LIPEZ, Circuit Judge.

Peter Dugas, convicted of arson in the New Hampshire Superior Court, asks us to review for a second time his petition for a federal writ of habeas corpus on the ground that he received constitutionally ineffective assistance of counsel. To succeed with his claim, Dugas must demonstrate both deficient performance by his attorney and prejudice, i.e., “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland v. Washington, 466 U.S. 668, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).

In our first review, we agreed with the district court that the performance of Du-gas’s defense counsel was deficient, but we concluded that the court should not have resolved the prejudice issue on a summary judgment record. Instead, further proceedings were needed in the district court to determine whether the deficiency in counsel’s performance resulted in prejudice within the meaning of Strickland. Dugas v. Coplan, 428 F.3d 317, 333, 341-42 (1st Cir.2005) ("Dugas IV”). On remand, the district court allowed further discovery by the parties, considered additional affidavits, and held an evidentiary hearing. The district court concluded again that Dugas had not established prejudice within the meaning of Strickland and denied his habeas petition. Finding no clear error in the district court’s prejudice determination, we affirm.

Dugas was convicted of arson for setting a fire that heavily damaged his family’s grocery store, the Dugas Superette, in Nashua, New Hampshire, on October 23, 1999. Dugas managed the store and owned a minority share of the business. He told police that, on the night of the fire, he had locked the store and left with another employee at approximately 10 p.m. He said he first learned of the fire when his wife called him at about 11:30 p.m. while he was picking up his daughter. State fire investigators found no signs of forced entry into the building, no likely cause from electrical or mechanical systems, and no accidental cause for the fire. The investigators concluded that the fire was intentionally set and that it had been started by igniting an accelerant on a pile of papers in the store’s basement.

The police interpreted enhanced videotape pictures from the store’s security camera to show that Dugas left at 10 p.m., as he had said, but then reentered the store a few minutes later. The tape showed Dugas reentering, proceeding to the rear office, then exiting the rear office and turning off the light. He then disappeared from view for about one minute, reappeared from the rear of the store, and moved quickly toward the front door. Du-gas initially denied that he had reentered the store, but at trial he recalled that he had returned to check on a cash drawer. He denied that he caused the fire.

At trial, Dugas’s defense counsel, Ray Raimo, attempted to present two defense theories: first, that the fire had not been *4 arson, and second, that even if the fire had been intentionally set, someone other than Dugas had set it. 1 Early in the trial, Raimo made it clear that he was, at least in part, pursuing a “not arson” theory of defense when he said to the jury: “[W]hat we’re going to be asking ourselves during this trial is how this fire started and why ... I want to just make it clear ... where the State is bringing in all of these witnesses ... we think they’re wrong.” To prepare for this line of defense, Raimo interviewed the state’s investigators and found them all to be credible and formidable witnesses. He toured the fire scene and concluded that, from a layperson’s perspective, the physical evidence seemed consistent with the state’s arson theory. However, he did not hire an arson expert to testify on Dugas’s behalf nor did he consult with an expert in preparing his cross-examination of the state’s experts.

In furtherance of the state’s theory that Dugas had set the fire when he returned to the store shortly after 10:00 p.m., as shown on the videotape, the state presented six expert witnesses who laid out the forensic evidence that the fire was intentionally set using an accelerant such as charcoal lighter fluid. They testified that the fire began in a stack of papers on the basement floor and initially flared up but then became oxygen-starved and smoldered, generating intense heat and smoke but little flame. Consistent with this theory, the firefighters who responded to the scene testified that they found the basement door closed and the basement full of heavy smoke and intense heat. The state also presented two experts who explained the techniques that had been used to enhance the exceptionally poor quality of the surveillance video, as well as an alibi witness for P.J. Kulas, the former store employee whom the defense theorized had set the fire, see supra note 1.

The state’s strongest evidence against Dugas was its expert testimony on arson. Raimo confined his challenge of this evidence to cross-examination of the state’s expert witnesses. He attempted to raise the possibility that the fire had started accidentally and pointed out some questionable evidence handling procedures. However, his cross-examination, lacking the aid of an arson consultant or an expert witness of his own, was problematic:

[T]he focus of Raimo’s cross-examination of the state’s experts was unclear, and many of the experts’ scientific conclusions went unchallenged. Raimo did not ask the kinds of questions that a trained fire investigator or forensic scientist would consider important. Instead, his questions amounted to an unfocused set of miscellaneous criticisms and evinced his lack of scientific knowledge. Despite his earlier statement to the jury that he believed that the state’s arson experts were wrong, Raimo presented no alternative theory of the fire.

Dugas IV, 428 F.3d at 324. After three days of deliberations at the completion of an eight day trial, the jury returned a guilty verdict. The court sentenced Dugas to five to ten years in prison. His convie *5 tion was affirmed on direct appeal. State v. Dugas, 147 N.H. 62, 782 A.2d 888 (2001)(“Dugas I”).

Dugas then moved for a new trial based on a claim of ineffective assistance of counsel. In support of his motion, Dugas offered a report from Michael Higgins, his proposed arson expert. The state trial court held a hearing on the motion, taking testimony from Higgins and Raimo, Dugas’s trial counsel. Raimo said at the hearing that he had been over-confident in the strength of Dugas’s defense and that he was “still shocked by the verdict.” He admitted that he had no scientific background or technical knowledge regarding arson. He explained that, although there was no financial impediment to hiring an expert, he had thought it might be a problem to do so because of a requirement, as he saw it, that he notify the state in order to have access to the fire scene. Then the state would have wanted to depose or talk with his expert. The state court concluded that Raimo had considered the benefits and perils of hiring an arson expert and made an appropriate strategic decision not to do so. The state trial court thus held that Raimo’s performance was not constitutionally deficient, and denied Dugas’s motion for a new trial. State v. Dugas, No. 98-S-1899 (N.H.Super.Ct. Aug. 12, 2002) (“Dugas II”). Because the court found that Raimo’s performance had not been deficient under the first prong of Strickland, it did not make any findings regarding prejudice. The New Hampshire Supreme Court declined to grant review.

Dugas then filed a petition in the district court for a federal writ of habeas corpus. The district court reviewed the trial record, affidavits, and hearing testimony from the state court proceedings. On summary judgment, the district court held that Raimo’s performance had been constitutionally deficient. Dugas v. Warden, No. 03-376-JD, slip op. at 9-12 (D.N.H. May 21, 2004) (“Dugas III”). The district court explained:

Based on the record, the state court’s conclusion that Raimo carefully investigated the case and made a tactical decision not to consult with or hire an expert is both an unreasonable determination of the facts and an unreasonable application of the federal standard.... Raimo apparently believed ... that he would have had to make his expert available to the state for questioning, whether or not he intended to call the expert to testify at trial, which would be detrimental if the expert agreed with the state’s experts. The state court recognized that ordinarily the defense would not have to disclose an expert who was used only for consultation but concluded, without explanation, that the necessary disclosure in this case presented a “peril” to the defense. Given the lack of legal support or explanation, the state court’s conclusion was unreasonable that the defense would have faced a “peril” if Raimo consulted an expert.

Dugas III, slip op. at 11-12 (citations omitted). The district court then turned its focus to the prejudice analysis required by Strickland. Reviewing the state court record de novo, the court concluded that “Higgins’s opinions do not undermine the court’s confidence in the outcome of the criminal trial.” Id. at 13-16. Therefore, the district court denied Dugas’s habeas petition. 2 Id. at 18.

On appeal, a majority of the panel agreed with the district court that Raimo’s performance was constitutionally deficient. *6 However, turning to the prejudice prong of the Strickland analysis, we held that the district court had erred in granting summary judgment in favor of the state. We remanded the case to the district court for further evidentiary development on the issue of prejudice, explaining:

We perceive a distinct possibility that, if Raimo had consulted an arson expert, the outcome of the trial would have been different. We cannot say, as a matter of law, that this possibility does not rise to the level of “a probability sufficient to undermine confidence in the outcome.” Neither can we say, as a matter of law, that it does rise to such a level.

Dugas IV, 428 F.3d at 341 (quoting Strickland, 466 U.S. at 694, 104 S.Ct. 2052) (internal citation omitted). Specifically, we noted three major issues that had been highlighted in Higgins’s report. First, he alleged that there had been flaws in the state forensic chemist’s analysis of samples taken from the fire scene. Second, he posited that the fire had been much faster-burning than the state alleged and pointed to potential sources of ventilation. Third, he identified evidence of smoke shadows, i.e., clean spots on the walls and floors, indicating Dugas’s alleged path to the basement had been blocked by boxes on the night of the fire. We held that “these three issues, taken together ... [were] sufficient to raise a genuine dispute of material fact concerning prejudice.” Id. at 335. As a result, we instructed the district court to order the state to turn over all of the evidence from the fire scene so that Higgins could conduct a more thorough analysis. We then asked the district court to “consider whether Higgins’s analysis would meaningfully challenge the state’s view of the evidence.” Id. at 342. We emphasized that “[t]he case for prejudice here is close; we do not conclude that there was prejudice, but only that, in the circumstances of this appeal, Dugas has raised sufficient doubts about the outcome to avoid summary judgment.” Id. at 343.

The district court followed our instructions, ordering the evidence turned over to Higgins for testing and analysis. 3 The court then considered Dugas’s petition for habeas corpus with the benefit of additional affidavits from Higgins and two state fire investigators. It also conducted an evidentiary hearing at which the experts from both sides testified in response to questions from the court and counsel. The district court concluded that “Higgins’s opinions would not have been helpful to the defense” because they are “speculative, inconsistent, contrary to the factual evidence, and significantly less credible and less persuasive than the opinions given by [the state’s experts] at the trial and by [the state’s experts] at the evidentiary hearing.” Dugas v. Warden, No. 03-cv-376-JD, 2006 WL 2463670 (D.N.H. Aug. 24, 2006), slip op. at 38 (“Dugas V’). Thus, the district court denied Dugas’s habeas petition and this appeal followed.

II.

The standards of review applied in the earlier stages of this case and in the appeal before us today require some explanation. Federal review of habeas petitions is normally governed by the high level of deference to state court findings set forth in the Antiterrorism and Effective Death *7 Penalty Act of 1996 (AEDPA), Pub.L. No. 104-132, 110 Stat. 1214. AEDPA allows federal courts to grant habeas relief after a final state adjudication only if the state court proceedings “resulted in a decision that was contrary to, or involved an unreasonable application of clearly established federal law, as determined by the Supreme Court” or was based on an “unreasonable determination of the facts.” 28 U.S.C. § 2254(d). However, this standard applies only when the claim being reviewed at the federal level was “adjudicated on the merits in State court proceedings.” Id. When the state court has never addressed the particular federal claim at issue, federal review is de novo. Pike v. Guarino, 492 F.3d 61, 67 (1st Cir.2007). As we have noted, a federal court “can hardly defer to the state court on an issue that the state court did not address.” Fortini v. Murphy, 257 F.3d 39, 47 (1st Cir.2001).

In its first review of Dugas’s habeas petition, the district court applied the AEDPA standard of deference to state court findings. That deference did not prevent the district court from concluding that “the state court’s conclusion that Rai-mo carefully investigated the case and made a tactical decision not to consult with or hire an expert is both an unreasonable determination of the facts and an unreasonable application of the federal standard.” Dugas III, slip op. at 11. Because the state court had never reached the question of prejudice, the district court applied a de novo standard to the state court record and concluded on summary judgment that Dugas had not been prejudiced by Raimo’s deficient representation.

At that stage in the proceedings, we reviewed the district court’s decision de novo. Dugas IV, 428 F.3d at 327. We recently articulated the rationale for our de novo review in this context in another case:

When the district court undertakes no independent factfinding in a habeas case, we are effectively in the same position as the district court vis-á-vis the state court record and have the ability to review that record from the same vantage point. Consequently, the district court’s recension of that record will engender de novo review.

Pike, 492 F.3d at 68. As noted, we found the state court record sufficient to “raise a genuine dispute of material fact concerning prejudice” and remanded to the district court for further evidentiary development. Dugas IV, 428 F.3d at 341. The district court, following our instructions, allowed Dugas additional discovery and held an evidentiary hearing at which both Dugas’s expert and the state’s experts provided new testimony. 4

The new fact-finding by the district court alters the standard of review that we must apply in this second appeal. Now that the district court has held its own evidentiary hearing, our review is comparable to the review that would apply if we were considering a direct criminal appeal or a § 2255 petition after a federal criminal conviction. See Dugas IV, 428 F.3d at 343; see also Pike, 492 F.3d at 68, 75; 2 Steven Aan Childress & Martha S. Davis, Federal Standards of Review § 13.06, at 13-50 (3d ed.1999). When the *8 district court has held an evidentiary hearing and made its own determinations as to the weight of the evidence and the credibility of witnesses, its findings are entitled to significant deference. Pike, 492 F.3d at 68; see also McNair v. Campbell, 416 F.3d 1291, 1297 (11th Cir.2005); McGregor v. Gibson, 248 F.3d 946, 951 (10th Cir.2001). Thus, we conduct a de novo review of legal issues, but disturb the district court’s factual conclusions only if they are clearly erroneous. Pike, 492 F.3d at 75; United States v. DiCarlo, 575 F.2d 952, 954-55 (1st Cir.1978)(applying clearly erroneous standard to district court findings in § 2255 proceeding). A finding of fact is clearly erroneous only when, upon a thorough assessment of the record, the reviewing court is left with “the definite and firm conviction that a mistake has been committed.” United States v. U.S. Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 92 L.Ed. 746 (1948); see also Pike, 492 F.3d at 75.

Before applying this standard of review in the present case, we have one further complexity to address&emdash;the interplay between the district court’s factual findings on disputed issues, all of which are subject to the clearly erroneous standard, and its ultimate determination on the prejudice issue, which is often characterized as a mixed question of law and fact. See, e.g., Scarpa v. DuBois, 38 F.3d 1, 9 (1st Cir.1994)(citing Strickland, 466 U.S. at 698, 104 S.Ct. 2052) (“[A]n inquiry into the effectiveness of counsel is almost always a mixed question of law and fact.”). The standard of review we apply to mixed questions of fact and law “depends, in the last analysis, on the extent to which a particular question is fact-dominated or law-dominated.” Pike, 492 F.3d at 68.

We remanded Dugas’s petition to the district court because we could not say, as a matter of law, that Raimo’s failure to consult an arson expert prior to trial rose to the level of prejudice required by Strickland. Dugas IV, 428 F.3d at 341. Instead, we characterized the prejudice inquiry as raising a “genuine factual issue.” Id. On remand, we instructed the district court to provide “the answer to a specific question&emdash;is there a reasonable probability that Higgins’s analysis of the chemical evidence and the use of evidence of smoke shadows and ventilation in cross-examining the fire investigators could have affected the outcome of the trial?” Id. at 342-43. The answer to this question required the district court to hear new testimony, consider its weight and credibility, and make numerous factual determinations that the trial judge is uniquely suited to make. See Anderson v. City of Bessemer City, 470 U.S. 564, 574-75, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985) (“[Ojnly the trial judge can be aware of the variations in demeanor and tone of voice that bear so heavily on the listener’s understanding of and belief in what is said.”); González-Soberal v. United States, 244 F.3d 273, 279 (1st Cir.2001) (remanding the prejudice question to the district court because that court “has a better perspective from which to evaluate the possible impact of [the omitted evidence] on the jury and its verdict”). Given the fact-dominated nature of the prejudice inquiry for which we ordered remand here, we review the district court’s decision on prejudice for clear error.

Dugas takes a decidedly different view of the role of the district court on remand. He argues that “[w]hen the district court found Higgins[’s] testimony to be admissible [because he was a qualified expert], it should have left the weight of that evidence to the jury and issued a writ of habeas corpus requiring a new trial.” However, this argument misunderstands the prejudice inquiry required by Strickland and the purpose of our remand. We did not ask the district court to make a deter *9 mination on the admissibility of Higgins’s testimony. Indeed, Higgins’s qualifications had never been questioned at any stage in these proceedings. If the mere admissibility of Higgins’s testimony had been determinative here, we would have simply granted Dugas’s petition in the first instance rather than remanding for additional factual development. The district court’s task on remand was to apply the prejudice prong of the Strickland test to a more developed record. In doing so, the district court was entitled — indeed, required — to make a judgment about the persuasiveness of Higgins’s testimony and then factor that assessment into the prejudice analysis. Moreover, Dugas does not argue, nor could he, that the failure to introduce a qualified expert’s testimony amounts to prejudice per se. See Scarpa, 38 F.3d at 13-14 (“[Ajttorney error, even when egregious, will almost always require analysis under Strickland’s prejudice prong.”).

In the Strickland analysis, prejudice exists when “ ‘there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.’ ” Dugas IV, 428 F.3d at 334 (quoting Strickland, 466 U.S. at 694, 104 S.Ct. 2052). “A reasonable probability is one ‘sufficient to undermine confidence in the outcome.’ ” González-Soberal, 244 F.3d at 278 (quoting Strickland, 466 U.S. at 694, 104 S.Ct. 2052). A showing of “ ‘some conceivable effect on the outcome’ ” is not enough. Id. (quoting Strickland, 466 U.S. at 693, 104 S.Ct. 2052). However, there is no requirement that “the defendant prove that the errors were more likely than not to have affected the verdict.” Id. Instead, Strickland requires that we focus on the “ ‘fundamental fairness of the proceeding.’ ” Id. (quoting Strickland, 466 U.S. at 696, 104 S.Ct. 2052).

“In weighing the prejudicial effect of counsel’s errors, we must consider the totality of the evidence before the judge or jury.” Stephens v. Hall, 294 F.3d 210, 218 (1st Cir.2002). In cases where defense counsel’s deficient performance resulted in a failure to introduce particular evidence or to challenge the credibility of the government’s witnesses on cross-examination, we have outlined three factors that need to be considered in the prejudice determination: first, the strength of the prosecution’s case; second, the effectiveness of the defense that was presented at trial; third, the potential value of the new evidence and new avenues for cross-examination “in undermining the credibility of the government witnesses’ testimony.” González-Soberal, 244 F.3d at 278.

In our first review of this case, when the issue was the appropriateness of the summary judgment disposition, we said that “[t]his case lay on a knife edge, and it would not have taken much to sway at least some jurors towards acquittal.” Dugas IV, 428 F.3d at 336. In that circumstance, we noted that “the threshold for prejudice is comparatively low because less would be needed to unsettle a rational jury” and, as a result, summary judgment was inappropriate. Id. at 336, 341. Nonetheless, this assessment did not alter the responsibility of the district court on remand to decide in the first instance if Higgins’s theories, now elaborated in supplemental affidavits and at an evidentiary hearing, “would have shaken the jury’s belief in the essential elements” of the government’s case at trial. See Stephens, 294 F.3d at 226.

III.

On remand, the district court considered two additional affidavits from Higgins. These affidavits were based on his initial inspection and testing of the fire scene in *10 August 2000 and on his examination in December 2005 and January 2006 of the store’s electrical system. Higgins also examined eleven samples taken from the fire that had been previously tested by the state’s expert and evidence held by the Nashua Police Department. The court also considered new affidavits from Richard Wood, a certified fire and explosion investigator, and Linda Bouchard, a Crimi-nalist II with the New Hampshire State Police Forensic Laboratory, filed by the state in response to Higgins’s analysis. Additionally, the court held an evidentiary hearing at which Higgins, Wood, and Bou-chard testified in response to questions from the court and counsel.

Higgins’s affidavits and hearing testimony challenged the state’s evidence on three grounds: the collection and chemical analysis of the samples from the fire scene, the origin and development of the fire, and the condition of the fire scene during the jury view. The district court concluded that Higgins’s testimony on these three points was insufficiently persuasive to undermine the evidence presented by the government’s witnesses and that, as a result, Dugas had not been prejudiced by his defense counsel’s deficient performance. As we explain below, we find no clear error with respect to this conclusion or the factual findings on which it is based.

A. Chemical Analysis

In our first review of this case, we were particularly concerned about the flaws Higgins had identified in the state forensic chemist’s analysis of the fire debris. Dugas IV, 428 F.3d at 334. We explained our concern by quoting Higgins’s state court affidavit:

[The applicable standard] requires the exact identification of flammable liquids (accelerants). [The state’s experts’] report ... gives a conclusion that they detected medium petroleum distillates ... As you can see they said they detected but nowhere does it say they identified. In other words, their conclusion says they might have a flammable liquid, but what they fail to say is that they may not have one.

Id. at 336. However, following our remand, Higgins himself examined the chro-matograms from the debris samples and identified the flammable liquid on two of them as charcoal lighter fluid. He testified at the evidentiary hearing that two of the chromatograms were “relatively identical” to the reference chromatograms the state had made from charcoal lighter fluid taken from the store. Thus, Higgins himself now seems to agree that even if the standards requiring specific identification of an accelerant had been followed, the resulting testimony would not have helped the defense.

In his testimony at the evidentiary hearing, Higgins instead pressed an argument that the chromatograms were abnormal in that they contained some of the lighter components of the accelerant. Lighter components, he explained, generally burn more quickly than the heavier ones and so the presence of the lighter components suggests that the sample had been contaminated or improperly collected. Bouchard effectively countered this contention at the hearing. She said that everything she saw in the chromatograms fell into the category of medium and not light hydrocarbons. Furthermore, if the debris had been thoroughly saturated with accelerant and not completely charred, some of the lighter hydrocarbons could be expected to remain in the sample. In addition, she explained that any minor discrepancies between the debris chromatogram and the reference chromatogram could best be explained by the fact that the charcoal lighter fluid sample could be tested directly while the fire *11 debris had to be heated in the lab to extract the chemicals in it and transfer them to a charcoal strip for testing. This means that “the chromatogram that you are going to get as a result of doing this kind of extraction is not going to be a true representation of that liquid if you ran the liquid straight.” Bouchard also indicated that, because of this extraction process, a direct examination of the charcoal strips, which had been lost by the time of our remand, would not have been helpful in determining whether some form of contamination of the evidence had occurred.

Weighing Higgins’s theory of contamination against this evidence, the district court concluded that Higgins’s critique of the chemical analysis by the state is “unpersuasive and insufficient to support a prejudice determination.” Dugas V, slip op. at 33. This finding is not clearly erroneous.

B. Origin and Development of the Fire

At trial, the state’s theory was that the fire at the Dugas Superette had been lit by Dugas a few minutes after 10 p.m. and that by 10:44 p.m. it had grown hot enough to short circuit the wires powering an electric clock in the store. The state’s experts concluded that the fire had begun quickly, generating substantial heat and smoke, but that, lacking an oxygen source, the fire had been reduced to a smolder when firefighters arrived just before midnight.

Both Higgins and the state appear to agree that if the fire had had a source of oxygen and had begun at roughly 10 p.m., it would have consumed the entire store before the firefighters arrived. The state explained that the entire store was not engulfed because of the lack of ventilation in the basement. In his affidavits, Higgins asserted that there was, in fact, a source of oxygen for the fire. He theorized that the fire was actively burning and growing when the firefighters arrived, but that the basement was too dark and the smoke too thick for the firefighters to properly assess this fact. Therefore, he contended, the fire must have started much later than the state argued at trial, and long after Dugas had left the building. 5

To support his theory, Higgins asserted that the basement door was partially open during the fire. The door to the basement at the Dugas Superette has two halves split vertically, each about eighteen inches wide, which may be opened independently. Higgins contended that the physical evidence shows that the right side of the door was open during the fire. The left side of the doorjamb is clean, indicating that it was closed and therefore protected from the smoke and heat, while the right side of the jamb is stained. He also pointed to a piece of melted plastic adhered to the basement side of the right door as evidence that it was open during the fire, allowing the plastic to melt onto it.

*12 Wood, the state’s fire investigator, asserted at the evidentiary hearing that the physical evidence shows that both sides of the door were closed when firefighters arrived. He explained the staining on the right side of the doorjamb by noting that one side of the door was propped open by firefighters’ hoses after they began to battle the fire, allowing smoke, heat, and soot to stain the jamb on that side. Additionally, he noted that the melted plastic banner extended across both sides of the door, indicating that the door was fully closed when the banner first melted.

Higgins’s ventilation theory rests on his contention that the basement door was open, acting as a chimney and allowing fresh air to be drawn in through a basement vent. As the district court noted, this theory directly contradicts the testimony of Lieutenant Keith Anderson of the Nashua Fire Department, who asserted at trial that the door to the basement was closed when he arrived and that the smoke and heat conditions worsened as soon as the basement door was opened. To counter this testimony, Higgins speculated that the firefighters did not realize that the door had two halves and that one half was already opened when they arrived.

The district court found that Higgins’s speculation would be unlikely to persuade the jury to discredit Anderson’s clear and specific testimony that the door was closed when firefighters arrived, and found Wood’s explanations for the staining on the doorjamb and the position of the banner more persuasive. As a result, the court found that “Higgins’s [ventilation] theory is not supported by the facts and is otherwise not credible.” There is no clear error in this finding.

C. Condition of the Fire Scene During the Jury View

Higgins also testified that the state had misled the jury during its view of the fire scene by not placing boxes precisely where they had been on the night of the fire. Higgins contended, based on his observation of smoke shadows, i.e. clean spots on the otherwise soot-covered walls and floor, that boxes of cooking oil were stacked three high in the narrow passage between a sandwich display case and a fish cooler, leaving only about four and a half inches between the boxes and the case. He insisted that, if these boxes had been present during the jury view, the jury would have understood that “the state’s theory of how Peter Dugas committed arson could not possibly be correct.” He asserted that it was “physically impossible for Dugas to have taken the path within the store the state claimed he took the night of the fire.”

However, Wood explained at the eviden-tiary hearing that even with the boxes stacked as Higgins claims, passing between the boxes and the cooler would still not be “physically impossible.” Wood described the fish cooler as “tapered from the base up to the top as most deli coolers are” so the passage was much narrower at the floor than at waist height. Wood testified: “Certainly you could put your foot in that space between the box and the fish cooler and then fit your upper body through the upper portion.”

At trial, Dugas was questioned about whether he could have fit through this narrow opening between the fish cooler and the display case. He answered, rather equivocally, “Not very well.” The prosecutor asked, “And were there things in the way?” Dugas responded, “Yeah, there’s a whole bunch of oil boxes there ...” Dugas was even asked by the prosecutor if the boxes were still where they had been on the night of the fire. Dugas responded, “I don’t remember. Mostly everything is still there. I know my dad took a box of oil out *13 of there at some point but-” Then the prosecutor asked again, “Could you have squeezed through there if you wanted?” Dugas again replied, “Not very easily. It’s a tight space and I’m not a small guy.” As the district court noted, this exchange illustrates that the jury already had before it the issue of whether the passageway between the cooler and the display case was too narrow for Dugas to pass through. Even when asked directly, Dugas never asserted that passing through with the boxes in place would have been “physically impossible.”

Weighing Higgins’s assertions regarding the location of the boxes against this testimony by Wood and Dugas, the district court concluded:

If Higgins’s opinion about the stacked boxes had been available to present at trial, it might have bolstered Dugas’s testimony that he could not have easily squeezed through that space, assuming Dugas would agree that the boxes were stacked as Higgins suggests. Higgins’s opinion, even if it were accepted as true, however, does not prove that the state’s theory was impossible. As best, it is cumulative of Dugas’s own testimony that he could not have fit through that area easily or very well. As such, Higgins’s opinion does little, if anything, to undermine confidence in the jury’s verdict.

Dugas V, slip op. at 37. There is no clear error in this finding.

IV.

We remanded this case to the district court for further proceedings that would permit the district court to determine whether “Higgins’s analysis would meaningfully challenge the state’s view of the evidence” and whether “ ‘there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.’” Dugas IV, 428 F.3d at 342 (quoting Strickland, 466 U.S. at 694, 104 S.Ct. 2052). At the conclusion of these proceedings, with a thoughtful explanation of its reasoning on each issue before it, the district court determined that Higgins’s theories were not persuasive enough to meaningfully undermine the government’s case against Dugas and affect the outcome of the trial. Therefore, the district court concluded that Raimo’s deficient representation did not prejudice Dugas within the meaning of Strickland. There was no clear error in this determination.

Affirmed.

. To support this second defense theory, Rai-mo attempted to show that P.J. Kulas, a former employee at the Superette, may have set the blaze. At trial, he sought to present evidence that Kulas was nearing the end of a criminal trial on aggravated sexual assault charges and that he would want to settle old scores he had against the Dugas family before he was convicted and sent to prison. Raimo presented testimony regarding Kulas’s resentment toward the Dugases and cross-examined Kulas’s wife regarding his whereabouts on the night of the fire. However, the trial judge did not allow Raimo to ask about the charges against Kulas or the trial. Dugas’s initial habeas petition challenged the exclusion of this evidence, but he did not appeal the district court’s ruling on that issue.

. The district court also rejected Dugas’s Confrontation Clause claim. Dugas did not appeal that decision.

. The charcoal strips and carbon disulfide vials prepared by Morris Boudreau, the state's forensic analyst, during his testing of the fire debris were not found following the remand. As such, Higgins never examined them. However, as we explain below, testing the original charcoal strips would not have materially aided Higgins in his contention that the fire debris had been contaminated through improper collection procedures. Thus, the loss of the strips and vials is immaterial to our analysis.

. AEDPA limits the availability of evidentiary hearings during federal habeas proceedings when a petitioner has failed to develop the factual basis of the claim at the state level. 28 U.S.C. § 2254(e)(2). However, the lack of factual development on the prejudice issue here was due to the state court's decision not to reach that issue and not because of any lack of diligence by Dugas. Thus, Dugas is not required to meet the stringent requirements of § 2254(e)(2) to obtain such a hearing. See Williams v. Taylor, 529 U.S. 420, 431-33, 120 S.Ct. 1479, 146 L.Ed.2d 435 (2000).

. A surveillance video introduced at trial showed Dugas leaving the store at 10 p.m., briefly reentering, and then leaving again a few minutes later. The state's theory at trial was that he set the fire during the brief reentry. The surveillance video system used by the store was outdated and barely functional. The video speed had to be adjusted to correspond to real time by a forensic video analyst, and the image had to be enhanced to make it more legible. Nonetheless, it was barely viewable and undated. Despite these flaws, Dugas admitted at trial to the sequence of events as depicted in the tape. Thus, for the purpose of this prejudice review, the weakness in the videotape evidence is immaterial.

Although Higgins never stated a particular time at which he theorizes the fire began, he seemed to suggest that the fire began sometime after 10:30 p.m. This timing, he claimed, undermined the state’s case by showing that the fire started well after the videotape evidence showed Dugas leaving the store at a few minutes past 10 p.m.