OPINION
WALLACE, Senior Circuit Judge:
Petitioners McDaniels and Jenkins appeal from the separate district court judgments denying their 28 U.S.C. § 2254 ha-beas petitions. We consider their appeals together.
Petitioners were tried and convicted together in the Alameda County Superior Court of California on a charge of first degree murder, among others. Here we consider only their argument, based on
Batson v. Kentucky,
476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), that the prosecutor in their case excluded African-American jurors based on race during jury selection. In a separately filed unpublished disposition we consider their arguments that their counsel each provided ineffective assistance.
We have jurisdiction under 28 U.S.C. § 2258. Applying de novo review,
see Mitleider v. Hall,
391 F.3d 1039, 1046 (9th Cir.2004), we affirm.
I.
We need not recount the details of the crime, because we only consider Petitioners’ contention that the prosecutor excluded African-American jurors based on their race.
The state trial judge limited voir dire to thirty minutes total. He explained that this was because jurors filled out questionnaires, the purpose of which was to do away with the need for extensive voir dire.
During the voir dire, the prosecutor challenged seven out of ten African-Americans called as potential jurors. Petitioners argued that the prosecutor excluded four of those jurors based on their race. During the
Batson
hearing in the state court, the trial judge held that Petitioners had established a prima facie case of discrimination and asked the prosecutor to offer race-neutral reasons for the challenges. The prosecutor gave his reasons, and the trial court concluded that there “didn’t appear ... to be any type of racism going on.”
Petitioners appealed to the California Court of Appeal (CCA), arguing that the record did not support the prosecutor’s reasons. Petitioners also contended that, but for a few exceptions, only African-American jurors were asked whether they were sympathetic to the defendants, although the CCA stated that six non-African-American jurors were also asked that question.
The trial court held that it was not required to engage in comparative juror analysis because, under then-controlling California law, appellate courts were not to perform comparative juror analysis when the argument was not raised in the trial court. The first day of the voir dire transcript, as well as the questionnaires for stricken jurors, were not included in the CCA record. Citing the significant deference it owed to the trial court where that court had undertaken a sincere effort to evaluate the prosecutor’s reasons, the CCA affirmed.
The California Supreme Court affirmed without discussion. Petitioners then separately petitioned the district court. In both proceedings, the state filed the first day of the voir dire transcript as an exhibit. The state also produced the questionnaires of the seated jurors and alternates. The remaining questionnaires had been destroyed.
District Judge Phyllis J. Hamilton heard McDaniels’s petition. She held that the state court’s finding that the prosecutor did not have discriminatory intent was not unreasonable. She further held that, although the CCA was incorrect that it was not required to perform comparative juror analysis because the state trial court had not done so, comparative juror analysis did not uncover any discriminatory intent because, as the CCA observed, six non African-American jurors were also asked whether they were sympathetic to Petitioners.
District Judge Marilyn H. Patel heard Jenkins’s petition, and also held that the CCA’s conclusion was not based on an unreasonable interpretation of the facts.
We review de novo a district court’s denial of a petition under 28 U.S.C. § 2254. Under the Antiterrorism and Effective Death Penalty Act (“AED-PA”), however, a federal court will [order] habeas relief only if the state court decision was (1) contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States, or (2) based on an unreasonable determination of the facts in light of the
evidence presented in the State court proceeding.
Mitleider,
391 F.3d at 1046 (internal citations and quotation marks omitted).
Petitioners argue (1) that the CCA unreasonably applied Supreme Court law by failing to augment the record sua sponte to include all juror questionnaires and the complete voir dire transcript so as to allow for a comprehensive “comparative juror analysis,” and that we should therefore give no deference to the state courts; and (2)even if deference is due, the CCA’s decision to credit the prosecutor’s nonracial justifications for challenging African-American jurors was objectively unreasonable. We consider both arguments in turn.
II.
Petitioners’ first argument is based on the proposition that
Batson
requires a state appellate court to perform a comparative juror analysis, and that failing to do so constitutes an unreasonable application of Supreme Court law and negates the deference usually due state courts in federal habeas proceedings. Although the CCA performed a version of comparative analysis, Petitioners contend that it was insufficient and that the CCA should have augmented the trial court record so that an acceptable comparative analysis was possible.
Ordinarily, Petitioners’ failure to raise this issue before the state trial court would be decisive. The usual rule is that, absent plain error, we would not fault a trial court for not ruling on an issue never raised, so that the trial court would have the opportunity to consider the issue.
See Johnson v. United States,
520 U.S. 461, 466-67, 117 S.Ct. 1544, 137 L.Ed.2d 718 (1997). But our court, sitting en banc, has held otherwise.
See Kesser v. Cambra,
465 F.3d 351, 377 (9th Cir.2006) (en banc) (Rymer, J., dissenting) (describing majority’s position). Our majority decision in
Kesser
held that comparative juror analysis is not waived “even when it was not requested or attempted in the state court.”
Id.
at 361;
see also Boyd v. Newland,
467 F.3d 1139, 1148 (9th Cir.2006) (amending prior opinion because it held that the CCA was not required to perform comparative juror analysis because it was not requested in the trial court). We thus first consider whether, in 2003 when the CCA issued its opinion, it was clearly established that comparative juror analysis was required such that we cannot give deference to the state court here.
“A
Batson
challenge involves a three-part test. 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. Third, the court must determine whether the defendant has shown purposeful discrimination.”
Kesser,
465 F.3d at 359 (internal quotation marks omitted).
The third of these determinations requires “a sensitive inquiry into such circumstantial and direct evidence of intent as may be available.”
Id.
at 361 (emphasis removed) (internal quotation marks omitted). The Supreme Court has recognized the utility of comparative juror analysis in completing that inquiry since long before the decision in this case by the CCA in 2003.
Id.
at 360 (“The Court’s holding means that the principles expounded in
Miller-El [v. Dretke,
545 U.S. 231, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005) ] [which applied comparative juror analysis] were clearly established Supreme Court law for AEDPA purposes at least by the time of the last reasoned state court decision in
Miller-El,
handed down in 1992, before Kesser’s 1993 trial”).
Following the Supreme Court’s lead, we have stated that comparative juror analy
sis is a “centerpiece' of the
Batson
analysis.”
Boyd,
467 F.3d at 1150. However, as we recently clarified,
“Batson
and the cases that follow it do not require trial courts to conduct a comparative juror analysis.”
Murray v. Schriro,
745 F.3d 984, 1005 (9th Cir.2014). Instead, “comparative juror analysis is an important means for federal courts to review a trial court’s ruling in a
Batson
challenge.”
Id.
In particular, we stated in
Murray
that
in order for
us
to discharge our responsibility under AEDPA to review a
Bat-son
claim under section 2254(d)(2), we often will have to conduct a formal comparative juror analysis, and our responsibility to conduct a comparative juror analysis is not contingent on whether the state court previously performed or did not perform a formal comparative juror analysis.
Id.
Moreover, we have not refused to accord AEDPA deference in a habeas proceeding based solely on a state court’s failure to apply comparative juror analysis. In fact, in
Cook v. LaMarque,
593 F.3d 810 (9th Cir.2010), relying on
Ali v. Hickman,
584 F.3d 1174 (9th Cir.2009), we explicitly refused to do so.
In
Cook,
neither the state trial nor appellate courts had performed comparative juror analysis. The majority applied the deferential review required by AEDPA, section 2254(d)(2), to petitioner’s claim. 593 F.3d at 816. The dissent argued that the failure to apply comparative juror analysis was “contrary to federal law” such that the state courts’ factual findings were entitled to no deference.
Id.
at 831 (Hawkins J., dissenting). The majority rejected this position because, even “[ajssuming for the sake of argument that this statement was correct prior to our opinion in
Ali,
it is no longer accurate.”
Id.
at 816 n. 2.
Ali
considered whether section 2254(e)(1) should apply rather than section 2254(d)(2) when a petitioner is arguing comparative juror analysis for the first time in federal habeas proceedings. We explained that, although the state courts had not performed comparative juror analysis, we nonetheless would review the case under subsection (d)(2) deference because the relevant evidence was in the state court record. 584 F.3d at 1180 n. 4. If the information was not entirely in the record, we would have reviewed the new evidence under section 2254(e)(1).
Id.
(citing
Taylor v. Maddox,
366 F.3d 992 (9th Cir.2004) (explaining that a court first reviews evidence before'the state court under section 2254(d)(2) and then moves on to apply the clear and convincing evidence standard under section (e)(1) to evidence introduced for the first time in habeas proceedings)). The important point, for our purposes, is that refusing deference under subsection (d)(1) was not an option that the court even considered. Accordingly,
Cook
and
Ali
undermine Petitioners’ “no deference” argument. As we recently explained, following
Cook,
a “state court’s finding that the prosecutor did not engage in purposeful discrimination is reviewed under the deferential standard set forth in [section 2254(d)(2) ].”
Jamerson v. Runnels,
713 F.3d 1218, 1224
&
n. 1 (9th Cir.2013).
Miller-El
and
Kesser,
on which Petitioners rely heavily, only bolster our approach. In
Miller-El,
despite the state courts’ failure to perform comparative juror analysis, the Supreme Court granted the deference required under section 2254(d)(2). 545 U.S. at 240, 125 S.Ct. 2317. We followed the same course in
Kesser,
465 F.3d at 358.
Two cases Petitioners cite,
Green
and
Boyd,
do vary from our chosen approach. However, both are distinguishable.
In
Green v. LaMarque,
we criticized the state courts for failing to “undertake a sensitive inquiry into such circumstantial and direct evidence of intent as
may be available, including a comparative analysis of similarly situated jurors, as required by clearly established Supreme Court law at the time of the trial.” 532 F.3d 1028, 1030 (9th Cir.2008) (internal quotation marks and citation omitted). We further stated that we would review do novo whether the prosecutor’s reasons were race neutral.
Id.
at 1031. However, our de novo analysis was not based solely on the state court’s failure to perform comparative analysis, but rather on our broader conclusion that the courts had simply failed “to reach step three in the
Batson
analysis.”
Id. Green
therefore does not require us to forego deference based only a failure to perform comparative juror analysis. If it did, it would be in direct conflict with
Cook.
Indeed, as we have recently explained, following both
Green
and
Cook,
even if the “state court declined to perform” a comparative juror analysis, “AEDPA deference still applies, and the state court decision cannot be upset unless it was based upon an ‘unreasonable determination of the facts.’ ”
Jamerson,
713 F.3d at 1225 (citing
Cook,
593 F.3d at 816 & n. 2, and
Green,
532 F.3d at 1031).
In
Boyd,
petitioner moved to supplement the record on appeal to include the entire voir dire transcript and requested a copy of the full transcript to assist in his development of his
Batson
argument. The CCA denied part of his request:
Petitioner filed three requests to supplement the record to include the entire voir dire transcript. The California Court of Appeal granted Petitioner’s requests in part and required that he be provided the voir dire of the excused African-American juror plus his counsel’s argument under
Batson.
But the court of appeal denied Petitioner’s requests for the entire voir dire transcript because it concluded that he did not comply with a California local rule that requires a defendant to establish with some certainty how5 the requested materials may be useful on appeal.
467 F.3d at 1142-43 (internal quotation marks omitted). Petitioner was therefore never actually provided with a full voir dire transcript. Because comparative juror review is important in
Batson
analysis and the petitioner was actively seeking to develop a comparative argument, we held that the state court had unreasonably applied Supreme Court law by
refusing
to allow the petitioner to have a copy of the transcript, thereby preventing comparative juror analysis.
Id.
at 1151. Because we did not have a full record, we remanded with instructions to the state appellate court to either provide a copy of the transcript to petitioner or grant the writ.
Id.
at 1152.
The key difference here is that the CCA did not prevent Petitioners from having access to the voir dire transcript in order to develop a comparative juror argument. Instead, Petitioners simply failed to bring a motion requesting that the CCA include the complete voir dire and questionnaires, which was their burden under California law.
In California, voir dire transcripts and jury questionnaires are not automatically included in the record that is before the CCA.
See People v. Goldberg,
110 Cal.App.2d 17, 242 P.2d 116, 121 (1952)
(“[V]oir dire
examination is not part of a normal record”). California law puts the burden squarely on counsel to move to augment the record:
[Cjounsel has a duty to insure that there is an adequate record before the appellate court from which those contentions may be resolved on their merits. Where the appropriate record is missing or incomplete, counsel must see that the defect is remedied, by requesting augmentation or correction of the appellate
record ... or by other appropriate means.
People v. Barton,
21 Cal.3d 513, 146 Cal.Rptr. 727, 579 P.2d 1043, 1047 (.1978) (citations omitted). Neither counsel fulfilled this obligation here, as they conceded at oral argument. The CCA did, on its own motion, augment the record to include parts of the voir dire, but omitted the first day and did not include the questionnaires. If anything, this should have prompted counsel to see that “the defect [was] remedied” — it did not relieve counsel from fulfilling their burden.
While we refused to adopt an inflexible requirement of comparative juror analysis in
Boyd,
we did not then specify what would be required to deviate from the general rule requiring the analysis. Here, based on Petitioners’ failure to augment the record, it is clear that an exception is merited. Our only alternative would be to hold that the CCA was unreasonable when it did not sua sponte perform what was otherwise counsel’s job. While Supreme Court law may have clearly established so strong a preference for comparative juror analysis that it was erroneous for the state court in
Boyd
to prevent affirmatively petitioner from making the argument, it was not clearly established in 2003 that the preference for comparative juror analysis is so unbending that it eviscerates the California law that places the burden on a petitioner to augment any deficits in the record.
Petitioners contend that moving to augment the record would have been futile in light of then — controlling California law, which, as we have discussed, did not allow an appellate court to perform comparative juror analysis when the trial court had not. But we do not allow litigants to escape their responsibilities on the basis of perceived futility.
See Engle v. Isaac,
456 U.S. 107, 130, 102 S.Ct. 1558, 71 L.Ed.2d 783 (1982) (“If a defendant perceives a constitutional claim and believes it may find favor in the federal courts, he may not bypass the state courts simply because he thinks they will be unsympathetic to the claim”).
We therefore hold that the CCA did not unreasonably apply
Batson
when it did not sua sponte augment the record so as to allow for comprehensive comparative juror analysis.
III.
Petitioners’ second argument is that, even granting deference, the CCA’s decision upholding the trial court’s finding that the prosecutor did not exclude jurors based on race was unreasonable.
At the outset, we must address the effect of the state’s decision to supplement the record in these habeas proceedings with the first day of voir dire and the jury questionnaires for the seated jurors. These materials, were we able to include them in our review, would be instrumental in determining whether the prosecutor’s reasons were pretextual. However, we cannot incorporate them into our analysis under either section 2254(d)(2) or (e)(1).
The Supreme Court’s decision in
Miller-El
provides some guidance on this issue. In
Miller-El,
there was no dispute that the voir dire transcript was in the record before the state courts. There was, however, some question about the juror questionnaires. 545 U.S. at 241 n. 2, 125 S.Ct. 2317. Justice Thomas, writing in dissent, argued that the questionnaires could not be part of the analysis because section 2254(d)(2) allows review only “in light of the evidence presented in the State court proceeding.”
Id.
at 280, 125 S.Ct. 2317 (Thomas, J., dissenting). In a footnote, the majority explained why it could incorporate the questionnaires:
So far as we can tell from the voluminous record before us, many of the juror questionnaires, along with juror information cards, were added to the habeas record after the filing of the petition in the District Court. The State raised no objection to receipt of the supplemental material in the District Court or the Fifth Circuit, and in this Court the State has joined with
Miller-El
in proposing that we consider this material, by providing additional copies in a joint lodging ... Neither party has referred to the provision that the reasonableness of the state-court determination be judged by the evidence before the state court, 28 U.S.C. § 2254(d)(2), and it is not clear to what extent the lodged material expands upon what the state judge knew; the same judge presided over the
voir dire,
the Swain hearing, and the
Batson
hearing, and the jury questionnaires were subjects of reference at the
voir dire.
The last time this case was here the State expressly relied on the questionnaires for one of its arguments, and although it objected to the Court’s consideration of some other evidence not before the state courts, it did not object either to questionnaires or juror cards. This time around, the State again relies on the jury questionnaires for its argument that the prosecution’s disparate questioning was not based on race. We have no occasion here to reach any question about waiver under § 2254(d)(2).
Id.
at 256 n. 15, 125 S.Ct. 2317 (citations omitted).
Because the majority expressly declined to base its consideration of the questionnaires on waiver by the state, its reason for relying on them must have been that because the trial judge had access to them, they were in fact part of the “evidence presented in the state court proceedings.” It is unclear how the trial court’s familiarity with the questionnaires related to the record before the Texas Court of Criminal Appeals, which, as the last state court to issue a reasoned opinion, was where the Supreme Court’s review focused.
See Kes-ser,
465 F.3d at 379. However,
Miller-El
originated in Texas state court. The majority’s conclusion as to the propriety of considering the questionnaires therefore does not speak to whether, when a petitioner fails to meet the burden imposed upon him by California law to augment the record, a habeas court may nonetheless conclude that the questionnaires were in some way before the CCA.
Because we must review the last reasoned state court decision, and because
Miller-El
is not clear as to what went into the record before the Texas appellate court, Miller-El’s consideration of the questionnaires does not require us to include them in our analysis here. We can only review the CCA’s decision under section 2254(d)(2) in light of the evidence before the CCA, and because it is undisputed that the first day of voir dire and the questionnaires were not in the record, we cannot include them in our analysis of whether the CCA made unreasonable factual findings.
However, section 2254(e)(1) may provide another avenue for considering the questionnaires. That section allows us to consider evidence produced for the first time during habeas proceedings after working through the section 2254(d)(2) analysis.
See Taylor,
366 F.3d at 1000. But if the additional evidence was not developed before the state court due to Petitioners’ failings, we cannot consider the evidence unless it relates to a new rule of constitutional law or “could not have been previously discovered through the exercise of due diligence.” 28 U.S.C. § 2254(e)(2)(h). Here, as discussed above, the fault lies with Petitioners for not augmenting the record. Because “the District Court made no finding that [Petitioners] had been dili
gent in pursuing [the evidence] (and thus that § 2254(e)(2) was inapplicable) or that the limitations set forth in § 2254(e)(2) were met,”
Holland v. Jackson,
542 U.S. 649, 653, 124 S.Ct. 2736, 159 L.Ed.2d 683 (2004), we will not consider the additional evidence under section 2254(e).
We thus turn to the partial voir dire and the
Batson
hearing transcript, as the “circumstantial and direct evidence of intent” that was before the CCA, to determine whether the CCA made an unreasonable factual determination under subsection (d)(2). “Here our standard is doubly deferential: unless the state appellate court was objectively unreasonable in concluding that a trial court’s credibility determination was supported by substantial evidence, we must uphold it.”
Briggs v. Grounds,
682 F.3d 1165, 1170 (9th Cir.2012), ce
rt. denied,
— U.S. -, 133 S.Ct. 894, 184 L.Ed.2d 694 (2013). “AED-PA imposes a highly deferential standard for evaluating state-court rulings and demands that state-court decisions be given the benefit of the doubt.”
Felkner v. Jackson,
— U.S. -, 131 S.Ct. 1305, 1307, 179 L.Ed.2d 374 (2011) (internal quotation marks omitted). “Additionally, it is widely acknowledged that the trial judge is in the best position to evaluate the credibility of the prosecutor’s proffered justifications.”
Briggs,
682 F.3d at 1171.
Petitioners focus on the prosecutor’s challenge of Jurors Andrews, Reeves, Hilton, and Woods. In reviewing their arguments, we keep in mind that
counsel is entitled to take account of the characteristics of the other prospective jurors against whom peremptories might be exercised; to reevaluate the mix of jurors and the weight he gives to various characteristics as he begins to exhaust his peremptories; and to take into account tone, demeanor, facial expression, emphasis — all those factors that make the words uttered by the prospective juror convincing or not. In short, counsel is entitled to exercise his full professional judgment in pursuing his client’s legitimate interest in using [peremptory] challenges ... to secure a fair and impartial jury.
Burks v. Borg,
27 F.3d 1424, 1429 (9th Cir.1994) (internal quotation marks omitted).
A. Juror Andrews
The prosecutor challenged Juror Andrews because she seemed hesitant, intimidated, weird, and inattentive. He also stated at the hearing that he wouldn’t be able to tell that Juror Andrews was African-American by looking at her.
Petitioners first argue that the prosecutor’s reference to Andrews’s race indicates discriminatory intent. The prosecutor stated: “[T]he only way we would even know she’s African-American is because she put on her questionnaire that she’s of Caucasian, African-American, [and] I think American Indian [sic]. But physically to look at her, you would not be able to tell she’s any parts African-American.” Because the prosecutor would have looked at Andrews’s questionnaire and known how she identified herself before seeing her, Petitioners argue that her appearance at voir dire was irrelevant and the prosecutor’s comments indicate a fixation on race. But Andrews’s appearance was not offered as a
reason
for challenging Andrews — it was an ultimately unpersuasive attempt to dispel the inference of racial motivation. We will not translate the prosecutor’s unpersuasive argument, which had nothing to do with his actual reasons for challenging Andrews, into evidence of racial animus.
Relying on
Kesser,
Petitioners also argue that the prosecutor’s “hesitant, intimidated, weird” rationale is insufficient. In
Kesser,
the prosecutor explained that he
challenged a juror because she was “misty” and “emotional.” 465 F.3d at 364. We were suspicious of the rationale because it was “so underdeveloped that it likely falls short of Batson’s mandate for a clear and reasonably specific explanation of [the] legitimate reasons for exercising the challenges.”
Id.
(internal quotation marks omitted). However, we only rejected the rationale after comparing the challenged juror with others and examining the record to determine whether the juror was actually misty and emotional.
Id.
at 364-65.
Kesser
therefore does not stand for the proposition that we can categorically reject the rationale here based on the fact that it is “underdeveloped,” even if we were to accept that characterization.
Furthermore, even if “[t]aken individually, these factors might seem so innocuous they would not support a peremptory challenge,” when considered together, “it is plausible that an unbiased prosecutor would be concerned by the juror’s overall demeanor.”
Cook,
593 F.3d at 819 (considering prosecutor’s explanation that he challenged a juror because the juror was “weird in appearance,” and made other comments that the prosecutor found strange or objectionable).
Petitioners also argue that the transcript does not show that Andrews was weird, hesitant, intimidated, or inattentive. However, we will not overturn the trial court’s credibility finding absent “extraordinary circumstances,” and Petitioners’ contrary interpretation of a cold transcript, without more, does not qualify.
[T]he best evidence [of discriminatory intent] often will be the demeanor of the attorney who exercises the challenge. In addition, race-neutral reasons for peremptory challenges often invoke a juror’s demeanor (e.g., nervousness, inattention), making the trial court’s firsthand observations of even greater importance. In this situation, the trial court must evaluate not only whether the prosecutor’s demeanor belies a discriminatory intent, but also whether the juror’s demeanor can credibly be said to have exhibited the basis for the strike attributed to the juror by the prosecutor. We have recognized that these determinations of credibility and demeanor lie peculiarly within a trial judge’s province.
Snyder v. Louisiana,
552 U.S. 472, 477, 128 S.Ct. 1203, 170 L.Ed.2d 175 (2008) (internal quotation marks and citations omitted). The transcript itself does not so clearly show that Andrews was not hesitant, intimidated, or inattentive such that we could here hold that the state court was objectively unreasonable in crediting the prosecutor. In fact, these are precisely the sort of behavioral nuances that a cold transcript is ill-suited to reveal.
We will not disturb the trial court’s findings unless the “facts in the record are objectively contrary to the prosecutor’s statements,”
McClain v. Prunty,
217 F.3d 1209, 1221 (9th Cir.2000). Petitioners’ ability to make that showing is quite limited in light of the very short voir dire we have to consider. Nonetheless, Petitioners advance the same comparative juror analysis argument here that they made before the CCA: namely, that a comparative analysis shows that all African-American jurors were asked whether they would be sympathetic to Petitioners, while only white jurors with relatives who had committed crimes or could “be expected to sympathize with defendants” were asked the same question.
The problem with this argument is that, without all the questionnaires, it is not possible to determine whether there was a non-racial common thread among the white and African-American jurors that explains the sympathy question. The Supreme Court has cautioned that
a retrospective comparison of jurors based on a cold appellate record may be very misleading when alleged similarities were not raised at trial. In that situation, an appellate court must be mindful that an exploration of the alleged similarities at the time of trial might have shown that the jurors in question were not really comparable.
Snyder,
552 U.S. at 483, 128 S.Ct. 1203. This hesitation must apply with even more force here where, because the voir dire was so limited and we cannot consider the questionnaires, we are not even able to perform a full comparison of jurors to begin with. To hold that the supposed disparate questioning supports a discriminatory intent so clearly as to render the CCA’s decision objectively unreasonable, without being able to compare fully the jurors, would be to substitute our half-informed judgment for the fully informed trial judge’s: “We have only a cold transcript to guide us while the trial judge was there to observe the jury selection-day in and day out for six months. Evaluation of the prosecutor’s state of mind based on demeanor and credibility lies peculiarly within a trial judge’s province.”
Burks,
27 F.3d at 1429 (internal quotation marks omitted). We must afford the state trial court the “benefit of the doubt” to which it is entitled on its credibility determination.
See Felkner,
131 S.Ct. at 1307.
Petitioners also argue that because the trial court did not explicitly credit these “demeanor based” justifications,
Snyder
forbids the CCA from relying on them. In
Snyder,
the prosecutor offered two reasons for a peremptory challenge of Juror Brooks: Brooks’ nervousness and conflicts with his schedule. 552 U.S. at 478, 128 S.Ct. 1203. The trial court allowed the challenge without further elaboration.
Id.
at 479, 128 S.Ct. 1203. While recognizing that deference is “especially appropriate where a trial judge has made a finding that an attorney credibly relied on demeanor in exercising a strike,” the Supreme Court held that it could not simply assume that the trial court had credited the prosecutor’s nervousness justification because the trial court had not specified which of the prosecutor’s two explanations it had relied on.
Id.
The Court therefore undertook an analysis of whether the other explanation was credible and determined it was not.
Id.
at 485, 128 S.Ct. 1203. Because if a peremptory strike is “shown to have been motivated in substantial part by discriminatory intent” it is invalid, and because the prosecution had described the scheduling reason as one of its “main con-cernís],” the court could not merely credit the nervousness justification and deny the petition.
Id.
at 485, 128 S.Ct. 1203 (internal quotation marks omitted).
Here, none of the prosecutor’s demean- or-based rationales are contradicted by the transcript. We therefore are not presented with the
Snyder
problem: presuming one of two rationales was the basis for the court’s decision when one rationale was pretextual.
B. Juror Reeves
The prosecutor challenged Juror Reeves because (1) the prosecutor received from Reeves a look that the prosecutor interpreted as disdainful while both were in the hall and (2) Reeves hesitated when asked if he felt sympathetic towards the Defendants.
As to the first rationale, Petitioners argue that this is in fact a race-based reason and indicates a discriminatory mindset. But the prosecutor’s concern was that Reeves was hostile towards
him,
not that Reeves would favor Petitioners based on Reeves’s race. Hostile looks or a negative attitude can be a legitimate basis for a peremptory challenge.
See Burks,
27 F.3d at 1429 (holding that a prosecutor is entitled to make a challenge based on a hunch or suspicion created by a juror’s
demeanor, tone, and facial expressions);
Williams v. Rhoades,
354 F.3d 1101, 1109 (9th Cir.2004) (holding that a prosecutor’s challenge to a juror because she was “cold” and “evasive” toward him was legitimate, in reliance on
Burks).
Petitioners’ second argument, that the transcript does not show hesitation, suffers from the same defect as above. Petitioners have presented no objective evidence that would firmly convince us that the trial court was wrong to credit the prosecutor on this point, and their contrary interpretation of the transcript is not sufficient.
See Williams,
354 F.3d at 1109 (“We must be left with a firm conviction that the determination made by the state court is wrong and the one urged by [Petitioners] is correct”).
C. Jurors Hilton and Woods
Petitioners have also made arguments concerning the prosecutor’s challenges to Jurors Hilton and Woods. The government argues that these contentions should be ignored because they constitute uncertified issues. While it is true that the panel only granted a certificate of ap-pealability as to Andrews and Reeves, the treatment of other jurors can be a valid consideration in step three of
Batson
as part of the “circumstantial and direct evidence of intent as may be available.” 476 U.S. at 93, 106 S.Ct. 1712. Thus, if there was something about the challenge to Hilton and Woods that shows their dismissal was pretextual, the Andrews and Reeves analysis may be affected.
See Snyder,
552 U.S. at 478, 128 S.Ct. 1203 (“Here, as just one example, if there were persisting doubts as to the outcome, a court would be required to consider the strike of Ms. Scott for the bearing it might have upon the strike of Mr. Brooks”).
Virtually all of Hilton’s examination took place on the first day, the transcript of which we cannot consider. Petitioners’ only argument that does not depend in some way on a review of evidence that was not in the record is that the prosecutor once again revealed a fixation on race by observing that Hilton had put a question mark in the race box on his questionnaire. But the prosecutor’s statement can reasonably be construed as a general distrust of Hilton based on his refusal to answer questions, and we find nothing in the record that would require us to overturn the trial court’s credibility finding.
The prosecutor challenged Woods because Woods (1) was a victim of police brutality and (2) had a weird look on his face during questioning that made the prosecutor uneasy. Petitioners argue that the police brutality reason is not relevant because there was no police brutality in this case. However, the prosecutor was entitled to use his professional judgment to decide whether Woods’s experience would have soured into a general distrust of law enforcement, making him an unfavorable juror. Petitioners’ contention that they would make an opposite inference does not render the CCA’s holding objectively unreasonable.
Finally, Petitioners argue that, beyond the circumstances of the individual challenges to Andrews, Reeves, Hilton, and Woods, there are “global points that emerge from the record” showing that the prosecutor’s reasons were pretextual.
First, Petitioners argue that the raw numbers show that a much larger percentage of African-American potential jurors had been eliminated than the non-African-American potential jurors: 70% to 30%. While it is true that “seriously disproportionate exclusion of blacks from the jury venire is powerful evidence of intentional race discrimination,”
McClain v. Prunty,
217 F.3d 1209, 1223 (9th Cir.2000), the trial judge was aware of this disproportion and nonetheless credited the prosecutor’s reasons.
[T]he most generous reading [of this disparity] would suggest only that the trial court had reason to question the prosecutor’s credibility ... That does not, however, compel the conclusion that the trial court had no permissible alternative but to reject the prosecutor’s race-neutral justifications and conclude [Petitioners have shown] a
Batson
violation. Reasonable minds reviewing the record might disagree about the prosecutor’s credibility, but on habeas review that does not suffice to supersede the trial court’s credibility determination.
Rice v. Collins,
546 U.S. 333, 341-42, 126 S.Ct. 969, 163 L.Ed.2d 824 (2006).
Petitioners also argue that an inference of racial discrimination is clear from the fact that for seventeen of the twenty non-African-American jurors the prosecutor struck, the reason for the strikes was quite clear. In contrast, Petitioners argue, the reasons for striking the African-American jurors are only clear for three of the seven. But Petitioners’ views as to the relative clarity of reasons for excluding a juror are irrelevant. Both obvious reasons, like manifest bias on the part of a potential juror, and non-obvious reasons, like hunches and suspicions on the part of the prosecutor, are legitimate. The fact that, in Petitioners’ minds, the reasons for all the white jurors were obvious while those for the African-American jurors were non-obvious does not show that the trial court was objectively unreasonable in considering and crediting the non-obvious reasons.
In sum, Petitioners have not demonstrated that the CCA made an unreasonable determination of fact in light of the evidence before it.
AFFIRMED.