California state prisoners Daniel Adam Wade and Paul Christopher Buckley (“petitioners”) appeal the district court’s denial of their petitions for habeas corpus, in which they challenge their convictions for multiple counts of assault upon a police officer with a firearm. Petitioners assert that the prosecutor in their joint criminal trial exercised a peremptory challenge to exclude an African-American from the jury in violation of the Equal Protection Clause of the Fourteenth Amendment to the United States Constitution, and that the state trial court erred in finding that petitioners had not established a prima facie case of discrimination under Batson v. Kentucky, 476 U.S. 79 , 106 S.Ct. 1712 , 90 L.Ed.2d 69 (1986).

Evaluation of allegedly discriminatory peremptory challenges to potential jurors in federal and state trials is governed by the standard established by the United States Supreme Court in Batson . However, California courts evaluate such challenges, as they have done in this case, under the pre-Batson standard established by the California Supreme Court in People v. Wheeler, 22 Cal.3d 258 , 148 Cal.Rptr. 890 , 583 P.2d 748 (1978). We hold that the Wheeler standard, as currently interpreted by the California courts, does not satisfy the constitutional requirement laid down in Batson . Because the state court did not reasonably apply federal law as clearly established by the United States Supreme Court in evaluating petitioners’ claims in this ease, we do not defer to its findings as we would otherwise be required to do under the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”), 28 U.S.C. § 2254 (d).

Evaluating petitioners’ Batson claims de novo, we nonetheless find that petitioners have not established a prima facie case of discrimination that would have required the prosecutor to provide a race-neutral basis for his peremptory challenge. We therefore affirm the district court’s denial of habeas corpus.

I. Background

Petitioners were small-scale marijuana dealers operating out of petitioner Wade’s apartment in Richmond, California. On the evening of May 5, 1993, a team of police officers attempted to serve a search warrant at Wade’s apartment. The police alleged that they knocked on the door, *1193 identified themselves as police officers, and demanded entry. Petitioners alleged that they did not hear the police officers identify themselves before they attempted to enter the apartment, and that they believed the officers were members of the “Butcher Boys” gang who had recently threatened them. Petitioners fired their weapons through the apartment door as the officers were attempting to break through it with a sledgehammer. Two officers were wounded. Petitioners are African-American, as were virtually all of the defense witnesses. The races of the wounded officers and of the prosecution witnesses are not clear from the record before us.

Sixty-four prospective jurors were called, of whom only five were African-American. Of these five, one no longer lived in Contra Costa County and was not properly part of the venire. Another potential African-American juror was excused for cause after he indicated during voir dire that he could not judge the case impartially. Two African-Americans— first Vera Rutherford and then Joann James — were peremptorily challenged by the prosecution. One African-American ultimately served on the jury. Petitioners assert that the peremptory challenge to Ms. Rutherford, and the relevant surrounding circumstances of the challenge, constitute a prima facie case of discrimination under Batson .

Prior to the voir dire, Ms. Rutherford completed a 70-question jury questionnaire. She indicated that she was fifty-one years old and living with two grandchildren. She was employed full-time as a legal technician for the United States Department of Labor and worked for an administrative law judge. She had completed some college and some vocational training. She was active in her church and liked to read the Bible and Stephen King novels. She had never had occasion to call the police, and she had not been the victim of any kind of crime. About ten years earlier, she had appeared in court as a witness in a DUI case in which she had been a passenger in the car of the driver who was charged. In partial answer to Question 28 of the questionnaire, she indicated that her experience being interviewed by the police was “not bad,” and -testifying in court was also “not bad.” (These two responses were crossed out on the questionnaire, as were all the other responses to Question 28. We nevertheless recite them here because Ms. Rutherford appears to have crossed them out when she later concluded that the DUI experience should have been described in response to Questions 29 and 80, rather than Question 28.) The DUI case was ultimately dropped by the prosecutor. Ms. Rutherford was concerned about burglaries and shootings in her neighborhood. Her family did not own any firearms. She did not know any law enforcement personnel, but she knew some lawyers and judges through her work. She had never used drugs, had never known anyone with a drug problem, and thought that drugs were bad.

As he did with other members of the venire during voir dire, the trial judge asked Ms. Rutherford follow-up questions directly rather than allowing the attorneys to ask the questions. She said that she “[got] the files together and the paperwork together” for the administrative law judge. She had no specialized legal training or paralegal training. None of the attorneys she knew did criminal work. She did not feel that anything about her job would affect her ability to be a fair juror. The DUI case occurred ten years earlier and the driver had been a friend of a friend. She had been a witness at a preliminary hearing and was asked questions by the district attorney. The charges were later dropped. She thought the police and the district attorney had treated her and the defendant fairly, and she felt that nothing about the experience would cause her any problems in serving as a juror.

Ms. Rutherford was the fifth juror peremptorily challenged during voir dire, and the third peremptorily challenged by the prosecution. Prior to challenging Ms. *1194 Rutherford, the prosecutor had peremptorily challenged two potential jurors who were not African-American. The prosecution did not attempt to challenge Ms. Rutherford for cause. After the prosecutor challenged Ms. Rutherford, defendants moved for a mistrial under Batson and Wheeler . After a brief hearing, the trial court denied the motion, finding that defendants had not made a prima facie showing of discrimination.

When the remaining two African-American venire members were called to the jury box later in the voir dire, the prosecutor peremptorily challenged one of them, Ms. Joann James. The defendants then renewed their motion for a mistrial. The trial court again held that the defendants had not made a prima facie case, noting that “there are things in the questionnaire, as well as in her answers [that would] cause concern ... for any reasonable attorney [from] the prosecutor’s viewpoint.” The second of the two remaining African-American venire members was ultimately seated as a juror.

After trial, the jury convicted petitioners on all charges. Wade was sentenced to a term in state prison of 14 years and 8 months, and Buckley was sentenced to a term of 10 years. Petitioners appealed within the state court system.

The California Court of Appeal panel asked for supplemental briefing on the question whether the standard for establishing a prima facie case under Batson is different from the standard under Wheeler , but the majority of the panel did not cite Batson , or indeed any other federal case, in its opinion. The majority applied Wheeler in affirming petitioners’ convictions, holding that “defense counsel failed to establish from all the circumstances of the case a strong likelihood that [the jurors in question] were being challenged because of their group association.” People v. Buckley, 53 Cal.App.4th 658, 663 , 61 Cal.Rptr.2d 860 (1997) (emphasis added). The dissenting Justice also applied Wheeler even though he recognized that Wheeler might have been “meant by our Supreme Court to impose a more stringent standard of bias than the 'inference’ of bias that suffices under Batson v. Kentucky [.]” Id. at 670 , 61 Cal.Rptr.2d 860 (Kline, P.J., dissenting). The California Supreme Court denied petitioners’ requests for discretionary review.

Petitioners each sought writs of habeas corpus in federal district court. They contended that the California courts used the wrong legal standard in applying Wheeler rather than Batson and that they had established a prima facie case of discrimination under Batson. Although the question was argued to it, the district court did not resolve the asserted inconsistency between the Wheeler and Batson standards. Giving some deference to the determination of the state trial judge, the district court held that petitioners had not established a pri-ma facie case of discrimination under Bat-son. The district court noted that, by striking Ms. Rutherford, the prosecutor had struck the sole African-American then on the panel, but it concluded that “more is required” to establish a prima facie case. The district court dismissed the petitioners’ evidence of a statistical disparity in the use of peremptory challenges on the ground that “[s]uch statistical analyses clearly require a larger sample size.” Further, the district court believed that Ms. Rutherford was not “quite so perfect” a prosecution witness as the petitioners suggested, in that a prosecutor might have felt that Ms. Rutherford’s “impartiality may have been tainted by her experience” as a witness in the ten-year-old DUI preliminary hearing.

The district court denied the petitions, and petitioners timely appealed to this court.

II. Standard of Review

We review de novo the decision of a district court to grant or deny a petition for writ of habeas corpus. See Bonin v. Calderon, 59 F.3d 815, 823 (9th Cir.1995). We “may affirm on any ground supported by the record, even if it differs from the rationale of the district court.” Id.

*1195 Petitioners’ habeas petitions are governed by AEDPA. 1 Under AEDPA, we may only disturb a state court’s determinations of law if they were “contrary to” or “involved an unreasonable application of’ clearly established federal law as determined by the United States Supreme Court. 28 U.S.C. § 2254 (d)(1); Furman v. Wood, 190 F.3d 1002, 1004 (9th Cir.1999). Although we have not yet defined the precise reach of these overlapping standards, it is clear that both are met when the state court has failed to follow the law as set forth by the Supreme Court. See id. If the state court’s determination is purely one of fact, we review it for clear error. See Bonin, 59 F.3d at 823 .

Normally, on a petition for a writ of habeas corpus, we review the state trial court’s fact-specific determination of whether a defendant has made a prima facie case of a Batson violation deferentially, with AEDPA’s “statutory presumption of correctness.” Tolbert v. Page, 182 F.3d 677, 685 (9th Cir.1999) (en banc). However, where the state trial court uses the wrong legal standard, this rule of deference does not apply. See United States v. Bishop, 959 F.2d 820, 827 (9th Cir.1992). We are therefore compelled to reach the question of whether the state trial court erred as a matter of law in applying Wheeler .

III. Comparison of the Batson and Wheeler Standards

The state trial court determined that petitioners had not established a prima facie case requiring the prosecutor to provide a race-neutral explanation of his challenge to Ms. Rutherford. In order to decide the degree of deference we owe the state trial court’s determination, we must decide whether it applied the correct legal standard in making that determination.

In Batson, the United States Supreme Court set out a three-step process in the trial court to determine whether a peremptory challenge is race-based in violation of the Equal Protection Clause. First, the defendant must make a prima facie showing that the prosecutor has exercised a peremptory challenge on the basis of race. See Batson, 476 U.S. at 96-97 , 106 S.Ct. 1712 . That is, the defendant must demonstrate that the facts and circumstances of the case “raise an inference” that the prosecution has excluded venire members from the petit jury on account of their race. Id. at 96 , 106 S.Ct. 1712 . If a defendant makes this showing, the burden then shifts to the prosecution to provide a race-neutral explanation for its challenge. See id. at 97 , 106 S.Ct. 1712 . The trial court then determines whether the defendant has established purposeful racial discrimination by the prosecution. See id. at 98 , 106 S.Ct. 1712 .

In Wheeler , decided eight years before Batson, the California Supreme Court presaged Batson by interpreting Article I, section 16 of the California Constitution to prohibit race-based peremptory challenges. 22 Cal.3d at 272 , 148 Cal.Rptr. 890 , 583 P.2d 748 . Just as the Batson Court would later do, the Wheeler Court outlined a process that the state trial courts should use to identify race-based peremptory challenges, the first step of which required the defendant to make a prima facie showing of discrimination. But unlike the Batson Court, which required only that the defendant “raise an inference” of discrimination, the Wheeler Court demanded that the defendant “show *1196 a strong likelihood” that the prosecutor had excluded venire members from the petit jury on account of their race. Id. at 280 , 148 Cal.Rptr. 890 , 583 P.2d 748 . Batson and Wheeler thus may be thought to prescribe different tests for establishing a prima facie case.

Upon closer examination, however, it appears that the Wheeler Court itself understood “a strong likelihood” to mean a “reasonable inference.” While the Wheeler Court, on page 280 of its opinion, required a defendant to “show a strong likelihood” that the prosecutor excluded venire members from the jury on the basis of race, the Wheeler Court phrased its central holding somewhat differently on the very next page: “Upon presentation of this and similar evidence — in the absence, of course, of the jury- — the court must determine whether a reasonable inference arises that peremptory challenges are being used on the ground of group bias alone.” Id. at 281 , 148 Cal.Rptr. 890 , 583 P.2d 748 (emphasis added). It is this language regarding a “reasonable inference” that the Supreme Court of the United States borrowed when it formulated the Batson test.

In 1982, the California Court of Appeal for the First District concluded that the two phrases used by the Wheeler Court, “strong likelihood” and “reasonable inference,” amounted to the same thing. “We believe that a fair reading of Wheeler requires only that the court find a reasonable inference of group bias once an appropriate foundation is laid in the first of the two stages.” People v. Fuller, 136 Cal.App.3d 403, 423 , 186 Cal.Rptr. 283 (1982). The Fuller court was unwilling to believe that the California Supreme Court “intended to create different options for trial judges within one page of each other in so carefully crafted an opinion as the Wheeler opinion.” Id. at 423 n. 25, 186 Cal.Rptr. 283 . We agree that when a court announces its central holding twice, using two different phrases, it logically follows that the two phrases are intended to mean the same thing. By subsequently adopting a variation of the “raise an inference” phraseology, the Batson Court therefore adopted a standard compatible with the Wheeler standard, as interpreted by Fuller . California state judges following Wheeler and Fuller thus conducted Wheeler/Batson inquiries by looking to the language of both tests; they saw that both tests required the defendant to raise a reasonable inference that jurors had been stricken on the basis of an impermissible criteria; and in so doing they successfully complied with the dictates of the California and Federal Constitutions.

This appears to have been the law in California until 1994, when another panel of the California Court of Appeal disavowed Fuller , thereby creating an inconsistency between Wheeler and Batson. The Court of Appeal in People v. Bernard, 27 Cal.App.4th 458 , 32 Cal.Rptr.2d 486 (1994), wrote:

Id. at 465, 32 Cal.Rptr.2d 486 . The Bernard court thus explicitly rejected the idea that “reasonable inference” and “strong likelihood” meant the same thing, turning its back on Fuller and Batson. It appears to us that from that point forward, the California state courts have applied a low *1197 er standard of scrutiny to peremptory strikes than the federal Constitution permits.

The California Supreme Court now routinely insists, despite Batson, that a defendant must show “a strong likelihood” of racial bias. Its consistent practice is to cite Batson and Wheeler together as controlling law but to quote the “strong likelihood” language from Wheeler rather than the “raise an inference” language from Batson. See, e.g., People v. Welch, 20 Cal.4th 701, 745 , 85 Cal.Rptr.2d 203 , 976 P.2d 754 (1999) (“Under Wheeler and Bat-son, if a party believes his opponent is using his peremptory challenges to strike jurors on the ground of group bias alone, .... he must show a strong likelihood that such persons are being challenged because of their group association.”) (emphasis added, internal quotation marks omitted); People v. Davenport, 11 Cal.4th 1171, 1199-1200 , 47 Cal.Rptr.2d 800 , 906 P.2d 1068 (1995) (“Under Wheeler and Batson, [i]f a party believes his opponent is using his peremptory challenges to strike jurors on the ground of group bias alone, .... he must show a strong likelihood that such persons are being challenged because of their group association.”) (emphasis added, internal quotation marks omitted); People v. Turner, 8 Cal.4th 137, 164 , 32 Cal.Rptr.2d 762 , 878 P.2d 521 (1994) (same). Batson is, of course, the law of the land. California law may give greater protection to criminal defendants than is required by the federal Constitution, but it cannot give less. Yet this is precisely what the California courts now do when they follow the Wheeler “strong likelihood” test in determining whether a prima facie case has been established.

In our view, the Wheeler “strong likelihood” test for a successful prima facie showing of bias is impermissibly stringent in comparison to the more generous Batson “inference” test. Indeed, when the California Court of Appeal resolved the direct appeal in the case now before us, it followed the literal language of Wheeler and characterized its test for a prima facie case as “not easy.” Buckley, 53 Cal.App.4th at 663 n. 17, 61 Cal.Rptr.2d 860 . It wrote, “The ‘strong likelihood’ phrase conveys the clear message that the test is not an easy one (a message we take to heart in the present case).... ” Id. The United States Supreme Court has explained that the Batson framework was intended significantly to reduce the quantum of proof previously required of a defendant who wished to raise a claim of racial bias in the jury selection procedure. See Georgia v. McCollum, 505 U.S. 42, 47 , 112 S.Ct. 2348 , 120 L.Ed.2d 33 (1992); see also Tolbert, 182 F.3d at 680. Accordingly, although the Batson prima facie case requirement cannot be “taken for granted,” it “is not onerous.” United States v. Escobar-de Jesus, 187 F.3d 148, 164 (1st Cir.1999) (quoting United States v. Bergodere, 40 F.3d 512, 516 (1st Cir.1994)). We therefore conclude that California courts in following the “strong likelihood” language of Wheeler are not applying the correct legal standard for a prima facie case under Batson.

AEDPA directs us to defer to state court determinations unless they involve an unreasonable application of clearly established federal law as determined by the United States Supreme Court. See 28 U.S.C. § 2254 (d). Where the California courts follow the “strong likelihood” language of Wheeler without any indication that they are actually applying a “reasonable inference” test consonant with Batson, they apply an incorrect legal standard. In such a case, we need not— indeed, should not — -give deference to their determination that a defendant has failed to establish a prima facie case of bias. Because the California courts followed the “strong likelihood” test of Wheeler rather than the “inference” test of Batson in adjudicating petitioners’ claims of racial bias in the prosecution’s use of its peremptory challenges, we review the petitioners’ Batson claims de novo.

IV. Application of Batson

As we noted above, in order to establish a prima facie case of a Batson *1198 violation, petitioners must show that the circumstances raise an inference that the prosecutor used a peremptory challenge to exclude a potential juror because of race. See Batson, 476 U.S. at 96 , 106 S.Ct. 1712 ; Tolbert, 182 F.3d at 680 . The petitioners contend that they made a prima facie showing that Ms. Rutherford was peremptorily struck because of her race. We disagree.

When the prosecutor challenged Ms. Rutherford, he had already exercised two peremptory challenges against prospective jurors who were not African-American. In United States v. Vasquez-Lopez, 22 F.3d 900 (9th Cir.1994), we reaffirmed the well-established principle that, in order to meet Batson’s prima facie burden, a petitioner need not show “that the prosecution had engaged in a pattern of discriminatory strikes against more than one prospective juror” because “the Constitution forbids striking even a single prospective juror for a discriminatory purpose.” Id. at 902 . However, we were also careful to explain that “the fact that the juror was the one Black member of the venire does not, in itself, raise an inference of discrimination.” Id. As the district court in our case properly stated, “More is required.”

Petitioners point out that when Ms. Rutherford was challenged, one of three (or 33%) of the prosecutor’s peremptory challenges had been exercised against an African-American, when only four of sixty-four (or 6%) of the prospective jurors in the venire were African-American. We have two responses. First, we agree with the district court that the sample is so small that the statistical significance of the percentages is limited. Second, we do not believe that the only relevant time at which to assess the would-be pñma facie case is the time of the challenge. If an African-American is the first person called, and thus the first person struck, all (or 100%) of the prosecutor’s peremptory challenges will have been exercised against African-Americans at the time of the challenge. But if that same African-American is called at the end of the voir dire, the percentage may be far lower.

In this case, by the end of the voir dire, two more prospective African-American jurors had been questioned. One of them, Ms. James, was challenged, but the record contains entirely plausible reasons, independent of race, why a prosecutor would not have wanted her as a juror. She had indicated that her brother had been prosecuted for burglary, that her brother used drugs, that she was somewhat familiar with the area in which the alleged crimes occurred, and that she thought that search warrants “do quite a bit of damage which is unnecessary.” Indeed, petitioners did not argue to the district court, and do not argue to us, that race was a factor in the challenge to Ms. James. The second African-American was seated on the jury.

Petitioners further argue that other jurors “similarly situated” to Ms. Rutherford were permitted to remain on the jury despite the fact they “appeared to be less favorable.” While this contention, if true, would allow us to infer racial bias and would thus support petitioners’ pnma fa-cie case, the district court “combed the record” and found no evidentiary support for this contention. We agree with the district court.

Finally, petitioners point to Ms. Rutherford herself, who, they contend, would have been a model juror from the prosecution’s prospective. We doubt our ability, or the ability of the parties, always to assess accurately whether a juror will be sympathetic to a particular side. To the degree that such an assessment can be made, it was not as clear to the district court (and is not as clear to us) as the petitioners claim it is to them that Ms. Rutherford would have been a model juror for the prosecution. Ms. Rutherford explained during voir dire that she had been in a car driven by “a friend of a friend” when the car was pulled over and the driver was arrested for drunk driving. Ms. Rutherford testified on the driver’s behalf at a hearing, during which she was cross-examined by a member of the Contra Costa County District Attorney’s Of *1199 fice, the same office that was prosecuting petitioners. Although she indicated that she believed the driver was fairly treated and that her experience with the police and in court was “not bad,” the district court concluded that it would be understandable if the prosecutor feared that her views might have been influenced by her experience with the criminal justice system in the same county in which she was now called to serve as a juror.

V. Conclusion

Because we conclude that the state trial court applied the wrong legal standard in determining whether defendants had established a prima facie case of discrimination under Batson, we decide that question de novo. After an independent review of the record, we conclude that petitioners did not establish a prima facie case. We therefore affirm the district court’s decision denying petitioners’ writs of habeas corpus.

AFFIRMED.

. As amended by AEDPA, 28 U.S.C. § 2254 (d) provides: