OPINION
WILKINSON, Circuit Judge:This habeas case appears before us for the second time. In Washington v. Murray, 952 F.2d 1472 (4th Cir.1991) (Washington I), we remanded for an evidentiary hearing on the sole issue of whether petitioner received ineffective assistance from his trial counsel because of counsel’s failure to investigate and present certain forensic evidence. The district court conducted a hearing, considered the evidence, and found that petitioner had not received ineffective assistance of counsel. Because we agree that the omission of this ultimately inconclusive evidence did' not prejudice the defendant, we affirm the district court’s judgment.
I.
The factual background of this appeal has been set forth in Washington I, 952 F.2d at 1475. To summarize, a jury convicted petitioner Earl Washington, Jr. of the capital murder of Rebecca Lynn Williams. Mrs. Williams had been raped and repeatedly stabbed while in a bedroom of her apartment. Defendant, who has a general I.Q. in the range of 69, confessed no fewer than three times to the murder, and was convicted on the basis of these confessions and his acknowledgment that he owned a shirt linked to the crime scene.
The district court initially dismissed Washington’s habeas petition in its entirety without a hearing. See id. at 1475-79. On remand, however, the court considered forensic evidence relating to five seminal fluid stains on a royal blue blanket found in Mrs. Williams’ bedroom.
Neither the blanket nor any evidence about the stains were introduced at Washington’s trial. Analysis of the stains indicated that four of 'them had been produced by a person with blood type A, PGM 1, while the fifth stain only could be identified as type A.1 Petitioner has type 0 blood, PGM 2-1, and he is a secretor.2 The victim was blood type A, PGM 1, and she was a secretor; her husband has type 0 blood, PGM 1, but he is a nonsecretor.
According to Washington’s theory of the evidence presented at the hearing, he could not have produced the seminal fluid stains on th& blanket because he is the wrong blood type. ■ He also maintains that the victim’s husband could not have left the stains because he too is the wrong blood type, thus indicating the presence of a third male in the bedroom at some point. Petitioner argues that it was ineffective assistance for his trial counsel not to have grasped the significance of this evidence and introduced it to support a defense that this, third male raped and murdered Mrs. Williams.
To buttress this theory, petitioner introduced at the hearing the testimony of his trial counsel and two forensic experts. Trial counsel testified that he received from the prosecution a report containing analysis of the various stains on the blanket, but that he did not realize from his review of the report that Washington could not have left the stains. Counsel never consulted with experts to have the report explained to him.
Petitioner’s first expert at the federal ha-beas hearing, Dr. Henry C. Lee, Director of the Connecticut State Police Forensic Science Laboratory, testified that if the stains were pure semen stains, they could not have *1287been left by either petitioner or Mr. Williams. Despite having examined only the forensic report, not the stains themselves, Lee also testified that it was equally possible that the stains were pure semen as that they were a mixture of semen and vaginal fluid. Dr. Lee stated that in his opinion the stains were not a mixture because there was no indication in the report that the stains contained vaginal epithelial (or skin) cells. On cross, Dr. Lee admitted that he could not state with certainty that the stains included no vaginal fluid, and that if the stains were a mixture, Mr. Williams could have contributed to them. Lee’s testimony also indicated that there was no evidence of blood in the five seminal fluid stains. Petitioner’s second expert, Dr. David Alan Stoney of the University of Illinois at Chicago, largely supported Dr. Lee’s testimony. Stoney testified, however, that even if a test had indicated a lack of epithelial cells in the stains, such a result did not preclude the presence of vaginal fluid.
The prosecution’s theory of the evidence was that the victim’s husband left the seminal fluid stains on the blanket. Even though Mr. Williams has blood type O, PGM 1, and the stains indicated blood type A, PGM 1, the seminal fluid still could have been his because, as a nonsecretor, his blood antigens do not appear in his semen for purposes of ABO typing. Under the prosecution’s theory, Mr. Williams would have left stains mixed with his wife’s vaginal fluid in the process of regular intercourse with his wife. Because Mrs. Williams was a secretor, her blood antigens would dictate what ABO blood type the stains indicated — her type A effectively “masking” his type O. Because Mr. and Mrs. Williams shared the PGM 1 blood type, the type A, PGM 1 result from the stains is consistent with Mr. Williams having contributed the semen.
To buttress its theory, the prosecution offered the testimony of Mr. Williams and two experts. Mr. Williams testified that he had been having sex with his wife three to five times a week for the three months prior to her murder and that the blue blanket was normally on their bed.
Deanne Dabbs of Virginia’s Division of Consolidated Laboratory Services, who analyzed the stains and prepared the forensic report, testified in support of the view that vaginal fluid from Mrs. Williams, not seminal fluid, dictated the blood type. She testified that based on her experience of examining “thousands” of such stains, these five stains appeared to be mixtures of vaginal and seminal fluids that were “distinctly different in appearance from a pure seminal fluid stain.” Dabbs testified that she routinely did not test stains for the presence of epithelial cells because such cells indicate only that a variety of bodily fluids could be present. She acknowledged that no scientific test could positively determine whether a stain contained vaginal as well as seminal fluid, but stated that there was no evidence that these stains were not a mixture of fluids. She also stated that she would never testify that these stains were pure seminal fluid because the amount of sperm cell material in them, as compared to the amount of sperm cells she might see in a pure semen stain, supported her conclusion of admixture. Ms. Dabbs further testified that a fluid sample from the victim’s vaginal vault indicated sperm in a “bloody mixture,” while as Dr. Lee and Dr. Stoney had indicated, no evidence suggested the presence of blood in the five seminal stains on the blanket. Dabbs also testified that in her experience it was “rather unusual” to find five semen stains produced in the course of a single sexual assault; more typically one or two stains might be found. According to Dabbs’ analysis, Mr. Williams could well have contributed the seminal fluid in these stains even though Mrs. Williams’ fluid ultimately determined the blood type. Both sides stipulated that the opinions and conclusions of Peter Milone, Assistant Director of the Division of Consolidated Laboratory Services, would have agreed with those of Dabbs.
Confronted with this scientific testimony, the district court found that the forensic evidence was “inconclusive.” The court noted that if the stains were pure seminal fluid, then neither petitioner nor Mr. Williams could have left them. The court, however, credited the testimony of Ms. Dabbs, the only expert actually to have seen the stains prior to testifying, that their appearance *1288“was consistent with a mixture of vaginal and seminal fluid.” Because the “presence of vaginal fluid” in the mixture could “mask the blood test results of the seminal stains,” the court found that the forensic evidence provided “no guidance as to the identity of the rapist,” but observed that “[i]f anything, the evidence leads to the conclusion that the stains were on the blanket prior to the murder and, in all probability, were produced by the victim and her husband.”
Applying the standards for determining ineffective assistance of counsel from Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2052, 2064, 80 L.Ed.2d 674 (1984), the district court held that petitioner suffered no prejudice from his trial counsel’s failure to pursue evidence about the stains because the evidence did not establish “a reasonable probability that petitioner was not the rapist.” The court also went on to hold that trial counsel’s performance was not deficient because counsel had made a “strategic choice” not to present this line of evidence. According to the district court, counsel chose not to present the evidence in order “to keep the victim’s husband from testifying” in rebuttal and thus to minimize “the jury’s sympathy toward the victim’s family.”
Finding that neither Strickland prong was met, the district court concluded that petitioner had not received ineffective assistance of counsel and dismissed his claim. This appeal followed.
II.
Petitioner argues that the district court erred in its analysis of both Strickland prongs and that the failure to present a defense based on this forensic evidence was ineffective assistance. As to the performance prong, petitioner maintains that the district court clearly erred in finding that trial counsel made a strategic decision not to pursue the evidence. As to prejudice, petitioner argues that the district court erred in substituting its own opinion of the significance of the seminal fluid stains for what a reasonable juror might have concluded from them.
We review de novo the district court’s holdings about trial counsel’s performance and prejudice to petitioner, as well as its conclusion that no ineffective assistance occurred, because these are all mixed questions of law and fact. See Strickland, 466 U.S. at 698,104 S.Ct. at 2070; Becton v. Barnett, 920 F.2d 1190, 1192 (4th Cir.1990). The district court’s findings of fact, however, are subject to the clearly erroneous standard of Fed. R.Civ.P. 52(a). Strickland, 466 U.S. at 698, 104 S.Ct. at 2070. We turn first to the district court findings on trial counsel’s performance, and then will take up the court’s holdings as to prejudice and overall ineffectiveness.
III.
Strickland makes plain that a lawyer’s performance will not be deemed deficient if it results from informed, strategic choices about how to mount a defense. 466 U.S. at 689-90, 104 S.Ct. at 2065-66. The district court found that Washington’s trial counsel made such a strategic decision here. During cross-examination at the evidentiary hearing, the government asked trial counsel what he would have done “if’ he had investigated the forensic evidence and “if’ it had been totally consistent with Mr. Williams having left the stains and not petitioner. Counsel responded that he probably would not have introduced the evidence. The district court concluded that trial counsel actually had confronted this dilemma and made a strategic choice.
Petitioner argues that this finding is at odds with the actual record, and we agree. Trial counsel testified that he never offered evidence about the forensic analysis of the stains because he was unaware of their potential significance. In fact, he consulted with no experts about the report, and never realized that Washington’s blood type was inconsistent with the seminal fluid on the blanket until Washington’s habeas counsel told him well after the trial. Prior to trial, counsel was completely in the dark about the import of the evidence, and therefore, could not have made a strategic choice against using it.
*1289Trial counsel’s post hoe hypothesis of what he might have done had the situation been different has a dubious bearing on what he actually did. His conduct should have been evaluated from his perspective at the time of trial, Strickland, 466 U.S. at 689, 104 S.Ct. at 2065, and the district court should not have constructed a tactical decision counsel might have made, but obviously did not. See Griffin v. Warden, Maryland Correc. Adjus. Ctr., 970 F.2d 1355, 1358 (4th Cir.1992). The district court thus erred by using a response to a hypothetical as the basis for a factual finding without support in the record.
IV.
A.
Petitioner next argues that the district court erred by adopting a higher standard of prejudice from counsel’s performance than that called for by Strickland. He maintains that the correct standard was whether a juror might reasonably have determined that the forensic evidence created enough doubt so as to prevent a finding of guilt. According to petitioner, instead of applying this standard, the district court substituted its own judgment as to whether the forensic evidence proved Washington had not been the rapist. Petitioner also points to our specific admonition on remand merely to assess “the reasonable probability that introduction of this forensic evidence would, at the least, have raised a reasonable doubt of Washington’s guilt in the jury’s mind, or have caused the jury to recommend life imprisonment rather than death” if it found Washington guilty. Washington I, 952 F.2d at 1477 (emphasis in original). Petitioner contends that the district court ignored this mandate.
We disagree. The district court followed both the Strickland prejudice standard and our specific instructions on remand. Under Strickland, prejudice is established by showing a “reasonable probability that, but for counsel’s unprofessional errors,” the outcome of the trial would have been different. 466 U.S. at 694, 104 S.Ct. at 2068. A reasonable probability is one “sufficient to undermine confidence in the outcome.” Id. On remand, we instructed the district court to apply this standard to petitioner’s claim. Washington I, 952 F.2d at 1479. We gave as an example of evidence that could undermine confidence “[sjcientifically reliable forensic evidence” which negated “to a high degree of probability Washington’s identity as the rapist.” Id. We also observed that this forensic evidence could fail to establish prejudice if “the raw data allegedly developed could be shown by credible scientific opinion not to be sufficiently probative on the identification issue.” Id.
The district court followed these directives precisely. It quoted and applied the Strickland standard. From its finding that the competing scientific evidence was inconclusive as to who might have left seminal fluid stains on the blanket, the court reasoned that the evidence did not negate the probability that Washington was the rapist, and therefore, that it was not the type of evidence we suggested would undermine confidence in Washington’s conviction.
B.
We cannot say the district court erred in concluding that petitioner was not prejudiced by the absence of the forensic evidence. To begin with, the prosecution never relied on any evidence from the stains on the blanket to link petitioner to the rape, but relied instead on Washington’s own statements that he compelled the victim to have sex with him. In this sense, the case differs markedly from the single case submitted and relied upon by petitioner in which a failure to present forensic evidence was held to be ineffective assistance. In that case, Sims v. Livesay, 970 F.2d 1575 (6th Cir.1992), the court ruled that it was ineffective assistance for defense counsel not to have presented evidence of powder burns on a quilt when the burns supported defendant’s claim that a shooting had been accidental. The Sixth Circuit observed that an FBI report about the quilt disclosed that the state’s theory of murder in the case was “easily refutable,” id. at 1580, and that the state “relied heavily” on expert testimony that the gun had been fired from a distance, id. at 1581. Here, the prosecution did not even introduce the blanket, *1290much less rely heavily on the stains as evidence, and the inconclusive forensic analysis of the stains does not begin to easily refute the prosecution’s case.
The problems with petitioner’s forensic evidence were considerable. First, the five stains on the blanket suggest that they were placed there, not in a single violent assault, but rather over a span of time during several instances of intercourse. Second, the absence of blood in the seminal fluid stains on the blanket and the presence of blood mixed with semen in the victim’s vaginal vault tended to indicate that the stains on the blanket were not caused by seepage from the victim during or after the assault. Third, Ms. Dabbs testified in detail, and from a substantial background in the field, both that the stains appeared mixed and that the victim’s vaginal fluid could have masked the presence of her husband’s seminal fluid in stains found on the blanket of a bed where the two regularly had intercourse. All of these problems — the inconsistency between the number of stains and an assault, the absence of blood in the stains, and the stains’ presence on a blanket that covered a bed where the Williamses regularly had intercourse — support the district court’s conclusion that “the stains were on the blanket prior to the murder and, in all probability, were produced by the victim and her husband.”
Petitioner does not dispute the district court’s assessment of the evidence as ultimately inconclusive; he merely argues that inconclusive evidence might have been enough to sway a jury. From such inconclusive evidence, however, it is difficult to shake confidence in either the prosecution’s theory of the stains or the final result of Washington’s trial. Undermining confidence in the trial result requires more than just showing that an error “had some conceivable effect” upon an outcome. Strickland, 466 U.S. at 693, 104 S.Ct. at 2067; see also Kenley v. Armontrout, 937 F.2d 1298, 1303 (8th Cir.1991). Petitioner must show a reasonable probability, not just a remote possibility, that introduction of this evidence would have altered the result of his trial.
C.
Even assuming that petitioner had presented the stained blanket and his experts at trial, the prosecution still had a strong case against petitioner. We determined in Washington I that the prosecution’s case was “obviously constitutionally sufficient to convict,” 952 F.2d at 1477; see also Jackson v. Virginia, 443 U.S. 307, 324, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979), and given the case’s strength, we cannot conclude that inconclusive forensic evidence would have overcome it. The prosecution relied on Washington’s confessions to the crime and corroborated them with his admitted ownership of a shirt linked to the crime scene. At trial, Washington’s primary defense was to deny these statements and suggest that the police officers must have made them up; the jury obviously credited the officers’ testimony more than that of petitioner. On appeal and habeas, Washington has repeatedly argued that the statements were given in violation of Miranda v. Arizona, 384 U.S. 436, 86 S.Ct. 1602, 16 L.Ed.2d 694 (1966), and that the officers coached him into making the statements because of his extreme suggestibility. We rejected these involuntariness claims in Washington I, 952 F.2d at 1482 & 1483, because in a pretrial suppression hearing, the state court determined that Washington knowingly gave the statements. Their volun-tariness is not at issue here. Washington’s confessions and acknowledgement of the shirt stand as the jury heard them.
The strength of the prosecution’s ease, however, lies beyond the mere voluntariness of the confessions. It rests in the numerous details of the crime that Washington provided to the officers as they talked with him. Our review of this evidence, as heard by the jury, indicates that petitioner knew so much about this crime that the jury could afford his confessions substantial probative weight. According to the testimony of the officers present and Washington’s own signed statement, it was Washington who volunteered that he “took [the victim] to the back bedroom” of the apartment to rape and kill her. Police officers had indicated to him only that the victim had been found outside her apart*1291ment.3 While officers mentioned only that the victim had been stabbed, Washington brought up his sexual assault of her. When asked what he had done after taking the victim to the back bedroom, Washington answered that he had “made her undress” be-' cause he “wanted to make love to her,” and that she had not wanted to make love and resisted but “I was holding a knife on her.” He said that he had sex with the victim one time.
When officers later asked whether he had left anything in the apartment, Washington responded that he had left his shirt, explaining that “[i]t had blood on it and I didn’t want to wear it back out.” When asked where he had left the shirt, Washington stated that he “[l]aid it on top of the dresser drawer in the bedroom.” Petitioner identified as his own a shirt that had been found with blood on it in a dresser drawer from the bedroom. He identified the particular shirt because of a patch that had been ripped off from the top of a pocket. In answers to other questions, Washington volunteered that a radio had been playing in the apartment “but it wasn’t too loud,” and that the victim was about 5'6" or 5'7" and weighed maybe 170 pounds. Police confirmed that when they entered the apartment, a radio was playing; the victim was 5'8" and weighed approximately 180 pounds.4
V.
We earlier remanded this case in order to have a trial judge examine the forensic evidence in the way that only a trial judge can do. Petitioner would now have us decide this appeal by placing our own spin upon this evidence rather than crediting the district judge who, under our earlier instruction, conducted an evidentiary hearing, listened to the experts on both sides, and pointedly indicated that he was unimpressed with the probative value of the evidence petitioner sought to introduce. Petitioner further attempts to intimate that the real defendant in this case ought to be one James Pendleton, an earlier police suspect whose blood was type A, PGM 1. That theory, however, simply fails to address the various grounds for the district court’s finding that the stains in all probability were produced by the victim and her husband. It also fails to account for the fact that the record shows that it was Mr. Washington, not Mr. Pendleton, who confessed to the commission of this crime. The investigation of Pendleton did not turn up grounds for prosecution, and Mr. Pendleton is not on trial in this appeal.
Petitioner attempts finally to discount the probative force of the evidence against Washington by pointing to Washington’s suggesti*1292bility. Testimony before the jury, however, indicated that Washington had supplied without prompting details of the crime that were corroborated by evidence taken from the scene and by the observations of those investigating the Williams’ apartment. He had confessed to the crime not in a general manner, but as one who was familiar with the minutiae of its execution. The district court thus did not err in determining that petitioner was not prejudiced by the failure to introduce the forensic evidence. Given the inconclusiveness of the seminal stain analysis and the strength of the prosecution’s case, the forensic evidence does not undermine confidence in the result of petitioner’s trial. Because both of Strickland’s requirements must be met to find ineffective assistance of counsel, the judgment of the district court dismissing petitioner’s claim is
AFFIRMED.
. Phosphoglucomutase (PGM) is an enzyme found in red blood cells that is used for blood typing. The general population falls into three PGM blood types — 1, 2, or 2-1. See Washington I, 952 F.2d at 1476 n. 3.
. Being a secretor means that an individual has blood antigens in their other bodily fluids, such as saliva or semen, that allow an ABO blood type to be determined from these other fluids. Approximately eighty percent of the population are secretors. Secretor status is irrelevant for determining PGM blood type.
. The victim had in fact been discovered in the front doorway of her apartment where she had apparently managed to drag herself after the attack in order to seek help.
. Petitioner points to other details he supplied that in his view lacked such corroboration. For example, Washington notes that he "initially wrongly identified the victim as having been black, and only corrected the statement on being re-asked the question.” He observes that he "said that he stabbed the victim two to three times.... She had been stabbed [thirty-eight] times.” Petitioner indicated that "he saw no one else in the apartment,” but the "victim’s two young children were present." Washington also states that officers drove him through the apartment complex where the murder occurred three times, and ”[o]n the third occasion, when asked to point out the scene of the crime, petitioner ‘pointed to an apartment on the exact opposite end from where the Williams girl was killed....”'
The Commonwealth contends that ”[t]he discrepancies in Washington’s confession are easily explained by his understandable reluctance to admit to police officers that he stabbed the victim almost forty times, or that the victim’s small children had witnessed what he had done to their mother. That the Petitioner did not readily identify the apartment in the apartment complex where the crimes were committed is explicable by the fact that he was taken to the complex one year after the crime was committed, and there was no evidence that he had ever had any familiarity with the apartment complex.” As to the victim’s race, petitioner changed his position himself merely when asked again if the victim was white or black.
Regardless of explanations, however, all of this information was before the jury, and Washington's trial counsel specifically crossexamined the police officers about their questioning regarding the victim's race and petitioner's identification of the apartment. Such questioning did not persuade the juty to discount the confessions. We fail to see how inconclusive forensic evidence would have caused the jury to take any different view of the case.