WILLIAMS, Circuit Judge.

Joyce Oliver-Thomas passed away sometime in the early morning of December 27, 2006. Her ex-husband and roommate Oscar Thomas was convicted of intentionally committing her murder (as well as first-degree sexual assault and false imprisonment). During the trial, the state's forensic pathologist testified that the autopsy findings were consistent with the application of intentional pressure to Oliver-Thomas's neck, resulting in manual strangulation and her death. Thomas argues that his trial counsel was ineffective for failing to consider and consult with an expert to review the pathologist's report and perhaps testify consistently with the defense's theory of the case, namely that Thomas unintentionally caused Oliver-Thomas's death by putting pressure on her neck for too long during sex. To show he received ineffective assistance, Thomas must demonstrate his counsel's performance was deficient and resulted in prejudice. Since the last reasoned opinion from the state courts did not address the performance analysis and applied the wrong standard to the prejudice analysis, we review Thomas's claim de novo. We agree with Thomas that a reasonable counsel would have consider and/or consulted with a forensic expert, especially when the state's expert testified there was no evidence of external bruising on Oliver-Thomas's neck but that the expert was still sure that this was intentional strangulation. Given the weakness of the state's case, especially as it relates to Thomas's intent, had counsel reached out to a forensic pathologist, or another expert similar to the habeas expert, and the expert testified, there is a reasonable probability the outcome of the trial would have turned out differently. Defendant's expert testimony would have highlighted the shortcomings in the medical evidence — the lack of external bruises on Oliver-Thomas's neck and lack of any signs of a struggle on either Thomas or Oliver-Thomas — and provided an expert, medical basis upon which the jury could have found reasonable doubt. Therefore we reverse the district court's denial of Thomas's petition and remand for proceedings consistent with this opinion.

I. BACKGROUND

A. Facts Surrounding Oliver-Thomas's Death

Thomas and Oliver-Thomas were married in 1990, had two children together, and divorced in 1999. Even after the divorce, Oliver-Thomas let Thomas live in her apartment, lent him money and helped him in other ways. The two would also occasionally have sex. They also fought, and there was testimony Oliver-Thomas threatened to kick Thomas out numerous times, including on December 26, 2006, but she never actually forced him out of the apartment.

Around 2 a.m. on December 27, 2006, Erica Cruz, the neighbor who lived directly below Oliver-Thomas woke to the sound of screaming. Cruz said she heard noises for about an hour. She testified that the noises included a woman screaming "Stop, stop, I love you. I love you" about three times, someone choking, and kicks and thumps on the ceiling. There was then ten minutes of silence, and then the sound of moving furniture, which she told police officers sounded like someone dragging a body a few feet. She then saw two men, including Thomas, leave the building. Cruz later heard someone go into Oliver-Thomas's apartment, walk around and say "oh my god move," either once (as testified at trial) or several times (as Cruz told police in statements).

Police received a call from Thomas at 3:24 a.m. saying Oliver-Thomas was unconscious. An officer arrived on the scene within minutes and found Oliver-Thomas unresponsive with her eyes open and without any pulse. She was pronounced dead in the hospital at 4:19 a.m.

While officers were tending to Oliver-Thomas, Thomas told two officers that he came back to the apartment and found Oliver-Thomas grabbing her neck in a choking manner before he called the police. Thomas then wrote a statement in which his story differed somewhat, and he said he discovered her unresponsive on the floor. A few hours later, Thomas voluntarily went to the police station to give another, more detailed written statement. He said around midnight he left the basement and went upstairs to Oliver-Thomas's apartment and began watching a pornographic movie. He went into the bedroom and he and Oliver-Thomas had sex, during which they fell off the bed and continued to have sex. Thomas left the apartment complex to get a cigarette. When Thomas returned, he found Oliver-Thomas on the floor.

After learning of the autopsy results — which we discuss in more detail later — the police confronted Thomas later that afternoon. He was interviewed for eight hours, towards the end of which he wrote a two-page statement. He reiterated his movie viewing and the ensuing sexual encounter, but this time added that he "had [his] left arm up around her neck, [his] right arm underneath her" while having sex. After they had sex, Thomas went out to the living room and watched more of the video. He then again went into Oliver-Thomas's room and:

Thomas did not say the two had a second sexual encounter. Thomas went to the basement and came back to find Oliver-Thomas "laying face down on the floor" making "gurgling sounds." The statement ends: "I do believe I was accidentally responsible for the death of Joyce."

B. The trial

In addition to presenting that evidence at trial, the state proposed two possible motives for Thomas's actions. First, it painted Thomas as desperate for money. It presented the testimony of two co-workers of Oliver-Thomas who said that Thomas called just hours after she died to ask about her paycheck. Oliver-Thomas's daughter also testified that the purse Oliver-Thomas used every day was missing. Second, the state suggested Thomas was upset about a relationship he perceived Oliver-Thomas as having with a co-worker. One of Oliver-Thomas's co-workers said Thomas was jealous of Oliver-Thomas and a male co-worker and at one point Thomas said "he was going to kill that mother fucker." The same witness admitted there was no evidence Oliver-Thomas and the co-worker were actually involved.

The state also put on the testimony of Dr. Mary Mainland, the coroner and the medical examiner for Kenosha County. Dr. Mainland testified that Oliver-Thomas had hemorrhages in her eyes and at least ten abrasions on her face. Dr. Mainland also found multiple hemorrhages inside Oliver-Thomas's neck and bruises to her thyroid and larynx. There were no marks around Joyce's neck, but Dr. Mainland stated, "It's possible that another part of the body [other than fingers] could have been used to inflict these injuries to her neck, such as an arm or a forearm." Dr. Mainland came to the conclusion that Joyce died from "strangulation and the strangulation was due to a physical assault." She testified, "This was not an accident," and estimated it would have taken roughly four minutes of continuous pressure to have caused Oliver-Thomas's death.

During its closing arguments, the state relied heavily on Dr. Mainland's testimony when arguing intent. It stated: "If there is any doubt about his intent or that his conduct was practically certain to cause her death," the jury should consider how long four minutes actually was. Defense counsel did not address the medical findings in his closing argument. In its rebuttal, the state said, "Here, we have that four-minute minimum where the Defendant was in fact choking the breath, the life, out of Joyce Marie Thomas. So there is no doubt about a long time in which he was reflecting, causing, continuing to kill her." Thomas was convicted and exhausted his direct appeals.

C. Post-conviction hearing

Thomas's state habeas counsel argued that Thomas was denied effective assistance by trial counsel, who did not reach out to a medical expert to review Dr. Mainland's findings. During the post-conviction proceedings, Thomas presented the report and testimony of Dr. Shaku Teas, a specialist in pathology and forensic pathology. Dr. Teas wrote, "it is my opinion that Joyce Oliver-Thomas died as a result of pressure on the neck and the autopsy findings are not inconsistent with Oscar Thomas' statement. There is no physical evidence that `intentional' pressure was applied to the neck."

Dr. Teas testified during the post-conviction hearing that some of the injuries indicative of strangulation were missing, e.g., external bruises on Oliver-Thomas's neck and the bone in the back of her neck that is often broken during strangulation was not in this case. "My issue," she testified, "is with the diagnosis of strangulation, whether it's intentional or unintentional." Dr. Teas testified that she could not "know a definite cause of death," but she was not saying she "cannot rule out strangulation." She also determined that there was "actually no evidence of manual strangulation" and none of the bruises or scratches on Oliver-Thomas's face are consistent with manual strangulation.

Trial counsel testified during the hearing that he considered retaining a pathologist to look into a sleep apnea defense, but Thomas stated that Oliver-Thomas had never been treated for the suspected illness and there was not any medical documentation, and so counsel did not pursue that defense. Counsel stated he did not "consider retaining a forensic pathologist to at least review Dr. Mainland's reports and findings to see if a forensic pathologist would have any disagreement with her findings."

The post-conviction court denied relief, finding no deficient performance or prejudice. The state appellate court affirmed, finding no prejudice, but did not address whether counsel's performance was deficient. The district court applied deference to the state post-conviction court's determination regarding trial counsel's performance and held that the state post-conviction court did not unreasonably apply United States Supreme Court precedent since "counsel had no reason to question Dr. Mainland's conclusion as to the cause of death" The court also found, under a de novo review, that Thomas was not prejudiced by counsel's performance since Dr. Teas "was unable to render an opinion on the cause of death" and "could not rule out strangulation." Thomas appeals.

II. ANALYSIS

Thomas argues he was denied effective assistance when trial counsel did not consult with an expert to review Dr. Mainland's findings or determine if his account was consistent with the medical evidence. To prevail on that claim, Thomas must show that (1) his counsel's performance was deficient, meaning it fell below the objective standard of reasonableness (the "performance prong"), and (2) that he was prejudiced by the deficient performance (the "prejudice prong"). Woolley v. Rednour, 702 F.3d 411, 420-21 (7th Cir.2012); see also Strickland v. Washington, 466 U.S. 668, 687, 688, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).

We review a district court's habeas decision de novo. Woolley, 702 F.3d at 420. We evaluate "the totality of the evidence — both that adduced at trial, and the evidence adduced in the habeas proceeding." Id. at 421. Under the Antiterrorism and Effective Death Penalty Act of 1996 ("AEDPA"), we are required to give deference to the "judgment of a State court" and will not grant habeas relief to "any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim — (1) resulted in a decision that was contrary to, or involved an unreasonable application of" clearly established Supreme Court precedent or "(2) resulted in a decision that 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); see also Williams v. Taylor, 529 U.S. 362, 376, 385-86, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000).

The parties first disagree on the deference owed on the performance prong. The post-conviction court addressed this prong but the appellate court, which issued the last reasoned opinion from the state system, did not. Clements argues that we nonetheless owe AEDPA deference to that prong and cites our decision in Atkins v. Zenk in which the trial court analyzed both prongs but the appellate court only analyzed one prong. 667 F.3d 939 (7th Cir. 2012). We stated: "Because both prongs have been addressed by Indiana state courts, in one form or another, the deferential standard of review set out in § 2254(d) applies to both." 667 F.3d at 944. Clements argues this means we should apply AEDPA deference to the performance prong even though the appellate court did not reach the issue. But, in Atkins, the standard of review was not subject to debate between the parties. Atkins conceded that his "habeas petition is subject to review under the new standards of 28 U.S.C. § 2254 as added by [AEDPA]" and said the question is "whether the Indiana Court of Appeals [sic] adjudication of those claims were [sic] contrary to or unreasonable application of Strickland." See Brief of Petitioner-Appellant at 5, Atkins v. Brown, No. 111891 (7th Cir. Aug. 8, 2011). The government agreed AEDPA deference applied to both prongs. Brief of Respondent-Appellee at 9-11, Atkins v. Brown, No. 11-1891 (7th Cir. Oct. 7, 2011). Without any adversarial challenge, there was no need for us to render a holding on the deference owed. See, e.g., McBride v. Houtzdale, 687 F.3d 92, 100 n. 10 (3d Cir.2012) (applying § 2254(d) deference to both prongs even though appellate court only ruled on prejudice prong because "McBride ha[d] affirmatively taken the position that AEDPA deference applies").

We do not read Atkins to alter our decisions before or after that have held that AEDPA deference is entitled to the "last reasoned opinion on the claim." Woolley, 702 F.3d at 421 (quoting Ylst v. Nunnemaker, 501 U.S. 797, 803, 111 S.Ct. 2590, 115 L.Ed.2d 706 (1991)). In Woolley, we held that "Unless a state-court opinion adopts or incorporates the reasoning of a prior opinion, `AEDPA generally requires federal courts to review one state decision.'" Id. (citing Barker v. Fleming, 423 F.3d 1085, 1093 (9th Cir.2005)). Since the "state appellate court declined to adopt the trial court's reasoning and instead remained silent on defense counsel's performance," and the appellate court's decision was the "last reasoned opinion," we gave deference only to the prong the appellate court did reach and reviewed the other de novo. Id. at 422. This is consistent with our precedent both before Atkins and since. See Ruhl v. Hardy, 743 F.3d 1083, 1091 (7th Cir.2014) ("In conducting federal habeas review under AEDPA, we look to the last reasoned state court opinion addressing each claim."); Quintana v. Chandler, 723 F.3d 849, 853 (7th Cir.2013) (noting the Woolley court "declined the state's request to apply 2254(d) deference to the state court on the ineffectiveness prong, holding that Strickland claims are divisible" and reviewing performance prong de novo); Earls v. McCaughtry, 379 F.3d 489, 492 (7th Cir.2004) (reviewing performance prong de novo when circuit court found deficient performance, but appellate court, in State v. Earls, 635 N.W.2d 905 (Wis.Ct.App.2001), decided case only on prejudice prong).

This conclusion about what deference must be given is supported by Supreme Court precedent. See Woolley, 702 F.3d at 421 (citing Ylst, 501 U.S. at 803, 111 S.Ct. 2590; Wiggins, 539 U.S. at 534, 123 S.Ct. 2527). It is also supported by AEDPA's plain language. The statute tells us to give deference to "any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim" falls under an exception in subpart (1) or (2).1 28 U.S.C. § 2254(d). AEDPA explicitly tells us deference is afforded to "the adjudication" — note the singular, rather than plural. Had Congress intended us to give deference to an amalgamation of adjudications, "it could have used different language." Cf. Grandberry v. Keever, 735 F.3d 616, 618 (7th Cir.2013) (interpreting statutory text). The exceptions also note that deference is not required when the state court's adjudication "resulted in a decision" either contrary to clearly established law or based on an unreasonable factual determination. 28 U.S.C. §§ 2254(d)(1) and (d)(2) (emphasis added). Again, the statute refers to a single decision, rather than multiple decisions. Based on Supreme Court precedent and the plain language of the statute, we reaffirm that we will give AEDPA deference to the "last reasoned opinion on the claim." Woolley, 702 F.3d at 421. Since the Wisconsin Court of Appeals did not adjudicate the deficiency prong, we review that prong de novo.

There is also disagreement over the deference we should give the appellate court's analysis of the prejudice prong. It said, "Thomas did not demonstrate that his trial counsel's failure to present Teas' testimony would have led to a different result at trial. [Citations omitted]." Clements concedes that the appellate court stated the wrong standard of review — Thomas has to show a "reasonable probability" that, but for counsel's unprofessional errors, the result would have been different. Strickland, 466 U.S. at 694, 104 S.Ct. 2052. He does not, as the appellate court said, have to show that counsel's performance would have led to a different result. Clements argues this prong is still entitled to AEDPA deference because the court "knew and applied the correct formulation of Strickland prejudice." But there is no evidence in the decision that it applied the proper standard. Cf. Sussman v. Jenkins, 636 F.3d 329, 360 (7th Cir.2011) (reviewing under 2254(d) deference even though court failed to cite "reasonable probability" language because "it is clear from the court's analysis that it did not believe that the note had a reasonable probability of altering the jury's verdict"); Charles v. Stephens, 736 F.3d 380, 392-93 (5th Cir.2013) (same). The state appellate court only used two sentences to address the prejudice prong and did not actually analyze why there was no prejudice, instead setting forth a matter-of-fact statement that there was no prejudice, all while applying the incorrect standard. The two sentences in the appellate court decision here cannot support Clements's argument. Thus, we find the appellate court's decision "involved an unreasonable application of" Strickland, and we review the prejudice prong de novo, making this completely de novo review. See Mosley v. Atchison, 689 F.3d 838, 850-51 (7th Cir.2012) (reviewing prejudice prong de novo when state appellate court "did not merely recite the wrong standard or use an inapt shorthand expression of the standard. It applied an incorrect and more onerous standard, and the difference may well have been decisive").

A. Thomas's Counsel Provided Deficient Performance

Thomas does not assert that trial counsel should have found an expert who would testify that Thomas did not cause Oliver-Thomas's death. Rather, Thomas argues that defense counsel was deficient in failing to consider and consult with a pathologist who would have reviewed the autopsy report and possibly testified. We agree with Thomas.

Time and again, the Supreme Court has highlighted how deferential we should be to the strategic and tactical decisions made by attorneys in performing their jobs. "[C]ounsel should be `strongly presumed to have rendered adequate assistance and made all significant decisions in the exercise of reasonable professional judgment.' To overcome that presumption, a defendant must show that counsel failed to act `reasonabl[y] considering all the circumstances.'" Cullen v. Pinholster, ___ U.S. ___, 131 S.Ct. 1388, 1403, 179 L.Ed.2d 557 (2011) (quoting Strickland, 466 U.S. at 690, 688, 104 S.Ct. 2052). There are "countless ways to provide effective assistance in any given case" and that is why counsel's decisions are afforded a "heavy measure of deference." Id. at 1407-08 (internal quotations omitted). To limit second-guessing, we must "judge the reasonableness of counsel's challenged conduct on the facts of the particular case, viewed as of the time of counsel's conduct." Strickland, 466 U.S. at 690, 104 S.Ct. 2052.

Were the state court's determination reviewed under the AEDPA deference, we might come out a different way. However, reviewing this case de novo, we find that counsel's performance in relation to a pathologist expert was deficient.

It is undisputed that counsel did not reach out to or even consider talking to a pathology expert to review Dr. Mainland's conclusion. In many cases, we would chalk such a decision up as strategic or tactical. See Hinton v. Alabama, ___ U.S. ___, 134 S.Ct. 1081, 1089, 188 L.Ed.2d 1 (2014) (per curiam) ("The selection of an expert witness is a paradigmatic example of the type of strategic choic[e] that, when made after thorough investigation of [the] law and facts, is virtually unchallengeable" (internal citations and quotation marks omitted)). But we cannot reach such a conclusion because counsel admitted his failure to reach out to an expert was not a conscious decision — he just did not think to do so. See Woolley, 702 F.3d at 423 (affording no deference to counsel's strategy choice because "[t]hough we often defer to an attorney's calculated decision to forgo a certain trial strategy, it is undisputed that there was no strategic rationale underlying these errors"); Carter v. Bowersox, 265 F.3d 705, 716 (8th Cir. 2001) ("The presumption that counsel's failure to raise the due process claim was a tactical decision, however, is undermined by counsel's affidavit that the instructional error was simply overlooked."). This inaction fell below the standard or professional norm since "counsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary." Strickland, 466 U.S. at 691, 104 S.Ct. 2052. Here, counsel admits to doing neither. While we appreciate counsel's candor in assessing his own performance and recognize that he was presented with a difficult case, that does not excuse his failure to even reach out to an expert under these circumstances, and thus, we give no deference to counsel's uncalculated actions. See Woolley, 702 F.3d at 423; Earls v. McCaughtry, 379 F.3d 489, 494 (7th Cir.2004) (affording no deference to counsel's decisions when, "We can think of no strategic reason why Earls' counsel would not have objected to the pieces of questionable testimony going to this issue; indeed, counsel admits such failures to object and redact were unintentional `oversights'").

The state had to prove that Thomas "cause[d] the death of another human being with intent to kill that person." Wis. Stat. 940.01(1)(a). To show cause, the state had to show that his act was a "substantial factor" in producing death. State v. Below, 333 Wis.2d 690, 799 N.W.2d 95, 101-02 (Ct.App.2011). "`With intent to' or `with intent that' means that the actor either has a purpose to do the thing or cause the result specified, or is aware that his or her conduct is practically certain to cause that result." Wis. Stat. 939.23(4); see also State v. Weeks, 165 Wis.2d 200, 477 N.W.2d 642, 644-45 (Ct.App.1991) (defining intent element). Here, counsel had a client who admitted to having his arm around the neck of the victim before she died and to causing her death. He also had a state pathologist's report finding the cause of death to be manual strangulation. Based on those facts, counsel made the wise decision to forgo arguing causation and instead argued that his client lacked the specific intent to cause Joyce's death. As counsel put it, "I didn't really see an issue in terms of cause of death. My issue was probably going to be manner of death, whether it was homicide — intentional homicide — or some other lesser form."

But that is what makes counsel's inaction deficient. By not reaching out to an expert to review or challenge Dr. Mainland's findings, counsel acquiesced to the state's strongest evidence of intent despite its perceivable flaws. Counsel knew or should have known that the state was going to use Dr. Mainland's testimony to show Thomas acted intentionally based on Dr. Mainland's pretrial testimony. She stated that, "The internal injuries were too severe and the pattern just simply doesn't fit" the possibility that this was anything but strangulation. She ended her testimony by saying that "I always keep an open mind while I'm doing an autopsy. But once I got to the neck organs, I was pretty certain as to what had happened." (App. R. 22-4, 44). In other words, counsel should have known Dr. Mainland was going to testify this was an intentional death.

Counsel also knew his client had said the death was unintentional and the result of what counsel later referred to as horseplay. Counsel knew there were no external marks on Oliver-Thomas's neck and no signs of any fight or struggle between Thomas and Oliver-Thomas. Counsel should have known there was reason to question a finding of intentional homicide. Based on those facts, a reasonable counsel would have at least reached out to a pathologist to see if the medical findings could be reconciled with Thomas's versions of the events. To not even contact an expert, however, was to accept Dr. Mainland's finding of intentional death without challenge and basically doom defense's theory of the case.

Nor was this a case where counsel's cross-examination of Dr. Mainland made up for the lack of expert. See Harrington v. Richter, 562 U.S. 86, 111, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011) (finding no deficient performance where defense counsel's cross examination of the state's experts "elicited concessions" from the experts and drew "attention to weaknesses in their conclusions"). In fact, it turned out to be the exact opposite. The state brought out during Dr. Mainland's direct examination the lack of fingerprint marks around Oliver-Thomas's neck, and it came out during Dr. Mainland's testimony that there was no evidence of a physical struggle between Oliver-Thomas and Thomas. Yet, without any medical testimony to tie those facts to an unintentional death, the best defense counsel could do was ask the state's expert whether she disagreed with her own diagnosis and thought the death could be an accident. This line of questioning fell flat:

It also prompted the state to follow up with:

If anything Dr. Mainland "repeatedly denied any alternative explanations," and "[w]ithout a countering defense witness, [the expert's] denials in the face of cross-examination only reconfirmed the one-sidedness of the expert opinion before the jury." Woolley, 702 F.3d at 424.

That is not to say reasonable performance requires retaining an expert every time the state does or every time the state presents the testimony of a forensic expert. Recently, the Supreme Court stated "[c]riminal cases will arise where the only reasonable and available defense strategy requires consultation with experts or introduction of expert evidence, whether pretrial, at trial, or both." Harrington, 562 U.S. at 106, 131 S.Ct. 770. However, "[t]here are ... countless ways to provide effective assistance in any given case," and "[r]are are the situations in which the `wide latitude counsel must have in making tactical decisions' will be limited to any one technique or approach." Id., 131 S.Ct. at 788-89 (quoting Strickland, 466 U.S. at 689, 104 S.Ct. 2052); Woolley, 702 F.3d at 424 ("[D]efendants [do not] enjoy an automatic entitlement to expert rebuttal witnesses whenever the government offers expert testimony in a trial," but there are times where the government's expert's conclusion require "expert illustration by the defense in order for the jury to weigh the evidence fairly."); Rogers v. Israel, 746 F.2d 1288, 1294 (7th Cir.1984) (collecting cases and noting that "[i]n several cases, the failure to investigate and present expert testimony has been found to be a matter of trial tactics within the range of reasonable performance"; "[y]et, under certain circumstances, `it may be vital in affording effective representation to a defendant in a criminal case for counsel to elicit expert testimony rebutting the state's expert testimony'").

We faced a somewhat analogous situation in Rogers v. Israel, when the petitioner fired one bullet in response to a feud with a fellow bar patron. 746 F.2d at 1289. The patron then charged at the petitioner and another shot was fired during the ensuing struggle. Id. The patron ultimately died from a gunshot wound. Id. The state's theory was that the petitioner intentionally shot the patron with the first bullet and the second shot went into the ceiling. Id. at 1290. The defense theory was that the first bullet went into the ceiling and the second, unintentionally fired bullet was the fatal one. Id. The prosecution called a pathologist who testified that, even if the patron were shot in the heart with the first bullet, he would have been able to engage in the ensuing struggle. Id. The defense presented no counter evidence to rebut this testimony. Id. After the petitioner was found guilty of first-degree murder, he alleged ineffective assistance of counsel on habeas review and presented the testimony of a pathologist who said that, in his experience, individuals who suffered heart wounds comparable to those of the patron were "immediately incapacitated upon receiving the wounds." Id. There was "no question" that petitioner caused the patron's death — the petitioner fired the bullet. Id. at 1292. Nonetheless, counsel's performance was deficient because he failed to "ask a qualified expert whether [the patron] would have been immediately incapacitated by his wound." Id. at 1295.

In other words, it was clear that the petitioner had caused the death, and the only issue was whether the death was intentional. The state presented an expert whose testimony was used to strongly support its theory of intentional death, and defense counsel never pursued any rebuttal expert. We found that performance deficient, and such is the case here. See also Dugas v. Coplan, 428 F.3d 317, 329-30 (1st Cir.2005) (finding counsel ineffective in arson case for not consulting expert when "much of Dugas's defense ... depended on [counsel's] ability to convince the jurors that the State's experts might be wrong," "the arson evidence was the cornerstone of the state's case," and counsel did not have technical proficiency to present defense without expert). As in Rogers, counsel's failure to even reach out to an expert was deficient.

B. Thomas Was Prejudiced By His Counsel's Deficient Performance

We next turn to the prejudice prong. To prevail, Thomas must show "there is a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different." Cullen, 131 S.Ct. at 1403 (citing Strickland, 466 U.S. at 694, 104 S.Ct. 2052). It is not enough to show that counsel's performance had an effect on the outcome or that "it is possible a reasonable doubt might have been established if counsel acted differently." Harrington, 562 U.S. at 111, 131 S.Ct. 770. "`A reasonable probability is a probability sufficient to undermine confidence in the outcome.' That requires a `substantial,' not just `conceivable,' likelihood of a different result." Cullen, 131 S.Ct. at 1403 (quoting Richter, 131 S.Ct. at 792). Our conclusion must keep in mind that "a verdict or conclusion only weakly supported by the record is more likely to have been affected by errors than one with overwhelming record support." Strickland, 466 U.S. at 696, 104 S.Ct. 2052.

Again, were the facts reviewed under the AEDPA deference, we might come out differently. However, reviewing this case de novo, we find that Oliver was prejudiced by counsel's deficient performance.

The state's case was not ironclad by any stretch of the imagination. Its theories of motive — that Thomas wanted money, that he was concerned that Oliver-Thomas would finally kick him out after all the previous threats, or that Thomas was jealous of Oliver-Thomas's relationship with a coworker — were all weak. Its case also suffers from a very serious flaw in terms of timing. The state's theory was that there was an altercation between Thomas and Oliver-Thomas. Cruz, the neighbor, testified that she heard noises (including screaming) above her for an hour. Yet Dr. Mainland said strangulation would likely result after four minutes of pressure, meaning there were roughly fifty-five minutes in which Oliver-Thomas was not being choked to death. Had there been an altercation, one may expect to find signs of a struggle after fifty-five minutes of conflict, but all parties admit there is no evidence of external marks on either Thomas or Oliver-Thomas. The state presents no explanation for this. Nor does the state explain how a woman being choked to death can scream "Stop, stop, I love you, I love you" loud enough to be heard one floor below her.

Dr. Teas's testimony,2 combined with Thomas's statement to the police, does explain what happened. Dr. Teas stated that the lack of external bruising on Oliver-Thomas's neck and lack of signs of struggle are not inconsistent with Thomas's story that he had his arm around her neck during sex and the rough housing, and that could have caused her death. This theory of the case could also explain the noises for an hour — the sex and then the rough housing, albeit with a break in between — rather than for just five minutes. Dr. Teas's testimony, therefore, provides a reconciliation of these facts and a medical foundation for the defense's argument that even if Thomas physically caused her death with a part of his body (which no one contests on appeal), he did not do so intentionally. As Dr. Teas admitted and the law requires, she could not testify as to Thomas's state of mind, but her testimony reconciles Thomas's statement and the evidence, which leads to the logical conclusion that he did not commit the act intentionally. See Steele v. State, 97 Wis.2d 72, 294 N.W.2d 2, 13-14 (1980) ("What we bar from introduction at the guilt phase of the trial is expert opinion testimony tending to prove or disprove the defendant's capacity to form the requisite criminal intent.").

Thomas's intent was one of the linchpins of the case, if not the key point, and yet defense counsel presented no affirmative evidence that Thomas did not have the requisite intent to commit the crime. It is true that defense counsel did draw out some evidence that Thomas did not commit the act intentionally, but that effort was not effective. For example, the lack of external bruises was discussed during Dr. Main-land's testimony; however she quickly rejected that absence as inconclusive and stated twice that this could not have been an accident. Conversely, Dr. Teas noted in her report that there is "no anatomical evidence to classify this death as `manual strangulation'" and stated affirmatively that "Oliver-Thomas died as a result of pressure on the neck and the autopsy findings are not inconsistent with Oscar Thomas's statement. There is no physical evidence that `intentional' pressure was applied to the neck." Her conclusion, therefore, undermines the state's already weak case on Thomas's intent. Had the jury been presented with this testimony, instead of just an argument unsupported by expert testimony as it was, it is substantially likely that Thomas could have raised at least a reasonable doubt and had a different outcome at trial. Therefore, Thomas has shown that there is a reasonable probability the outcome of the trial would have been different had counsel provided adequate representation. See, e.g., Rogers, 746 F.2d at 1295 (finding prejudice since post-conviction expert would have rebutted testimony of state expert that physical evidence supported conclusion of intent); Showers v. Beard, 635 F.3d 625, 631-34 (3d Cir.2011) (finding counsel's performance deficient and prejudice when defense failed to hire expert to determine if the taste of the drug could be masked and therefore whether death was a result of suicide or intentional homicide).

There is a shortcoming that weakens Dr. Teas's report. Namely, she equates Dr. Mainland's testimony of manual strangulation as only strangulation by the hands, but Dr. Mainland clearly testified at trial she used the phrase "manual strangulation" to also include strangulation by other parts of the body, e.g., the forearm. Nonetheless, the de novo review of the record leads to the conclusion that Dr. Teas's ultimate determination that the facts are consistent with an accidental death is sufficient to raise a reasonable doubt and therefore show prejudice for ineffective assistance of counsel purposes. See, e.g., Strickland, 466 U.S. at 695, 104 S.Ct. 2052 ("[T]he question is whether there is a reasonable probability that, absent the errors, the factfinder would have had a reasonable doubt respecting guilt.").

III. CONCLUSION

For these reasons, we REVERSE the district court's denial of Thomas's petition and REMAND for proceedings consistent with this opinion.