dissenting:
I respectfully dissent.
At a family gathering the morning after his birthday party, defendant General Waiters shot at the son of his girlfriend; but, shooting wildly, he wounded his target, hit three others, and killed two, including a toddler. He did this in a drunken rage. He is not entitled to much indulgence; but he was entitled to a fair trial. He did not get one because his counsel failed to arrange entry into evidence of a hospital record showing his blood alcohol at a potentially lethal level.
The United States District Court for the Eastern District of New York (Gleeson, J.) granted a writ of habeas corpus on the ground that defense counsel was constitutionally ineffective, that prejudice was established, and that the constitutional violation survives such deference as is owed to the rulings of the state court.
The two elements of ineffectiveness under the federal Constitution are deficient performance by counsel below and prejudice. The majority opinion goes on the assumption that “counsel made errors so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment,” Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), but reverses anyway on the ground that the showing of prejudice is insufficient to withstand the deference we owe to state courts in habeas cases. Because I vote to affirm, I need to *485consider counsel’s performance as well as prejudicial impact.
I
Defense counsel failed to adduce the only effective evidence in support of the only defense that was pressed on summation. Waiters’s only conceivable defense was lack of intent, and the only conceivable way he could show lack of intent was by alcohol impairment. The jury charge allowed the jury to consider the impact of alcohol on the question of intent; defense counsel on summation argued alcoholic impairment of intent; and the prosecution joined issue on that point at closing.
On the issue of intoxication, defense counsel held a largely bad hand. The impairment defense was undercut at trial by the defendant’s girlfriend, whose family members were decimated by the gunfire. She testified that, on the morning of the shooting, she was drinking with him for half an hour but she could not say he had imbibed more than a single drink. Her son, whom Waiters shot, testified that Waiters’s words were “a little bit slurred,” but also testified that he had seen Waiters “way worse” at least ten other times. This was the only testimony bearing on intoxication as the case went to the jury.
The only hope of the defense (and a potent one) was a hospital record showing that Waiters’s blood alcohol content (“BAC”) was .39—five times the legal limit for driving; indicative (for a person of defendant’s 130 pounds) of sixteen drinks the morning of the shooting; and typically enough to cause “significant cognitive impairment” including “impairment on judgment,” “blackouts,” “amnestic effects” and “potentially even unconsciousness.” App’x at 81. And death. This was no mere incremental detail; it was powerful evidence of an otherwise forlorn defense.
But defense counsel inexplicably never played that trump card. He did attempt to enter into evidence the defendant’s post-shooting hospital records, but the state trial court allowed in just two lines of them, and excluded the BAC number on the (valid) ground that expert testimony was needed to explain what it meant. The court was more than happy to give defense counsel time to find an expert to introduce and explain the BAC record, and everyone in the courtroom (except defense counsel) seemed to appreciate the import of that evidence: the state trial judge repeatedly urged defense counsel to call an expert to introduce it, and the prosecution was willing to adjourn to allow the defense to find an expert. See infra at 489-90. Defense counsel dug in. When the court informed him that it would not allow the BAC information in unless he called an expert, defense counsel told the court he would not call any more witnesses.
Thus the state trial court allowed into evidence just two lines of the medical records, both of which blandly observed that Waiters was “intoxicated.” Neither line included the BAC level, and neither line could answer the crucial question of whether Waiters’s intoxication was sufficient to impair intent. Defense counsel seemed not to understand how much more significant the .39 BAC was than the reference to his client as “intoxicated”: in the evidentiary colloquy with the state trial judge, he failed to zero in on that number; and at the post-conviction hearing, he would not agree that it signified that Waiters had been “very drunk.” App’x at 129.
Unsurprisingly, the prosecutor’s closing statement on the intoxication defens'e focused on problems that would have been resolved by the BAC record. See infra at 487-88.
The refusal by defense counsel to introduce his only powerful piece of evidence is *486simply unaccountable. At trial, he offered no explanation. He seems not to have discussed the issue with his client before or during the evidentiary colloquy. At the hearing to vacate Waiters’s conviction under New York Criminal Procedure Law § 440.10, conducted by the same state trial judge, defense counsel could not recall why he let pass the offered opportunity to call an expert. A pity: I would be deeply curious to know.
Without deciding the adequacy of counsel’s representation, the majority nevertheless posits several reasons why a lawyer might do what defense counsel did here. None of it washes. It is true that cross-examination of the expert would elicit the concession that Waiters probably built up resistance by long-term alcohol abuse. But every piece of evidence is subject to cross-examination and attack. As Judge Gleeson observed:
No reasonable defense attorney would opt to keep the jury ignorant of Waiters’s astronomically high BAC on the off chance that it might not be astronomical enough ... Trial counsel’s decision simply was not a tactical choice; it was incompetence.
App’x at 20, 22.
The majority also speculates that defense counsel may have been afraid to ask for an adjournment because that would have given the prosecution time to seek additional evidence of Waiters’s intent. But the majority does not extend that speculation to come up with any particular piece of evidence the prosecution might have sought out. I can’t come up with one either.
The majority opinion further posits that the expert testimony might open the door to statements Waiters made to psychological evaluators, in which he denied that he had been drunk at the time of the shooting (and the state trial judge made a similar point).1 But the effect of the admissions is easy to discount. Many drunks profess perfect sobriety, and no rational juror would credit an individual’s statement that he was sober over scientific evidence that he had sixteen drinks in the prior hour.
II
The majority opinion relies solely on its conclusion that the showing of prejudice was insufficient even to establish a “reasonable probability” that the outcome would have been different—i.e., that the jury might have acquitted, or hung, or convicted on a lesser charge, Strickland, 466 U.S. at 695, 104 S.Ct. 2052—and that in any event we owe deference to the state trial court’s ruling after the post-trial hearing.
Consider how the prosecution would fare in this ease if it had withheld from the defense a document showing that, one hour after the shooting, the defendant had enough alcohol in his blood to kill a normal person. The prejudice would be seen as obvious and the effect palpable.
As a matter of law, the result here can be no different. The standard for gauging prejudice in a case of ineffectiveness is derived from Brady, 373 U.S. 83, 83 S.Ct. 1194 (1963); and the Supreme Court has told us that the standard for assessing *487prejudice is the same in Brady claims as in claims for ineffectiveness.2 Tellingly, the majority opinion does not assert that it would reach the same finding on prejudice if the Strickland and Brady standards were the same. Instead (proceeding by footnote), the majority opinion agrees that the Strickland standard derives from the Brady standard and uses the same words, but holds that the two standards are somehow different. In doing so, it cites no relevant law, and splits from published decisions by every other circuit in the country.3
Applying the Strickland standard (as derived from Brady), prejudice is clear because the evidence that Waiters had absorbed a potentially lethal intake of sixteen drinks was the vital corrective to Warren’s testimony that she had seen him have only one drink that morning in about half an hour. It is easy to conclude here, as we would in a Brady case, that the evidence the jury never got to see would have “alter[ed] the entire evidentiary picture,” such that acquittal was at least reasonably probable. Strickland 466 U.S. at 696, 104 S.Ct. 2052.4
*488This was not lost on the prosecutor. In summation, the prosecution recalled testimony that one victim smashed an aquarium over Waiters’s head, and discounted the hospital note that Waiters was “intoxicated” with the rhetorical question:
How are you able to tell intoxication from the effects of getting hit over the head with a fish aquarium? Who knows?
Supp. App’x at 86-87. “Who knows?” The prosecutor knows. He knew that the excluded BAC information would establish not only that the impairment was due to alcohol, but that Waiters was so intoxicated he could have fallen over dead.
The majority opinion advances two arguments to discount the possibility of an acquittal even if the BAC record had been allowed into evidence with an expert explanation. First, the majority opinion claims that the jury was “well aware ... that [Waiters] was intoxicated when he emptied his revolver in the direction of Warren’s family.” Maj. Op. at 481. A review of the record shows that there was actually very little evidence of Waiters’s intoxication before the jury when it went to deliberate. His girlfriend testified that her fight with Waiters began when she told him he had too much to drink, but (as set forth above) she also testified that she did not know whether he had more than one drink that morning, and her son testified that, though Waiters’s words were “a little bit slurred,” he had seen Waiters “way worse” at least ten other times. The only other evidence of intoxication was the two lines in the medical records saying that Waiters was “intoxicated.” But the prosecutor effectively attacked that medical evidence as unreliable and imprecise in his closing statement. See supra at 487-88.
None of that evidence would persuade a jury that Waiters was extremely intoxicated, as required for an intoxication defense. None of it has anything like the force of the BAC record, which (with an expert explanation) would prove that the 130-pound defendant must have had sixteen alcoholic drinks that morning before the shooting started.
The majority opinion also considers that there was scant hope for an intent defense because Waiters did a number of voluntary acts that show self-possession: he bought milk and cereal; he got the gun; he concealed it on his person; he taunted his target with having it. But when he then shot, he barely wounded the target with whom he stood face to face, and he accidentally shot three others, two of them fatally. In other words, the evidence that Waiters was sober enough to act with intent was vulnerable.
The evidence of intoxication that the jury saw did not approach the level deemed necessary in the jury charge (as set out in the margin).5 A jury that had the most powerful evidence in Waiters’s favor—a jury that knew he had a .39 BAC and that was told what that meant—could easily conclude that he was profoundly un*489der the influence, and had a diminished ability to form the requisite intent.6
Ill
The majority opinion is understandably reluctant to hold on the merits that it was not at least “reasonably probable” that evidence of the defendant’s prodigious intake might have changed the guilty verdict. Instead, the majority opinion leans on the concept of deference.
First, it is of course true that ineffectiveness claims generally involve deference. But the deference due under Strickland is deference to the tactical choices made by defense counsel—the “strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance.” Strickland, 466 U.S. at 689, 104 5.Ct. 2052. That goes to the “ineffectiveness” prong of Strickland, not to prejudice. Since the majority opinion punts on ineffectiveness and relies solely on prejudice, Strickland deference thus does no work for it. See Hardy v. Chappell, 849 F.3d 803, 825 n.10 (9th Cir. 2016); Evans v. Sec’y, Dep’t of Corr., 703 F.3d 1316, 1334 (11th Cir. 2013) (Jordan, J., concurring) (“[it] makes no sense to say that initial judicial review as to whether prejudice resulted from counsel’s deficient performance—on its own, before adding AEDPA deference—involves any deference”).7
Nor can the majority rely on deference to the state trial judge under AEDPA. The views of the trial judge are best manifested by her several efforts to focus defense counsel on the need to get an expert to give the testimony that would make admissible medical records containing the crucial .39 BAC. The judge went as far as she could go without committing advocacy:
THE COURT: I would agree with [the prosecutor] that the best way would have been to get someone from the hospital to certainly interpret the records as to those limited issues. Quite frankly, I don’t know why, you know, someone wasn’t called or at the very least to get an assessment by the doctor who appeared, even if it was by the people, something that would assist.
THE PROSECUTOR: Well, I’m not in a position of assisting [the defendant] your Honor.
THE COURT: I understand that. Nor should you be. But I’m saying that [expert testimony] is possibly one way that it could have been rectified.
•THE PROSECUTOR: My position is it’s the only way, your Honor.
What is [defense counsel] going to get up there and say? That this level means something? That this person’s observation is any better than any other person’s observation? That one person’s in-tox, what does it mean in a medical standpoint? Does it mean that he can’t form intent? I mean, this is all going to be speculation. What’s he going to do, testify? Then it becomes impossible for me to rebut. Nor should I have to.
*490Again, you say it might be the better course. We’re still not under a time crunch here, Judge. It can be done.
[]
THE COURT: So, if you’re seeking to have blood levels, the amount of ethanol alcohol in Mr. Waiters’s blood stream, then clearly there should be someone to explain it for the edification of the jury as to what, in fact, that would mean. Otherwise, there’s no point to seek to introduce those levels of ethanol alcohol in his blood stream.
DEFENSE COUNSEL: Your Honor, the defense at this point is not introducing any additional witnesses at this point.
We would request that the Court permit all the information in, and we’re prepared for the Court to make its ruling based on the defense not presenting any additional witnesses at this time.
THE COURT: May I see counsel at sidebar. (Off-the-record discussion held at the bench.)
It’s the Court’s understanding- that the defense is not going to be calling any medical witness in order to certainly enlighten the jury as to what the numbers mean contained in the medical records.
Supp. App’x 68-69, 71-72.
The majority opinion relies on the post-hearing conclusion of the state trial judge. She expressed her conclusion in a single clause, which does not help the majority at all:
[Tjhere is a little doubt the claims raised by the defendant would have served to change the jury’s verdict.
App’x at 56. That sentence is best read to reflect that the trial judge declined to rule on prejudice; after all, given her holding that defense counsel was effective, she did not need to decide prejudice. The majority opinion, however, urges that this sentence is afflicted by a pair of typographical errors: if the indefinite article “a” is removed and the little word “not” is inserted at a likely spot, the finding would coincide with the view of the majority opinion. (The majority opinion justifies its rewrite on the ground that the state habeas opinion—to which it urges deference—is full of typos.) I think it is possible that the state trial judge did intend to say the opposite of what she wrote. The context is ambiguous. But surely it is odd to rewrite a sentence that is grammatically sound in order to make it say something else for the purpose of giving it deference.
Finally, AEDPA deference does not mean that we use the rubber stamp. When a state court decision is unreasonable, we may grant habeas relief. Rompilla v. Beard, 545 U.S. 374, 380, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005). So even if the state court had held that there was no sufficient prejudice in this case, that decision would have been -unreasonable for reasons I need not repeat.
[[Image here]]
Defense counsel pursued an intoxication defense, but failed to introduce and explain evidence that his 130-pound client had sixteen alcoholic drinks before committing the crime. He never provided a reason for his omission, and none is conceivable. It is at least reasonably probable that a jury, hearing such potent evidence, would develop reasonable doubt as to the element of intent.
. After the post-trial hearing, the state trial court noted that introducing the .39 BAC into evidence would be inconsistent with Waiters's own protestations that he was sober the morning of the shooting. It may be that a defense lawyer is not obligated to argue an intoxication defense when his client denies being drunk; but once defense counsel chooses to argue a defense (and intoxication was the only defense he argued on summation), he has an obligation to do so effectively.
. Compare Strickler v. Greene, 527 U.S. 263, 289, 119 S.Ct. 1936, 144 L.Ed.2d 286 (1999) (holding that prejudice under Brady requires a " 'reasonable probability' that the result of the trial would have been different”) with Strickland v. Washington, 466 U.S. 668, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) (explaining that the test for prejudice in ineffective assistance claims is a "reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different,” and that the prejudice test in ineffectiveness cases “finds its roots in the test for materiality of exculpatory information not disclosed to the defense by the prosecution”).
. See Ruiz v. United States, 339 F.3d 39, 43 (1st Cir. 2003) ("Brady claims are subject to the same prejudice requirement as ineffective-assistance claims”); Marshall v. Hendricks, 307 F.3d 36, 85 n.37 (3d Cir. 2002) ("the Strickland prejudice standard is the same as the Brady materiality standard”); United States v. Higgs, 663 F.3d 726, 735 (4th Cir. 2011) ("[t]he standard for Strickland prejudice is the same as for Brady materiality”); Felder v. Johnson, 180 F.3d 206, 214 (5th Cir. 1999) (“the standard for prejudice under Strickland is identical to the standard for materiality under Brady” (internal quotation marks omitted)); Moreland v. Robinson, 813 F.3d 315, 330 (6th Cir. 2016) ("Strickland and Brady claims use the same 'reasonable probability’ standard to assess prejudice”); Harris v. Thompson, 698 F.3d 609, 645 (7th Cir. 2012) ("[wjhen a defendant is deprived of favorable evidence, the same 'reasonable probability’ standard applies to determining materiality under Brady and ... to determining whether the accused was prejudiced for the purposes of Strickland”); Clay v. Bowersox, 367 F.3d 993, 1000 (8th Cir. 2004) ("[t]he materiality standard under Brady is the same as the prejudice standard under Strickland”); United States v. Olsen, 704 F.3d 1172, 1187 (9th Cir. 2013) ("Brady materiality and Strickland prejudice are the same”); Romano v. Gibson, 239 F.3d 1156, 1172 (10th Cir. 2001) ("Brady's prejudice inquiry is equivalent to the prejudice analysis that applies to an ineffective assistance of counsel claim under Strickland”); Jennings v. McDonough, 490 F.3d 1230, 1243 (11th Cir. 2007) ("[t]he prejudice prong of Strickland incorporates the same standard used for assessing the materiality of evidence under Brady”); United States v. Baxter, 761 F.3d 17, 24 n.4 (D.C. Cir. 2014) (citing Montgomery v. Bobby, 654 F.3d 668, 679 n. 4 (6th Cir.2011) for the proposition that "[i]t is well settled that the test for prejudice under Brady and Strickland is the same.”).
.Waiters argues that his counsel was also ineffective for failing to impeach Warren on a crucial point. Her testimony that Waiters had had only one drink could have been impeached with her prior statement to the defendant’s investigator that Waiters drank a pint of liquor by himself that morning. "In evaluating prejudice, we look to the cumulative effect of all of counsel’s unprofessional errors,” and that error can only exacerbate the existing prejudice. Gersten v. Senkowski, 426 F.3d 588, 611 (2d Cir. 2005). The mandate in this case, which vacates and remands for consideration of this claim of ineffectiveness (and another), therefore leaves open the way to an ultimate ruling on remand that *488Waiters's counsel was ineffective and that prejudice resulted.
. Supp. App'x at 88:
“Now, jurors, under our law, intoxication is not as such a defense to a criminal charge. But evidence of the defendant’s intoxication may be considered by you whenever it is relevant to negative an element of the crime charged.
"So, in determining whether the defendant had the intent necessary to commit a crime, you may consider whether the defendant’s mind was affected by intoxicants to such a degree that he was incapable of forming the intent necessary for the commission of any of the crimes charged that I just submitted to you."
. The majority opinion also cites several New York state cases on the intoxication defense. None of them is relevant. The question here is not sufficiency of the evidence or entitlement to an intoxication charge; it is whether there is a "reasonable probability” that—if an expert had explained to the jury what a .39 BAC meant—the jurors would have convicted on the lesser charge, or acquitted, or hung. Strickland, 466 U.S. at 694, 104 S.Ct. 2052.
. The majority goes as far as is prudent in suggesting (without holding) that "double deference” may be owed as to prejudice, i.e., deference (uncontroversially) to the trial court as well as deference (oddly) to the defense counsel. But I concede there is a circuit split on this, and that some other courts have made the same error.