In 1993 Glenn Bradford was convicted in an Indiana state court of a murder and arson committed in Evansville the previous year, and was sentenced to 80 years in prison, where he remains. In 2013 he filed this federal habeas corpus suit, in which he claims that he can prove his innocence. He asks for a new trial, which the district judge denied, precipitating this appeal.
*903Bradford, an Evansville police officer, was involved in an extramarital affair with a woman named Tamara Lohr. After his wife discovered the affair, Bradford decided to end it; but Lohr resisted, emailing him that if he left her she’d tell his wife the affair was continuing.
At 6:35 a.m. on a day a month or so after Lohr’s threats and just after Bradford had finished a night shift, he reported a fire at Lohr’s house. He told the responders that he’d entered the house to try to extricate her but had been unable to do so and believed she was dead. A firefighter had driven by the house at 6:30 a.m. and seen no signs of fire, but shortly after Bradford reported it the firefighter entered the house, extinguished the fire, and found Lohr’s corpse on her bed. He estimated that the fire had started only a few minutes before he arrived. Fire investigators found in Lohr’s bedroom an empty gasoline can from which gasoline had been poured onto the mattress on which Lohr’s body lay and onto the floor between the mattress and the bedroom door. They inferred that the gasoline had been ignited by someone standing in the bedroom doorway. Lohr, it was discovered, had been stabbed to death and her body had been burned after she died. Her poodle was found dead, also with stab wounds, in the living room, which was just outside the bedroom, but we’ll see that the dog was still alive when the fire started. From the conditions of the house and grounds the police concluded that the arsonist-murderer had staged a burglary rather than broken into the house.
Although the exact time of Lohr’s death could not be determined, the state believed that she’d been stabbed during the night. Suspicion quickly focused on Bradford, who appears to have concealed his whereabouts on the fatal night and thus could have murdered Lohr sometime during his night shift before returning in the morning to set the fire. He had no alibi for the period from about 11 p.m. to midnight. He claimed in an entry in his police activity log to have responded during that period to a hit and run accident, but the officer working that accident testified that he hadn’t seen Bradford at the accident scene — while an Evansville police car that could have been Bradford’s had been seen outside Lohr’s house at about 11 p.m. The evidence presented at the trial included not only Bradford’s seeming attempt to invent an alibi but also an attempt by him to delete Lohr’s threatening emails to him a couple of hours after the fire.
Although only 65 seconds elapsed between when a bank camera revealed Bradford two blocks from Lohr’s house and driving toward it and when he reported the fire, an investigator testified at trial that that was enough time for Bradford to have driven to the house, entered, spread the gasoline, and lighted it.
At trial, rival fire experts testified for the state and the defense about whether the fire had to have burned for more than eight minutes, in which event Bradford would have had a solid alibi because the fire was extinguished by 6:43:19 a.m. and he could not have set it before 6:35 given that he was seen by the bank camera two blocks away at 6:34:04 and that it would have taken him about a minute to reach the house from there, start the fire, and then call emergency services (which he did at 6:35:09) — and he’d been on police duty until about 6:30. No gasoline had been found on Bradford’s uniform and the police had failed to investigate another possible suspect — a man who had lost his job as a jailer for the Sheriffs department after Lohr accused him of sexual harassment, and whom she had subsequently reported to the police for parking outside her house in the middle of the night. There is, howev*904er, no evidence connecting that man with the arson or murder.
Bradford was convicted by a jury of murder and arson, and his conviction was upheld in Bradford v. State, 675 N.E.2d 296 (Ind. 1996), and his request for state post-conviction relief denied in Bradford v. State, 988 N.E.2d 1192 (Ind. App. 2013). He principally argues in the present proceeding that he couldn’t have been the arsonist (and if not, then presumably not the murderer either) because his expert witnesses testified that the fire must have burned for more than eight minutes. If that’s true Bradford could not have been the person who set the fire because, the reader will recall, eight minutes was the maximum time that the fire if set by Bradford would have lasted. Although his expert witness at trial testified that the fire had burned for 15 minutes, his principal expert witness in the post-conviction proceedings, Douglas Carpenter, went further, testifying that the fire must have burned for at least 30 minutes before it was extinguished. He based that estimate on his inspection of burned and unburned wood and other materials in the house and on the level of carboxyhemoglobin (COHb) in the poodle’s blood — a house fire generates carbon monoxide (CO) that can combine with hemoglobin, a constituent of blood.
Bradford relies on three facts asserted by Carpenter to support his contention that the fire must have lasted more than eight minutes: (1) the time it would have taken the fire to burn through the upper panel of the bedroom door; (2) the time it would have taken for the poodle to have accumulated the amount of COHb found in its blood; and (3) the time it would have taken the fire to consume the mattress on Lohr’s bed. These facts don’t make a case. Carpenter was unable to prove that the fire would have taken too long to burn through the upper panel of the bedroom door for Bradford to have set it. He conducted experiments under conditions that simulated a maximum temperature near the bedroom ceiling of 350°C. But the fire might well have been hotter than that. He contended that because none of the 12 panes of glass in the front bedroom window was broken, the room temperature near the ceiling (and thus near the top door panel) must have been below 280°C, and at that temperature it would he claimed have taken at least 30 minutes for the bedroom door to char through. He based this temperature claim on a study reported by Vytenis Babrauskas, “Glass Breakage in Fires” 2-3, www.doctorfire. com/GlassBreak.pdf (visited Aug. 2, 2016), of the probability of glass breakage at different temperatures. That study had been limited to glass three millimeters thick, and' he admitted not knowing the thickness of Lohr’s windows, though he testified that three-millimeter glass is “typical.”
He claimed that because none of the 12 panes of glass was broken, the temperature near the ceiling of Lohr’s room must have been below 280°C. That was the temperature at which the three-millimeter glass in the study on which he relied had a one-in-twelve probability of breaking. But a one-in-twelve probability of breakage at a particular temperature does not mean that if there are twelve panes one of them is certain to break, and Carpenter failed to justify his assumption that the more panes in the unbroken window, the lower the temperature had to have been. Moreover, he had no evidence that all the windowpanes had been exposed to the same temperature as the ceiling and the top panel of the door. In a normal room fire the highest temperature is at the ceiling and the temperature near the floor is much lower. See T. J. Shields, et al., “Performance of a Single Glazing Assembly Exposed to Enclosure Corner Fires of Increasing Severi*905ty,” 25 Fire & Materials 123, 125-26 (2001).
His testimony that the top panel of the door would have required about 11 minutes to char through even if the fire’s temperature had reached (though it hadn’t) 600°C (measured just below the ceiling, where the fire would have been at its hottest) misrepresented his own report, according to which the 11-minute estimate was based on the heat flux (the rate at which heat energy passes through a given surface per unit of time) that would have been produced at a near-ceiling temperature of 350°C. And as we said, Carpenter didn’t test char-through times for heat fluxes corresponding to temperatures higher than 350°C, even though the fire in Lohr’s house may have been hotter than that.
There was also no basis for his conclusion that it would have taken 30 minutes for the fire to have caused the COHb level discovered in the poodle’s autopsy. That estimate was based on the rate of carbon-monoxide uptake for an adult human being at rest, and such an estimate is not valid for a fifteen-pound dog who may moreover as we noted have been moving around in the presence of the fire. (What is clear, however, is that the poodle was alive when the fire started, because the only possible source of the elevated COHb found in its blood was inhalation of carbon monoxide (CO) produced by the fire.) Carpenter’s own source explained that the uptake rate is significantly higher for smaller bodies and bodies in motion. David A. Purser, “Assessment of Hazards to Occupants from Smoke, Toxic Gases, and Heat,” The SFPE Handbook of Fire Protection Engineering 2-116 (4th ed. 2008).
The fire duration that he inferred from the poodle’s COHb level was further inflated by his use of carbon-monoxide concentrations measured outside closed doors. The door to Lohr’s bedroom was open by about an inch because of clothing jammed into the opening, and so the dog, who was a small distance outside the bedroom, would have breathed more carbon monoxide from the fire raging in the bedroom than had the door to the bedroom been closed.
Carpenter concluded that Lohr’s mattress must have been completely consumed before the door charred through, for had it been consumed afterward additional air ventilation caused by the fire burning a hole in the door panel would have interacted with the mattress (fuel for the fire) to cause the fire to reach flashover conditions. Flashover conditions occur when the upper layer of gas in a room becomes hot enough (approximately 500°C to 600°C) to ignite all available combustible surfaces. Since it is undisputed that the room did not reach flashover, Carpenter concluded that the mattress must have been completely consumed before the door charred through. He used studies of rates of mass loss for flexible polyurethane foam (the material in Lohr’s mattress) to predict what the mass loss rate would be in a room on fire with the door open one inch. Estimating the maximum duration of the fire as two hours and the minimum as 30 minutes, he concluded that the fire must have been started between 4:30 and 6:00 a.m.
Bradford infers from this that it would have taken two hours for Lohr’s mattress to be consumed, and so the fire must have lasted at least that long. But two hours was Carpenter’s estimate of the maximum time for the mattress to be consumed. His estimate of the minimum time (30 minutes) if it were correct would exonerate Bradford, but the reliability of that estimate was undermined by the errors, discussed earlier, in Carpenter’s testimony and report.
*906Two officers who appeared at the fire scene three minutes after Bradford had arrived reported that smoke was coming from the eaves of the house. Carpenter argues that smoke couldn’t have come from the eaves until after the door had charred through because until then smoke would have gone out through the roof vents rather than sideways through the eaves. After the door charred through, however, the roof vents would be inadequate to vent all the smoke emerging from the bedroom and therefore smoke would go through the eaves as well as the vents. If Carpenter is correct, then, for Bradford to have been the arsonist, the fire had to have started, the door must have charred through, and smoke from the eaves must have appeared — all within three minutes of Bradford’s arrival at the house, since responders to the fire arrived three minutes after Bradford and observed smoke upon arrival. But as we’ve said, the char-through time if the fire was hotter than 350°C is unknown and we also don’t know how fast the smoke would have moved through Lohr’s house and out the eaves and roof.
It’s true that two joggers saw smoke coming from Lohr’s house at 6:36:07, about a minute after Bradford called emergency services and two minutes after he was recorded by the bank camera driving toward the house. Carpenter claimed that the joggers’ testimony corroborated his conclusion that Bradford could not have been the arsonist, since it is unlikely that Bradford would have been able to set the fire in time for it to emit visible smoke in fewer than two minutes. But this argument does not appear in Bradford’s briefs and the state’s expert witness testified that smoke would have been visible outside the house within a minute after the fire was set.
Bradford further argues that the evidence that he could have reached Lohr’s house and set the fire within 65 seconds (between 6:34:04 when he was photographed by the bank camera and 6:35:09 when he called emergency services) did not account for the time required for the murderer-arsonist to carry the, gas can from the back porch to Lohr’s room, turn off the electrical breakers, and stab the poodle. Not so. Although Lohr normally kept her gas can on the back porch, there is no evidence that it was there when the arsonist arrived that morning. Had Bradford brought it inside during his visit to Lohr between 11 p.m. and midnight, he could have used it at 6:34 a.m. without having to fetch -it from the porch. And he would have needed to turn off the electrical breaker in the morning only if Lohr’s neighbor was correct that he saw her light on at 12:30 a.m. That neighbor acknowledged, however, that he was estimating the time and had not been looking at a clock, so Bradford may have turned off the breaker earlier. Nor would it take long to stab a small poodle twice. The jury thus could reasonably find that 65 seconds was enough time for Bradford to arrive at Lohr’s house and complete all the tasks that the evidence indicated had been completed by the murderer-arsonist.
Faced with “a request for an evidentiary hearing” to determine whether a convicted defendant should be given a chance to prove his innocence in a habeas corpus proceeding, “the District Court must assess the probative force of the newly presented evidence in connection with the evidence of guilt adduced at trial.” Schlup v. Delo, 513 U.S. 298, 331-32, 115 S.Ct. 851, 130 L.Ed.2d 808 (1995). Bradford had his chance and failed to present rehable evidence that would establish his innocence of the arson and murder.
Changing course, Bradford argues that his trial counsel was ineffective in deciding to retain a fire expert named *907Barker Davie, who co-owned a fire-investigation business, attended training programs and wrote articles, and had testified many times as a fire and arson expert for the state — experience that gave him particular credibility as a defense witness. Although Bradford claims that Davie was not an expert on fire duration, there is no support for that claim. According to Bradford, Davie’s testimony that the fire lasted about 15 minutes (and definitely more than 9 minutes, so that Bradford could not have set it), though favorable to Bradford, was not convincing, because he utilized a wood-charring rate that the National Fire Protection Association had warned was not universally applicable as it was “based on [only] one set of laboratory conditions.”
The Indiana Court of Appeals can’t be said to have erred in holding that Bradford trial counsel did not render ineffective assistance to his client by selecting Davie as a witness. See Strickland v. Washington, 466 U.S. 668, 687-91, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). He was a qualified and well-regarded fire expert who offered a defensible though not definitive estimate of the fire’s duration. It was not a case in which counsel made no meaningful investigation or failed to present a defense expert on a critical issue. See, e.g., Thomas v. Clements, 789 F.3d 760, 768-69 (7th Cir. 2015); Richey v. Bradshaw, 498 F.3d 344, 362-63 (6th Cir. 2007); Dugas v. Coplan, 428 F.3d 317, 328-34 (1st Cir. 2005).
Bradford’s final claim is that his right to a fair trial was violated because the jurors visited Lohr’s house and timed whether Bradford could have set the fire in 65 seconds. Assuming that such an experiment was conducted (different jurors made contradictory reports of the matter), it was lawful because the jury was merely using its common sense to evaluate a timing issue critical to the case, and there has been no showing that it created an unacceptable risk of an incorrect verdict. See Kurina v. Thieret, 853 F.2d 1409, 1413-14 (7th Cir. 1988).
The judgment denying habeas corpus is
AFFIRMED.