ORDER
The memorandum disposition filed on September 8, 2011 is hereby withdrawn. A new disposition will be filed.
The majority of the panel votes to deny the petition for rehearing. Judge Callahan votes to grant the petition for rehearing. Judge Berzon votes to deny petition for rehearing en banc and Judge Noonan recommends denying the petition for rehearing en banc. Judge Callahan votes to grant the petition for rehearing en banc.
The full court has been advised of the petition for rehearing en banc, and no judge of the court has requested a vote on whether to rehear the matter en banc. Fed. RApp. P. 35.
The petition for rehearing is DENIED and the petition for rehearing en banc is DENIED.
No further petitions for rehearing and for rehearing en banc will be entertained.
MEMORANDUM *
Richard Raymond Tuite appeals the denial of his petition for a writ of habeas corpus. Applying Brecht v. Abrahamson, 507 U.S. 619, 113 S.Ct. 1710, 123 L.Ed.2d *703353 (1993), and holding that there is at least “grave doubt” as to whether the confrontation clause error at issue had a substantial and injurious effect or influence on the verdict, we reverse and remand. See Pulido v. Chrones, 629 F.3d 1007, 1012 (9th Cir.2010).
FACTS AND PROCEEDINGS
On January 20, 1998, Stephanie Crowe, aged twelve, was stabbed nine times by knife in her own bed in her own bedroom. She had been on the phone about 10 p.m. and died before midnight. Her body was discovered by her grandmother about 6 a.m., the next day.
Within two weeks of her death, the Escondido police had identified three teenage boys as suspects, had interrogated them, and had obtained confessions. On May 22, 1998, these suspects were indicted for Stephanie’s murder. The district attorney of San Diego County prepared the case for trial, calling on an F.B.I. expert, Mary Ellen O’Toole, for help in developing the case. But in February 1999, the state Attorney General dismissed the indictments without prejudice. This court recently reversed the dismissal of a § 1983 lawsuit brought by the former suspects against the investigating police officers, so that case is going forward. See Crowe v. County of San Diego, 608 F.3d 406 (9th Cir.2010).
Here, however, we consider only the record as it developed in Tuite’s criminal trial. That record shows that the day after Stephanie’s death, the police picked up Tuite, an itinerant known to have been as close as one quarter of a mile to the Crowe house at one point on the fatal evening. They took hair samples and fingernail scrapings, photographed him and impounded some of his clothes, in particular a red turtleneck shirt and a white T-shirt. These items were examined on April 28, 1998 for bloodstains. Wetting the red shirt completely and using a fluoroscopic process, the police found no blood on the red shirt. The red shirt was subsequently photographed using a tripod that had previously been used at the scene of the crime without using protective coverings for the legs so as to avoid contamination. The white T-shirt had visible bloodstains and was sent to a laboratory for DNA testing. The test excluded Stephanie as the donor. Search of the Crowe house found no physical evidence of Tuite ever having been there.
Tuite was a mentally deficient person, without known employment or home, who had been living in the San Diego area. On the evening of January 20, he had bothered three residences as he sought “the girl» or “TraCy_» After Stephanie’s death, he continued this quest during the rest of January, February and March 1998. Sometimes he annoyed residents enough that they called the police. On no occasion was he violent or did he use a weapon.
The San Diego District Attorney re-cused himself. The state Attorney General took over the case. In May 2002, the Attorney General obtained the indictment of Tuite for murder in the first degree.
At trial, the state presented evidence that a criminologist in 1999 had retested the red shirt worn by Tuite and found Stephanie’s DNA in a stain on it; in April 2003, a second criminologist had retested Tuite’s white T-shirt and found her DNA on it. The defense countered these reports with expert testimony that the police could have inadvertently contaminated the shirts while they were in their custody as they investigated the case. O’Toole, the F.B.I. expert first retained by the district attorney to prosecute the boys, testified for the defense that in her judgment “the crime scene” reflected organization, that *704is, control of the victim and of surrounding events and circumstances so that the murder could be brought off without alarming the family members sleeping nearby, and without the murderer leaving fingerprints or the murder weapon. The prosecution rebutted O’Toole with its last witness, Gregg O. McCrary, who had been an agent of the F.B.I. for nearly thirty years, and was now in the business of consulting on criminal behavior. He testified that the crime scene was, on the whole, “disorganized,” reflecting a random attack.
The defense moved to impeach McCrary by cross-examining him on a letter he had written, attacking O’Toole’s analysis of the crime scene, accusing her of undermining the prosecution of this case, suggesting that she had acted unethically, and expressing a strong desire that O’Toole be persuaded not to testify at Tuite’s trial. After a hearing outside the presence of the jury, the court excluded the letter and cross-examination based on the letter.
McCrary’s excluded letter, dated February 24, 2004, was written to the International Criminal Investigative Analysis Fellowship (ICIAF) about O’Toole’s proposed testimony for the defense, and what he termed “ethical issue[s]” that testimony raised. He wrote that Tuite was the “true killer,” and that he was hopeful O’Toole would not testify. He went on:
Neither the San Diego County Sheriffs Office nor the Office of the Attorney General for the State of California has requested the assistance of the NCAYC [National Center for the Analysis of Violent Crime] or the ICIAF in this matter. Both agencies are shocked and dismayed that Mary Ellen O’Toole, a representative of both the FBI and [the] ICIAF, has injected herself into this case in what they view as an attempt to obstruct justice and undemine the successful prosecution of Richard Tuite.
(emphasis added). At the hearing, McCrary admitted that he had not spoken to anyone from the Sheriffs Office. He also admitted that no one had accused O’Toole of obstructing justice or undermining Tuite’s prosecution.
A jury found Tuite guilty of manslaughter. The court sentenced him to thirteen years imprisonment. He appealed to the California Court of Appeal for the Fourth District, which found that the trial court had committed constitutional error in excluding McCrary’s letter and cross-examination on it, but held the error to be harmless and affirmed. The Supreme Court of California denied review. The federal district court denied Tuite’s petition for habeas.
Tuite’s appeal to us focuses on the California’s court’s ruling of harmlessness. That court held:
Under the above-stated legal principles, we agree Tuite’s counsel should have been allowed to cross-examine McCrary about his February 24, 2004 letter, and the trial court violated Tuite’s constitutional right to confront adverse witnesses when it precluded such cross-examination.
The letter was relevant because it demonstrated bias and impacted McCrary’s credibility in a manner that could lead a reasonable jury to question the reliability and validity of his testimony. (See Evid.Code, §§ 210, 780, subd. (f).) The letter bore directly on McCrary’s credibility and reliability by indicating McCrary had a personal interest in convicting Tuite, whom he referred to as “the true killer.” The letter also demonstrated McCrary had prejudged Tuite’s case and was acting more as an advocate for the prosecution than as a forensic expert. Moreover, McCrary’s unusual attempt to dissuade O’Toole from testifying revealed a bias *705in favor of the prosecution and a bias against O’Toole. The letter also revealed McCrary’s tendency to exaggerate; his statement that the sheriffs office and the prosecuting agency viewed O’Toole as obstructing justice was not only a gross overstatement but was also unreliable because no one from the sheriffs office had talked to him about O’Toole’s upcoming testimony. For all these reasons, it is likely that a reasonable jury would have received “a significantly different impression of [the witness’s] credibility had [the excluded cross-examination] been permitted.” (Delaware v. Van Arsdall, supra, 475 U.S. [673] at p. 680 [106 S.Ct. 1431, 89 L.Ed.2d 674 (1986) ]).
Nonetheless, the court of appeal held this constitutional error to be harmless beyond a reasonable doubt.
The court of appeal weighed the harm of the constitutional error. It held that the jury itself was informed of the elements making up “the crime scene”; the jury was not dependent on the experts in determining whether the crime scene was organized or disorganized. The court of appeal added: “Further, the error in excluding McCrary’s letter had no impact on the central evidence against Tuite — the DNA evidence that Tuite had Stephanie’s blood on his clothing.”
In assessing the harmlessness of the constitutional error, however, the court of appeal left out of the account these considerations: (1) the improbability of the prosecution’s case — namely, that a stranger could enter a house he did not know without leaving any signs of forced entry, make his way to a bedroom occupied by a girl without being seen or heard by the five other family members in the house, be moved to kill a girl he did not know, and depart from the house leaving no trace of any kind of his having been there to commit a motiveless murder; (2) the lack of explanation regarding how Tuite could have entered or exited the home, given the fact that the door commonly used to enter and exit the home was found deadbolted from the inside the morning after the murder, and the other doors did not appear to have been used to enter or exit the house; (3) the jury’s view of the evidence as shown by the length of deliberations and the jury’s reported deadlock after a week of deliberation; (4) the jury’s compromise verdict of voluntary manslaughter rather than murder, reached after another week of deliberation, a verdict virtually impossible to square with the evidence of a series of deliberate knife wounds inflicted on a helpless victim. The court of appeal did note the strength of the DNA evidence. But if that evidence had been seen as dispositive, the jury would neither have reported deadlock nor compromised in reaching a manslaughter verdict in a case in which the victim was stabbed nine times with a knife. In fact, members of the jury could well have regarded the DNA evidence as not entirely persuasive, given the contradictions between the earlier and later test results and the alternative, contamination explanation offered by the defense. No view of the evidence was advanced by the prosecutor or by the defense that would have justified mitigating murder, deliberately inflicted on a defenseless child, into a manslaughter conviction. Only a deeply-divided jury could have reached a compromise agreement of that sort.
All these factors increase the likelihood that the Confrontation Clause error with regard to the expert testimony had a substantial and injurious effect on the verdict. Moreover, the court ignored the weight that McCrary’s testimony likely carried, given its strategic presentation. Coming at the end of a long trial, McCrary was *706presented by the prosecution as an essential rebuttal witness to O’Toole. He was her superior in experience and in his past position at the F.B.I. He was a government witness no longer in government, speaking with apparent impartiality on his speciality while in government. Absent some reason not to do so, the jury would trust his honest evaluation of the scene. To establish that he was jealous and resentful of O’Toole, that he was seeking to dissuade her from testifying by sullying her reputation, that he had developed a personal interest in the case, and that he was willing to misrepresent the stance of two governmental agencies to maintain his position was to destroy the impartial, measured image of him presented by the prosecution.
Without that evidence in the record, the prosecution was able to emphasize that McCrary was providing his testimony for free, so he had no motive to lie:
But really interesting, it was fascinating to me, Gregg Me Crary was criticized by the defense for coming into this courtroom and sharing with you his knowledge, his experience, his expertise and not charging us for it; while at the same time, Dr. Leo comes in and testifies and charges for it and he gets hammered for charging for it. What is it? Which is it? You do it for free, you got to be a creep; if you do it for money, you’re still bad, evil.
You know, Gregg Me Crary explained to you when he testified here he does things sometimes pro bono, for no charge, to the extent that he can. He does have to pay for a mortgage and he does have childi-en and a family and obviously expenses. And I would submit to you that Gregg Me Crary when he testified in this case was completely credible, it was an honor to have him in the courtroom, it was an honor to have him share his knowledge with us, and it was a courageous thing for him to do. He is a lifetime F.B.I. agent.
Having promoted him to “lifetime F.B.I. agent,” the prosecution emphasized McCrary’s independence and integrity and lack of personal animus in challenging O’Toole:
Why would he be willing to do that? Because he could see how honestly wrong and inaccurate Ms. O’Toole was.
It was courageous for him to do that. And for him to be criticized for not charging the prosecution, I feel like calling the Attorney General’s Office, calling Attorney General, or calling the Governor and saying, you know, I know we don’t have that much money, but isn’t it nice we got Mr. Me Crary to testify and not charge.
The prosecutor returned to attack O’Toole and to praise McCrary:
I want to stay on the subject of means for a second. Let’s talk about crime scene assessment. There were two witnesses; the defense called. Mary Ellen O’Toole and we called Gregg Me Crary.
And you recall Mr. Me Crary described to you his experience with the F.B.I. and crime scene assessment, had I would say phenomenal, extensive, international, the number of cases he’s been involved with, the number of crime scene assessments he’s been involved with, and the number of times that he has testified.
Ms. O’Toole doesn’t have the benefit of that, of that experience. And I would submit to you that it was reflected in the work done and the conclusions drawn.
If you recall, Gregg Me Crary said for a crime scene assessment to be done, there are four things that you look at in combination. The four things are: all of the evidence, and we’ll talk about that in *707a second, victimology, the crime scene location, and suspect evidence.
And he said something that I did think really kind of stands out; that is, if any one of those is ignored, then the findings become then unreliable. And unfortunately, that is what Ms. O’Toole did. She ignored evidence in the evidence prong, if you will, and she ignored all of the suspect information. As a result, her findings were, honestly, flat-out inaccurate. As we were going through her findings, undoubtedly, as you were seated there, you probably were thinking, well, that is wrong, that is not true. You know organized versus disorganized. When you see the crime scene itself, there is nothing about the crime scene that is organized at all.
And then luckily, Mr. Me Crary can describe for you what an organized crime scene is, was actually involved in doing an organized case. It is not reflected in this case whatsoever.
One of the things that stood out, I think, in Mr. Me Crar/s testimony is simply the reality of it as he is touching on different factors. And, you know, we asked — asked him, well, you know, does luck have anything to do with basically in the commission of a crime? Oh yeah, of course it does. And you know in real life it does; good luck, bad luck, always has some — some play of what is going on. In this case there is a ton of both. Depends on who you are and which way you see it. With regard to, for example, Stephanie, it was all bad luck completely-
In contrast to the prosecutor’s argument, as the California Court of Appeal recognized, “[t]he letter bore directly on McCrary’s credibility and reliability”; it showed that “McCrary had a personal interest in convicting Tuite”; that McCrary’s own letter showed that he “was acting more as an advocate for the prosecution than as a forensic witness.” His own letter “could lead a reasonable jury to question the reliability and validity of his testimony.” Moreover, the letter actually contained unsubstantiated accusations— that the San Diego Sheriffs office and the office of the Attorney General of California viewed O’Toole as “injecting herself into this case” in “an attempt to obstruct justice.” These were accusations made up by McCrary without foundation in fact. These accusations revealed a witness with a passionate animus against O’Toole. Cross examination on the letter would have significantly undermined his credibility to the jury.
Given the lack of evidence tying Tuite to the crime, the problems with the DNA evidence, the jury’s deadlock and compromise verdict, and the weight and strategic position of McCrary’s testimony, this case is one of those “unusual” circumstances in which we find ourselves “in virtual equipóse as to the harmlessness of the error.” O’Neal v. McAninch, 513 U.S. 432, 435, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995). We must treat the error as affecting the verdict, and are compelled to grant the writ. Id.
The state argues that, instead of or in addition to our analysis under Brecht concerning whether the confrontation clause error caused actual prejudice, AED-PA requires us to analyze whether the state appellate court’s determination was unreasonable under Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), again asking if the confrontation clause error was harmless beyond a reasonable doubt. In other words, the state, persistently and passionately, asks us to apply a standard of review less favorable to the state than the one we have employed. As the Supreme Court has recognized, the Brecht standard “obviously *708subsumes” “the more liberal AED-PA/Chapman standard.” Fry v. Pliler, 551 U.S. 112, 120, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007). Therefore, “it certainly makes no sense to require formal application of both tests.” Id. We note that — as Fry predicts — were we to apply the AED-PA/Chapman standard, we would conclude, for the same reasons stated in the text with regard to the Brecht standard, that the state court Chapman harmless error ruling was unreasonable.
For the foregoing reason, the judgment of the district court is REVERSED and REMANDED with instruction to issue a conditional writ of habeas corpus.
This disposition is not appropriate for publication and is not precedent except as provided by 9th Cir. R. 36-3. The panel finds that a published opinion might cast unjust aspersions upon those not before the court.