leading from Hugo to the front door lent strong support to this theory. But Officer Horn’s testimony suggesting that paramedics or other officers could have left the shoeprints gave the jury a reason to disregard strong evidence raising questions of reasonable doubt. Had the prosecution disclosed Bra-non’s observation about the shoeprints, the source of several bloody shoeprints would remain a mystery. This means the jury would have been left with powerful evidence that Browning was not the killer.
In its briefing, the state responds that Officer Branon’s observation was not so helpful for Browning’s defense because the *465shoeprints could have been made by Josy Elsen or Debra Coe. But that is pure speculation. The prosecution had the opportunity to offer at trial evidence that Josy or Coe made the shoeprints, but either chose not to do so or did not have such evidence. We cannot now assume such evidence exists. See Miller-El v. Cockrell, 537 U.S. 322, 345, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003) (“Had there been evidence obtainable to contradict and disprove the testimony offered by petitioner, it cannot be assumed that the State would have refrained from introducing it.” (internal quotation marks omitted)). In any event, as the district court noted, it is unlikely that either Josy or Coe was the source of the prints because there is no evidence that either of them went anywhere near the store’s front door after Hugo was stabbed. There are also photos in the trial record of the shoeprints next to a ruler. Examining the photos as if we were the jury viewing them as exhibits at trial, the shoeprints appear to us larger than those of a typical woman’s shoe.9
At oral argument, counsel for the state proffered a different theory. He asserted that there were in fact two sets of prints: one set around Hugo Elseris body, and another leading from the body to the front door. See Oral Arg., Browning v. Baker, No. 15-99002 (Mar. 16, 2017) at 23:00-23:47, https://www.youtube.com/watch?v= 8va4fhOsWZ8. Counsel argued that the shoeprints Officer Branon referred to in his testimony were only those immediately surrounding Hugo’s body, which Josy or Coe may have created while giving aid to Hugo. Id. at 23:50-24:20. According to his theory, the other prints—which led from the body to the front door—were left later by paramedics. Id. at 24:22-24:55. But the state did not raise this argument in its briefing before this court (or apparently in any court); it has thus long been forfeited. See Harger v. Dep’t of Labor, 569 F.3d 898, 904 n.9 (9th Cir. 2009); Fed. R. App. P. 28(b).10
A final possibility, one the parties do not discuss but that the jury might have considered, was that Browning made the shoeprints, but then changed his shoes before being arrested. But like the theory that Josy or Coe left the prints, there is no evidence to support this possibility. The prosecution never found any such shoes in the Wolfes’ or Browning’s apartments, nor any other shoe that matched the prints.
The bloody shoeprints were strong evidence that Browning was not the killer. Had the prosecution disclosed Branon’s observation, that strong evidence would have gone unrebutted.
*466Benefit for Randy Wolfe’s Testimony, Randy and Vanessa Wolfe were the prosecution’s most important witnesses. They were the original accusers, the source of the alleged murder knife, and the source of Browning’s alleged confession. It is fair to say that had the jury not credited the Wolfes’ testimony, Browning would not have been convicted. . .
The jury had plenty of reasons not to believe the Wolfes. The Wolfes admitted that they used heroin and cocaine,11 that Vanessa was a prostitute, and that Randy stole property. Randy had prior convictions, including a recent conviction for which he would soon be sentenced. Randy admitted to keeping some of the stolen jewelry and lying about it, at Browning’s preliminary hearing. Vanessa testified that she used to “bilk people out of them money.”
Given this mountain of evidence providing potential reasons to doubt the Wolfes’ credibility, getting just one juror to change his or her mind about the truth of the Wolfes’ testimony likely would not have required much. See United States v. Agurs, 427 U.S. 97, 113, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976) (“[I]f the verdict is already of questionable validity, additional evidence of, relatively minor importance might be sufficient to create a reasonable doubt.”).
Evidence suggesting that Randy was expecting leniency in'his sentencing in exchange for his testimony against Browning could have accomplished this task. And, it could have'done so without being merely cumulative of the impeachment evidence already in the record. Evidence that Randy was expecting a benefit for his testimony would have revealed that the Wolfes had a “direct, personal stake in [Brówn-ingj’s conviction.” Bagley, 473 U.S. at 683, 105 S.Ct. 3375. The other impeachment evidence concerning the Wolfes’ criminal activity and penchant for lying suggested to the jury that the Wolfes were untrustworthy. But evidence suggesting-that Randy had a personal stake in Browning’s conviction would have shown the jury why the Wolfes would lie in this particular case. See Maxwell v. Roe, 628 F.3d 486, 510 (9th Cir. 2010) (“The undisclosed benefits that the informant received added significantly to the benefits that were disclosed and certainly would have cast a shadow on the informant’s credibility. Thus, their suppression was material,” (internal quotation mark's and alterations omitted)). Such evidence would have been uniquely impeaching, and if disclosed, may have broken the camel’s back and persuaded the jury to disbelieve the Wolfes. Without the Wolfes, the prosecution had no case.12
Hugo Elsen’s Description of the Assailant’s Hair. Had the prosecution disclosed the precise words Hugo used to describe his assailant’s hair, the prosecution’s argument that the source of the description must not have known the difference be*467tween a Jheri Curl and an Afro would have failed, leaving the jury with no reason to disregard Hugo’s description. Hugo’s precise description—wet, shoulder length, and loosely curled—significantly undermines the case against Browning. . The description was markedly different from Browning’s hair on the day of the murder—an Afro— and, according to Branon, Hugo was lucid when he gave it. Moreover, it was unlikely that Hugo was mistaken about his description in light of Branon’s “meticulous” questioning. Hugo also had the closest and most accurate view of the assailant’s hair, while, as discussed below, every other eyewitness identification was seriously flawed. Hugo’s Vivid description of a hairstyle so different from Browning’s presented substantial reason to doubt that Browning was-the one who stabbed Hugo. If the jury no longer had reason to reject that description, and the jury knew that the description came from the victim, it would have raised grave doubt about the prosecution’s theory of the case.
Identifications, The identifications presented at trial were significantly flawed. Two of the three original positive identifications were equivocal at best, and the officers’ presentation procedures were textbook examples of suggestive techniques. See United States v. Wade, 388 U.S. 218, 228, 87 S.Ct. 1926, 18 L.Ed.2d 1149 (1967) (“A major factor contributing to the high incidence of miscarriage of justice from mistaken identification has been the degree of suggestion inherent in the manner in which the prosecution presents the suspect to witnesses for pretrial identification.”).
' Josy Elseri saw only the side of the assailant’s head after waking from a nap. She told officers it was unlikely that she would be able to identify the assailant because she saw him only for a moment, and never saw his face. During the photo lineup, Josy pointed to three photos of men that were not Browning, shying only that their hair resembled the assailant’s. While Josy identified Browning at trial, it was only after seeing Browning at more than a dozen preliminary hearings, and at each he was presented as the accused. Josy’s -identification deserves, at most, minimal weight.13
Coe’s identification was not much better. An officer presented Coe with Browning— who was shirtless and ‘ handcuffed—and said, “We think we have a suspect. Is this him?” At this point, Browning’s appearance, the officer’s question,- and the form of the showup rendered the procedures highly suggestive and any resulting identification of little evidentiary value. See Stovall v. Denno, 388 U.S. 293, 302, 87 S.Ct. 1967, 18 L.Ed.2d 1199 (1967) (“The practice of showing suspects singly to persons for the purpose of identification, and not as part of a' lineup, has - been widely condemned.”), abrogated on other grounds by Griffith v. Kentucky, 479 U.S. 314, 107 S.Ct. 708, 93 L.Ed.2d 649 (1987). Coe then said that she only “thought” Browning was whom she had seen running by her office. This-was despite initially telling police that the man she saw was white and did not look as if he came from the Elsens’ store. *468Coe also told police that “when [she] see[s] a black person, [ ] they all look the same,” giving less reason for confidence in her already uncertain identification. Coe said at trial that after having had “time to think about it,” she was sure that it was Browning that she saw. But her in-court identification says far less than her equivocal contemporaneous one. Cf. Gilbert v. California, 388 U.S. 263, 272, 87 S.Ct. 1951, 18 L.Ed.2d 1178 (1967) (tying admissibility of in-court identification in part to the constitutionality of a pretrial identification).
Woods gave a positive identification of • Browning at the scene. But police used the same suggestive procedure they did with Coe. They told Woods that they had a suspect and then presented Browning wearing only pants, and not as part of a multi-person lineup. Again, these procedures cast doubt on the answer they produced.14
Not counting the far-after-the-fact and suggestive in-court identifications, the only unequivocal identification of Browning at the time of the crime was Woods’s, in response to a suggestive, one-person show-up. But even crediting Woods’s identification, it tells us almost nothing about who killed Hugo Elsen. Woods stated that he saw Browning jogging towards him (and away from the Elsens’ store), getting within touching distance. Browning was not carrying anything and had no blood on him. We do not understand how Woods’s identification and description of Browning—no blood on him and nothing in his hands—would support an inference that Browning had just brutally stabbed a man to death and stolen 70 pieces of jewelry. The only thing that Woods’s identification, if credited, proves is that Browning was a few blocks from his residence around 4:30 p.m. on November 8,1985.
The Jewelry. When police arrested Browning in the Wolfes’ apartment, they found many pieces of the stolen jewelry in the same room as Browning. However, the rest of the jewelry was turned over later by the Wolfes (except for the items Randy kept for himself). There was no evidence at trial that Browning’s fingerprints were on any of the stolen jewelry. The jewelry evidence is just as consistent with the Wolfes framing Browning for the murder as it is with Browning being guilty.
Browning’s Appearance When Arrested. The Elsens, Coe, Hoffman, and Woods gave somewhat similar descriptions of the clothes on the man they saw. Josy and Hugo Elsen both described a man wearing a blue cap. Coe described a blue cap, Levi’s, and a dark blue jacket. Woods described dark pants, a light-colored shirt, and a “darker color” hat—a “beret sort of thing.” Hoffman described Levi’s, a shirt that he vaguely recalled as plaid, and a blue baseball cap. When police arrested Browning, he was not wearing any jacket, any shirt, or any hat. The shoes Browning was wearing did not match the bloody shoeprints at the crime scene, and there was no evidence that Browning had any blood on him. Police recovered a blue hat in the dumpster outside the Normandy Motel, but again, this discovery was just as consistent with the Wolfes being responsible for the murder as Browning.
*469Fingerprints. Officer Horn lifted “twenty some odd” fingerprints from the scene, two of which matched Browning’s prints. One of Browning’s prints was from the top glass of one of the disturbed counters, and the other was from a fragment of the counter’s broken sliding-glass door. However, the store was only two blocks from Browning’s residence, and there was no evidence as to how long the prints had been present. See Mikes v. Borg, 947 F.2d 353, 361 (9th Cir. 1991) (“[I]n a case resting upon the premise that the defendant impressed his fingerprints on an object at the time of the commission of the crime and supported solely by evidence that the defendant’s fingerprints were found on that object, the record must contain sufficient evidence ... that the defendant touched the object during the commission of the crime.”). Browning could have visited the store at some earlier point. Or he could have been involved in the robbery, while someone else—say Randy Wolfe or the Wolfes’ Cuban friend—committed the murder. Cf. Wearry, 136 S.Ct. at 1006 (“[T]he evidence the dissent cites suggests, at most, that someone in Wearry’s group of friends may have committed the crime, and that Wearry may have been involved in events related to the murder after it occurred.”).
Steven Scarborough, who examined the fingerprints, also testified that he compared the approximately twenty fingerprints lifted from the scene against only Browning’s and Hugo Elsen’s. This left no evidence about whether any of the prints matched Randy or Vanessa Wolfe, or any other possible suspect. Moreover, the prosecution did not present any evidence that the fingerprints were bloody or that blood was on the glass. Had Browning stabbed Hugo and then broken the case and stolen the jewelry as the prosecution suggested, the fingerprints likely would have had blood on them.
The fingerprint evidence is probably the strongest evidence against Browning. But it is by no means decisive, and we conclude that it is not enough to avoid the otherwise substantial reasonable doubt created by the shoeprint evidence, the evidence that Randy expected a benefit for his testimony, and Hugo’s description of the assailant’s hair.15
Blood-Spotted Jacket. Vanessa Wolfe identified a tan windbreaker found in her apartment as belonging to Browning. Criminalist Minoru Aoki testified that blood found on the jacket was type B, the same blood type as Hugo Elsen’s. The prosecution argued at trial that the jacket proved that Browning was the killer. But Aoki’s testimony only showed that the blood on the jacket was the same type as Hugo Elsen’s (out of four types), not that there was a DNA match.16 The jacket did not match any of the descriptions given by the identification witnesses, who all said the jacket they saw was blue. And even if *470the killer had worn the jacket, there was no reason to believe that Browning was wearing the jacket on November 8, 1985. The jacket was found in the Wolfes’ apartment, tying it just as easily to Randy or his Cuban friend as to Browning. The jacket was a zero-sum for the prosecution’s case. •
The Knife. Vanessa Wolfe testified that she saw Browning shaking water off of a knife in her apartment, and that he asked her to help get rid of it, But there was no blood or fingerprints found on the knife, and apart .from Vanessa’s testimony, no,, other evidence tying the knife to the murder. Dr. Giles Green testified, that Hugo’s wounds could have been made by the knife, but nothing made him think that that particular knife was the murder weapon.17 Even if there had been some physical evidence connecting the knife to the murder, there was still no such evidence that Browning had even touched it. Indeed, the knife, like the tan jacket, is just as consistent with the Wolfes or a friend of theirs committing the murder as with Browning being the killer. Outside of Vanessa’s testimony, the knife adds nothing to the prosecution’s case.
In sum, the jewelry, the knife, and the tan jacket all failed to tie Browning to the murder. The identifications were flawed and mostly equivocal. There were endless reasons to distrust the Wolfes. When arrested, Browning was not wearing any clothing described by the eyewitnesses,' and there was no evidence that Browning had any blood on him. All the prosecution had left was the fingerprints. But even with those, there was no evidence about how long the prints had' been present, and no evidence that the other prints from the scene did not mátch either of the Wolfes or their Cuban friend. The upshot is that the prosecution presented a fundamentally weak case. Add Officer Branon’s observation of the shoeprints, thus leaving unanswered significant evidence that someone besides Browning, Josy, Coe, a paramedic, or a police officer was in the store with Hugo while he was bleeding; add evidence that the prosecution was planning to help its best witness, in an unrelated sentencing, suggesting a motive for the witness and his wife to lie at trial; add an unrebutted closeup description from the victim that did not match the defendant; and that fundamentally weak case collapses- under the weight of -its reasonable doubt. “Even if the jury—armed with all of this new evidence—could have vested1 to convict [Browning], we have no confidence that it would have done so.” Wearry, 136 S.Ct. at 1007 (internal quotation marks omitted). And yet Browning sits on death row. We conclude that there is a reasonable probability that had the concealed evidence not been withheld, the jury would have reached a different result.
We also hold that this result is the only objectively reasonable conclusion. Whatever confidence the Supreme Court of Nevada found in Browning’s verdict, it was not a confidence that was* objectively reasonable. The strength of the undisclosed evidence is too great, and the" remainder of the trial record too weak. “‘[F]airness’ cannot be stretched to the point of calling this1 a fair trial.” Kyles, 514 U.S. at 454, 115 S.Ct. 1555. The district court should have granted habeas, relief on Browning’s Brady, claims.
IV
Browning also asserts that he was denied his right to effective assistance of *471trial counsel. To show a violation of that right, Browning must demonstrate that (1) Pike’s performance was deficient, and (2) that deficiency -prejudiced Browning. Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2062, 80 L.Ed.2d 674 (1984). Because AEDPA guides our review, we ask whether the Supreme Court of Nevada “applied Strickland to the facts of [t]his case in an objectively unreasonable manner.” Bell v. Cone, 535 U.S. 685, 699, 122 S.Ct. 1843, 152 L.Ed.2d 914 (2002). We conclude that it did.
A
We first clarify the scope of Browning’s IAC claim on appeal. In its order, the district court limited the CÓA to particular “claims” that Pike’s failure to investigate particular avenues of evidence wére deficient. The district court granted COAs on whether Pike’s failure to investigate (1) the source of the bloody shoeprint, (2) the Wolfes’ credibility as witnesses, and/or- (3) Hugo Elsen’s actual description of the assailant to Officer Branon.each constituted individual instances of ineffective assistance of counsel. Limiting.the.COA in this manner was error.
Browning is entitled to a COA if he “has made a substantial showing of the denial of a constitutional right.” 28 U.S.C. § 2253(c)(2) (emphasis added). Browning’s habeas petition asserts that he was denied the constitutional right of effective trial counsel. This right is a guarantee of effective counsel in tota—it promises that counsel will perform reasonably. While an'individual claiming IAC “must identify the acts or omissions 'of counsel that are alleged not to have been the result of reasonable professional judgment,” Strickland, 466 U.S. at 690, 104 S.Ct. 2052, the court considers counsel’s conduct as a whole to determine whether it was constitutionally adequate, see, 'e.g., id; Wong v. Belmontes, 558 U.S. 15, 17, 130 S.Ct. 383, 175 L.Ed.2d 328 (2009) (“In light of the variety of circumstances faced by defense counsel and the range of legitimate decisions regarding how best to represent a criminal defendant, the performance inquiry necessarily turns on whether counsel’s assistance was reasonable considering all the circumstances.” (internal, quotation marks and • alterations omitted)). The district court distorted this inquiry by separating Browning’s IAC argument into individual “claims” of IAC corresponding to particular instances of Pike’s conduct. This approach was misguided. -Rather, the IAC portion of the COA should have been crafted at a higher level of generality.
Browning asks us to expand the COA to include whether he “was denied effective assistáhee of counsel by his trial lawyer’s wholesale failure to investigaté and prepare for trial.” Because this articulation more appropriately frames the constitutional right Browning’s petition contends was violated, and because—as explained below—he “has made a substantial showing of the denial” of that, right, 28 U.S.C. § 2253(c)(2), we GRANT his m.otion to expand the COA to include that issue.
B
The first element of an IAC claim requires Browning to show that his counsel’s performance was “deficient,” or more precisely, “below an objective standard of reasonableness.” Strickland, 466 U.S. at 688, 104 S.Ct. 2052. Browning may not rely on generalities in making this showing; he must point us to specific instances of Pike’s conduct that demonstrate incompetent performance. Id. . at 690, 104 S.Ct. 2052. Because-Browning asserts that Pike failed to adequately investigate the case, Browning must show that Pike violated his “duty to make reasonable investigations or to make a reasonable decision that makes *472particular investigations unnecessary.” Id. at 691,104 S.Ct. 2052. The ABA Standards for Criminal Justice in effect during Browning’s trial, to which the Supreme Court has “long ... referred as guides to determining what is reasonable,” Wiggins v. Smith, 539 U.S. 510, 524, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) (internal quotation marks omitted), help clarify Pike’s investigatory obligations. They included “the duty ... to conduct a prompt investigation of the circumstances of the case ... in-clud[ing] efforts to secure information in the possession of the prosecution and law enforcement authorities.” ABA Standards for Criminal Justice, Standard 4-4.1 (2d ed. 1980); see also Summerlin v. Schriro, 427 F.3d 623, 629-30 (9th Cir. 2005) (en banc) (“The standards in effect at the time of Summerlin’s trial [which occurred prior to Browning’s] clearly described the criminal defense lawyer’s duty to investigate ...”).
We examine Pike’s performance in a “highly deferential” manner, “indulging] a strong presumption that [Pike’s] conduct falls within the wide range of reasonable professional assistance.” Strickland, 466 U.S. at 689,104 S.Ct. 2052. Moreover, because we are operating under AEDPA deference, our review is “doubly deferential^] ... tak[ing] a highly deferential look at counsel’s performance, through the deferential lens of § 2254(d).”. Cullen v. Pinholster, 563 U.S. 170, 190, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011) (citation and internal quotation marks omitted). Browning therefore must show more than “a strong case for relief’—he must demonstrate that “there is no possibility fair-minded jurists could disagree that the state courts decision conflicts with” Supreme Court precedent. Harrington v. Richter, 562 U.S. 86, 102, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011). We conclude Browning has made this showing.
Browning first points us to Pike’s failure to interview Officer Branon prior to calling Branon to testify at trial. The Supreme Court of Nevada concluded that this was deficient performance, Browning, 91 P.3d at 46, and we agree. We have previously assumed without deciding that “it ordinarily falls below the Strickland level of required competence to put a witness on the stand without interviewing him.” Jackson v. Calderon, 211 F.3d 1148, 1160 (9th Cir. 2000). Because the government does not challenge the Supreme Court of Nevada’s conclusion on this point, we need not decide whether failing to interview a witness before calling him to the stand invariably constitutes objectively unreasonable representation, and do not disturb the Supreme Court of Nevada’s conclusion that it did in this case.
Browning also argues that Pike’s failure to investigate the source of the bloody shoeprints constituted deficient performance. At the state habeas hearing, Pike explained this decision by noting that if he attempted to determine the source of the shoeprints and discovered that the source was paramedics or responding officers, he would disprove his own theory that the prints were left by the assailant (who was someone other than Browning). In other words, Pike apparently chose not to investigate the source of the shoeprints because he thought Browning was guilty, and thus assumed the shoeprints had been left by paramedics and other responders. The Supreme Court of Nevada held that this was a reasonable tactic, explaining, “[a]s long as the source of the prints was unknown, counsel could argue to the jury that the actual murderer had left them.” Browning, 91 P.3d at 46.
On this issue, the Supreme Court of Nevada unreasonably applied Strickland’s deficiency standard by blindly accepting Pike’s strategy. “Counsel cannot *473justify a failure to investigate simply by invoking strategy. ... Under Strickland, counsel’s investigation must determine strategy, not the other way around.” Weeden v. Johnson, 854 F.3d 1063, 1070 (9th Cir. 2017). Pike’s invocation of strategy here is an extreme instance of strategy determining investigation. If a defense attorney’s “fear of learning the truth” rendered every decision not to investigate a reasonable tactic, even the most egregious failures to investigate a client’s case would be protected from constitutional scrutiny. Even worse, under the Supreme Court of Nevada’s reasoning, a criminal defendant’s entitlement to a reasonable investigation would depend on his attorney’s uninformed, gut-based intuition about his client’s guilt. In other words, according to the Supreme Court of Nevada, if your criminal attorney does not believe your story, your attorney need not investigate your case. The Sixth Amendment required more in 1986, and still does today.
To be sure, a decision not to investigate particular facts may be reasonable when the attorney has reason to believe doing so would reveal inculpatory evidence. In Richter, for example, the defendant argued that his attorney should have had a blood expert test a pool of blood from the crime scene to determine whether it was a mixture of two victims’ blood. 562 U.S. at 108,131 S.Ct. 770. Such a result would have dramatically bolstered the defense’s theory. See id. But the test could have also disproved the defense’s theory by only detecting a single blood source. See id. The Court explained that the defendant’s attorney decided not to test the blood because he “had reason to question the truth of his client’s account” in light of the client’s prior false statements. Id. Because of the “serious risk” that the test would expose the defense’s theory “as an invention,” defense counsel’s decision was reasonable. Id.
Here, the facts are the opposite. Pike had no reason to disbelieve Browning’s assertions that he had been framed by the Wolfes. And more importantly, contrary to Pike’s fears, there was little risk that investigation into the source of the shoeprints could damage Browning’s defense theory. Had Pike interviewed Bra-non before calling him to the stand, Pike could have asked him whether paramedics or police officers had entered the store before Branon’s arrival. If Branon’s answer was “no,” this would have bolstered Browning’s theory. But even if Branon had responded “yes,” Pike could have decided then to inquire no further, and still would have inflicted no harm on his theory. Pike thus had no reason to fear that any inquiry into the source of the shoeprint would damage his case. While “[a]n attorney need not pursue an investigation that would be fruitless, much less one that might be harmful to the defense,” id. at 108,131 S.Ct. 770, the reverse is also true: the obligation to investigate, recognized by Strickland, exists when there is no reason to believe doing so would be fruitless or harmful.18
Browning also asserts that Pike’s performance was deficient because Pike never interviewed the Wolfes before trial. When asked about this at the state habeas hearing, Pike explained that he had a policy of never personally interviewing witnesses to prevent becoming a witness himself. Rath*474er, he.had Martin Schopp conduct all interviews. The Supreme Court of Nevada concluded that this was a reasonable strategy. Browning, 91 P.3d at 46.
There can be no doubt that Pike’s policy of not personally interviewing witnesses was reasonable. But that policy in no way explains why Pike rejected Schopp’s request to interview the Wolfes. Merely articulating a reasonable strategy in response to a deficiency argument-does not end the inquiry when that strategy does not explain the decision- itself. See Wayne R. Lafave, et al., Criminal Procedure § 11.10(c), at 797 (6th ed. 2016) (“Of course, a decision apparently based on a tactical judgment is not therefore rendered immune from an incompetency challenge.”). Pike gave no explanation for why Schopp could ■ not' conduct the interview himself. Yet the Supreme Court of Nevada concluded that Pike’s no-personal-interviews strategy explained his decision to not subject the Wolfes to interviews. That conclusion makes no sense and is objectively unreasonable.
Finally, in arguing for an expansion of the COÁ, Browning lists á number of other alleged deficiencies in Pike’s representation.19 Because we find that Browning’s ineffective assistance of counsel claim succeeds on other grounds, we do not here assess these other alleged deficiencies.
Our IAC analysis is based on the fundamental obligations of each attorney, and is not a product of hindsight. See Bell, 535 U.S. at 702, 122 S.Ct. 1843. Pike had “countless ways” to investigate adequately Browning’s case. We do not limit ■ him to just “one technique or approach.” Richter, 562 U.S. at 106, 131 S.Ct. 770. And to be sure, Pike’s “poking holes” and “casting shadows” strategy could have been appropriate under , the right circumstances. See id. at 109, 131 S.Ct. 770' (“To support -a defense argument that the prosecution has not proved its case it sometimes is better to try to cast pervasive suspicion of doubt than to strive to prove a certainty that exonerates.”). But to reach the conclusion that this strategy was reasonable, Pike first had an obligation “to. make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Strickland, 466 U.S. at 691, 104 S.Ct. 2052. Here, Pike did neither. His failure to investigate what happened on November 8, 1985, “so undermined the proper functioning of the adversarial process that [Browning’s] trial cannot be relied on as having produced a just result.” Id. at 686,104 S.Ct. 2052.
We conclude that Pike unreasonably failed to investigate Browning’s case, and that the Supreme Court of Nevada unreasonably concluded that Browning failed to prove just that.
C
We now consider whether the unprofessional deficiencies identified above prejudiced Browning. Despite its differing terminology, prejudice in the IAC context mirrors the materiality standard under Brady. We ask whether there is a “reason*475able probability that, but for counsel’s unprofessional errors, the result of the proceedings would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052; see Gentry v. Sinclair, 705 F.3d 884, 906 (9th Cir. 2013) (“Brady materiality and Strickland prejudice are the same.”). To meet this standard, Browning must demonstrate a reasonable probability that had Pike conducted an adequate investigation, “at least one juror would have harbored reasonable doubt about” Browning’s guilt. Buck v. Davis, — U.S. -, 137 S.Ct. 759, 776, 197 L.Ed.2d 1 (2017). Because we defer to the Supreme Court of Nevada’s decision under AEDPA, our ultimate inquiry is whether that court’s conclusion that any deficient performance by Pike did not prejudice Browning was objectively unreasonable. We conclude that it was.
We have already explained in detail why the prosecution’s case against Browning was quite weak. In fact, because the standards of materiality for Brady and Strickland are the same, our materiality analysis above is in large part identical to the assessment of the prejudicial effect of Pike’s ineffective assistance. But while we will not repeat that analysis here, we cannot just incorporate the materiality section above in its entirety. The evidence to consider in the IAC context differs in one important aspect: even if Schopp had interviewed the Wolfes, there is no reason to believe that Randy would have made any mention of his expectation that he would receive leniency in exchange for his testimony against -Browning. After all, Randy stated explicitly at Browning’s trial that he was not anticipating any such benefit. We therefore consider, for purposes of assessing prejudice in the context of Browning’s IAC claim, the following facts that would have been available to Browning had Pike engaged in an adequate investigation: (1) that the shoeprints could not have been created by a paramedic or responding officer, and (2) that Hugo Elsen described his assailant not as having a Jheri Curl, but as having, shoulder-length, wet, loosely-curled hair. As discussed above, this evidence would have had a significant impact on Browning’s case. The bloody shoeprints were the only evidence left in the store during the period between the robbery and the arrival of the first-responders, and the evidence does not support that the shoe-prints were left by Josy Elsen or Debra Coe. This leaves Browning with evidence that someone else—not Browning, not Josy, not Coe—was in the store with Hugo Elsen before the arrival of the first-responders. Such evidence would have been significantly and uniquely exculpatory. Cf. Richter, 562 U.S. at 112, 131 S.Ct. 770 (rejecting petitioner’s prejudice claim because it “established nothing more than a theoretical possibility” that petitioner’s defense theory was true); Pinholster, 563 U.S. at 200-01, 131 S.Ct. 1388 (rejecting petitioner’s prejudice claim because the evidence at issue was duplicative of other evidence presented to the jury).
And Hugo’s description of his assailant’s hair is powerful evidence of Browning’s innocence. As stated above, Hugo’s view and description of the assailant suffers from none of the flaws inherent, in each of the other eyewitness accounts involved in this case. Officer Branon said it was not possible that Hugo was mistaken about his description of the assailant. And Woods’s description of Browning running towards him on the sidewalk away from the Elsens’ store—with not a drop of blood on him or a piece of jewelry in his hands—is, if anything, supportive of Browning’s innocence.
As described in great detail above, the prosecution’s evidence was far from overwhelming. There is a strong possibility that had Pike offered the evidence he would have obtained if he had made a reasonable investigation, at least one juror *476would have harbored reasonable doubt. Browning would have so substantially ben-efitted from that evidence that it was objectively unreasonable for the Supreme Court of Nevada to conclude to the contrary.
V
Our conclusions above regarding Browning’s claims under Brady and Strickland involve only his convictions relating to the robbery and murder of Hugo Elsen. They do not affect the validity of his escape conviction. It is not clear from Browning’s habeas petition whether he challenges his escape conviction. But even assuming he means to challenge that conviction, he has identified no exculpatory evidence withheld that would have affected the jury’s decision to convict him of escape under Nev. Rev. Stat. § 2j2.090. And while Pike’s investigation into Browning’s case was deficient, Browning points to no evidence that Pike would have obtained had he reasonably investigated the case that would have affected the jury’s decision on the escape count. In sum, while Browning has demonstrated that he is entitled to habeas relief from his murder- and robbery-related convictions, he is not entitled to relief from his escape conviction.
VI
The Supreme Court of Nevada’s denial of Browning’s claims under Brady and Strickland constituted an unreasonable application of clearly established Supreme Court precedent. Browning is entitled to a writ of habeas corpus with respect to his convictions of burglary, robbery with the use of a deadly weapon, and murder with the use of a deadly weapon. Because Browning has offered no reason to call the validity of his escape conviction into question, he is not entitled to habeas relief as to that conviction.
We AFFIRM in part, REVERSE in part, and REMAND to the district court for further proceedings consistent with this decision.
. The prosecution also presented evidence that once Browning was brought to the police station on the night of November 8, 1985, he escaped from the interview room where he was being held. He was caught before he left the police station building, As explained below, see Section V, infra, because Browning's petition provides no basis for challenging his escape conviction, our analysis focuses only on his robbery- and murder-related convictions.
. Browning has moved for expansion of the COA to include three additional issues. For the reasons set forth below, see Section IV.A, infra, we GRANT Browning’s motion in part and expand the COA to include the issue of whether Browning’s trial counsel was ineffective because’of his overall failure to investigate Browning's case. Browning also seeks to expand the COA to include: (1) whether the trial court improperly instructed the jury on the element of deliberation, and (2) whether the prosecutor's statements during closing argument violated Browning’s rights under the Due Process Clause. Because Browning has not "made a substantial showing of the denial of a constitutional right” for either issue, we DENY the motion in part as to those claims. 28 U.S.C. § 2253(c)(2),
. As the dissent notes, this deferential AÉDPA standard is occasionally described as allowing habeas relief only when the state court’s conclusions .are so unreasonable that there is no "possibility of fairminded disagreement." Davis v. Ayala, — U.S. -, 135 S.Ct. 2187, 2199, 192 L.Ed.2d 323 (2015). The existence of cases post-dating AEDPA where the Supreme Court has granted habeas relief over dissent, however, suggest that this language is not to be construed as requiring unanimity, or as suggesting that jurists who disagree with a grant of habeas relief are not fair-minded. • See, e.g., Panetti v. Quarterman, 551 U.S. 930, 127 S.Ct. 2842, 168 L.Ed.2d 662 (2007) (a 5-4 decision holding that the state court unrea- . sonably applied Ford v Wainwright, 477 U.S. 399, 106 S.Ct. 2595, 91 L.Ed.2d 335 (1986)) and Abdul-Kabir v. Quarterman, 550 U.S. 233, 127 S.Ct. 1654, 167 L.Ed.2d 585 (2007) (a 5-4 decision holding that the . state court unreasonably applied clearly established Supreme Court precedents requiring a sentencing jury in a capital case to be able to consider all mitigating evidence),
. We expand the COA to cover this claim because we conclude that reasonable jurists could disagree with the district court's ruling that not disclosing Hugo’s precise description of the hair did not violate Brady. See Buck v. Davis, — U.S. —, 137 S.Ct. 759, 773, 197 L.Ed.2d 1 (2017). As we explain in this opinion, the prosecution’s failure to disclose that evidence was in violation of Brady.
The state argues that we lack jurisdiction to expand the COA in this manner because Browning did not explicitly include the hair description issue in the section of his brief labeled "uncertified issues.” We disagree. When the content of a brief covers an uncertified issue, "we may treat it as a request to expand the scope of the certificate of appealability.” Robertson v. Pichon, 849 F.3d 1173, 1187 (9th Cir. 2017) (quoting Delgadillo v. *461Woodford, 527 F.3d 919, 930 (9th Cir. 2008)). Though Browning did not style his hair-description arguments as a request to expand the COA, he nonetheless thoroughly discussed the issue. We construe that discussion as a request to further expand the COA.
. The Supreme Court of Nevada held that while impeachment evidence was withheld, that information was not material, Browning, 91 P.3d at 55.
. The dissent reasons that this determination by the Supreme Court of Nevada was "not based on any facts in the record,” and that the Supreme Court of Nevada therefore "engaged iri an unreasonable determination of the facts” in violation of 28 U.S.C. .§ 2254(d)(2). But at no point in this case has the state challenged the Supreme Court of Nevada’s ruling on that point. We also do not see how it could. The federal habeas statutes provide a mechanism by which state prisoners can challenge on federal grounds the authority behind their detention by state officers. They do not. provide a means for federal courts to engage in error correction of state court rulings that favor defendants* The statutory language makes this plain: 28 U.S.C, § 2254(d)(2) states that "[a]n application for a writ of habeas corpus .,. shall not be granted with respect to any claim ... unless the adjudication of the claim ... resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” Also, 28 U.S.C. § 2254(e)(1) states "a determi- • nation of a factual issue made by a State court shall be presumed to be correct, The applicant shall have the burden of rebutting the presumption of correctness by clear and convincing evidence.” These statutory principles are limitations on federal courts’ power to grant habeas relief. We do not understand how the dissent wrings from these provisions an affirmative power to rule that a state court erred in doing too little to justify its detention of a petitioner.
. The dissent calls this a "novel view” of Brady. According to the dissent, our analysis "extends the state’s obligations into the murky zone of interpretations of otherwise neutral facts.” But facts do hot exist in’ a vacuum. Their exculpatory value invariably depends on the interpretations offered, and the theories pressed, by the parties. Consider, for example, one of the pieces of Brady material in Kyles. 514 U.S. 419, 115 S.Ct. 1555. The prosecution in that case did not disclose to the defense a list of the cars parked in the parking lot where the victim was killed, a list which did not include the defendant’s car. Id. at 450, 115 S.Ct. 1555, The Supreme Court held that the list was exculpatory in part because the prosecution "argued to the jury[ ] that the killer drove to the lot and left his car there.” Id. Had the prosecution instead argued that the killer walked to the scene of the crime, the list of cars would have had less exculpatory value to the defense. Likewise here, the prosecution’s argument that the speaker did not know the difference between a Jheri Curl and an Afro affected the exculpatory value of Hugo’s precise words, We offer a straightforward application of clearly established Brady principles, not a "novel interpretation.”
. In the state habeas proceeding, Browning argued the same point but with expert testimony. He submitted a report from forensics examiner Michael Sweedo that stated that the shoeprints were too big to have been made by the typical woman’s tennis shoe. This evidence was not part of the original trial record, and we are doubtful that we may consider it in determining materiality under Brady. Nevertheless, we need not decide the issue because the Brady violations here were material without considering post-trial revelations.
. Moreover, the record does not support this new theory. Counsel seemed to derive the theory from the particular words chosen by Officers Branon and Horn in describing the location of the shoeprints. At the state habeas hearing, Branon testified that the shoeprints were "near” Hugo, while Michael Sweedo, reading from Horn’s police report, stated that the prints led from Hugo’s body to the front door. We are unpersuaded. A description of shoeprints "near” a body could easily mean shoeprints leading from that body to another part of the room. The state also presents no other evidence, such as different tread patterns in the shoeprints, to support the two-sets theory. And we have already explained why the evidence does not support Josy or Coe having left any of the shoeprints.
. The Wolfes' use of drugs is relevant to their credibility because it. indicates that., in the past they engaged in illegal activity, However, that the Wolfes were physically addicted to drugs is not here relevant, to their credibility. See United States v. Kizer, 569 F.2d 504, 506 (9th Cir. 1978).
; The dissent contends that, when engaging in this analysis, "our role is not to reweigh ' the evidence and make fresh credibility determinations.” We disagree that this argument is controlling. To determine whether the jury would have been swayed by additional evidence, we 'must püt ourselves in the shoes of the jurors and ask whether the same result would be reached if presented with this new, hypothetical trial record. There is no way to do that without making fresh credibility' determinations, particularly when the new evidence is impeachment evidence, and is therefore relevant only because of its tendency to affect credibility,
. Browning argues that Josy never positively identified Browning at trial, and that the state conceded as much during state habeas proceeding. However, Browning has not provided the Court with any documentation regarding the state’s supposed concession. The prosecution did significantly overstate Josy’s identification, saying that she pointed at Browning and said, "That’s the man who was kneeling over my husband,” when in fact Josy-had merely said that Browning was in the courtroom. The identification was not, as Prosecutor Seaton maintained, "as good as you can ask for.” Nevertheless, .we agree with the trial court that- Josy identified Browning at trial.
. Hoffman was subject to the same suggestive procedure as well, but curiously, no party at trial squarely asked him whether Browning was the same man he had seen earlier walking towards the Elsens' store. Nevertheless, Hoffman’s trial testimony hints at what his answer would have been. He testified that the man he saw was wearing a hat, and that Browning did not appear as though he had recently been wearing a hat. Hoffman also told police that the man he saw was "Cuban,” supporting Browning's defense theory that the Wolfes and their Cuban friend framed him for the murder.
. In the state habeas proceeding, Browning presented expert testimony that it was possible that the print on top of the glass case could have come from someone leaning over the case, and that the print found on the shard of glass on the floor behind the counter could be consistent with someone pushing the glass door open. Defense counsel Pike also testified at the habeas hearing that he had planned to call Browning’s girlfriend, Marsha Gaylord, to testify that she had been in the jewelry store with Browning prior to the day of the crime, but that Pike was unable to do so because Gaylord had disappeared. As already discussed, we grant relief without deciding whether such post-trial evidence bears on materiality under Brady.
. In the state habeas proceeding, the parties stipulated that post-trial DNA testing revealed that the blood on the jacket conclusively did not belong to Hugo Elsen. Again, we do not rely on this post-trial evidence.
. In the state habeas proceeding, Browning introduced a forensics report indicating that Hugo Elsen’s wounds "do not coherently coincide” with the knife found in the Wolfe’s apartment. We do not rely on this evidence.
. Browning offers a different but related argument: the Supreme Court of Nevada’s conclusion that Pike’s performance was deficient due to his failure to interview Branon at all requires a holding that Pike was deficient for not asking Branon about the shoeprints. Because there was a deficiency in not interviewing Branon at all, we need not decide if any particular questions were needed to be asked.
, These include: Pike's not offering evidence at trial that Browning had no blood, cuts, or scrapes on his- body when he was arrested; ' failing to secure Gaylord's testimony; not .challenging the prosecution’s assertion that when they entered .the Wolfes’ .apartment Browning was "surrounded” by jewelry; not objecting to the prosecutor's improper closing statements; not bringing out at trial that Browning had no heroin in his body when he was arrested; not objecting to Seaton’s claim . that the tan jacket had Hugo Elsen’s blood on it; not presenting evidence demonstrating that Gaylord was not in jail on November 8, 1985; never" obtaining, a forensic evaluation of the . ■ knife; and never getting a witness to testify ■ that Randy Wolfe tried to sell the jewelry.
. I concur in the majority’s rejection of Browning’s Napue claim and in its affirmance of his escape conviction.