Opinion by Judge REINHARDT; Dissent by Judge BYBEE
REINHARDT, Circuit Judge:To ... not prepare is the greatest of crimes; to be prepared beforehand for any contingency is the greatest of virtues.
—Sun Tzu, The Art of War 83
(Samuel B. Griffith trans., Oxford University Press 1963)
At the heart of an effective defense is an adequate investigation. Without sufficient investigation, a defense attorney, no matter how intelligent or persuasive in court, renders deficient performance and jeopardizes his client’s defense.
Here, counsel did not meet his basic obligation to his client. Much was riding on his performance in this case: his client, Joshua Richter, was accused of murder, among other charges, and faced life imprisonment without parole. Yet, counsel failed to undertake the most elementary task that a responsible defense attorney would perform in a case of this nature, and consequently provided representation that fell well below a reasonable standard of professional competence. Although it was apparent that an issue critical to the outcome could best be resolved through the presentation of forensic evidence, counsel failed at each stage of the case to consult with a forensic expert of any type and thus failed to conduct the rudimentary investigation necessary in order to (1) decide upon the nature of the defense to be presented, (2) determine before trial what evidence he should offer, (3) prepare in advance how to counter damaging expert testimony that might be introduced by the prosecution, and (4) effectively cross-examine and rebut the prosecution’s expert witnesses once they did testify during the course of the trial. There was in fact no strategic reason for counsel’s failure to do so. As it turned out, these repeated fail*947ures to investigate were prejudicial: available forensic testimony would have contradicted the prosecution’s explanation of the events that transpired and would have strongly supported the defense’s version.
We conclude that, singly and collectively, counsel’s failures rise to the level of ineffective assistance of counsel under the Sixth Amendment. There is nothing novel about our holding. Rather, we arrive at the only reasonable conclusion that can be reached, given the facts of the case and the well-established applicable law. We therefore reverse the district court and remand with directions to grant the writ of habeas corpus.
I.
On December 18, 1995, a California jury convicted Joshua Richter (“Richter”) and Christian Branscombe (“Branscombe”) of the murder of Patrick Klein (“Klein”) and attempted murder of Joshua “Gunner” Johnson (“Johnson”), as well as of burglary and robbery. Both young men were sentenced to life in prison without the possibility of parole. Richter and Branscombe were each twenty at the time of the offense and twenty-one years old at the time of conviction.
Almost precisely a year prior to the date of their conviction, on the evening of December 19, 1994, Richter and Branscombe drove to the Sacramento home of Johnson, a close friend of Richter’s and an acquaintance of Branscombe’s. The two young men had earlier completed their final day of work at a nearby Christmas tree lot, and Richter’s boss had paid him approximately $800 in cash. He had also given him the wages owed to Johnson’s housemate, Tony,1 a friend and co-worker who had failed to show up for work that day. Richter and Branscombe drove to Johnson’s house so that Richter could pay Johnson some money he owed him out of his newly received earnings, deliver Tony’s wages, and buy some “head stash” from Johnson, who was a major marijuana dealer.
When Richter and Branscombe arrived at the house, no one was home. They waited in the driveway, in Richter’s girlfriend’s car, until Johnson returned shortly thereafter, accompanied by Klein and another friend. Johnson did not immediately recognize Richter’s girlfriend’s car, and approached it with his .380 caliber M-12 handgun loaded and drawn. Upon seeing the defendants, Johnson put the weapon away. Johnson, Klein, Richter, and Branscombe went into the house, where they socialized for several hours. While they talked and Johnson, Klein, and Richter smoked marijuana, Branscombe cleaned a .32 caliber handgun that he had recently acquired from Johnson as a means of protection when he worked late nights at the tree lot. Richter and Branscombe left Johnson’s residence shortly after 2:30 a.m. Klein decided to spend the night.
At trial, Richter and the State of California (“the State”) presented dramatically different accounts of the ensuing events. According to the State, after Richter and Branscombe left, Johnson went to sleep in his bedroom and Klein lay down on the couch in the living room. Johnson awoke somewhere between 4:00 and 5:00 in the morning to find Richter and Branscombe in his bedroom, in the act of stealing his gun-safe, which he said was located in his bedroom closet. Branscombe then twice shot Johnson, who fell back wounded onto the bed. Soon thereafter, Johnson heard gunshots coming from the living room. After Richter and Branscombe left the house, Johnson got out of bed, found Klein lying on the living room couch bleeding, and discovered that his gun-safe, his .380 *948caliber M-12, and a hip sack that contained $6000 in cash were all missing.
Richter told a markedly different story. He testified that after leaving Johnson’s house around 2:30, he and Branscombe decided to go back to the Christmas tree lot where they had worked. Their boss had instructed them to clear out their belongings from the trailer on the property before morning, and, not wanting to go to sleep for a few hours only to wake up early to finish the job, he and Branscombe got his pick-up truck and cleared out the trailer, including the belongings of Johnson’s housemate, Tony. The two young men then returned to Johnson’s residence around 4:30 a.m., in order to see whether Tony had come home and to drop off his belongings, along with his pay. Branscombe also intended to return the .32 caliber handgun to Johnson if he was still awake. Richter stayed in his truck in Johnson’s driveway, smoking a cigarette, while Branscombe knocked on the door and was let into the house by Klein.
Shortly thereafter, Richter heard gunshots. He headed toward the house and heard yelling and more gunshots as he approached the front door. He found Klein lying in the doorway to Johnson’s bedroom, saw Johnson lying twisted on the bed, and found Branscombe “totally freaked out,” standing in the middle of the bedroom holding a firearm, and shouting, “They tried to kill me.” According to the defense, Johnson, who had earlier been drinking and smoking marijuana, had drawn his .380 caliber M-12 and fired at Branscombe when he entered the room. After firing one bullet, the gun — which Johnson had attempted to modify from a semi-automatic to a fully automatic — jammed, and Johnson threw it down. He then took out a second handgun — apparently a .22 caliber — that he kept under his mattress and shot at Branscombe, but hit Klein instead. Branscombe then fired three shots with the .32 caliber handgun and hit both Klein and Johnson.2 After Richter arrived inside the house, Branscombe picked up Johnson’s M-12 from the floor and ran outside to try to start the truck. Richter panicked and ran back out to the truck as well, and the two young men drove away.
Sometime after the shootings, Johnson made a 911 call to the police.3 Johnson testified that, before the police arrived six minutes later, he made a phone call to his girlfriend’s father and took two trips through the house and into the yard to hide his marijuana plants. When the police arrived, they encountered a “hysterical” Johnson, who had blood on his cheeks, shirt, hands and right shoulder. The police found Klein lying on top of a sleeping bag on the living room couch, near death.
The initial investigation of the scene uncovered two spent .32 casings in the bedroom, as well as blood on the bed where Johnson said he had been shot. There was a large pool of blood in the doorway to Johnson’s bedroom, where Richter testified that Klein had been shot. The pool had been disturbed by a foot stomp or *949some other external force. Investigators determined that Klein had been shot twice, by a .22 and a .32 caliber bullet. They found a spent .32 and a spent .22 casing in the living room near the couch where Klein was lying, although the prosecution’s expert testified at trial that it is impossible to determine with certainty how casings end up where they do.
A significant amount of blood was found throughout the house. Investigators photographed and videotaped the scene, collected a few blood swabs, and took some fingerprints, but primarily determined that the scene appeared to be consistent with Johnson’s account of what had happened and did not pursue an in-depth forensic investigation. The police took a sample of blood from the door molding above the pool of blood in the doorway, but not from the pool itself. They did not perform any blood typing or blood spatter analysis and did not attempt to discover whose blood formed the pool in the doorway until the middle of the trial, nearly a year later, in the belief that the case was sufficient without such evidence.
Richter testified that, upon fleeing the scene in a panic, he stopped briefly at his girlfriend’s house, while Branscombe waited outside in the truck, and after a conversation about Johnson and Klein drove back to Johnson’s house to check on his friends. He and Branscombe saw two sheriffs cars in the driveway, however, and decided not to stop. Instead, they drove out to a remote area, the Yolo bypass, where Richter, Branscombe, Johnson, and their friends frequently went to hang out and shoot their guns. There, they threw away the .380- and .32-caliber weapons and discussed what to do next. They slept for awhile. When they awakened, they discovered that their truck was stuck in the mud, and after obtaining assistance to get it moving, they eventually returned to town and were arrested shortly thereafter, a day and a half after the shooting.
Following his arrest, Richter, his attorney, and a police detective went out to the Yolo bypass in an attempt to recover the discarded weapons. They could not locate the firearms, but did find a $100 bill in the vicinity. Later, Richter’s lawyer returned to a nearby area and managed to recover the .380 M-12, which he then turned over to the police. Investigators also found a spent .380 casing in Richter’s truck which they determined had been ejected from Johnson’s M-12. Richter testified that, driving to the Yolo bypass, Branscombe had dislodged a shell that had been jammed in the receiver by jerking on the lever. The weapons used to kill Klein and injure Johnson were never found.
In the aftermath of the shooting, Johnson told the police that his missing gun-safe would be at Richter’s house. Sure enough, while searching Richter’s residence, investigators found Johnson’s gun-safe lying on a pile of assorted items in Richter’s garage. According to the defense, the safe had been at Richter’s house with Johnson’s permission all along; Johnson frequently stored some of his belongings, including scuba gear and firearms, with Richter. Johnson admitted that he had stored the safe at Richter’s residence but stated that he and a friend had moved it to Johnson’s house a couple of weeks before the shooting. Although Richter denied owning a handgun, investigators also found .22 caliber cartridges and a loaded magazine in Richter’s garage. The cartridges were consistent with the .22 caliber bullet found in Klein’s body, as well as with the casing found in Johnson’s living room. In closing argument, Richter’s lawyer argued that the cartridges were Johnson’s, and that Johnson was storing the cartridges there along with the rest of his belongings.
*950Johnson owned a number of firearms, including two AK-47 assault rifles and a .22 caliber shotgun. Neither Richter nor Johnson admitted to owning a .22 caliber handgun, and Richter denied owning any handguns at all. He testified at trial, however, that Johnson owned, and had previously shown him, a .22 caliber pistol.
After the trial began, the State conducted two significant forensic tests that it had not previously performed and presented related testimony at trial. First, the State asked Detective Robert Bell to perform a blood spatter analysis of the crime scene, based on photographs that had been taken shortly after the shootings. Bell testified as an expert in blood spatter that it was unlikely that Klein could have been killed in the doorway and carried to the couch, primarily because of blood flow patterns on his head and the lack of smearing around the blood pool.4 This suggested that Johnson would have had to have lifted Klein straight up. Bell acknowledged, however, that although there was some evidence of high velocity blood spatter near the couch, suggesting a shooting in that location, there was no visible blood spatter on the sleeping bag or on the couch itself, and no blood spatter that corresponded with two shots being fired. Second, the State asked Jill Spriggs, a serologist, to determine the blood type of a sample collected from spatter four inches above the pool of blood in the bedroom doorway, and to compare that sample to Johnson’s and Klein’s blood. Spriggs testified, based on the PGM subtyping of the samples, that none of the blood was Klein’s. The defense offered no expert testimony in response.
After a trial lasting over three weeks, a California Superior Court jury found Richter and Branscombe guilty of all charges. Following their convictions, Richter and Branscombe appealed to the California Court of Appeal, which affirmed the judgments of the trial court. Richter and Branscombe next filed a petition for review in the California Supreme Court, which was summarily denied, and then sought writs of habeas corpus in the California Supreme Court, which were also summarily denied. Richter and Branscombe timely petitioned for writs of habeas corpus in the United States District Court for the Eastern District of California, which rejected the petitions. They then sought certificates of appealability from this court, which we granted. A panel of our court affirmed the district court’s denial of Richter’s petition. We now sit en banc to rehear the appeal, having granted his petition for rehearing en banc.
II.
Under the Antiterrorism and Effective Death Penalty Act of 1996 (“AED-PA”), a federal court may grant habeas relief only if the state court decision “was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” or “was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” 28 U.S.C. § 2254(d). A state court decision is “contrary to” clearly established federal law “if the state court ... arrives at a conclusion opposite *951to that [of the Supreme Court] on a question of law,” or decides the case differently than the Supreme Court on a set of materially indistinguishable facts. Williams v. Taylor, 529 U.S. 362, 405, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). A state court decision is an “unreasonable application of’ clearly established federal law “if the state court identified] the correct governing legal rule ... but unreasonably applied] it to the facts” at hand. Id. at 407, 120 S.Ct. 1495. For a state court decision to be an unreasonable application of “clearly established Federal law” under § 2254(d)(1), the Supreme Court’s prior decisions must provide a “governing legal principle or principles” applicable to the issue before the state court. Lockyer v. Andrade, 538 U.S. 63, 71, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003). Circuit precedent is relevant under AEDPA when it illuminates whether a state court unreasonably applied a general legal standard announced by the Supreme Court. See Crater v. Galaza, 491 F.3d 1119, 1126 n. 8 (9th Cir.2007).
When reviewing a state court’s summary denial of a habeas petition, we “look through” the summary disposition to the last reasoned state court decision. Plascencia v. Alameida, 467 F.3d 1190, 1198 (9th Cir.2006). Where, as here, no state court has explained its reasoning on a particular claim, we conduct an “independent review of the record to determine whether the state court’s decision was objectively unreasonable.” Sass v. Cal. Bd. of Prison Terms, 461 F.3d 1123, 1127 (9th Cir.2006).5 We review a district court’s decision to grant or deny a writ of habeas corpus de novo, Lewis v. Mayle, 391 F.3d 989, 995 (9th Cir.2004), and the district court’s findings of fact for clear error, Bonin v. Calderon, 59 F.3d 815, 823 (9th Cir.1995).
As we conduct our independent review of the record, we apply the Supreme Court law on ineffective assistance of counsel claims — in particular, that “counsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Strickland v. Washington, 466 U.S. 668, 691, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984); see also id. at 688, 104 S.Ct. 2052 (“More specific guidelines are not appropriate. The Sixth Amendment refers simply to ‘counsel,’ not specifying particular requirements of effective.... The proper measure of attorney performance remains simply reasonableness under prevailing professional norms.”). Under AEDPA, we apply Strickland to each individual case, irrespective of whether the precise fact pattern at issue has been considered previously by the Supreme Court. See Williams, 529 U.S. at 391, 120 S.Ct. 1495 (“That the Strickland test ‘of necessity requires a case-by-case examination of the evidence’ ... obviates neither the clarity of the rule nor the extent to which the rule must be seen as ‘established’ by this Court.”). “[B]ecause the Strickland standard is a general standard, a state court has even more latitude to reasonably determine that a defendant has not satisfied that standard.” Knowles v. Mirzayance, - U.S. -, 129 S.Ct. 1411, 1420, 173 *952L.Ed.2d 251 (2009) (citing Yarborough v. Alvarado, 541 U.S. 652, 664, 124 S.Ct. 2140, 158 L.Ed.2d 938 (2004)). We do not, however, afford the state courts a blank check to determine, at their whim, whether an attorney’s conduct was reasonable or unreasonable:
That the standard is stated in general terms does not mean the application was reasonable. AlEDPA does not require state and federal courts to wait for some nearly identical factual pattern before a legal rule must be applied. Nor does AEDPA prohibit a federal court from finding an application of a principle unreasonable when it involves a set of facts different from those of the case in which the principle was announced. The statute recognizes, to the contrary, that even a general standard may be applied in an unreasonable manner.
Panetti v. Quarterman, 551 U.S. 930, 127 S.Ct. 2842, 2858, 168 L.Ed.2d 662 (2007) (citations and internal quotation marks omitted). Where counsel’s failure to investigate was both objectively unreasonable and prejudicial, and where the state court acted unreasonably in finding to the contrary, we will grant a petition for habeas corpus.
III.
Richter challenges his conviction primarily on the ground that he received ineffective assistance of counsel and thus was deprived of his rights under the Sixth Amendment. Strickland sets forth the well-established two-part standard for ineffective assistance of counsel claims. “First, the defendant must show that counsel’s performance was deficient.” Strickland, 466 U.S. at 687, 104 S.Ct. 2052. To be “deficient,” counsel’s performance at trial must be objectively unreasonable — it must be “outside the wide range of professionally competent assistance.” Id. at 688, 690, 104 S.Ct. 2052. We “evaluate the conduct from counsel’s perspective at the time” to “eliminate the distorting effects of hindsight,” and we are “highly deferential” in judging counsel’s performance, affording counsel a strong presumption of adequacy. Id. at 689, 104 S.Ct. 2052.
Next, “the defendant must show that the deficient performance prejudiced the defense.” Id. at 687, 104 S.Ct. 2052. We find prejudice if “there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Id. at 694, 104 S.Ct. 2052. “It is clear ... that [Richter] need not show that [counsel’s] deficient conduct more likely than not altered the outcome in the case. This ‘preponderance’ standard was explicitly rejected in Strickland.” Sanders v. Ratelle, 21 F.3d 1446, 1461(9th Cir.1994) (citing Strickland, 466 U.S. at 693, 104 S.Ct. 2052) (emphasis in original).
Finally, because AEDPA applies, if we determine that the defendant did not receive effective assistance of counsel, we must decide whether the state court’s decision that he did constitutes an unreasonable application of clearly established Supreme Court law. 28 U.S.C. § 2254(d).
Richter contends that his counsel rendered ineffective assistance under Strickland by (1) failing to investigate and present expert testimony on blood spatter, serology, and pathology in order to explain the source of the pool of blood by the doorway; (2) failing to investigate the accessibility of the floorboard that the state investigator dropped into the crawl space below the house, and to present expert ballistics testimony that the bullet hole was made by a .380-caliber, rather than a .22-caliber, bullet; (3) failing to investigate and present expert testimony on ballistics to show that Johnson’s .380 M-12 might have jammed; and (4) failing to *953investigate and present testimony from lay witnesses. On state and federal habeas review, Richter submitted declarations from experts in blood spatter, serology, pathology, and ballistics, as well as from lay witnesses, to support his ineffective assistance claim.
We consider here only Richter’s first argument — counsel’s failure to investigate and present expert testimony on blood evidence — and conclude that he is correct. Even affording counsel the “strong presumption that [his] conduct falls within the wide range of reasonable professional assistance,” Strickland, 466 U.S. at 689, 104 S.Ct. 2052, we hold that there was no reasonable basis for his failure, before deciding on the nature of his defense, as well as while preparing for trial and during its course, to investigate the availability of forensic evidence that might corroborate Richter’s testimony regarding the source of the pool of blood by the bedroom doorway. We so conclude, fully aware, as counsel should have been, that establishing the source, or at the least raising substantial doubts as to the prosecution’s evidence on that point, was critical to the success of Richter’s defense. Finally, we conclude that Richter was prejudiced by counsel’s deficient representation in this respect, and thus did not receive effective assistance of counsel. The state court holding to the contrary constitutes an objectively unreasonable application of Strickland. We therefore reverse the district court’s denial of the writ of habeas corpus.
A. Deficient Performance
. As we have explained, a central dispute between the prosecution and the defense was Klein’s location when he was shot. Investigation of the crime scene revealed a pool of blood in the doorway between the bedroom and the living room. The State claimed that Klein was shot as he lay on the couch in the living room and that the pool of blood in the doorway was Johnson’s: that the pool formed as Johnson stood, dripping blood on the floor, waiting for the police to arrive. Richter, however, although acknowledging that some of the blood might have been Johnson’s, insisted that it was at least partly Klein’s, as Klein had been shot near the doorway in a shootout and not in cold blood on the couch. This factual dispute was, indeed, the single most critical issue in the case, at least from the standpoint of the defense.6 If the jury was persuaded beyond a reasonable doubt that Klein was killed on the couch, the defense would have no hope of succeeding. If, however, the jury believed that there was a reasonable possibility that Klein had been shot in the doorway, as Richter testified, Johnson’s version of the events — and the prosecution’s case — would be severely undermined, and the shootout version offered by the defense would in all likelihood preclude a verdict of guilty. Under the circumstances, any competent defense counsel would have immediately recognized the critical importance of investigating the source of the pool of blood, and of attempting to obtain forensic evidence that would support Richter’s account of the fa*954tal evening’s occurrences. In fact, knowing that forensic evidence could provide support for either version of the events, and could corroborate or discredit his client’s account, competent counsel would have consulted with an expert in blood evidence before settling upon his theory of the ease and his decision to attempt, unsuccessfully, to turn it into a pure credibility contest. This is indeed precisely what Strickland requires.7
Despite the centrality of the source of the pool of blood to Richter’s defense, counsel consulted no forensic expert as to what type of expert testimony or evidence might be available, conducted no forensic investigation whatsoever with respect to the blood pool, and in the end, as a result, offered no expert testimony to explain its source. Nor did he seek to obtain a forensic consultant who could assist him in evaluating the testimony of the prosecution’s experts or guide him in developing effective cross-examination of those witnesses. Other than a conversation in the hallway with the prosecution’s serology expert in the midst of trial, at no point either before or during the trial did defense counsel consult any expert on blood evidence in an attempt to determine the nature of the defense that could be presented or to discover how expert testimony might be used to bolster his client’s case or undermine the prosecution’s. He made no effort to find or develop expert testimony on such important questions as the blood type of the person or persons whose blood formed the pool in the doorway; whether Johnson’s injuries were severe enough to create such a large pool of blood; and, most important, whether the blood spatter around the pool of blood indicated how or from whom the blood had been deposited in the pool.
Counsel’s failure to consult any forensic expert constituted a threefold abrogation of his duty under Strickland. Counsel’s first failing lay in his inadequate investigation prior to settling on a trial strategy; his second, in failing to conduct the necessary investigation before trial to determine what forensic evidence might be available for him to introduce or how expert testimony might support the theory he had chosen; and his third, in failing to consult experts who could assist during trial when the State, foreseeably, introduced damaging expert testimony regarding the blood evidence in the case.8
*955l.
Strickland obligates defense attorneys to make reasonable investigations before settling on a trial strategy or, at the least, to conduct sufficient inquiries to make an informed decision about whether further investigation is needed. See Wiggins v. Smith, 539 U.S. 510, 525, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) (holding counsel must make an “informed choice” among possible defenses); Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052 (“[Strategic choices made after less than complete investigation are reasonable precisely to the extent that reasonable professional judgments support the limitations on investigation. In other words, counsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.”); see also Jennings v. Woodford, 290 F.3d 1006, 1014 (9th Cir.2002) (“[AJttorneys have considerable latitude to make strategic decisions about what investigations to conduct once they have gathered sufficient evidence upon tvhich to base their tactical choices.” (emphasis in original)). Until a reasonable investigation is conducted, counsel is not in a position to make critical strategic decisions or settle on a trial strategy — certainly including the decision to rest on his client’s testimony irrespective of the forensic facts. We have repeatedly held that “[a]n uninformed strategy is not a reasoned strategy,” Correll v. Ryan, 539 F.3d 938, 949(9th Cir.2008), cert. denied sub nom. Schriro v. Correll, — U.S. -, 129 S.Ct. 903, 173 L.Ed.2d 108 (2009), and we have followed the Supreme Court’s holding that “the traditional deference owed to the strategic judgments of counsel is not justified where there was not an adequate investigation ‘supporting those judgments,’ ” id. at 948-49 (quoting Wiggins, 539 U.S. at 521, 123 S.Ct. 2527).
Counsel’s strategy, as explained in his deposition, was to pit his client’s credibility against Johnson’s and to attack the evidentiary gaps in the police investigation of the crime scene. In light of Strickland, it was not reasonable to settle upon this strategy, as counsel did, without investigating whether it was a sound one. Counsel made no inquiry into what expert testimony might be available to support, or to contradict, his client’s version of what occurred; by what means the perceived evidentiary gaps in the State’s investigation might be filled at or before trial; how any testing to fill those gaps would affect either his client’s or Johnson’s credibility; or how he could respond to, or minimize *956the effect of, the damaging testimony that might emerge from forensic blood testing. Counsel’s decision not to consult any forensic expert in blood evidence before settling upon a defense strategy that excluded the use of expert testimony, when the defense so clearly depended upon the source of the pool of blood in the doorway, indubitably fails to meet Strickland’s standard of care. See Strickland, 466 U.S. at 690-91. Indeed, this failure, in itself sufficient to require a holding of deficient performance, marked only the beginning of counsel’s inexplicable series of ineffective acts and omissions.
2.
Having chosen his trial strategy without conducting an adequate investigation, counsel then continued to perform deficiently by failing to consult an expert on blood evidence while preparing for trial. A reasonably competent attorney would have made inquiries of such an expert regarding what evidence might be available that would strengthen the chances that his “he said-he said” strategy would succeed or might cause him to broaden that strategy were he to locate an expert whose testimony would support his client’s statements. For some unexplained reason, counsel again failed to consult any blood expert who could provide advice as to what forensic evidence might be adduced to support Richter’s version of the events' — or what forensic evidence the State might adduce, either prior to or during trial, that might seriously undermine the defense, and how that evidence might be countered.9 Instead, counsel willfully pursued his credibility theory, determined to rely on Richter’s testimony alone, knowing all the while that his client was not a witness likely to appear particularly credible to a jury. In doing so, counsel not only neglected the opportunity to discover the availability of objective corroborating forensic evidence in support of his client’s version of the facts, but also rendered the defense vulnerable to impeaching expert testimony that he was not prepared to rebut.
Counsel is obligated to conduct a reasonable investigation in order to present the most persuasive case that he can. Counsel must conduct a pretrial investigation into the availability of independent, objective sources to support the part of his client’s testimony that he knows or can reasonably expect will be challenged, and subsequently to present to the jury any evidence he finds that tends to show his client’s innocence, tends to undermine the prosecution’s case, or raises a reasonable doubt as to his client’s guilt, unless he *957makes an informed, strategic decision that the risks of introducing such evidence outweigh its benefit to the defense. See, e.g., Hendricks v. Calderon, 70 F.3d 1032, 1040 (9th Cir.1995) (“An attorney ‘must provide factual support for the defense where such corroboration is available.’ ... Failure to pursue such corroborating evidence with an adequate pretrial investigation may establish constitutionally deficient performance.” (quoting United States v. Tucker, 716 F.2d 576, 594(9th Cir.1983)) (internal alterations omitted) (emphasis- added)); Jones v. Wood, 114 F.3d 1002, 1013 (9th Cir.1997) (granting “an evidentiary hearing on the issues of his attorney’s failure to investigate ... before trial and his lawyer’s failure to test the evidence”). In some circumstances, such corroboration may take the form of eyewitness testimony or character witnesses. In this case, a forensic explanation of the source of the pool of blood at the bedroom doorway could have supported the defendant’s account. Counsel made no effort to obtain any evidence or expert assistance that would provide such an explanation. “This court has repeatedly held that ‘a lawyer who fails adequately to investigate and introduce evidence that demonstrates his client’s factual innocence, or that raises sufficient doubt as to that question to undermine confidence in the verdict, renders deficient performance.’ ” Duncan v. Ornoski, 528 F.3d 1222, 1234 (9th Cir.2008) (quoting Hart v. Gomez, 174 F.3d 1067, 1070 (9th Cir.1999)) (internal alterations omitted), cert. denied sub nom. Duncan v. Ayers, — U.S. —, 129 S.Ct. 1614, 173 L.Ed.2d 1001 (2009). The obligation to investigate only grows more imperative where the evidence at issue is the “only forensic evidence” that could reasonably support the defense theory. Duncan, 528 F.3d at 1236 (“[T]he central role that the potentially exculpatory blood evidence could have played in [the defendant’s] defense increased [counsel’s] duty to seek the assistance of an expert.”).
We therefore strongly reject the district judge’s conclusion that counsel’s investigation was understandably hamstrung by the element of surprise — that his failure to consult experts was reasonable because he did not learn until mid-trial that the State would call a blood spatter or serology expert, and that he was at this point struggling to deal with significant time constraints and the limitations of his co-counsel. See Richter v. Hickman, No. S-01-CV-0643-JKS, 2006 WL 769199, at *8 (E.D.Cal. Mar.24, 2006). This argument truly misperceives the duty of counsel to investigate. The unfortunately harried circumstances in which counsel found himself in the midst of trial, a year after the events in question took place, do not excuse his failure to investigate prior to trial- — -indeed prior to determining the defense strategy — whether expert testimony would corroborate or undermine his client’s explanation of how the pool of blood came to lie on the floor by the bedroom door. Whether or not the prosecution presented any expert testimony about the pool of blood, defense counsel had an obligation to investigate and, should his investigation prove successful, introduce expert testimony regarding its source in order to provide crucial — indeed, potentially outcome-determinative — corroboration for his client’s story.10
*958Finally, we reject the district court’s conclusion that counsel’s failure to consult experts in preparation for trial was reasonable because counsel thought that the case was, at bottom, a credibility contest. Richter, 2006 WL 769199, at *7-8. Precisely the opposite is true. Where the defense strategy is to win a credibility contest, the importance of corroborating the accused’s testimony with physical evidence is paramount. Leaving the jurors to believe or disbelieve defendants solely on the basis of their own testimony, without supporting evidence, where such evidence could be obtained with diligent investigation, is objectively unreasonable. See, e.g., Riley v. Payne, 352 F.3d 1313, 1319-20 (9th Cir.2003) (“[W]ithout any corroborating witnesses, [Riley’s] bare testimony left him without any effective defense.” (internal quotation marks omitted)); Hart, 174 F.3d at 1070 (“Defense counsel failed to investigate or introduce into evidence the records that fully corroborated [the witness’s] statements. Thus, the jury was left to decide, without benefit of supporting or corroborative evidence, whether [the witness’s] testimony was truthful and accurate, or whether it was unreliable or offered simply in an effort to assist a former lover.”); Lindstadt v. Keane, 239 F.3d 191, 203 (2d Cir.2001) (“[I]n a credibility contest, the testimony of neutral, disinterested witnesses is exceedingly important.” (quoting Williams v. Washington, 59 F.3d 673, 681-82 (7th Cir.1995))). Relying on a pure credibility contest without seeking to investigate the availability of corroborating evidence is unreasonable regardless of the witness; it is at its most unreasonable, however, when the witness is the defendant himself — especially when he is an admitted drug offender who stands accused of committing murder and attempted murder in the course of a burglary.
If counsel’s decision not to consult experts in preparation for trial had been a “strategic judgment,” it would have been an unreasonable one, as there was no adequate investigation to support it. See Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052. But, even more fundamental, counsel’s performance in this respect was not “strategic” at all. There was absolutely no reason not to consult a blood expert and counsel never suggested that there was. In his deposition, counsel was unable to provide any reasoned explanation for failing to consult forensic experts or to seek *959expert testimony in order to corroborate his client’s testimony or prepare to rebut the prosecution’s case.
Certainly, he did not suggest that any strategy explained that failure. When counsel offers no strategic reason for failing to perform what would otherwise constitute the duty of a reasonably competent counsel, we may not invent such a strategy by engaging in “a post hoc rationalization of counsel’s conduct” in lieu of relying on “an accurate description of [counsel’s] deliberations prior to [trial],” Wiggins, 539 U.S. at 526-27, 123 S.Ct. 2527.11 Counsel’s chosen trial strategy, as he explained it, was to convince the jury of an alternative set of facts to those presented by the prosecution — and to rely on his client’s testimony to carry the day in a pure credibility contest. The defense testimony would be that Klein was killed in the doorway, in a shootout, and the alleged robbery of the gun-safe never took place. Counsel provided no reason why he did not consult experts to bolster that story. Although he had an obligation to adequately investigate so as to support the strategy that he had adopted, he failed to do so. It is perfectly clear, and indeed it is the only reasonable conclusion under Strickland, that his performance was constitutionally deficient.12
3.
Counsel’s failure to conduct the necessary pretrial investigation was compounded finally by his failure to consult experts as the trial unfolded. After counsel presented his opening argument laying out the defense strategy and pointing out the gaps in the prosecution’s case, including its incomplete forensic investigation, the State acted quickly to fill some of these gaps and to determine whether it could prove its case or at least disprove the defense theory through the use of forensic evidence. The State then introduced “surprise” testimony from two experts on blood evidence: a blood spatter expert and a serology expert. The thrust of their testimony was that the sole source of the blood in the pool was Johnson and that none of the blood was Klein’s.
The prosecution’s blood spatter expert, Detective Robert Bell, testified that it was highly unlikely that Klein could have been killed in the doorway and carried to the couch, primarily because of blood flow patterns on his head and high velocity blood spatter near the couch. After this damag*960ing testimony, defense counsel considered consulting a blood spatter expert of his own. He discussed with Richter whether to call such an expert, and opted to wait until after cross-examination to decide. Counsel researched and gathered names of potential blood spatter experts, but does not remember contacting anyone. He failed to request a continuance in order to obtain expert testimony to counter the prosecution’s evidence, or even to obtain assistance in preparing for cross-examination of the prosecution’s witnesses. Indeed, he did not consult an expert at any point after Bell testified, even though Bell was recalled twice, including once on rebuttal, during a trial that lasted nearly a month.
The prosecution also called a serology expert, Jill Spriggs, who tested the blood type of a sample of blood on the door molding above the blood pool in the doorway and testified she had “no reservations. [Klein] is excluded as a possible donor of that sample.” Counsel again considered the possibility of consulting an expert, this time on blood typing, but he did not do so. After speaking with the State’s serology expert in advance of her testimony during trial, counsel was prompted to do some reading on his own, and again to compile a list of potential experts, but he could not recall actually contacting anyone. When the State announced its intention to call Spriggs as an expert witness, counsel raised an objection. According to counsel, the judge then “said that [counsel] may need a continuance and that [he] may need to call an expert now [him]self.” Counsel offered conflicting testimony as to whether he did, in fact, request a continuance this time. It is clear, however, that again he did not consult any blood expert whatsoever.13
In the end, counsel again failed to consult a single expert in the field of blood evidence, this time to help him prepare for his cross-examination of the State’s two expert witnesses once he knew that they would testify. He, further, made no effort to consult forensic experts so as to present testimony in rebuttal.14 When the need *961for expert advice was abundantly clear, counsel still failed to take reasonable steps mid-trial to remedy his deficient pretrial investigation and preparation. That failure was itself deficient, and the state court’s decision to the contrary was objectively unreasonable.
4.
In short, the Supreme Court has held that the touchstone of our inquiry must be the reasonableness of counsel’s conduct. The failure here to consult forensic experts regarding critical issues in the case was unreasonable and constitutes deficient performance. Other circuits, also under deferential AEDPA review, have come to similar conclusions. See Draughon v. Dretke, 427 F.3d 286, 296 (5th Cir.2005) (finding deficient performance under AEDPA where “the failure to investigate the forensics of the fatal bullet deprived [the defendant] of a substantial argument, and set up an unchallenged factual predicate for the State’s main argument that [the defendant] intended to.... [The defendant] became the sole source of evidence available to counter the prosecution’s theory.”); Gersten v. Senkowski, 426 F.3d 588, 607-08 (2d Cir.2005) (finding deficient performance under AEDPA where “defense counsel failed to call as a witness, or even to consult in preparation for trial and cross-examination of the prosecution’s witnesses, any medical expert on child sexual abuse”). Here, the source of the pool of blood was critical. If the blood came in whole or in part from Klein, defendant’s version of the events would be strongly corroborated. If it came from Johnson alone, the prosecution’s case would be strengthened immeasurably. We conclude that counsel’s failure to consult any forensic experts regarding how to establish that source (1) before settling on his defense, (2) when preparing for trial, or (3) during the course of trial when he was surprised by the prosecution’s expert testimony cannot be deemed to be reasonable under any standard, and that the state court’s failure to so determine unquestionably constituted an objectively unreasonable application of Strickland.
B. Prejudice
Richter must show more, however, than his counsel’s deficient performance. He must also show that he was prejudiced by that performance — that “there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. “A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Id. Applying this standard, we hold that counsel’s deficient performance was prejudicial, and that the state court’s failure to so conclude was an objectively unreasonable application of Strickland.
Counsel’s pretrial failure to consult experts who could have interpreted the blood evidence at the crime scene and advised him of its significance had direct, and damaging, consequences. Counsel developed a strategy for trial to treat the case as a pure credibility contest, to call only the defendant as a witness regarding the shooting, and to criticize the state’s incomplete investigation of the crime scene, without any knowledge whether forensic evidence, if introduced, would corroborate or contradict his Ghent’s testimony. He assumed that the prosecution would not offer any expert forensic evidence to sup*962port the testimony of its principal witness and did not prepare for the possibility that it would. In the end, the prosecution did introduce damaging expert testimony on blood spatter and blood typing that undermined Richter’s account and the defense theory as a whole. At this point, counsel was unequipped to respond to it. These negative consequences, in turn, were exacerbated by counsel’s failure during the trial to make reasonable efforts to compensate for his inadequate pretrial investigation; even at this point, he did not consult a single expert who could advise him whether exculpatory evidence existed that he might present or how he might successfully cross-examine the prosecution’s experts or counter their testimony. Critically, as a result, counsel failed to introduce any expert testimony — testimony that could likely have raised reasonable doubts in the minds of the jurors.
Counsel’s failure to present expert testimony at trial did not go unnoticed. In closing, he emphasized the significance of the pool of blood in the doorway, citing it as evidence of the location at which Klein had been killed. The prosecutor, in response, drove home the fact that the defense had presented no evidence to support its version of the events, other than Richter’s testimony. He then ridiculed defense counsel’s failure to present any evidence to support his account of the source of the pool of blood, and especially his failure to call any expert witnesses:
Bob Bell, [the state blood spatter expert,] he’s 22 years as a blood spatter expert, all that stuff means nothing. Hey, [defense counsel] says, the blood be here. Bob Bell, hey he’s wrong, trust me. 1 am not going to go get an expert. I am not going to bring somebody in here to tell you because I don’t need to do it. I will just do it in closing argument. I will just say it. If you are willing to believe me, hey, that will work.
I am not going to worry about Jill Spriggs[, the state serology expert,] because, hey, her seven years as a biochemist and a criminalist, and the fact that she went to college to learn this stuff doesn’t mean anything, because I am a lawyer. I went to California, Berkeley, in 1975 and U.O.P. and played ball in Mexico, and I know more than Jill Spriggs. But I am not going to pay and bring in an expert to show you. I am not going to put anyone up here to tell you that. I will just tell you that. I am a lawyer. I can do it.
(Emphasis added). Because expert testimony that contradicted that offered by the prosecution was in fact available, and strongly persuasive, the harm caused by counsel’s failure to make any effort to obtain it is readily apparent, and devastating.
Had counsel bothered to conduct the requisite investigation, his efforts would have been highly productive. On state and federal habeas review, Richter’s new counsel conducted the investigation that trial counsel had failed to perform, and with significant effect. Habeas counsel submitted declarations from four experts in blood evidence, all of whom would have offered testimony that supported the defense theory of the case. Taken together, they would have established reasonable doubt.
Most significant, and most damaging to the prosecution, Ken Moses, an expert in blood spatter analysis with over thirty years of experience in the field, would have directly refuted the prosecution’s explanation as to the source of the critical pool of blood. Specifically, he would have offered his opinion that “[t]he lack of a large number of satellite dropletts [sic] surrounding the pool eliminates the prosecution’s theory that Mr. Johnson was standing into [sic] the doorway dripping into the pool below” (emphasis added). Moses would have *963been a most persuasive witness. He had established the Crime Scene Investigations Unit of the San Francisco Crime Laboratory in 1983 and, over the course of his career “[a]s an Inspector-Sergeant and senior crime scene investigator with the San Francisco Police Department ... investigated hundreds of violent crimes ... in which [he] applied the science of bloodstain pattern interpretation,” and had on numerous occasions been qualified as an expert.
Additionally, James Thornton and Brian Wraxall, experts in blood typing, would have testified to the scientific possibility that the pool contained some of Klein’s blood, because the relative intensity of the 2+ and 1+ bands found in the blood sample could not exclude the possibility that the blood pool contained a mixture of Johnson’s (subtype 2+ 1+) and Klein’s (subtype 1+) blood. Finally, Dr. Paul Herrmann, a pathology expert, would have opined that,
[g]iven Mr. Johnson’s testimony and the amount of blood likely to result from the wounds which he received, and the short time between Mr. Johnson being shot and Detective Wright making his observations of Mr. Johnson, it is highly unlikely that the blood pool found in the doorway between the bedroom and living room was caused by Mr. Johnson’s wounds.
No such expert testimony was presented to the jury.15
1.
Although the failure to consult any forensic expert with respect to blood evidence constituted deficient performance, the investigation conducted during the habeas proceedings reveals that the primary source of prejudice lay more narrowly in counsel’s failure to consult, and subsequently to call, an expert in blood spatter.
To explain the significance of the blood spatter testimony, it is necessary to reiterate the importance of the pool of blood— that it was, in fact, the linchpin of the defense. If Klein was killed while lying on the couch in the living room, there was no possibility that Richter’s account was correct. If, in contrast, Klein was killed in the doorway to the bedroom, Johnson’s account of the events in question could not possibly be true. Indeed, if any portion of the blood in the pool in the doorway came from Klein, he could not have been shot, as Johnson claimed, while he was asleep in another location. Accordingly, had defense counsel been able to raise a reasonable doubt in the minds of the jurors as to whether Klein was a source, in any part, of the pool of blood, there was more than a reasonable chance that “the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. The introduction of forensic evidence that had the effect of showing that Johnson’s version of the events was false, and that was consistent with Richter’s explanation, would therefore have been invaluable to the defense.
Had counsel conducted an adequate pretrial investigation and, as a result, offered Moses’ blood spatter testimony at trial, the jury would have been presented with testimony from two expert witnesses on blood spatter, one for the prosecution (Bell) and one for the defense (Moses), each of whom would have contradicted key pieces of the theory presented by the opposing side. On the one hand, the spatter evidence *964presented by Richter’s habeas counsel would have conclusively refuted the prosecution’s explanation of the source of the all important pool of blood: that it contained only Johnson’s blood, which dripped to the floor as he stood awaiting the police’s arrival. Indeed, Moses’ expert testimony that the pool could not have been formed in that manner remains undisputed. On the other hand, the prosecution’s unrebutted testimony about high velocity spatter near the couch and the blood flow on Klein’s face would have supported the conclusion that Klein was killed on the couch. Each expert’s testimony was material to resolving the central dispute in the case: Klein’s physical location when he was shot and, by extension, the circumstances of Klein’s death and the credibility of the key participants.
In the absence of Moses’ testimony, the only expert explanation of the evidence that the jury heard was consistent with the prosecution’s version of the case. Had the jury also heard expert testimony from Moses, it would have been faced with highly credible forensic evidence from an experienced former law enforcement officer that directly contradicted the prosecution’s version and raised substantial doubt as to the testimony regarding the circumstances of Klein’s death offered by the key prosecution witness. At this point, to reconcile either Johnson or Richter’s narrative with the forensic evidence, the jury would have had to conclude that one blood spatter expert or the other was wrong or untruthful. We cannot, of course, determine which expert the jurors would have believed, or if they would have been able to determine beyond a reasonable doubt that one or the other was correct. We can determine, however, that, under these circumstances, counsel’s failure to present blood spatter expert testimony undermines our confidence in the result and, accordingly, was prejudicial.16 Moreover, because the prosecution’s case rested so heavily on Johnson’s testimony, impeaching his credibility on a central point through forensic evidence would have likely precluded the jury from returning a verdict of guilty.
The expert blood spatter testimony was also essential in light of substantial weaknesses in the prosecution’s case, that rendered it highly likely that expert testimony would determine the outcome.17 Here, *965neither explanation of the events — neither the story presented by the prosecution nor that presented by the defense — was wholly satisfactory. The prosecution provided no convincing explanation why Richter and Branscombe would have returned to a house that they knew to be occupied to steal a gun-safe that had, only weeks before, been left for safekeeping in Richter’s garage. If Richter desired to obtain its contents, it would have been far easier for him to tell Johnson, while it had been left in his possession, that a third party had stolen it. Nor did the prosecution offer any explanation for the numerous inconsistencies in Johnson’s account of the events in question, including his initial report that there were four or five assailants, then that there were three, and then that there was only one. Nor did it explain Johnson’s inability to accurately describe his assailants, even though he later asserted that they were two individuals he knew extremely well. It would appear that at the least Johnson’s mental processes were not functioning normally at the time of the shooting — perhaps because of a mixture of drugs and alcohol — which might explain why, in his befuddled state, not recognizing Branscombe, he may have shot what appeared to him to be an intruder when he awoke and saw an unknown person unexpectedly in his house. There were other less important puzzling items, such as Johnson’s contradictory testimony about where he kept his .380 caliber gun. Also, the prosecution could not adequately explain why there was no blood spatter on the sleeping bag where Klein allegedly was lying when he was shot, and why there was no blood on the light switch in the bedroom, although Johnson claims that he turned on the light after touching his bloody face. Moreover, Johnson was overall a less-than-credible witness. He was a big-time marijuana and psylocibin mushroom dealer who sold approximately a pound of marijuana per day. He twice received immunity from the prosecution in exchange for his testimony. Johnson was also a prolific gun collector who had at least on one occasion purchased a stolen handgun — the .32 caliber pistol that Branscombe ultimately used to shoot him. Earlier in the evening in question, Johnson had apparently been ready and willing to use his firearm: He had drawn a loaded weapon and pointed it at Richter and Branscombe when he didn’t recognize their car in his driveway.
Of course, the defense’s explanation, too, had holes. The story Richter told raised a number of questions: Why were there .22 caliber cartridges found in his garage? How had Johnson carried Klein to the couch with so little evidence of blood flow? "Why was the safe thrown haphazardly on a pile of other belongings in Richter’s garage, rather than standing upright as a witness had seen it previously? Why had Richter and Branscombe taken off for the Yolo bypass shortly after the killing and not returned until the next day? Why did Richter initially tell the police that his truck was not at Johnson’s house on the night in question? Even aside from these questions, Richter suffered from credibility problems. He was a marijuana and mushroom user himself, as well as an occasional dealer.18
*966In light of the weaknesses in both sides of the case, we cannot say definitively that Richter would have been acquitted, even had the defense’s blood spatter expert testified that blood dripping from Johnson while he stood and waited in the doorway for the police to arrive could not have formed the pool near the bedroom door. There are a number of factual inconsistencies in this case pointing in both directions, and it is impossible to say what verdict the jury might have reached had it heard all the relevant evidence. But certainty has never been the standard to which we hold ourselves, even on AEDPA review. The existence of factual inconsistencies in a ease does not absolve defense counsel of the responsibility to investigate; nor does it eliminate the likelihood of prejudice when he fails to do so. Rather, introducing forensic evidence that confirms the defendant’s version of the event or casts doubt upon the prosecution’s is even more critical in such circumstances and even more likely to establish reasonable doubt in the minds of the jurors. Our obligation is to determine whether there is a “reasonable probability that, absent [counsel’s] errors, the factfinder would have had a reasonable doubt respecting guilt.” Strickland, 466 U.S. at 695, 104 S.Ct. 2052. A “reasonable probability” does not require certainty, or even a showing that it is “more likely than not” that a different outcome would have resulted. See Sanders, 21 F.3d at 1461. Rather, the probability must simply be “sufficient to undermine [our] confidence in the outcome.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. Here, our confidence is unquestionably undermined. We cannot say — and we do not believe that any reasonable adjudicator could — that the defense expert’s spatter testimony, when considered along with all the evidence previously introduced, would not have resulted in reasonable doubt.19
According the state court the deference mandated by AEDPA, we nonetheless hold that it was an unreasonable application of Strickland for that court to conclude that counsel’s failure to conduct an adequate investigation into the source of the pool of blood and to adduce expert blood spatter testimony that corroborated the defense’s version of the facts, did not constitute prejudice and thus did not violate Richter’s Sixth Amendment right to effective assistance of counsel.
2.
As we have explained, the blood spatter testimony alone was sufficient to give rise to reasonable doubt in the minds of the jurors. We note, however, that such doubt would have been reinforced had counsel also offered expert testimony on blood typing to. rebut that of the State’s serology expert.20 The expert declarations from *967Thornton and Wraxall, presented during habeas proceedings, do not prove definitively that Klein’s blood was in the pool, but they tend to counter the testimony by the State’s serology expert that Klein was excluded as a source of the blood sample taken from the door molding above the pool.21 The testimony of the two defense experts, if credited, serves to establish the possibility that the pool contained a mixture of Johnson’s and Klein’s blood and raises a reasonable doubt as to the prosecution’s contrary view. If that expert testimony had been presented along with Moses’ blood spatter testimony, the expert evidence as a whole would have allowed the defense to argue that both the blood spatter evidence and the serology evidence supported — or, in the case of the serology evidence, at least did not conflict with — the theory that the pool contained Klein’s blood.
Although we base our prejudice holding on counsel’s failure to introduce blood spatter evidence, that holding is bolstered by the detrimental effects of counsel’s failure to consult a serology expert — both because he could have introduced expert testimony that would have supported his case and because an expert could have advised him as to how to cross-examine the prosecution’s witnesses.22 Strickland requires us to assess the aggregate impact of counsel’s deficient actions when evaluating whether such failures are prejudicial. 466 U.S. at 695-96, 104 S.Ct. 2052. Here, the jury would have been even more likely to credit Moses’ testimony that the blood pool could not have been formed by the dripping of Johnson’s blood while he stood *968waiting for the police to arrive if the prosecution’s serology testimony was also cast into doubt.23 While it would not have been unreasonable for the state court to conclude that, by itself, the failure to present serology evidence was not prejudicial, the absence of that expert testimony contributed to the prejudice that resulted from the failure to offer expert blood spatter testimony.
3.
To reiterate, we conclude that Richter was prejudiced by counsel’s deficient investigation with respect to the critical question of the source of the blood in the pool by the bedroom doorway. The resultant failure to introduce exculpatory expert testimony on blood spatter, as well as other expert testimony that would, together with the spatter testimony, have raised reasonable doubt in the minds of the jurors with respect to that question undermines one’s confidence in the outcome of Richter’s case. The state court’s conclusion to the contrary constituted an objectively unreasonable application of Strickland.
IV.
In so holding, we need not reach Richter’s other claims. Because we have con-eluded that counsel provided ineffective assistance by failing to investigate the source of the pool of blood, and then to introduce expert testimony on that point, and because we have also concluded that the state court decision to the contrary constituted an unreasonable application of Strickland, we do not consider here whether counsel’s failure to investigate the caliber of a bullet hole found in the floor of Johnson’s bedroom was also ineffective. Similarly, we need not decide whether counsel’s failure to present lay testimony and expert testimony on firearms was ineffective, nor whether the prosecution violated Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), in misrepresenting the size of the bullet hole and the inaccessibility of the piece of floorboard containing it that fell into the crawl space below the house. We do not, moreover, decide whether Richter’s Eighth and Fourteenth Amendment rights were violated when the state court provided an incorrect or inaccurate answer to a request for clarification that the jury submitted to the trial court during deliberations.
V.
Counsel’s ineffectiveness in failing to investigate and present expert testimony re*969garding the source of the pool of blood is evident. The state court’s decision to the contrary constituted an unreasonable application of Strickland. The decision of the district court denying a writ of habeas corpus is reversed, and the matter is remanded with directions to enter judgment granting a writ of habeas corpus ordering that Richter be released unless the State elects to re-try him within 90 days of its entry.
. The record reflects uncertainty regarding Tony’s last name.
. The dissent suggests the defense theory rested on the unlikely scenario that Klein attacked Branscombe ''immediately” after letting him into the house. See Dissent at 975, 993-95. This is incorrect. The defense theory was that Johnson (not Klein) initiated the attack from inside the bedroom when Branscombe approached the bedroom door— which is not implausible if Johnson, a major drug dealer, impaired by drugs and alcohol, was startled awake and saw a figure he did not recognize in his bedroom doorway, especially given his demonstrated willingness to use his M-12 earlier in the evening.
. The defense contended that Johnson delayed for approximately half an hour before calling the police — time enough to manipulate the evidence at the scene and hide the .22 caliber firearm.
. The State's blood spatter expert drew his conclusions from an analysis of photographs, nearly a year after the events in question. He explained, "I did not see the victim. He was transported prior to my arrival, and all I have had is these photos which are lacking. I mean, they are dark and not clear, but I was not able to see in viewing the photos anything consistent with [Klein falling to the floor] after having been asked to look at them.” In analyzing the blood flow patterns on Klein’s face, Bell admittedly could not see the right side of Klein's head.
. Here, the California Supreme Court denied Richter's habeas petition in one sentence, without providing any reasoning for its decision. No other state court commented on Richter's claim that counsel provided ineffective assistance by failing to investigate the source of the pool of blood, consult expert witnesses, or call any such witnesses who could testify as to its source.
Because we would grant the writ whether we reviewed the state court's decision de novo or for objective unreasonableness, we apply the stricter unreasonableness standard. Thus, we need not determine whether or when an unreasoned state court decision warrants AEDPA deference.
. Although later stating that the defense's "entire case” depended on the plausibility of Johnson shooting at Branscombe during a struggle, Dissent at 992, the dissent takes issue with our conclusion that the dispute over Klein's location when shot was the most important issue for the defense. See Dissent at 969-70. We must disagree; there can be no doubt of its importance to Richter's account of self-defense. Blood evidence, which obviously might prove or disprove the location of the shooting, was therefore critical. That the State, which was unaware of Richter's version of the events before trial began, did not immediately understand the significance of the pool of blood in the bedroom doorway has no bearing upon the relative importance of the source of the pool to the defense’s theory of the case.
. Counsel need not, of course, "consult an expert on every conceivable evidentiary issue." Dissent at 969-70. Nor must he adopt a "scorched earth" investigative strategy. Id. at 970, 980 n.15. In the circumstances of this particular case, however, counsel was obligated to consult with forensic experts on blood evidence in order to provide adequate representation to his client. The point here is that the pool of blood was critical to the defense, and that counsel therefore was required to investigate the availability of forensic evidence that might prove its source — not only "because it was possible that the State might introduce similar experts at trial,” id. at 988-89, but more important, because counsel had an independent obligation to develop his client's case and to present as strong a defense as possible.
. We do not, as the dissent complains, "force counsel to seek expert advice at every stage of the proceedings, even when counsel believes that it will not be helpful and will detract from the other issues counsel must confront." Dissent at 969. Rather, we fault counsel for failing to conduct the requisite investigation about a critical issue in the case at the first stage — before settling on his defense strategy — and then for failing to rectify his deficient performance at any of the subsequent times he had the opportunity to do so.
Counsel had three separate opportunities to consult with forensic experts about the source of the pool of blood. Had he consulted an expert at the outset, it would have been unnecessary to repeat his investigation at each subsequent stage; he would have only had to follow through on the information obtained from his initial investigation as appropriate— in this case, by presenting expert testimony at trial and by employing expert advice as to how he could effectively cross-examine the *955prosecution’s witnesses or counter their expert testimony with expert testimony of his own. Counsel failed, however, to investigate at the first opportunity. Having done so, he had a continuing constitutional obligation to perform that duty at each subsequent opportunity. Each time, he failed to do so. This clearly constitutes deficient performance.
Moreover, the dissent fails to explain how counsel was in a position to make a reasonably informed decision as to whether expert assistance would be helpful at any of these three stages, given his admitted lack of knowledge about the relevant fields of forensic science. See Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052 (holding counsel’s strategic choices are entitled to deference to the extent they are supported by reasonable professional judgments). Counsel admitted that he had no independent knowledge of blood type or blood spatter science before he began working on this case. He further admitted that he did not conduct independent research to educate himself about either science until mid-trial, when he read a single book about blood typing the night before Spriggs' testimony, and his deposition testimony is silent on any research or investigation he conducted that would have allowed him to make an informed decision as to whether a blood spatter expert would be helpful in interpreting the forensic evidence. Thus, the record demonstrates through counsel’s own testimony that he could not have made an informed decision leading up to trial that seeking expert assistance would be unhelpful or counterproductive.
. We agree with the dissent that counsel's decision not to ask the prosecution to test the blood sample before trial in order to determine the blood type was not unreasonable, as such a test might have eliminated existing ambiguities that were useful to the defense and instead have produced definitive results that were damaging. However, there could have been no negative consequence to consulting a blood spatter expert or a serology expert prior to trial. An adequate pretrial investigation would not have alerted the State to counsel’s strategy, or somehow "cooked his own client.” Dissent at 969. Had counsel consulted a blood spatter expert, the State would have been unaware that he had done so until the point at which counsel presented expert testimony at trial, and then only if he made an informed decision that doing so would be helpful to the defense. Had counsel consulted a serology expert, the State would similarly have been unaware of such pretrial consultations unless and until counsel made an informed decision to ask the expert to test the blood type, or until he presented the expert as a witness to counter the testimony of the State's serology expert. Such consultations would not, therefore, have posed any risk. They would, however, have provided crucial assistance to counsel in deciding whether or not to present expert testimony, as well as in preparing for the possibility that the State would introduce adverse forensic evidence.
. We also reject the excuse that counsel believed that the prosecution would not present any expert testimony. That the State would present expert forensic testimony was entirely predictable. To the extent that counsel believed otherwise, relying on the element of surprise in telling a story of self-defense that the prosecution had not anticipated, in which Klein was killed not on the couch but near the doorway to the bedroom, his performance was unreasonable. Any competent counsel would have known that, whatever element of surprise he might have counted on would be dissipated when he disclosed in his opening *958statement that he would attack the prosecution for failing to conduct necessary forensic investigations to prove that Klein was killed on the couch rather than by the doorway. That is precisely what occurred. Following counsel's opening statement, the prosecution investigated the availability of such forensic evidence forthwith, leaving the defense as the only party without expert testimony. Counsel failed completely to prepare for the possibility that the prosecution would obtain such evidence and, consistent with his performance up to that point, failed to consult any expert to determine how to handle the State's likely response to his client’s version of the events. As a result, once he was confronted at trial by forensic evidence that undermined the theory he had presented to the jury, he had to hastily try to educate himself in order to prepare for cross-examination, had no informed basis for evaluating whether his cross-examination of the prosecution experts would be effective, and failed, once again, given the time constraints, to retain and call experts to develop alternate testimony, attack the prosecution expert's methodology as unsound, or even advise him as to how to perform an adequate cross-examination.
As it turned out, then, counsel guessed wrong when he thought, unreasonably, that the source of the blood would be determined entirely through a credibility contest rather than expert testimony. Guessing wrong does not usually amount to incompetent representation; even the best lawyers guess wrong on occasion. But being unprepared for the consequences of a wrong guess, when the issue is critical to the defendant’s case, may well amount to ineffective performance.
. Despite the dissent’s allegation that we have retroactively imposed our own theory of the case, Dissent at 970, we in no way suggest a new or different strategy for counsel. Rather, it is our dissenting colleague who has suggested a strategic rationale for counsel's actions that finds no support in the record. See Dissent at 990 (suggesting that counsel did not consult a blood spatter expert because he “had good reason to believe that a blood spatter expert would not be able to confirm his client's story”). Our colleague's valiant but wholly unpersuasive explanation indubitably constitutes a “post hoc rationalization of counsel's conduct.”
. The present case is therefore unlike Knowles v. Mirzayance, - U.S. -, 129 S.Ct. 1411, 173 L.Ed.2d 251 (2009), in which counsel made a careful, strategic, and reasonable decision not to pursue a hopeless defense. In that case, there was no failure to investigate; counsel, after investigating, had drawn a well-informed conclusion that it was futile to present a defense of not guilty by reason of insanity to a jury that had convicted the defendant of first-degree murder after considering the same evidence of mental illness. Counsel's failure here to present expert testimony on the source of the pool of blood was not, by contrast, a well-considered decision; rather, it was the result of his failure to investigate in order to seek to discover available evidence that might be critical to his case. Nor would presenting such forensic evidence have been an exercise in futility. As will be discussed with regard to prejudice, it would instead have significantly undermined a critical aspect of the prosecution’s case.
. The research that counsel conducted on his own after learning about Spriggs' upcoming testimony was more modest than the dissent suggests: at his deposition, counsel testified that a librarian helped him find “several different books, one of which [he] took home and read.”
. The dissent excuses this mid-trial failure by explaining that counsel acted reasonably by relying solely on his cross-examination of the State’s serology expert instead of introducing expert testimony of his own that might have been available. Dissent at 982-83. That position is simply erroneous. When defense counsel obtains concessions’ from the witness on cross-examination, that may on occasion be sufficient to establish reasonable doubt. It is not ordinarily, however, a substitute for affirmative evidence that would directly prove the point to be established. Here, equivocal statements by a prosecution witness in no respect serve the same function as positive affirmations by defense experts. That would seem to be an elementary tenet known to all trial lawyers. Moreover, in this case, Spriggs repeatedly disavowed any possibility that Klein's blood was in the pool. She did admit that if the blood sample was degraded, it would be possible that the 1 + band would be less intense than she expected, but she saw no indication that in this case the sample was degraded. On redirect, she repeated conclusively that she had “no reservations. [Klein] is excluded as a possible donor of that sample.” From the standpoint of a jury, the cross-examination of Spriggs was in no way comparable to the evidence that could have been introduced by an expert witness for the defense.
Moreover, where an attorney lacks expertise in a scientific subject such as blood typing, he is in a poor position to determine whether his cross-examination has touched upon the most important weaknesses in the expert’s testimony. If counsel in fact decided that another expert would not "interpret the serology evidence any differently than the State’s expert,” Dissent at 983, as the dissent without any support claims that he did, he made that assessment based not upon a rea*961sonable investigation but entirely on speculation. In fact, counsel quickly realized after trial that he had failed to question Spriggs adequately about reasons for the relative intensity of the 1+ band other than degradation, important evidence that an expert, had one been consulted, would have been able to quickly point out.
. The parties stipulated that the defense experts would testify to the information sworn to in their declarations. In their declarations the experts stated that they would have been available to testify as to their opinions if they had been called at trial. We therefore consider the statements in these declarations as the relevant testimony that could have, and would have, been presented at trial.
. The dissent advances a curious argument why the failure to introduce blood spatter evidence could not have been prejudicial. Because Bell's unrebutted testimony that the high velocity spatter on the couch meant that someone had to have been shot there, the dissent reasons — derisively—that the jury could accept Richter's story that Klein had been shot by the bedroom door only by creating a third shooting victim who had been shot by the couch. See Dissent at 990-91, 993. The dissent fails to acknowledge, however, that for a precisely analogous reason, the jury could accept Johnson's story that Klein had been shot on the couch only by inventing a third shooting victim who created the pool of blood by the bedroom. Given Moses’ unrebutted testimony that the blood pool could not have come from someone standing by the bedroom, and given that Johnson did not testify that he squatted, sat, or stretched out on the floor by the bedroom, the jury could not, without a similar degree of creativity, accept the prosecution's explanation of the events. The important fact is that a jury — -just like the dissent — could not reasonably conclude that both experts were correct while also accepting the prosecution's version of the events. Because we cannot determine which expert the jury would have believed and certainly cannot determine that in light of Moses’ testimony it would have found beyond a reasonable doubt that the prosecution’s explanation of the events was accurate, our confidence in the outcome is necessarily undermined.
. Indeed, it appears likely that it was in part because of the limitations of the State's case— in particular, the credibility problems of its main witness — that it undertook a last-minute effort to gather forensic evidence and then relied so heavily upon that evidence during *965the remainder of the trial and in closing argument. Refuting the State's forensic case would, therefore, have been critical for Richter.
. Despite the questions raised by his testimony, we must note that Richter's account was not, on the whole, implausible. The most damaging evidence at trial was Detective Bell’s blood spatter analysis. In many other ways, Richter’s version of the events, which the dissent deems "highly improbable," Dissent at 10754, was internally consistent and provided a coherent explanation for the evidence introduced at trial. For example, Richter testified that he had been paid in cash on the night of Klein’s death; that he dropped a $100 bill at the Yolo bypass was therefore perfectly understandable in light of his testi*966mony. Richter also explained, moreover, why he and Branscombe returned to the Christmas tree lot late at night, rather than waking up a couple of hours later to clean out their belongings from the trailer by the following morning and why they returned to Johnson's house. Also consistent with Richter’s explanation, Johnson had in fact tried to modify his M-12, attempting to convert it into a fully automatic weapon, thus explaining why it might have jammed.
. Of course, contrary to what the dissent appears to believe, the evidence need only raise a reasonable doubt as to the defendant's guilt, not conclusive proof of the defendant’s innocence.
. Taken by itself, the failure to present expert testimony on blood typing did not sufficiently prejudice the defense such that habeas relief would be warranted. However, in combination with the failure to present expert blood spatter testimony, which was prejudicial in itself, the failure to present expert testimony that could have undermined the damaging testimony offered by Spriggs contributed to the overall showing of prejudice.
. Indeed, the State’s serology expert adhered strongly to this position throughout. On direct examination, Spriggs stated definitively, "I would include Johnson as a possible donor of the human blood on item 14-A, and I would exclude Klein as a possible donor of the human blood on 14-A.” On cross-examination, she did not retreat from that position. Prompted by the Court, she stated, "In my opinion, [the sample] does not contain more than one person’s blood.” When asked by defense counsel whether she could say that to a scientific certainty, she responded, "Based on the testing that I perform with the PGM sub typing it does not look like there are two people contributing to that stain. It looks like one person.” Upon further questioning she continued to insist, “So I am going by based on my testing, based on the intensity of the bands, I do not think there are two people there.” Spriggs did acknowledge that it was possible that due to degradation a 1+ band might be less intense than expected. On redirect, however, she further explained that the sample did not appear to be degraded, and that if a sample was degraded, one would get an inconclusive answer, not a wrong answer. On re-cross, her unequivocal final conclusion was that she had "no reservations. [Klein] is excluded as a possible donor of that sample.” In light of Spriggs’ testimony, and in light of the obvious fact that the questions posed by counsel on cross-examination are not evidence and may not form the basis for a jury’s conclusions, there would have been a significant benefit to the defense had the jury heard directly from an expert that the possibility of blood mixing existed in this particular case, and that a source of the blood in the pool could well have been Klein.
. Counsel acknowledged that he missed crucial information when conducting his cross-examination. He ”remember[ed] kicking [him]self” after trial because he failed to elicit from Spriggs the possibility that, even without degradation, the mixture of Johnson’s and Klein’s blood could have produced the 2 + 1 + band it did — due to the proportions of the blood in the pool, or because Johnson had a stronger 2 + band and a weaker 1 + band. Such testimony certainly would have been significant, particularly because Spriggs testified that it did not appear that the sample was degraded. Counsel acknowledged that “where you have experts like this that you can go to so you don’t make those mistakes like I made.... Sure, you use them whenever you can. They’re a great tool.” Nevertheless, without the benefit of such a consultation to confirm whether or not he had adequately questioned Spriggs on cross-examination, "at the time I thought I did an okay job.”
. Similarly, it would have been helpful to introduce expert medical testimony regarding the amount of blood that Johnson could have deposited in the pool, although it would not have been unreasonable to conclude that, standing alone, the absence of that testimony in this case was not prejudicial. The dissent dismisses Dr. Herrmann's expert opinion, in part because in its view this testimony would have been inadmissible at trial. Dissent at 986-87. We cannot agree, however, particularly in light of the fact that the parties "agree[d] and stipulated] that the ... declarations of petitioners' expert witnesses are admissible for consideration on the merits, and that [the experts], if called to testify, would testify in conformance with their declarations. ...” We believe, moreover, that this expert testimony would have been both relevant and significant for the jury to consider. Johnson, we must remember, claimed that he immediately called 911 after finding Klein on the sofa and, contended that in the six minutes before the police arrived, he ran around the house, hid the scale on which he measured marijuana, took two trips outside to throw two marijuana plants over the fence, and called his girlfriend's father. That blood was found in various locations around the house does not negate the medical expert's opinion that the pool in the doorway was too large to have come from Johnson. There was little time for such a pool to form, and further corroborative expert testimony on blood pooling could have been persuasive.