Opinion by Judge FISHER; Dissent by Judge McKEOWN.
OPINION
FISHER, Circuit Judge:In 1989, a California jury convicted Steven Crittenden of two murders and sentenced him to death. Crittenden, who is African-American, filed a federal habeas petition, arguing the prosecutor excluded an African-American prospective juror on account of her race, in violation of the Equal Protection Clause of the Fourteenth Amendment, as interpreted in Batson v. Kentucky, 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986). The district court initially denied Crittenden’s petition. The court found, although race played a significant part in the peremptory challenge, the prosecutor would have made the challenge even if race had played no role, because of the prospective juror’s opposition to the death penalty. We remanded in light of Cook v. LaMarque, 593 F.3d 810 (9th Cir.2010), which clarified that a peremptory challenge violates the Equal Protection Clause if it is “motivated in substantial part” by race, id. at 815, “regardless of whether the strike would have issued if race had played no role.” Crittenden v. Ayers, 624 F.3d 943, 958-59 (9th Cir.2010) 0Crittenden I) (emphasis added).1 On remand, the district court found the prosecutor was substantially motivated by race, and granted Crittenden’s petition.
The state presents several challenges on appeal: (1) under Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989), the district court was prohibited from retroactively applying the standard articulated in Cook; (2) the district court failed to apply deference to decisions by the California Supreme Court and the state trial court, as required under the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA); (3) the district court improperly rejected the magistrate judge’s credibility determination without conducting its own evidentiary hearing; and (4) the district court clearly erred by finding the prosecutor was substantially motivated by race.
We have jurisdiction under 28 U.S.C. § 1291, and we affirm. First, Cook merely clarified the standard of proof for Batson claims; it did not set forth a new rule for purposes of Teague. Second, as we held in Crittenden I, the California Supreme Court’s decision is not owed deference under AEDPA, because it was contrary to clearly established federal law, and the presumption of correctness afforded to the state trial court’s factual findings is rebutted by clear and convincing evidence. Third, the district court was not required to conduct its own evidentiary hearing, because it did not reject the magistrate judge’s credibility determination. Finally, the district court’s finding that the prosecutor was substantially motivated by race was not clearly erroneous.
The Supreme Court has eloquently explained a jury selected without regard to race is a critical constitutional right:
The jury acts as a vital check against the wrongful exercise of power by the State and its prosecutors. The intrusion of racial discrimination into the jury selection process damages both the fact *1003and the perception of this guarantee. Jury selection is the primary means by which a court may enforce a defendant’s right to be tried by a jury free from ethnic, racial, or political prejudice, or predisposition about the defendant’s culpability. Active discrimination by a prosecutor during this process condones violations of the United States Constitution within the very institution entrusted with its enforcement, and so invites cynicism respecting the jury’s neutrality and its obligation to adhere to the law.
Powers v. Ohio, 499 U.S. 400, 411-12, 111 S.Ct. 1364, 113 L.Ed.2d 411 (1991) (citations and internal quotation marks omitted). Accordingly, it is well established that a Batson violation is structural error. See Williams v. Woodford, 396 F.3d 1059, 1069 (9th Cir.2005).
Given the district court’s careful analysis of the record and its consequent findings, Crittenden is entitled under Batson to a new trial before a properly selected jury. The district court’s judgment granting Crittenden’s habeas petition is affirmed.
BACKGROUND
Jury selection in the state trial court took place between November 1988 and February 1989.2 Initially, a pool of over 60 prospective jurors completed questionnaires asking them about their backgrounds and beliefs. Question 56 asked about their feelings regarding the death penalty. Manzanita Casey was the only African-American prospective juror. In answer to question 56, she wrote, “I don’t like to see anyone put to death.” She also wrote that she could set aside her personal feelings regarding what the law should be and follow the law as the court explained it.
After filling out the questionnaires, the prospective jurors appeared one-by-one for voir dire. During her voir dire, Casey reiterated her opposition to the death penalty. She also said, however, that her opposition would not prohibit her from voting for a first-degree murder conviction or the death penalty. At the conclusion of Casey’s voir dire, the prosecutor challenged her for cause, “based upon her answer that she doesn’t believe in the death penalty.” The court denied the for-cause challenge.
After each prospective juror completed voir dire and passed for-cause challenges, the prosecutor wrote a rating on his copy of that juror’s questionnaire. He gave favorable jurors one to four ‘Vs,” four being the most favorable, and gave unfavorable jurors one to four “X”s, four being the most unfavorable. The prosecutor rated Casey XXXX, the most unfavorable rating possible, and a rating he gave to only one other prospective juror of the over 50 who went through voir dire. The prosecutor later testified that, although he did not remember the basis for individual ratings, his general practice was to rate prospective jurors primarily based on their position regarding the death penalty— “Xs were ... I would say, to a person, you were opposed to the death penalty and strongly stated it.... Checkmarks were people who either were for the death penalty or medium ground that I thought to some degree I would be able to tolerate having on the jury.” He testified he also considered “people’s backgrounds, whether they’re employed, homeowners, what they had to lose. I wanted people who had something to lose in society, who might be victims of crime, solid citizens, preferably well educated.”
*1004A pool of over 40 prospective jurors who had gone through voir dire — including Casey — returned in February 1989 for the exercise of peremptory challenges. The court seated an initial group of 12 jurors. The prosecution and defense were allowed 26 peremptory challenges each. When a prospective juror was challenged, the court would seat another prospective juror who had gone through voir dire. The prosecutor based his challenges primarily on his ratings. He challenged all jurors who received one or more Xs.
Casey was seated after the prosecution’s 13th challenge. The prosecutor used his 14th challenge against a juror who had received one /. He then used his 15th challenge against Casey. At the time Casey was challenged, only one other seated juror had received an unfavorable rating (¿a, one or more Xs). After Casey was challenged, Crittenden immediately moved for a mistrial under People v. Wheeler, 22 Cal.3d 258, 148 Cal.Rptr. 890, 583 P.2d 748 (1978), arguing the peremptory challenge was motivated by race. Wheeler is the California procedural equivalent of Batson, and serves as an implicit Batson objection for purposes of preserving a Batson claim. See Crittenden I, 624 F.3d at 951 n. 2. A Batson/Wheeler claim has three steps: “first, ‘the defendant must make a prima facie showing that a challenge was based on race;’ second, the prosecution must offer a race-neutral basis for the challenge; and third, the court must determine whether the defendant has shown ‘purposeful discrimination.’ ” Cook, 593 F.3d at 814 (quoting Ali v. Hickman, 584 F.3d 1174, 1180 (9th Cir.2009)).
In moving for a mistrial, Crittenden argued he had made a prima facie showing because: (1) Casey was the only African-American prospective juror; (2) she' was a “solid member of the ... community in terms of age, family composition, employment, length of residence, and so forth”; (3) the prosecutor examined her “at greater length[] than he’[d] examined other jurors”; and (4) in a different capital case a year earlier, the same prosecutor struck the only African-American prospective juror because he “was the President [of] the Student Law Union of Minorities,” which indicated to the prosecutor that the individual was “active in law problems involving minorities” and had “sympathy for minorities.”
The trial court denied the motion, finding Crittenden had not made a prima facie showing that the challenge was based on race. The trial court said it “would have expected a peremptory challenge” against Casey because she had expressed opposition to the death penalty and “couldn’t decide whether or not she would be able to follow the law.” Because the trial court denied the motion at step one of the Bat-son/Wheeler test, it did not request an explanation for the challenge from the prosecution at step two. Ultimately 12 jurors were selected, each of whom had received one or more /s.
The California Supreme Court affirmed Crittenden’s subsequent conviction and sentence on direct review in 1994. See People v. Crittenden, 9 Cal.4th 83, 36 Cal.Rptr.2d 474, 885 P.2d 887 (1994). It affirmed the trial court’s finding that Crit-tenden had failed to make a prima facie showing, holding:
Casey’s apparent opposition to, uncertainty about, and repeatedly contradictory responses pertaining to the death penalty, her indication she might be unable to apply the law in that regard, her apparent general apprehension at serving on a jury for the first time, as well as her concern over her transportation to the court for trial, indicate there were legitimate, race-neutral grounds upon *1005which the prosecutor reasonably might have challenged her.
Id. 36 Cal.Rptr.2d 474, 885 P.2d at 904.
Between 1994 and 2000, Crittenden filed multiple state and federal habeas petitions. The California Supreme Court dismissed his state habeas petitions in 1994 and 1999. The federal habeas petition was referred to a magistrate judge, who conducted an evidentiary hearing in 2002. At the hearing, the state produced through discovery new evidence not considered by the state courts — specifically, the prosecutor’s rating of each potential juror after the voir dire. The prosecutor also described his general practice of rating jurors, and testified that generally he “would try to get people who were threes and fours with checkmarks, to sit on the jury.” He testified the jury selection “took place over a long period of time,” and the ratings reflected his “gut feeling at the time that I spoke with jurors and was present when they were examined. And it was at that time that I made a decision.”
In 2005, the district court denied Crit-tenden’s federal habeas petition. The district court disagreed with the state trial court and the California Supreme Court, finding that, although their step one finding was presumed correct, Crittenden had rebutted the presumption and made a pri-ma facie showing that the challenge was based on race. The district court based that finding on several facts, including: (1) a comparative juror analysis; (2) the prosecutor used “charged” terms when questioning Casey; (3) Casey was the only prospective juror the prosecutor challenged for cause based on general objections to the death penalty, and it was well established that such objections did not warrant a for-cause challenge; and (4) the prosecutor challenged the sole African-American prospective juror in the previous capital case. The court, therefore, proceeded to step two of the Batson inquiry and found the state met its burden to proffer a race-neutral justification for the challenge — Casey’s opposition to the death penalty.3 The court then proceeded to step three, and found Crittenden had not proven purposeful discrimination. Although the court found that “race played a significant part” in the peremptory challenge, and race-neutral factors could not “justify Casey’s XXXX rating, especially when compared to other venire members,” the court concluded the prosecutor “would have made the challenge in the absence of the- improper motivation” because of Casey’s opposition to the death penalty. Crittenden appealed.
On appeal, we held the California Supreme Court’s decision with respect to Crittenden’s Batson claim was not entitled to deference under AEDPA because, contrary to clearly established federal law, at step one it “required Crittenden to show a ‘strong likelihood’ that the prosecutor’s challenge had been racially motivated.” Crittenden I, 624 F.3d at 954. We affirmed the district court’s determinations at Batson step one and step two that Crittenden had made a prima facie showing and the state had articulated a race-neutral justification for the challenge. See id. at 956-58.
At Batson step three, we declined to determine whether Crittenden had proven the challenge was based on race, because the district court had decided the question *1006prior to Cook, which clarified that “the proper analysis ... is whether the peremptory strike was ‘motivated in substantial part’ by race ... regardless of whether the strike would have issued if race had played no role.” Id. at 958 (quoting Cook, 593 F.3d at 815). Because the district court “was operating under the erroneous impression” that it could apply so-called “mixed-motives” analysis, such that the presence of a race-neutral, but-for cause for the challenge would defeat a Batson claim, we remanded “to give the court an opportunity to apply the proper standard, as articulated in Cook.” Id. at 958-59.
On remand, the case again was referred to the magistrate judge, who issued new factual findings in light of both the district court’s previous factual determinations in Crittenden I and other undisputed facts in the record. The magistrate judge recommended the Batson claim be denied because “the bias shown by the prosecutor ... although significant, was not substantial in terms of the prosecutor’s motivation.” Reviewing those findings de novo, the district court disagreed with the magistrate judge’s ultimate recommendation and instead found the prosecutor was substantially motivated by race for four reasons. First, the prosecutor rated Casey far more negatively than comparable white jurors. Second, Casey was the only prospective juror the prosecutor challenged for cause based on a general objection to the death penalty, and it was well established that such objections did not warrant a for-cause challenge. Third, the prosecutor asked Casey a provocative question regarding the death penalty, and twice used the charged term “gas chamber,” whereas “no other juror was questioned in this manner with use of the same charged term.” Fourth, “even if it is not given great weight, [the prosecutor’s] strike of another black juror in a prior trial suggests that he took account of race in assessing how a juror would vote.” The court granted Crittenden’s petition. The state appeals.
STANDARD OF REVIEW
We review de novo a district court’s grant of habeas corpus relief. See Gallego v. McDaniel, 124 F.3d 1065, 1069 (9th Cir.1997). A district court’s factual findings in granting a habeas petition are reviewed for clear error. See Fed.R.Civ.P. 52(a)(6); Lambert v. Blodgett, 393 F.3d 948, 964 (9th Cir.2004). At Batson’s first step, whether the defendant has made a prima facie showing is a mixed question of law and fact accorded a presumption of correctness in the habeas context. See Tolbert v. Page, 182 F.3d 677, 681 n. 6, 685 (9th Cir.1999) (en banc) (applying 28 U.S.C. § 2254(e)(1)). At Batson’s third step, it is “settled in this circuit” that “[w]hether the defendant has satisfied the ultimate burden of proving purposeful discrimination is, of course, a question of fact reviewed for clear error.” Id. at 680 n. 5.
Notwithstanding this authority, the dissent argues we should review de novo the district court’s factual finding at Batson step three because the district court relied solely on a cold record, rather than testimony before the district judge. That argument is squarely foreclosed by Federal Rule of Civil Procedure 52(a)(6), which says, “[findings of fact, whether based on oral or other evidence, must not be set aside unless clearly erroneous.” (emphasis added). “[I]t is impossible to trace the [dissent’s] theory[ ] ... back to the text of Rule 52(a),” which applies “even when the district court’s findings do not rest on credibility determinations, but are based instead on physical or documentary evidence or inferences from other facts.” Anderson v. City of Bessemer City, N.C., 470 U.S. 564, 574, 105 S.Ct. 1504, 84 *1007L.Ed.2d 518 (1985). The rationale for deference “is not limited to the superiority of the trial judge’s position to make determinations of credibility,” but also reflects that “[duplication of the trial judge’s efforts ... would very likely contribute only negligibly to the accuracy of fact determination at a huge cost in diversion of judicial resources.” Id. at 574-75, 105 S.Ct. 1504; see Fed.R.Civ.P. 52(a)(6) advisory committee’s notes (1985 amendment) (explaining that permitting de novo review of findings based on documentary evidence would “tend to undermine the legitimacy of the district courts in the eyes of litigants, multiply appeals by encouraging appellate retrial of some factual issues, and needlessly reallocate judicial authority”).4 Accordingly, we properly review for clear error the district court’s finding of purposeful discrimination at Batson step three.
DISCUSSION
I. Teague does not prohibit the retroactive application of the standard for Batson claims articulated in Cook
The state argues Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989), prohibits the retroactive application of the standard for Batson claims articulated in Cook. “Teague held that federal habeas corpus petitioners cannot rely on new constitutional rules of criminal procedure that took effect after their convictions became final.” Boyd v. Newland, 467 F.3d 1139, 1145 (9th Cir.2004), as amended on denial of reh’g (Oct. 26, 2006). It is undisputed that Crittenden’s conviction became final several years before Cook, and that the relief requested does not “fall[ ] within one of two exceptions to nonretroactivity on habeas review.” Leavitt v. Arave, 383 F.3d 809, 816 (9th Cir.2004). The question, then, is whether Cook announced a new rule for purposes of Teague. We hold it did not.
“In general ... a case announces a new rule when it breaks new ground or imposes a new obligation on the States or the Federal Government.” Teague, 489 U.S. at 301, 109 S.Ct. 1060. “To put it differently, a case announces a new rule if the result was not dictated by precedent existing at the time the defendant’s conviction became final.” Id.
Here, our holding that Cook did not announce a new rule follows from Boyd, which rejected a Teague challenge under analogous circumstances. Boyd held the Supreme Court’s decision in Johnson v. California, 545 U.S. 162, 125 S.Ct. 2410, 162 L.Ed.2d 129 (2005), did not establish a new rule for purposes of Teague. See 467 *1008F.3d at 1146. Like Cook, Johnson clarified the standard for Batson claims. Johnson rejected the California Supreme Court’s holding that, to establish a prima facie case of discrimination at step one, a defendant must show it is more likely than not that a challenge was based on race. See 545 U.S. at 168, 125 S.Ct. 2410. Instead, Johnson held, “a defendant satisfies the requirements of Batson’s first step by producing evidence sufficient to permit the trial judge to draw an inference that discrimination has occurred.” Id. at 170, 125 S.Ct. 2410.
Boyd held Johnson “merely clarified],” or “explained]” Batson. 467 F.3d at 1146. The same is true of Cook. Whereas Johnson clarified the standard at Batson step one, Cook clarified the standard at Batson step three. Further, Boyd recognized Johnson was “an example of the Supreme Court’s consistent interpretation of Batson to date.” Id. Like Johnson, Cook’s clarification of Batson’s standard is consistent with existing precedent, as neither the Supreme Court nor this circuit had previously adopted mixed motives analysis. See Snyder v. Louisiana, 552 U.S. 472, 485, 128 S.Ct. 1203, 170 L.Ed.2d 175 (2008); Kesser v. Cambra, 465 F.3d 351, 358 (9th Cir.2006) (en banc). Thus, like Johnson, Cook neither “breaks new ground [n]or imposes a new obligation on the States or the Federal Government.” Teague, 489 U.S. at 301, 109 S.Ct. 1060.
This conclusion also is consistent with Tanner v. McDaniel, 493 F.3d 1135 (9th Cir.2007). Tanner held the Supreme Court’s decision in Roe v. Flores-Ortega, 528 U.S. 470, 120 S.Ct. 1029, 145 L.Ed.2d 985 (2000), did not establish a new rule for purposes of Teague. See 493 F.3d at 1142-44. Flores-Ortega held, in relevant part, that “counsel has a constitutionally imposed duty to consult with the defendant about an appeal when there is reason to think either (1) that a rational defendant would want to appeal (for example, because there are nonfrivolous grounds for appeal), or (2) that this particular defendant reasonably demonstrated to counsel that he was interested in appealing.” Flores-Ortega, 528 U.S. at 480, 120 S.Ct. 1029. Flores-Ortega held breach of this duty constituted ineffective assistance of counsel under the Sixth Amendment.
Tanner concluded this holding in Flores-Ortega was not a new rule, but merely an “application of’ the “circumstance-specific reasonableness inquiry” dictated by Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Tanner, 493 F.3d at 1143. Tanner observed that “the general nature of the Strickland standard requires courts to elaborate upon what an ‘objective standard of reasonableness’ means for attorney performance in innumerable factual contexts,” and “[e]aeh time that a court delineates what ‘reasonably effective assistance’ requires of defense attorneys with respect to a particular aspect of client representation, it can hardly be thought to have created a new principle of constitutional law.” Id. at 1143 — 44. (citations omitted). Similarly here, the general nature of the Batson standard requires courts to elaborate upon what constitutes “purposeful discrimination,” and Cook’s explanation that “purposeful discrimination” may exist even when there is also a'race-neutral, but-for cause of a prosecutor’s decision to challenge a juror did not create a new principle of constitutional law. See Wright v. West, 505 U.S. 277, 309, 112 S.Ct. 2482, 120 L.Ed.2d 225 (1992) (Kennedy, J. concurring) (“Where the beginning point is a rule of this general application, a rule designed for the specific purpose of evaluating a myriad of factual contexts, it will be the infrequent case that yields a result so novel that it forges a new rule, one not dictated by precedent.”). Therefore, we hold *1009Teague did not prohibit the district court from applying the standard articulated in Cook.
II. The district court was not required to apply AEDPA deference to the California Supreme Court’s decision or the state trial court’s findings
A. The California Supreme Court’s decision was not entitled to AEDPA deference because it was contrary to clearly established law
The state next argues the district court failed to afford the necessary deference under AEDPA to the California Supreme Court’s decision rejecting Crit-tenden’s Batson claim on direct review. In Crittenden I, we held the California Supreme Court’s decision was not entitled to AEDPA deference because, contrary to clearly established federal law, “it required Crittenden to show a ‘strong likelihood’ that the prosecutor’s challenge had been racially motivated” in order to establish a prima facie case. Crittenden I, 624 F.3d at 954. That holding is the law of the case. See Hanna Boys Ctr. v. Miller, 853 F.2d 682, 686 (9th Cir.1988). We have discretion to reconsider our prior decision when “intervening controlling authority makes reconsideration appropriate” or “the decision is clearly erroneous and its enforcement would work a manifest injustice.” Jeffries v. Wood, 114 F.3d 1484, 1489 (9th Cir.1997) (en banc) (footnote omitted), overruled on other grounds by Gonzalez v. Arizona, 677 F.3d 383 (9th Cir.2012) (en banc). Neither circumstance exists here.
The state contends two cases decided after Crittenden I-Harrington v. Richter, 562 U.S. 86, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011), and Johnson v. Williams, — U.S. -, 133 S.Ct. 1088, 185 L.Ed.2d 105 (2013) — establish a presumption that the California Supreme Court applied the correct federal standard. Neither case stands for that proposition. 'Richter held, “when a state court issues an order that summarily rejects without discussion all the claims raised by a defendant, including a federal claim that the defendant subsequently presses in a federal habeas proceeding, the federal habeas court must presume (subject to rebuttal) that the federal claim was adjudicated on the merits.” Williams, 133 S.Ct. at 1091 (discussing Richter). Williams extended that presumption to cases in which a state court addresses “some issues but does not expressly address the federal, claim in question.” Id. Neither Richter nor Williams addressed whether, when a state court does address a federal claim on the merits, it should be presumed to have applied the correct federal legal standard. Thus, neither provides a basis to reconsider our prior holding.
In any event, any such presumption would not aid the state here. At the time it decided this case, the California Supreme Court had erroneously concluded the “terms ‘strong likelihood’ and ‘reasonable inference’ state the same standard.” Johnson, 545 U.S. at 166, 125 S.Ct. 2410. As we held in Crittenden I, the California Supreme Court relied on that erroneous conclusion when deciding Crittenden’s appeal, conflating the two standards in its decision. See 624 F.3d at 952 (citing People v. Crittenden, 36 Cal.Rptr.2d 474, 885 P.2d at 902).
The state also contends our holding is inconsistent with our earlier decision in Boyd. We disagree. In Boyd, although the state court first applied the “strong likelihood” standard for a prima facie case of discrimination recognized by the California Supreme Court, the state court “also held that Petitioner ‘clearly did not establish a prima facie case of group dis*1010crimination, even under federal precedent.’ ” 467 F.3d at 1144 (emphasis added). Because the state court “recognized the difference between the two standards,” and held the petitioner had “failed to establish a prima facie case, under either state or federal law,” Boyd held the court’s “determination deserves deference.” Id.
Here, in contrast, the California Supreme Court did not separately address the federal standard. It cited and discussed only the erroneous “strong likelihood” standard, and incorporated its discussion of the facts under that standard as the basis for its denial of the Batson claim. See People v. Crittenden, 36 Cal.Rptr.2d 474, 885 P.2d at 902-06. As a result, the California Supreme Court’s decision was contrary to clearly established federal law, and the district court properly considered the Batson claim “without the deference AEDPA .otherwise requires.” Crittenden I, 624 F.3d at 954 (quoting Panetti v. Quarterman, 551 U.S. 930, 953, 127 S.Ct. 2842, 168 L.Ed.2d 662 (2007)).
B. The state trial court’s factual findings were rebutted by clear and convincing evidence
We also reject the state’s contention that Crittenden I failed to afford a presumption of correctness under 28 U.S.C. § 2254(e)(1) to the state trial court’s finding that Crittenden did not establish a prima facie violation at Batson step one. We said in Crittenden I that “[w]e presume the state court’s factual findings to be correct, a presumption the petitioner has the burden of rebutting by clear and convincing evidence.” 624 F.3d at 950. The district court found, and Crit-tenden I affirmed, that Crittenden rebutted that presumption as to the Batson step one finding. His clear and convincing evidence included that the crime was racial in nature, Casey was the only African-American juror in the venire and the only juror subject to a meritless fof-cause challenge, and there was a disparity between the prosecutor’s rating of Casey and his ratings of comparable white jurors. That ratings disparity, discussed in further detail below, is new evidence not presented to the state trial court.
We disagree with the dissent that Cullen v. Pinholster, 563 U.S. 170, 131 S.Ct. 1388, 1398, 179 L.Ed.2d 557 (2011), precludes Crittenden I’ s consideration of that new ratings evidence to rebut the trial court’s factual finding at Batson step one. Pinholster precludes the consideration of new evidence only for the purpose of determining whether the last reasoned state court decision was contrary to or an unreasonable application of clearly established law or an unreasonable determination of the facts under 28 U.S.C. § 2254(d). See Pinholster, 131 S.Ct. at 1398 (“We now hold that review under § 2254(d)(1) is limited to the record that was before the state court....”). We have since clarified — after Pinholster and the cases cited in the dissent — “If we determine, considering only the evidence before the state court,” the petitioner has satisfied § 2254(d), “we evaluate the claim de novo, and we may consider evidence properly presented for the first time in federal court.” Hurles v. Ryan, 752 F.3d 768, 778 (9th Cir.2014) (citing Pinholster, 131 S.Ct. at 1401); see also Johnson v. Finn, 665 F.3d 1063, 1069 n. 1 (9th Cir.2011) (holding Pinholster did not preclude the district court from conducting an evidentiary hearing after concluding the state court of appeal’s decision was contrary to clearly established law under § 2254(d)(1)).5
*1011In reviewing the merits of a habeas petitioner’s claim after § 2254(d) is satisfied, we still defer to a state court’s factual findings under § 2254(e) in two ways. First, those findings are presumed to be correct, a presumption that can be overcome only by clear and convincing evidence. See 28 U.S.C. § 2254(e)(1). Second, with limited exceptions, new evidence cannot be considered if “the applicant has failed to develop the factual basis of a claim in State court proceedings,” id. § 2254(e)(2) — which is not the case here. The dissent would introduce a third layer of deference, confining review of a state trial court’s factual findings under § 2254(e)(1) to the record before the state trial court. But nothing in Pinholster or Murray v. Schriro, 745 F.3d 984 (9th Cir.2014), requires limiting the record on review once a federal court has found unreasonable the last reasoned state court decision — here, that of the California Supreme Court. The state does not argue otherwise.
Here, Crittenden I held the California Supreme Court’s decision was contrary to clearly established law under § 2254(d)(1) because it applied an improper legal standard at Batson step one. Having made that determination, Crittenden I properly turned to the merits of Crittenden’s Bat-son claim, while affording a presumption of correctness to the state trial court’s factual findings under § 2254(e). Thus, contrary to the argument advanced by our dissenting 'colleague, Crittenden I properly considered new evidence in rejecting the state trial court’s Batson step one finding under § 2254(e).
III. The district court was not required to conduct its own evidentiary hearing
The state next argues the district court erred by rejecting the magistrate judge’s recommendation without conducting its own evidentiary hearing, in violation of Johnson, 665 F.3d at 1063. Johnson held, in the Batson context, “a district court may not ... reject a magistrate judge’s proposed credibility determination without hearing and seeing the testimony of the relevant witnesses.” Id. at 1074. This case is distinguishable because the magistrate judge did not make — and hence the district court did not reject — any credibility determination related to the prosecutor’s explanation for striking Casey, because the prosecutor offered none.
At the evidentiary hearing, which took place over a decade after the trial, the prosecutor could not articulate a race-neutral explanation for his peremptory challenge. Instead the state reconstructed a race-neutral justification based on the evidence in the record. As the district court stated, “the prosecutor’s credibility as to his articulated race-neutral reason was never at issue in this case.” Thus, the district court did not reject any credibility determination by the magistrate judge, but instead, based on the1 court’s independent review of the record, drew different inferences and reached different conclusions than the magistrate judge. The court was *1012not required to conduct a new evidentiary hearing.
IV. The district court’s finding that the challenge was substantially motivated by race was not clearly erroneous
Turning to the merits of Crit-tenden’s Batson claim, we hold the district court’s finding that the prosecutor’s challenge of Casey was substantially motivated by race was not clearly erroneous. A finding is clearly erroneous if it is “(1) ‘illogical,’ (2) ‘implausible,’ or (3) without ‘support in inferences that may be drawn from the facts in the record.’ ” Unites States v. Hinkson, 585 F.3d 1247, 1262 (9th Cir.2009) (en banc) (quoting Anderson, 470 U.S. at 577, 105 S.Ct. 1504). The court’s finding of purposeful discrimination is supported by the facts in the record. First and foremost, the court found the XXXX rating of Casey was evidence of racial bias. A comparative juror analysis shows the XXXX rating on which the prosecutor based his challenge cannot be explained by Casey’s death penalty views or other race-neutral factors. The prosecutor’s merit-less for-cause challenge provides additional support for the district court’s finding that he was substantially motivated by race. Further, because we conclude the district court found only purposeful discrimination, we reject the dissent’s contention the court found or relied on unconscious bias.
A. Comparative Juror Analysis
“Comparative juror analysis is an established tool at step three of the Batson analysis for determining whether facially race-neutral reasons are a pretext for discrimination.” Crittenden I, 624 F.3d at 956 (citing Miller-El v. Dretke, 545 U.S. 231, 241, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005)). The prosecutor’s ratings of prospective jurors provide a useful Basis for a comparative juror analysis because he closely adhered to the ratings when issuing challenges. Indeed, he testified that, when he assigned the ratings after each voir dire, “it was at that time that I made a decision.”
1. Comparison with Juror Smith
The district court found a comparison of Casey and another juror, Lois Smith, weighed in favor of finding racial bias. Smith was the only other prospective juror rated XXXX. The court found Smith was a far worse juror for the prosecution than Casey. It stated Smith’s “distinctly strong and unshakable death penalty views and experiences with the justice system made her the antithesis of a prosecution juror.” The court observed it was “puzzling why Casey received the same rating as Smith, when no similarly glaring evidence for prosecutorial disfavor of Casey exists.”
The court’s findings from the comparative analysis of Casey and Smith were not clearly erroneous. Smith also recounted what she described as a “horrendous” experience with law enforcement in which her husband was wrongly implicated in a crime by an eyewitness who had identified him notwithstanding that he is white and the suspect was African-American. She said she “would be extra cautious” because of that experience. The district court reasoned this experience would have been of particular concern to the prosecution because “a key element of [the] evidence against [Crittenden] at trial was that eye witnesses had seen a black man matching [Crittenden’s] description near the victims’ home when the murders occurred.”
Casey, in contrast, presented no similar negative experiences with law enforcement. And, as the court correctly noted, aside from her death penalty views, Casey was a “model prosecution juror according to [the prosecutor’s] own criteria.” The *1013prosecutor testified he looked for jurors who were “employed, homeowners ... people who had something to lose in society, who might be victims of crime, solid citizens, preferably fairly well educated.” Casey had been married for 42 years, had two adult children, went to church on Sundays, had lived in the same home for 17 years, and said she was concerned about drugs and street gangs.
Although Casey opposed the death penalty, she repeatedly affirmed that her opposition would not prohibit her from following the court’s instructions, applying the proper standard of proof or voting to impose the death penalty. The trial judge asked “whether your feelings concerning the death penalty would influence your vote to the extent ... that you would not vote for a first degree murder conviction,” and she answered, “No.” The judge then asked whether her death penalty views would cause her to refuse to vote for special circumstances that would implicate the death penalty, and she answered, “No.” The judge then asked whether her death penalty views would cause her to “automatically and in every case vote against the imposition of the death penalty,” and she answered, “No.”
When questioned by defense counsel, Casey again said she could conceive of a situation in which the death penalty might be appropriate, she would be willing and able to vote for the death penalty if a crime were “awful bad” and she had no “qualms” about applying the court’s instructions regarding the proper standard of proof. The prosecutor then began his examination by asking Casey, “Now, I gather[ ] ... that you do not believe in the death penalty?” Casey answered:
I really don’t. But if it is bad, ... really bad and I felt that, you know — I hate death. I don’t know how to express myself, really. But I really hate to see anybody be put to death. And I hate to see someone take a life. I don’t care whose it is. So — it is — it is hard for me to express it. But I could, if proven to me, to, no doubt, that it was a crime, then I don’t think I would hesitate.
Upon further questioning, Casey expressed some hesitation, saying, “if it is proven to me, truly proven to me, and I feel deep down inside that he did it, I could. I think I could.... I have to say I think I could. This is all new to me. So I am very upset with it.” She also said she thought her feelings about the death penalty would make it difficult for her to make a decision regarding the death penalty, and she did not know whether it would substantially impair her ability to fairly evaluate the evidence. She then reaffirmed, though, that her feelings about the death penalty would not cause her to “lean[] toward life instead of death,” and that she could vote for the death penalty if she “heard facts and circumstances which warranted it.”
Although Casey and Smith both expressed opposition to and reservations about imposing the death penalty, the voir dire transcripts support the court’s conclusion that Smith was the worse juror for the prosecution. Smith arguably expressed stronger opposition to the death penalty than did Casey — she said she found the prospect of serving on a jury in a death penalty case “horrifying” — and recounted a “horrendous” experience with law enforcement caused by mistaken eyewitness identification. The court did not clearly err by concluding the comparison of Casey and Smith supports the conclusion that the prosecution’s challenge of Casey was substantially based on her race.’6
*10142. Comparison with Clark and Krueger
The district court also found a comparison of Casey with jurors Gisela Clark and Mary Krueger provided “strong evidence of discriminatory motive.” The court found, although Clark and Krueger “are demographically similar to Casey, except they are both white,” and although they “expressed death penalty views generally unfavorable to the prosecution,” they were both rated at least //, and selected to serve on the jury.
The court’s comparative analysis of Casey and Clark is supported by the evidence. The prosecutor rated Clark ///. Aside from, race, Clark was demographically comparable to Casey. She had been married for 34 years before her husband passed away, lived in the same home for 21 years and identified as Catholic. On her questionnaire, Clark wrote that she was “against” the death penalty. At voir dire, she expressed strong opposition to the death penalty and serious reservations about her ability to vote for it:
My opinion is this. First of all I am Catholic, and I have been brought up no matter what, I cannot take somebody’s life, I don’t feel that I am better than the next person. Or another reason I think that I am not quite sure which' is the worse thing a person can do. Whether the worse thing is murder or the worse thing is defrauding someone of their life savings. And I always felt — even voting for it, I felt if I am for it, I should be the one that should execute it more or less. My feelings. But, I shouldn’t ask someone else to do it for me. And, therefore, I feel opposed to it. I just don’t feel I should take someone’s life.
The trial judge then asked Clark whether, no matter what the circumstances might be, she would ever vote to sentence someone to his death, and she answered, “Probably not.” On further questioning from the judge, she provided a more equivocal answer: “I have never been in that predicament. I am not quite sure how I would react. I feel the person should be punished for their crime. Maybe I could. I am not sure what would, happen.” After a few more questions, the judge again asked, “would you in every case automatically vote for life imprisonment without the possibility of parole, and would you never vote to impose the death penalty,” and she again equivocated, stating, “I don’t — I am really not sure.” In contrast, when the trial judge asked Casey whether she would automatically, in every case, vote against the imposition of the death penalty, Casey answered, unequivocally, “No.”
To be sure, in response to questioning from defense counsel, Clark expressed a greater willingness to vote for the death penalty than she had earlier. Defense counsel asked, if “information indicated that not only were the crimes bad, but there were aggravating matters about this defendant that were also brought to your attention — if those matters that you heard indicated to you that death was the appropriate verdict, would you be able to vote for such a verdict?” Clark answered, “Yes.” Later, though, Clark again equivocated. The prosecutor asked, “Are you telling us now that your feelings about the death penalty are not so strong and that you could actually fairly and impartially decide on the penalty in the case?” Clark replied:
Probably. I have, like I said, I have never been in this kind of predicament, so I am sure if the law would be, have to be applied, I would, probably could. *1015But I am not a hundred percent sure. I would have to see what happens during the whole trial to be convinced. I really don’t know. I can’t really tell you how I feel about it. All my life I felt I can’t take someone’s life. But that doesn’t mean — I have never be in this kind of predicament. I mean in this kind — so I don’t know. It is possible I could be completely convinced.
In sum, Clark and Casey were demographically similar, apart from race, and both similarly equivocated regarding their ability to vote for the death penalty. But the prosecutor rated Clark ///, and selected her to serve on the jury, whereas he rated Casey XXXX, attempted to strike her for cause and then used a peremptory challenge against her. As we noted in Crittenden I, even if Clark was clearer than Casey about her ability to vote for a death verdict and to be decisive, “both expressed hesitancy or uncertainty,” and “the wide difference between [the prosecutor’s] rating of Ms. Casey and Ms. Clark is evidence from which an inference of discrimination could have been drawn.” 624 F.3d at 956 & n. 3. That wide difference in ratings provides strong additional support for the district court’s finding that the prosecutor was substantially motivated by race. See Miller-El, 545 U.S. at 241, 125 S.Ct. 2317 (“If a prosecutor’s proffered reason for striking a black panelist applies just as well to an otherwise-similar non-black who is permitted to serve, that is evidence tending to prove purposeful discrimination to be considered at Batson’s third step.”).
The district court’s comparative analysis of Casey and Krueger also is supported by the evidence. The prosecutor rated Krueger // and % /. On her questionnaire, Krueger wrote, “no one should receive [the] death penalty.” At voir dire, the trial judge asked Krueger about her death penalty views, and she said, “I think I would have to be absolutely sure that the person were without a doubt guilty of the crime before I would be able to say a death penalty.” The judge also asked Krueger whether she would “automatically and in every case vote for life without the possibility of parole and never vote for a death penalty,” and Krueger answered, “No. I think I would be able to vote for the death penalty.”
The prosecutor then asked Krueger again about her “general feeling about the death penalty,” and she answered, “Well, I feel very strongly that someone, who is in life — in prison without any parole is — to me is close to death. I mean, there’s nothing that they can do, but be there. But at the same time, I — I, myself, would have to have no doubt in my mind, that the individual was guilty, before I would be able to vote for the death penalty.”
Thus, like Casey, Krueger expressed opposition to the death penalty and some hesitation about applying it, but also said multiple times she could follow the law and vote for the death penalty in certain circumstances.- We held in Crittenden I the “marked difference” in the ratings of Casey and Krueger “adds to the evidence from which ... an inference [of discrimination] could be drawn ... given the demographic similarity and somewhat analogous views on the death penalty.” 624 F.3d at 956. Although Krueger arguably expressed less hesitancy than Casey about her ability to vote for the death penalty, the marked difference in their ratings provides additional support for the district court’s finding that race substantially motivated the challenge of Casey. See Miller-El, 545 U.S. at 241, 125 S.Ct. 2317.
3. Comparison with Sullivan ,and Tennies
Finally, the district court found a “comparison of [other] anti-death penalty jurors *1016reveals that anti-death penalty views were not X or / determinátive.” The court found prospective juror Suzanne Tennies, who was rated //, had “repeatedly expressed strong anti-death penalty views during voir dire, ... [the prosecutor] noted on her questionnaire ‘probably wouldn’t vote for DP.’ ” Similarly, the court found prospective juror Frances Sullivan, who was rated //, “stated he would ‘automatically and in every case vote for life without possibility of parole and never vote for the death penalty.’ ” Although neither Tennies nor Sullivan was selected for the jury, the court found their positive ratings provided further evidence that the state’s proffered race-neutral reason for Casey’s XXXX rating and challenge — her opposition to the death penalty — was pretextual.’7
The court’s findings with respect to Ten-nies and Sullivan are supported by the evidence. Tennies wrote on her questionnaire, “I’m not for the death penalty.” At voir dire, the trial judge asked whether Tennies would absolutely and in every case refuse to vote for the death penalty, and she answered, “No.” When defense counsel then asked her the same question, though, Tennies equivocated, saying, “I really feel strongly against the death penalty. But I — on the other hand, I would have to hear the whole facts of the case. So, it is kind of a hard question for me to answer.”8 The prosecutor then asked her what she meant by her statement that she did not believe in the death penalty, and she explained:
Well, I just kind of feel that — there have been cases, you know, I have read — not lately, but, you know, years and years ago, where they found the person they put to death was innocent, so somebody came forth and said they committed the crime. I can’t remember exact, you know, things. So I just always felt that, you know, that would be wrong. You know, unless you knew a hundred percent that person was guilty.
Tennies then said she would not have difficulty signing the jury verdict to impose the death penalty if she thought the defendant deserved it, and her feelings against the death penalty would not interfere with her ability to make a decision as a juror.
Although Tennies said she could vote for the death penalty, she also expressed clear opposition to and hesitancy about imposing it. Therefore, Tennies’ // rating supports the district court’s finding that “anti-death penalty views were not X or / determinative,” and its resulting inference that the state’s proffered race-neutral justification — opposition to the death penalty — for rating Casey XXXX and challenging her, was pretextual.
With respect to Sullivan, although he wrote on his questionnaire that he was “for” the death penalty, at voir dire he said he would automatically and in every case vote for life without possibility of parole and never vote for the death penalty. The trial judge asked him to clarify, and he said, “I have reservations about the death penalty. I can’t see a person sitting around ten years on death row and *1017then putting them to death after he had been punished already for ten years. It don’t make sense to me.” Sullivan went on to say, however, that he would not have a problem following the law, and his reservations would not affect his ability to determine whether to impose the death penalty. ■ Although Sullivan expressed less hesitation than Casey, given his stated concerns about the death penalty, his // rating also supports the district court’s finding that “anti-death penalty views were not X or / determinative.”
4. Comparison with other seated jurors when Casey was challenged
At the time Casey was challenged, only one other seated juror had received an .unfavorable rating. The magistrate judge found the makeup of the jury at that time was “critical” because, although Casey should not have been rated XXXX, she should have been rated with at least one X, given her opposition to and equivocation regarding the death penalty. Had she been rated with one X, the prosecutor likely still would have challenged her. As a result, the magistrate judge reasoned, the makeup of the jury at the time Casey was challenged shows that “she would have been stricken regardless of her race.”
The district court properly declined to grant this factor substantial weight. Initially, as the court explained, we cannot assume, even if “Casey objectively deserve[d] to be in the X category, [the prosecutor] himself actually was motivated to put her in that category for nondiscriminatory reasons.” As discussed above, white prospective jurors expressed similar views regarding the death penalty yet were rated with multiple /s. Therefore, Casey, too, might have been rated with /s if race had not been a factor.
More significantly, even assuming there were a race-neutral justification for at least a single X rating of Casey, and such a rating would have led to the challenge, that is not the dispositive question. Under the standard articulated in Cook, the question is whether race was a substantial motivating factor. See Cook v. LaMarque, 593 F.3d 810, 814-15 (9th Cir.2010); Crittenden I, 624 F.3d at 958. Independent, race-neutral reasons for the challenge do not preclude a finding that race also was a substantial motivating factor. See Kesser v. Cambra, 465 F.3d 351, 359 (9th Cir.2006) (en banc) (“A court need not find all nonracial reasons pretextual in order to find racial discrimination”). Here, as the district court reasoned, the XXXX rating “virtually assured Casey would be struck at some point.” Because the prosecutor essentially predetermined at the outset that Casey would be challenged, the makeup of the jury at the time she was in fact challenged is entitled to little weight.9
In sum, because jurors comparable in their death penalty views and otherwise were rated with /s, and some were selected for the jury, the comparative juror analysis significantly weakens the- government’s race-neutral explanation for Casey’s XXXX rating and challenge. That analysis is strong evidence in support of the district court’s finding that the challenge of Casey was substantially motivated by race.10
*1018B. The for-cause challenge
The district court also found the prosecutor’s for-cause challenge of Casey based on her general opposition to the death penalty was evidence he was substantially motivated by race. In Crittenden I, we held:
the circumstances of the prosecutor’s for-cause challenge of Ms. Casey also add to the evidence from which an inference of improper discrimination could be drawn. The prosecutor said he challenged her for cause because she did not believe in the death penalty; however, it was well established law at the time that challenges for cause based on a juror’s general objections to the death penalty were improper.
624 F.3d at 956-57 (citing Wainwright v. Witt, 469 U.S. 412, 424, 105 S.Ct. 844, 83 L.Ed.2d 841 (1985)).
The state contends the for-cause challenge was made in good faith because Casey’s voir dire responses suggested her opposition to the death penalty “substantially impaired her ability to be a juror.” The prosecutor, though, did not make the for-cause challenge on that ground. Instead, he stated “[t]he People voice a challenge for cause based upon her answer that she doesn’t believe in the death penalty.” Not only was it clearly established at the time of the trial that general opposition to the death penalty did not provide a basis for a for-cause challenge, but also as discussed above, Casey repeatedly said she would be able to follow the instructions of the court and vote for the death penalty.
The magistrate judge agreed that “the meritless cause challenge ... evidenced an ulterior motive to remove” Casey from the jury, but gave this evidence little weight because he found the prosecutor also had challenged prospective juror Jonell More-nO’ for cause based on her general objections to the death penalty. The district court reasonably concluded, however, that Moreno had voiced more than general objections to the death penalty. Moreno said she would not want to be the foreperson of the jury because she would not want to sign the death verdict. As the district court concluded, “Moreno’s stronger and more specific Objection to the death penalty materially distinguishes Moreno from Casey.” The for-cause challenge thus provides some additional support for the district court’s finding that the prosecutor’s challenge of Casey was substantially motivated by race.
Viewed cumulatively, Casey’s XXXX rating, which essentially predetermined that she would be challenged, the wide disparity between her rating and the ratings of comparable white jurors and the meritless for-cause challenge provide sufficient evidence from which the district court logically could find the prosecutor’s decision to challenge Casey was substantially motivated by race. See United States v. Hinkson, 585 F.3d 1247, 1262 (9th Cir.2009) (en banc).11 The district court did not clearly err.
*1019C. Unconscious Bias
The dissent contends the district court erroneously based its finding of a Batson violation in part on unconscious bias. The district court’s decision refutes that reading. As the court explained, it undertook “a sensitive inquiry into [the] circumstantial and direct evidence of intent” and found the prosecutor engaged in “purposeful discrimination.” Its order repeatedly articulated the court’s holding that the prosecutor’s strike “was motivated in substantial part by race,” and affirmatively rejected the proposition that that holding was “based on the existence of unconscious discrimination.” According to the district court, the evidence thus left “no doubt” that a conscious, “racially discriminatory impetus” motivated the prosecutor’s strike of Casey.
The dissent makes much of the district court’s passing comment that “[t]he [side-by-side juror] comparisons demonstrate that ... [the prosecutor] was motivated, consciously or unconsciously, in substantial part by race.” But all the court meant was, whatever the explanation for the prosecutor’s racial motive, that motive was a substantial reason for his use of a peremptory strike. As the court clarified:
[T]he court cannot, and does not, address why [the prosecutor] was motivated by race. The court cannot say whether [he] thought Casey would be partial to petitioner “because of their shared race,” Batson, 476 U.S. at 97, 106 S.Ct. 1712, or if he was influenced solely by “conscious or unconscious racism,” id. at 106, 106 S.Ct. 1712. And it need not. The court’s reference to the potential for unconscious racism ... clarified] that the court in no way sought to impugn [his] character as it úndertook the Bat-son inquiry....
In other words, why the prosecutor had a conscious racial motive to strike Casey in the first place — whether or not “unconscious racism” partly explained that motive — was simply irrelevant to the Batson inquiry.12
We agree. And because we uphold the district court’s finding of a conscious racial motive, we do not — and need not — address whether unconscious bias can establish a Batson violation.
CONCLUSION
We affirm the judgment of the district court.
AFFIRMED.
. We noted in Crittenden I that “the Supreme Court and this court have used the words 'significant' and 'substantial' interchangeably in analogous contexts,’’ but we did not assume the district court's finding of "significant” bias necessarily was sufficient under Cook. 624 F.3d at 959 n. 6.
. A detailed account of the crime and the evidence underlying the conviction and sentence is set out in Crittenden I. See 624 F.3d at 948-49.
. Because the prosecutor could not remember why he challenged Casey, the state reconstructed from the record “what the prosecutor would have said had he been asked his reason for exercising the peremptory challenge,” relying on Johnson v. Love, 40 F.3d 658, 667 (3d Cir.1994), and United States v. Nicholson, 885 F.2d 481, 482-83 (8th Cir.1989).
. The dissent contends Rule 52(a)(6) does not apply here because the district court made few true factual findings. That is belied by the district court’s 11-page review of the magistrate judge’s factual findings on remand and de novo review of the record. Those findings, which underpinned the district court’s ultimate factual conclusion at Batson step three, are hardly insignificant. In any event, we are bound by Rule 52(a)(6). As Anderson makes clear, "Rule 52(a) 'does not make exceptions or purport to exclude certain categories of factual findings from the obligation ... to accept a district court’s findings unless clearly erroneous.’ ” 470 U.S. at 574, 105 S.Ct. 1504 (quoting Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 72 L.Ed.2d 66 (1982)). The dissent’s sole authority, Holder v. Welborn, 60 F.3d 383, 388 (7th Cir.1995), cites no authority and does not even mention Rule 52(a)(6) or Anderson. It is therefore neither binding on us nor persuasive. See Anderson, 470 U.S. at 573, 105 S.Ct. 1504 ("In applying the clearly erroneous standard to the findings of a district court sitting without a jury, appellate courts must constantly have in mind that their function is not to decide factual issues de novo.” (quoting Zenith Radio Corp. v. Hazeltine Research, Inc., 395 U.S. 100, 123, 89 S.Ct. 1562, 23 L.Ed.2d 129 (1969))).
. This case law is consistent with authority in the Fifth and Sixth Circuits. See Harris v. Haeberlin, 752 F.3d 1054, 1057 (6th Cir.2014) ("Pinholster is inapplicable to this case be*1011cause it precludes consideration of evidence introduced in federal court only when determining whether a state [appellate] court's adjudication of a claim involved an unreasonable federal-law error. Here, by contrast, the evidence introduced in federal court was not considered for the purpose of ascertaining whether the state [appellate] court had unreasonably applied clearly-established federal law, because we had already concluded that the state court had done so.” (citation omitted)); Smith v. Cain, 708 F.3d 628, 635 (5th Cir.2013) (“Because the district court appropriately and correctly concluded that the state court had unreasonably applied Batson under section 2254(d)(1) based solely on the state court record, Pinholster is inapplicable.”).
. In reaching this conclusion, the district judge on remand agreed with the district *1014judge who reviewed the case prior to the first appeal, as well as the magistrate judge.
. The dissent maintains the district court’s comparison to Sullivan and Tennies was “misplaced” because neither was allowed to serve. We disagree. The district court properly relied on all of the prosecutor’s ratings because the ratings all were made before Casey was struck, all were relied on when the prosecutor exercised his peremptory strikes, and all indicate — when examined together, in their entirety — the prosecutor’s initial assignment of ratings was substantially motivated by race.
. Both Sullivan and Tennies, like Casey, and like Smith, Clark and Krueger, also later said they could follow the law and, in some circumstances, vote for the death penalty.
. The dissent errs by focusing on the magistrate judge’s supposition that Casey would have been stricken regardless of her race. That is not now, and has never been, the Batson standard, as Cook makes clear.
. We are not persuaded by the state’s argument that “the record reveals a number of non-discriminatory factors that are more plausible reasons for the [XXXX] rating than racial bias,” including Casey’s concern about transportation to and from the court, her gen*1018eral indecision and her reluctance to serve on a jury. As discussed earlier, Clark and Ten-nies both expressed indecision about their ability to vote for the death penalty, yet were rated with at least two checkmarks. Further, as the magistrate judge found, "[ajlthough Casey raised a concern that she would need to make arrangements to get to the court, after being questioned and reassured by the judge, her transportation issue appeared resolved.” Finally, Casey's statement that serving on a jury was "scary” is not sufficient to account for the significant difference in juror ratings.
. The district court also found the prosecutor’s challenge of an African-American prospective juror in a prior capital case provided "some evidence” of discriminatory motive. In that case, the prosecutor said the “most *1019important! ]” reason for striking the juror was that "he was the President of the Student Law Union of Minorities ... which indicates to me a sympathy for minorities, and in this case, since the Defendant is a minority, People feel that there would be a bias in that regard.” In Crittenden I, we held that “[t]he probative value of this information is weak because it is a single instance and the trial court denied the Batson objection in that case.” 624 F.3d at 957 n. 4. The district court found this evidence was entitled to "slight weight.” We agree this evidence does not add significantly to Crittenden's case. The district court also relied on the prosecutor's allegedly disparate mode of questioning Casey, as compared to other prospective jurors. For the reasons we stated in Crittenden I, we do not find that factor to "add significantly to [Crittenden's] prima facie case.” 624 F.3d at 957 n. 4.
. It was relevant, of course, to the prosecutor’s reputation. The district court's reference to "unconscious racism” spared him from being found a racist. By suggesting the prosecutor may have had more benign racial motives for the strike, or that his racial motive may have been influenced by unconscious racism, the court hoped to shield the prosecutor from possible disrepute. As the court made clear, however, this effort was not designed to — and did not — detract from the court’s key finding that the strike was consciously motivated by race.
. Crittenden's other evidence does not add significantly to the analysis. As the prior panel opinion and the majority note, the prosecutor’s other case involving an unsuccessful Wheeler challenge adds little; the same is true with respect to the prosecutor's "gas chamber” voir dire question. See Crittenden v. Ayers, 624 F.3d 943, 957 n. 4 (9th Cir.2010) ("Crittenden I"). The prosecutor’s earlier for-cause challenge of Casey is similarly unpersuasive and wasn't even mentioned by Crittenden’s counsel in making his Wheeler motion. The prosecutor also unsuccessfully challenged a white juror for cause based on similar anti-death penalty statements.