dissenting:
This is a troubling case, but after considering all of the petitioner’s arguments and applying the standard of review prescribed by the federal habeas statute, I see no ground for reversing the decision of the district court. The majority holds that the petitioner’s rights under Batson v. Kentucky, 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), were violated, but I do not believe that there is a proper basis for disturbing the credibility findings made by the conscientious state judge. The majority also holds that comments made by the prosecutor in closing argument at the penalty phase of the trial violated Caldwell v. Mississippi 472 U.S. 320, 105 S.Ct. 2633, 86 L.Ed.2d 231 (1985), but in my view the majority misinterprets that decision as in effect embodying a per se prohibition against any mention of the availability of appellate review of a death sentence, a procedure of which virtually all jurors are surely aware. Because I cannot agree with the majority’s analysis of either of these issues, I must respectfully dissent.
I.
A.
I turn first to the argument that the prosecution violated Batson by using peremptory challenges to strike three African Americans from the jury panel. In Bat-son, the Supreme Court held that it is a violation of the Equal Protection Clause for a prosecutor to strike a juror because of race. The Court also set out a three-step process for adjudicating a claim that a particular peremptory was racially based.
[0]nce the opponent of a peremptory challenge has made out a prima facie case of racial discrimination (step one), the burden of production shifts to the proponent of the strike to come forward with a race-neutral explanation (step two). If a race-neutral explanation is tendered, the trial court must then decide (step three) whether the opponent of the strike has proved purposeful racial discrimination. Hernandez v. New York, 500 U.S. 352, 358-359, 111 S.Ct. 1859, 114 L.Ed.2d 395 (1991)(plurality opinion); id., at 375, 111 S.Ct. 1859 (O’CONNOR, J., concurring in judgment); Batson, [476 U.S.] at 96-98, 106 S.Ct. 1712.
Purkett v. Elem, 514 U.S. 765, 767, 115 S.Ct. 1769, 131 L.Ed.2d 834 (1995)(per curiam).
*318In this case, the Superior Court found that the petitioner, William Riley, made out a prima facie case, see Riley v. State, No. 200, 1988 (Super. Ct. April 21, 1989) at 2, and the state does not dispute this point. The state offered race-neutral justifications for its contested strikes, and the state courts accepted those explanations and found that the disputed peremptories were not racially based. Id. at 3-6; Riley v. State, 585 A.2d 719, 725 (Del.Sup.Ct.1990). The majority however, rejects the state courts’ findings regarding two of the state’s peremptories and substitutes its own contrary findings. I will discuss each of the challenges on which the majority relies.1
B: Ray Nichols
1. The prosecutor testified that he struck Nichols because he was uncertain that Nichols would be able to vote for a death sentence. See App. 797-99. According to the prosecutor’s testimony, “there was a pause and a significant pause in [his] answering [the trial judge’s] inquiry and that to me was enough to suggest that he might not be able to return a death penalty.” Id. Having heard the prosecutor’s testimony, the judge who presided over the Batson hearing [hereinafter “the hearing judge”], concluded: “I find the State provided a credible, race-neutral reason for exercising its peremptory challenge after appraising the demeanor and credibility of the juror. The State’s exercise of its peremptory challenge was nondiscriminatory. I am satisfied that the peremptory challenge was not made on the ground of the juror’s race.” Id. at 889.
Riley suggests that it is not believable that the prosecutor was able to remember at the time of the evidentiary hearing in 1988 that Nichols had paused while answering a question during voir dire six years earlier. In addition, Riley contrasts the prosecutor’s ability to remember this pause with his inability to remember another potentially significant aspect of the jury selection process, and Riley notes that the prosecutor was a friend and neighbor of the victim. These facts were highlighted during the cross-examination of the prosecutor at the Batson hearing, see App. 820-29, and I agree that they were important factors to be considered in assessing the prosecutor’s credibility. The hearing judge was aware of these facts and had the opportunity to observe the prosecutor testify on the witness stand. Despite these facts, however, the hearing judge found that the prosecutor’s testimony was credible.
Our standard of review of the hearing judge’s finding is narrow. In Batson, the Supreme Court took pains to note that “[s]ince the trial judge’s findings in the context under consideration here will largely turn on evaluation of credibility, a reviewing court ordinarily should give those findings great deference.” Batson, 476 U.S. at 98 n. 21, 106 S.Ct. 1712. In a later case applying Batson, the plurality elaborated:
In the typical peremptory challenge inquiry, the decisive question will be whether counsel’s race-neutral explanation for a peremptory challenge should be believed. There will seldom be much evidence bearing on that issue, and the best evidence often will be the demeanor of the attorney who exercised the challenge .... [Evaluation of the prosecutor’s state of mind based on demeanor and credibility lies ‘peculiarly within a trial judge’s province.’
*319Hernandez v. New York, 500 U.S. 352, 353, 111 S.Ct. 1859, 114 L.Ed.2d 395 (1991) (plurality) (citation omitted).
Because the present case is a proceeding under the federal habeas statute, our scope of review is, if anything, even narrower. Under 28 U.S.C. § 2254(d)(8)(1988 & Supp.1990), any state-court factual finding that is “fairly supported by the record” is entitled to a presumption of correctness. Discussing this provision, the Supreme Court wrote in Marshall v. Lonberger, 459 U.S. 422, 434, 103 S.Ct. 843, 74 L.Ed.2d 646 (1983), that “28 U.S.C. § 2254(d) gives federal habeas courts no license to redetermine credibility of witnesses whose demeanor has been observed by the state trial court, but not by them.” Accord, Rushen v. Spain, 464 U.S. 114, 122 n. 6, 104 S.Ct. 453, 78 L.Ed.2d 267 (1983)(per curiam). The Marshall Court elaborated:
In United States v. Oregon Medical Society, 343 U.S. 326, 72 S.Ct. 690, 96 L.Ed. 978 (1952), commenting on the deference which this Court gave to the findings of a District Court on direct appeal from a judgment in a bench trial, we stated:
“As was aptly stated by the New York Court of Appeals, although in a case of a rather different substantive nature: ‘Face to face with living witnesses the original trier of the facts holds a position of advantage from which appellate judges are excluded. In doubtful cases the exercise of his power of observation often proves the most accurate method of ascertaining the truth.... How can we say the judge is wrong? We never saw the witnesses.... To the sophistication and sagacity of the trial judge the law confides the duty of appraisal.’ Boyd v. Boyd, 252 N.Y. 422, 429, 169 N.E. 632.” Id., at 339, 72 S.Ct. 690. We greatly doubt that Congress, when it used the language “fairly supported by the record” considered “as a whole” intended to authorize broader federal review of state court credibility determinations than are authorized in appeals within the federal system itself.
459 U.S. at 434, 103 S.Ct. 843. See also Burkett, 514 U.S. at 769, 115 S.Ct. 1769.
Under the very limited scope of review that applies here, I do not see how the hearing judge’s finding that the prosecutor testified truthfully regarding the reason for challenging Nichols can be overturned. The hearing judge heard the prosecutor testify. He was aware of the factors noted above that provided grounds for doubting his testimony, but he nevertheless found that the prosecutor was truthful. I would sustain that finding.
2. The majority rejects that finding (as well as the state courts’ finding with respect to another prospective juror whom I discuss below) in large part because, in the majority’s view, “there is no basis to determine if the state courts undertook, or even were aware of, the required Batson step three inquiry.” Maj. Op. at 291. In making this argument, the majority (a) misunderstands what Batson requires, (b) ignores what the Delaware courts did in this case, (c) imposes novel and unwarranted procedural requirements on the state courts, and (d) awards relief that extends beyond what its own logic warrants.
What step three of Batson requires. Although the majority makes step three seem elaborate and elusive — so elusive that, according to the majority, the Delaware courts may not have “fully appreciated the requirement” (Maj. Op. at 289)— step three, is neither conceptually difficult nor procedurally complicated. Step three simply requires the judge to make a finding of fact — “to determine if the defendant has established purposeful discrimination,” *320Batson, 476 U.S. at 98, 106 S.Ct. 1712 (footnote omitted). See also Purkett, 514 U.S. at 767, 115 S.Ct. 1769; Hernandez, 500 U.S. at 358-59, 111 S.Ct. 1859 (plurality opinion); id at 375, 111 S.Ct. 1859 (O’Connor, J., concurring in judgment). Neither Batson nor any subsequent Supreme Court or Third Circuit case has added to this requirement.
What the Delaware courts did. The Delaware courts did exactly what step three requires. With respect to potential juror Nichols, the hearing judge, whose analysis the state supreme court endorsed, see 585 A.2d at 725, noted that the state had provided a race-neutral reason for the challenge, stated that he found the explanation credible, and concluded: “I am satisfied that the peremptory challenge was not made on the ground of the juror’s race.”2 This is precisely the finding that step three of Batson mandates. See Batson, 476 U.S. at 98, 106 S.Ct. 1712 (the court has “the duty to determine if the defendant has established purposeful discrimination”).
What the majority requires. Although the majority opinion is loathe to admit it, what the majority really finds wanting in the opinions of the Delaware courts is not a failure to make the finding mandated by step three of the Batson inquiry but a failure to comment on the record regarding evidence that seems, in the majority’s view, to undermine the prosecution’s proffered explanations for the disputed per-emptories. See Maj. Op. at 287 (“the state courts in this case rejected Riley’s Batson claim without discussing any of the ample evidence that throws into question the explanations offered by the prosecutor for striking two of the black jurors.... ”).3 This approach is inconsistent with the federal habeas statute and Supreme Court precedent.
The provision of the federal habeas statute on which the majority relies provides that if a state court’s “determination after a hearing on the merits of a factual issue” is “evidenced by a written finding, written opinion, or other reliable and adequate written indicia,” that determination “shall be presumed to be correct” unless it is not “fairly supported by the record” as a whole. 28 U.S.C. § 2254(d)(8)(1988 & Supp.l990)(amended 1996). Under this provision, the state court’s factual determination must simply be evidenced by “a written finding, written opinion, or other rehable and adequate written indicia,” and it is not even necessary that a state court “specifically articulate its credibility findings.” LaVallee v. Delle Rose, 410 U.S. 690, 692, 93 S.Ct. 1203, 35 L.Ed.2d 637 (1973). See also Marshall v. Lonberger, 459 U.S. at 433, 103 S.Ct. 843. Thus, 28 U.S.C. § 2254(d) plainly does not authorize us to disregard a state court’s factual finding on the ground that the state court failed to discuss all the evidence or to explain why it was not persuaded by a particular piece of proof.
*321I do not question that a judge, in making the factual finding required by step three of Batson, should consider all of the relevant evidence that has been adduced. But neither Batson nor any later Supreme Court or Third Circuit case4 suggests that a federal habeas court is free to reject the factual findings of a state court if the state court does not comment on all of the evidence or provide what the federal court regards as a satisfactory explanation for its finding.5 The majority confuses the obligation to consider all of the relevant evidence (something that a court should always do in making findings of fact) with the obligation to comment on all of the evidence (an obligation that we are not free to impose on state courts).
In this case, as I have noted, there is no question that the state courts did precisely what step three of Batson required — they made findings as to whether Riley had established purposeful discrimination. And they did so only after discovery and a thorough hearing. There is no reason to believe that the state courts did not consider all of the relevant evidence, including all of the evidence that the majority now finds persuasive. The Delaware courts simply did not comment on all of this evidence. (Judgments about credibility based on a witness’s demeanor often do not lend themselves to such explanation). But the Delaware courts were not obligated to comment on all of the evidence. The majority in this case reviews the decisions of the Delaware courts as if they were decisions of a Social Security administrative law judge, who must, we have held, “give some reason for discounting the evidence she rejects.” Plummer v. Apfel, 186 F.3d 422, 429 (3d Cir.1999). The Delaware courts, however, are not to be treated as if they were federal administrative agencies.
The majority’s relief. Even if the majority were correct that the Delaware courts were obligated to explain on the record why they accepted the prosecution’s explanations for its strikes and were not persuaded by Riley’s evidence, that would hardly justify the relief that the majority orders — the granting of the writ unless Riley is re-tried. When a decision is found to be faulty for failure to provide an adequate explanation, the logical remedy is to remand so that an adequate explanation can be supplied. See, e.g., Florida Power & Light Co. v. Lorion, 470 U.S. 729, 744, 105 S.Ct. 1598, 84 L.Ed.2d 643 (1985); Motor Vehicle Mfrs. Assn. v. State Farm Mut., 463 U.S. 29, 57, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983). Here, however, the majority does not give the Delaware courts a chance to comply with the majority’s newly found procedural requirements. Nor does the majority remand to the District Court for a hearing and findings specifically addressing the points that the majority views as important. Instead, the majority orders that the writ be granted unless Riley is re-tried.
*322How the state courts’ failure to address Riley’s evidence on the record can justify such relief is baffling. If the lack of explanations on the record is important to the majority’s analysis, then the case should be remanded so that explanations may be provided. If no explanations on the record could satisfy the majority-and I believe that to be the case-then the majority’s lengthy discussion of the inadequacy of the Delaware courts’ opinions is beside the point.
3. The majority’s remaining reasons for rejecting the hearing judge’s finding require little response. The majority notes that, although the prosecutor testified that Nichols paused, “the record reflects no such pause.” Maj. Op. at 278. I have read many trial transcripts, and I do not recall any in which the court reporter noted that a witness had or had not paused before answering a question.
The majority finds it significant that “despite Nichols’ alleged pause, the prosecutors did not ask the trial court to remove Nichols for cause.” Maj. Op. at 279. The majority raises the question “why if Nichols actually did pause ‘a significant pause,’ the state did not seek to have him removed for cause.” Id. Does the majority seriously believe that a prospective juror who pauses before answering a question about the death penalty may properly be removed/or cause?
The majority observes that “[t]he record does not show ... that any of the contemporaneous notes kept by the prosecutors as to some of the jurors reflected either the existence of a pause or the concern about which[the prosecutor] testified six years later.” Maj. Op. at 279. The prosecutor’s notes, however, consist of a handwritten sheet with a few words or abbreviations scrawled next to the names of some of the prospective jurors. Nichols’s name is not even on this sheet. The notes by no means record the reasons for all of the prosecution’s strikes.
I wish there were some scientific test that could determine with complete certainty whether Nichols paused and whether the prosecutor told the truth. Unfortunately, there is no such test. We must rely to a substantial degree on the ability of the judge who heard the prosecutor’s testimony to make an accurate assessment of his credibility. There is no question that the hearing judge took his responsibility seriously and made his finding in good faith. Our role under the federal habeas statute is to determine whether that credibility finding is “fairly supported by the record.” It is.
C: Charles McGuire.
1. Riley’s strongest Batson claim concerns the prosecution’s strike of Charles McGuire. At trial, the prosecutor first used a peremptory challenge against McGuire and then immediately made the following application to the trial judge:
[THE PROSECUTOR]: Your Honor, may I ask the Court to reconsider charging the State for that strike. This Mr. McGuire came to chambers yesterday and expressed his belief that he didn’t know if he could last the two weeks [the estimated length of the trial], there was some problem with work. He was an inspector or something for the Department of Labor. I know he came in yesterday.
THE COURT: I will not strike him for cause for that reason. He asked to be excused yesterday and I decided not to excuse him.
App. 250.
At the evidentiary hearing held before the hearing judge, the prosecutor testified that he struck McGuire because McGuire “had previously requested to be excused *323from jury service” and because the prosecutor “wanted attentive jurors” who were not worried about missing other obligations or activities while the trial took place. App. 801.
The defense called McGuire as a witness at the evidentiary hearing. McGuire testified that he was employed by the State of Delaware as a Social Security “disability adjudicator,” App. 846-47; that he had been reporting for jury duty in the courthouse in Dover for two to three weeks before he was questioned in connection with the Riley case but had not been seated on a jury, id. at 852-53; that while he was away from work, the disability claims assigned to him would “just sit[ ],” id. at 850; that the director of his office had told him that he was going to make a “formal request” that McGuire be excused, id. at 860; that such a request was sent, id. at 853, 856; and that the request had been discussed in chambers with the judge. Id. at 849-50, 856. McGuire said, however, that he himself had never expressed an unwillingness to serve on the jury and had been willing to do so. See id. at 850.
The hearing judge accepted the prosecutor’s explanation of the reason for striking McGuire. The hearing judge found:
The State peremptorily challenged Charles McGuire because [the prosecutor] believed he requested to be excused from jury duty and, therefore, may have been unable or unwilling to serve for the entirety of the trial.... McGuire’s employer sent a letter requesting he be released from jury duty because he could not be replaced at his job if he was chosen for jury duty. The letter by McGuire’s employer clearly gave the State reason to question whether McGuire would give his full time and attention to the trial and whether he would be able to serve for the entirety of the time projected for the trial. Whether McGuire, in fact, did not request relief from jury duty and did wish to serve is of no consequence.
Riley v. State, No. 200, 1988 at 4-5 (emphasis added). The hearing judge then noted that the state’s explanation for striking McGuire, was “entirely unrelated to the juror’s race,” and the judge credited that explanation. Id. at 5. Obviously, by crediting an explanation that was “entirely unrelated to the juror’s race,” the hearing judge necessarily found that Riley had not “established purposeful discrimination,” Batson, 476 U.S. at 98, 106 S.Ct. 1712, and the hearing judge thus fully complied with Batson’s step three.6
Several factors provide substantial support for this finding. It is apparent that McGuire’s work situation was on the prosecutor’s mind when McGuire was peremptorily challenged because, as noted, immediately after striking McGuire, the prosecutor. asked that McGuire’s dismissal be deemed for cause since he had “expressed his belief that he didn’t know if he could last the two weeks.” App. 250. In addition, a reasonable prosecutor might well have wondered whether McGuire’s work situation would adversely affect his attentiveness at trial. As noted, McGuire’s supervisor had made a “formal request” that he be excused “because he could not be replaced at his job if he was chosen for jury duty.”7 Whether or not McGuire himself in fact wished to serve on the jury, the impression apparently was conveyed that *324McGuire wanted to be excused and to return to work, since the trial judge commented: “He asked to be excused yesterday and I decided not to excuse him.” See App. 250. Under these circumstances, a reasonable prosecutor could have been concerned that McGuire might have been inattentive at trial due to worry about missing work, leaving his duties unattended, and perhaps incurring his supervisor’s displeasure.
Riley attacks the hearing judge’s finding on two grounds. First, he points out that, according to McGuire’s testimony at the post-conviction relief evidentiary hearing, McGuire himself did not ask to be excused. This argument is unpersuasive. Although McGuire testified that he did not ask to be excused, the trial judge, as noted, stated at the time of McGuire’s dismissal: “He asked to be excused yesterday and I decided not to excuse him.” App. 250 (emphasis added). Thus, McGuire, who was unable to remember many details at the time of the post-conviction relief evidentiary hearing, see id. at 853, 857-62, may have been mistaken, or he may have conveyed the impression at the time of trial that he personally wanted to be excused.
Second, Riley points out that the handwritten sheet prepared by the prosecutors during voir dire contains the following notation next to the name of a white juror, Charles Reed, whom the prosecution did not peremptorily strike: “works Lowe’s— wants off.” One of the prosecutors was questioned about this notation by Riley’s attorney at the post-conviction relief evi-dentiary hearing, but the prosecutor testified that he had no recollection of Reed. See App. 823-24.
The notation by Reed’s name and the prosecutor’s testimony at the evidentiary hearing are certainly factors that the hearing judge could have viewed as tending to undermine the credibility of the prosecutor’s explanation for striking McGuire, but the notation and the prosecutor’s testimony are insufficient to show that the hearing judge’s finding is not “fairly supported by the record.” 28 U.S.C. § 2254(d). It is reasonable to infer from the notation “wants off’ that, at some point in the jury selection process, Reed expressed a desire to be excused for some reason. As far as I am aware, however, the record does not establish why8 or how strongly Reed wanted to be excused. The transcript of the voir dire shows that, at the final stage of the jury selection process, the members of the venire were asked whether there was “any reason why [they] absolutely [could not] serve,” App. 223; that members of the venire then successfully asked to be released for reasons such as a previously planned vacation, id. at 253; but that Reed made no request to be excused at that time. See id. at 229-30. Thus, as far as the record appears to reveal, Reed may have had a relatively weak desire and reason to be excused, and his situation may not have been at all comparable in this respect to McGuire’s.9
*325As I have noted, our scope of review of the hearing judge’s finding is narrow. Although it would be satisfying to know why Reed was not stricken, that unanswered question is not enough, in view of the “great deference”10 owed the hearing judge’s credibility determination, to demonstrate that the hearing judge’s finding is not “fairly supported by the record.”11 28 U.S.C. § 2254(d)(8) (1988 & Supp.1990).
In an effort to bolster its unusual decision to overturn the hearing judge’s credibility finding, the majority points to a statement contained in the brief filed by the state in Riley’s direct appeal. The majority writes:
When Riley’s direct appeal came before the Delaware Supreme Court in 1984, the State justified the use of race in selecting jurors in criminal trials. On that occasion, which was the State’s first opportunity to defend the use of its peremptory challenges in Riley’s trial, the State did not offer a single race-neutral explanation, not even as an alternate argument; instead, it claimed that it was permissible — even socially desirable — to exclude jurors based on what it called ‘group association.’ ”
Maj. Op. at 284 (quoting App. 896).
This argument is not well taken. Responding to Riley’s suggestion that the Delaware Supreme Court should hold that individual race-based peremptories were unconstitutional, the state’s brief argued as follows:
Because the Sixth Amendment does not support [Riley’s argument] and the decision in Swain v. Alabama, 380 U.S. 202, 85 S.Ct. 824, 13 L.Ed.2d 759 (1965), more appropriately recognizes how peremptory challenges, even those exercised on the basis of group association, foster the constitutional goal of an impartial jury, the state asserts that no reversal is required here.
App. 896-97 (footnotes omitted). Thus, the state’s brief — which the lead trial prosecutor did not even sign — merely urged the state supreme court to follow the reasoning of the United States Supreme Court in what was then the governing federal precedent. It is far-fetched to interpret the state’s reliance on Swain as a tacit admission that its peremptories in this case were based on race — particularly since, in a footnote to the sentence quoted above, the state was careful to deny that its challenges were racially based.12
It is also unreasonable to draw an adverse inference against the state for not providing race-neutral explanations for its challenges in its appellate brief. Since *326there was no evidence in the record regarding the reasons for the strikes, the state could hardly have expected the state supreme court to base a decision on explanations provided without record support. The majority’s arguments regarding the state’s brief are insubstantial.
The majority’s reliance on statistical evidence is even worse. In the Batson proceeding before the hearing judge, Riley made a proffer that no African American had served on any of the three other first-degree murder trials that had occurred in Kent County within a year of his own and that in those cases the prosecution had peremptorily challenged five African Americans. The three other trials were those of Andre Deputy, an African American, and two whites, Daniel Pregent, who was acquitted, and Judith McBride, who was convicted. With respect to these cases, no information was provided at the time — and none has been provided since— about the identities of the prosecutors who participated in jury selection, the racial makeup of the venire, or the race of jurors who were dismissed for cause or peremptorily challenged by the defense.
In the trial of Andre Deputy, who was convicted and ultimately executed, the state struck four whites, one African American, and one person listed as “Indian.” Deputy argued that the prosecution’s peremptory challenge of the African American venireperson violated Batson. See Deputy v. Taylor, 19 F.3d 1485, 1492 (3d Cir.1994). Deputy’s Batson argument was rejected in the district court decision denying his petition for a writ of habeas corpus, and our court affirmed. See id. at 1492. Since it has been held that no Bat-son violation was shown in Deputy, it is difficult to see how that case can be viewed as supporting Riley’s argument here.
In Pregent’s case, the state struck four whites and one black. There is nothing before us to indicate that any Batson objection was made, and it is doubtful that the pattern of strikes exercised by the prosecution sufficed to make out a prima facie case.
The remaining case is the prosecution of Judith McBride for murdering her husband. See McBride v. State, 477 A.2d 174 (Del.1984). The state exercised a total of 10 strikes, of which three were against potential jurors identified as black.13 There is nothing to indicate that any Bat-son objection was made. Without in effect holding a Batson hearing, there is no way of determining whether any prosecution peremptories were based on race.
Although Riley was represented at the Batson hearing by a professor of law and has been represented in the federal habe-as proceeding by attorneys from one of the nation’s leading law firms, no expert analysis of these statistics has ever been offered.14 According to the majority, however, the “sophisticated analysis of a statistician” is not needed to interpret the significance of these statistics. Maj. Op. at 280. “An amateur with a pocket calculator,” the majority writes, can calculate that “there is little chance of randomly selecting four consecutive all white juries.” Id.
Statistics can be very revealing — and also terribly misleading in the hands of “an *327amateur with a pocket calculator.” The majority’s simplistic analysis treats the prospective jurors who were peremptorily challenged as if they had no relevant characteristics other than race, as if they were in effect black and white marbles in a jar from which the lawyers drew. In reality, however, these individuals had many other characteristics, and without taking those variables into account, it is simply not possible to determine whether the prosecution’s strikes were based on race or something else.
The dangers in the majority’s approach can be easily illustrated. Suppose we asked our “amateur with a pocket calculator” whether the American people take right— or left-handedness into account in choosing their Presidents. Although only about 10% of the population is left-handed, left-handers have won five of the last six presidential elections.15 Our “amateur with a calculator” would conclude that “there is little chance of randomly selecting” left-handers in five out of six presidential elections. But does it follow that the voters cast their ballots based on whether a candidate was right— or left-handed?
Whether even a careful multiple-regression analysis of peremptory challenge statistics in other cases would suffice to show that a Batson violation occurred in this case is unclear. Cf. McCleskey v. Kemp, 481 U.S. 279, 107 S.Ct. 1756, 95 L.Ed.2d 262 (1987). Here, however, we have not been presented with any expert statistical evidence.
The majority and the concurrence argue that an adverse inference should be drawn against the state for failing to come forward with data to rebut Riley’s statistics. I see no basis for this approach. Whether an adverse inference should be drawn under particular circumstances based on a party’s failure to produce evidence in a state proceeding is in the first instance a question of state law, and unless a state court’s failure to draw such an inference in a particular case denies due process or a fair and adequate hearing, a federal habe-as court should be reluctant to reject the state court’s ruling. Cf. 28 U.S.C. § 2254(6) and (7) (1988 & Supp.1990) (amended 1996). Moreover, even if this were a collateral attack on a federal conviction, we would defer to the decision of the judge who conducted the hearing as to whether the circumstances justified the drawing of adverse inference and would reverse only if the judge committed an abuse of discretion. See, e.g., Bouzo v. Citibank, N.A., 96 F.3d 51, 60 (2d Cir.1996).
Here, the state courts’ failure to draw such an inference certainly did not constitute an abuse of discretion. The state was never given notice that it had any obligation to provide additional data, and it is not at all clear what sort of evidence the majority expects the state to have provided. The information that is most critically lacking — the prosecutors’ reasons for striking the five African American venire members in the Deputy, McBride, and Pregent cases — probably could not have been obtained without in effect conducting retrospective Batson hearings in those cases. Does the majority think that such a hearing would have been practical? Or does the majority think that the state should have retained am expert to analyze the state’s use of peremptory challenges in some other set of cases? In order to make such an analysis, the expert probably *328would have needed detailed information about the prospective jurors whom the state did and did not strike — e.g., their ages, marital status, education, occupations, and past experiences with law enforcement, to name just a few of the myriad variables that often figure in decisions about peremptory challenges. We have no indication that such information was available, and in any event, compiling and analyzing the data concerning a reasonable sample of cases could have been a massive undertaking. In my view, it is entirely unwarranted to hold that the state courts abused their discretion because they did not draw adverse inferences from the state’s failure to volunteer to conduct such a study in response to the statistics that Riley proffered.
In sum, I see no ground for overturning the hearing judge’s credibility findings. I would thus hold that the presumption of correctness has not been overcome and would reject Riley’s Batson argument. The majority — by in effect making its own credibility findings on the cold state court record — seriously errs. See Marshall v. Lonberger, 459 U.S. at 434, 108 S.Ct. 843.
II.
I now turn to the majority’s holding that a remark made by the prosecutor in closing argument at the penalty phase of the trial violated Caldwell v. Mississippi supra. In Caldwell, the defense attorney’s closing argument asked the jury to “confront both the gravity and responsibility of calling for another’s death.” 472 U.S. at 324, 105 S.Ct. 2633. In response, the prosecutor took strong exception to the defense attorney’s comments and stated:
Now, they would have you believe that you’re going to kill this man and they know — they know that your decision is not the final decision. My God, how unfair can you be? Your job is reviewable. They know it.... For they know, as I know, and as [the judge] has told you, that the decision you render is automatically reviewable by the Supreme Court.
Id. at 325-26, 105 S.Ct. 2633.
By a vote of five to three, the United States Supreme Court reversed the defendant’s death sentence. The plurality opinion approved by four justices concluded that the prosecutor’s comments were improper for two reasons: first, because the prosecutor’s description of the state scheme of appellate review was not “accurate” and, second, because the availability of appellate review was “wholly irrelevant to the determination of the appropriate sentence.” Id.
Justice O’Connor, who cast the deciding fifth vote for reversal, refused to endorse the principle that “the giving of nonmis-leading and accurate information regarding the jury’s role in the sentencing scheme is irrelevant to the sentencing decision.” 472 U.S. at 341, 105 S.Ct. 2633 (opinion of O’Connor, J.) (emphasis added). However, she agreed that the prosecutor’s statements were improper because they “creat[ed] the mistaken impression that automatic appellate review of the jury’s sentence would provide the authoritative determination of whether death was appropriate,” whereas in fact the state supreme court exercised only a narrow scope of review. Id.
In subsequent cases, the Court has clarified the holding in Caldwell. In Romano v. Oklahoma, 512 U.S. 1, 9, 114 S.Ct. 2004, 129 L.Ed.2d 1 (1994), the Court wrote as follows:
As Justice O’CONNOR supplied the fifth vote in Caldwell, and concurred on grounds narrower than those put forth by the plurality, her position is controlling. See Marks v. United States, 430 *329U.S. 188, 193, 97 S.Ct. 990, 51 L.Ed.2d 260 (1977).... Accordingly, we have since read Caldwell as “relevant only to certain types of comment — those that mislead the jury as to its role in the sentencing process in a way that allows the jury to feel less responsible than it should for the sentencing decision.” Darden v. Wainwright, 477 U.S. 168, 184, n. 15, 106 S.Ct. 2464, 91 L.Ed.2d 144 (1986). Thus, “[t]o establish a Caldwell violation, a defendant necessarily must show that the remarks to the jury improperly described the role assigned to the jury by local law.” Dugger v. Adams, 489 U.S. 401, 407, 109 S.Ct. 1211, 103 L.Ed.2d 435 (1989), see also Sawyer v. Smith, 497 U.S. 227, 233, 110 S.Ct. 2822, 111 L.Ed.2d 193 (1990).
The Romano Court rejected the Caldwell argument advanced in that case because “the jury was not affirmatively misled regarding its role in the sentencing process.” Id. at 10, 114 S.Ct. 2004.
Riley’s argument is based on a statement made by the prosecutor near the very beginning of his summation at the sentencing phase of the trial. The prosecutor stated:
As the Judge has explained to you we have a specific statute with regard to what occurred in a penalty hearing on a capital case.
Let me say at the outset that what you do today is automatically reviewed by our Supreme Court and that is why there is an automatic review on the death penalty. That is why, if you return a decision of death, that is why you will receive and have to fill out a two-page interrogatory that the Court will give you. This is an interrogatory that specifically sets out the questions that the State request and whether or not you believe it beyond a reasonable doubt and if you want in your determination, if you believe the sentence should be death then each and every one of you has to sign this. This goes to the Supreme Court. That is why it is concise and we believe clear and it should be looked carefully on and answered appropriately.
App. 393 (emphasis added). Riley argues that the highlighted words quoted above violated Caldwell.
In its decision on direct appeal, the Delaware Supreme Court responded to this argument as follows:
[T]he prosecutor’s remarks in no way suggested that responsibility for ultimately determining whether defendant faced life imprisonment or death rested elsewhere. The prosecutor’s passing comment to the jury that its decision would be “automatically reviewed” was fairly made in the context of the prosecutor’s preceding reference to the “specific statute[controlhng] a penalty hearing on a capital case.” 11 Del.C. § 4209. Since subsection (g) of § 4209 mandates the “Automatic Review of Death Penalty by Delaware Supreme Court”, the prosecutor in the instant case was simply quoting the statute. In no sense may it reasonably be said that the prosecutor was either misstating the law, misleading the jury as to its role, or minimizing its sentencing responsibility.
496 A.2d at 1025 (alteration in original). I agree with this analysis.
The prosecutor’s remarks in Caldwell were “quite focused, unambiguous, and strong.” 472 U.S. at 340, 105 S.Ct. 2633. The clear message was that, contrary to the suggestion of defense counsel that the jury should “confront both the gravity and responsibility of calling for another’s death,” id. at 324, 105 S.Ct. 2633, the jury need not shoulder that responsibility because “the authoritative determination of whether death was appropriate” would be *330made by the state supreme court. Id. at 343, 105 S.Ct. 2633 (Opinion of O’Connor, J.). It was in this sense that the remarks “‘improperly described the role assigned to the jury by local law1 ” 16 and thus “ ‘allowed the jury to feel less responsible than it should for the sentencing decision.’ ” 17
The prosecutor’s remarks in this case were very different. Here, the prosecutor made accurate, unemotional, passing remarks in the context of describing the state statute and explaining why the jury would have to “fill out a two-page interrogatory” if it returned a capital sentence. These remarks did not convey the message that the jury should not confront the gravity of returning a death verdict, and thus the mere mention of the fact that there would be an automatic appeal to the state supreme court did not mislead the jury as to its role in the sentencing process. In this connection, it is noteworthy that after the closing arguments, the trial judge instructed the jury on its role using language that left no doubt about its responsibility. The trial judge stated: “Where the jury submits such a finding and recommendation, the Court shall sentence the defendant to death.” See 585 A.2d at 731 (emphasis added). A “recommendation of death, supported by the evidence, shall be binding on the Court.” Id. (emphasis added). “Your unanimous recommendation for the imposition of the death penalty, if supported by the evidence, is binding on the Court.” Id. at 734 (emphasis added). In light of the substantial factual differences between Caldwell and this case, and in light of the Supreme Court’s subsequent explanation of the meaning of Caldwell, I would reject Riley’s Caldwell claim.
The majority appears to hold that a Caldwell violation occurred simply because the prosecutor accurately stated that there would be an automatic appeal to the state supreme court without attempting to explain the scope of review that the state supreme court would exercise. I do not agree with this reading of Caldwell. Neither Justice O’Connor’s controlling opinion in Caldwell nor the Court’s subsequent explanation in Romano took the position that an unadorned reference to automatic judicial review of a capital verdict is enough to violate the Constitution. And such a holding would make little sense. As the Seventh Circuit has noted:
Everyone knows that after a death sentence is imposed, there are tiers of appellate review designed to catch errors; the prosecutor wasn’t telling the jurors anything they didn’t know already. Appellate review is a fact of almost all criminal cases that are tried. Knowledge of this does not cause jurors to take lightly their sentencing responsibilities.
Fleenor v. Anderson, 171 F.3d 1096, 1098 (7th Cir.1999). What Caldwell forbids is not a simple reference to automatic appellate review, but the suggestion that the scope of review is broader than it is in fact. The remarks in Caldwell conveyed such a suggestion; the comments here did not. I would therefore hold that no Caldwell violation occurred.
III.
Reviewing habeas decisions in capital cases is one of the most important and difficult responsibilities of this court. Our role is.vital-but limited-and is not to be confused with that of the jury or the vari*331ous branches of state government. Applying the legal standards that are applicable to us in the present context, I believe that the decision of the district court must be affirmed.
The majority cites two federal habeas cases in which courts of appeals found great fault with the procedures used by state judges in adjudicating Batson objections. In Jordan v. Lefevre, 206 F.3d 196 (2d Cir.2000), the court of appeals held that the trial judge “could not properly decide the third Batson step” because he “resisted counsel’s efforts to make arguments regarding the peremptory strikes so as to create a full record” and instead “ruled summarily” after “an extremely brief colloquy.” Id. at 201. Likewise, in Coulter v. Gilmore, 155 F.3d 912 (7th Cir.1998), the court of appeals spent several pages describing the bizarre nature of the procedure used by the state trial court in ruling on Batson objections. Id. at 915-lb, 918. The court of appeals ultimately concluded that the state court had not considered the totality of the relevant circumstances and thus ordered that the petitioner be released if the state court did not conduct a Batson hearing using “the proper methodology.” Id. at 922. The procedures used by the Delaware courts in the present case bear no resemblance to the procedures found deficient in Jordan and Coulter.
The majority also cites two appeals in which the Sixth Circuit remanded cases for the district courts to provide more complete explanations of Batson rulings. See United States v. Harris, 192 F.3d 580, 588 (6th Cir.1999); United States v. Hill, 146 F.3d 337 (6th Cir.1998). These, however, were direct federal appeals, not habeas proceedings initiated by state prisoners, and the relief ordered — remands for fuller explanation — goes far beyond what the majority ordered here.
. Riley also contends that a third member of the venire, Lois Beecher, was peremptorily challenged by the state because of race. The majority, however, does not rely on this strike, and accordingly I do not discuss it in this opinion.
. The state courts' findings regarding the other potential juror at issue, Charles McGuire, are discussed below. See infra at 322-24.
. See, e.g., Maj. Op. at 280 ("With regard to both Nichols and McGuire, the state courts failed to mention in their opinions the weaknesses in the State's explanations ...."); id, at 281 ("[TJhe hearing judge discussed neither the statistics nor the State's failure to explain them[,3 ... overlooking and ignoring a significant segment of Riley’s evidence ....''); id. at 282 (hearing judge made no "reference to, or analysis of, Riley’s evidence of pretext”); id. at 286 ("Here, the state courts failed to examine all of the evidence to determine whether the State's proffered race-neutral explanations were pretextual. Not only is there no indication on the record that the hearing judge engaged in the required analysis, but there is no indication that the Delaware Supreme Court did so.”).
. The only Third Circuit case cited in this connection by the majority (see Maj. Op. at 290) is Jones v. Ryan, 987 F.2d 960 (1993). Jones, however, was very different from the present case (see footnote nine, infra) and does not support the proposition that a finding of a state court is not entitled to the presumption of correctness if the state court did not explain why the court was not persuaded by particular items of evidence.
. Nor does the majority cite much other authority to support its position. See Maj. Op. at 287, 287, 289, 290. The majority (at 286) quotes dictum in a footnote in United States v. McMillon, 14 F.3d 948, 953 n. 4 (4th Cir.1994)(emphasis added), to the effect that at step three of Batson "the court then addresses and evaluates all evidence introduced by each side.” The reasons for not attaching too much weight to one word ("addresses”) in this statement are too obvious to require mention.
. The majority, however, incorrectly suggests (Maj. Op. at 289) that finding “the prosecutor to be credible” is different from finding that purposeful discrimination was not proved.
. Riley v. State, No. 200, 1988 at 4. See also App. 860 (McGuire's testimony at the eviden-tiary hearing).
. Although the notation “wants off" appears after the words "works at Lowe's,” it is not clear that Reed’s desire to be excused was related to his employment. The prosecutor's notes appear to contain notations of the employment of other jurors.
. Many decisions have held that Batson is not contravened simply because two jurors exhibit similar characteristics and one is excluded while the other is retained. See, e.g., Matthews v. Evatt, 105 F.3d 907, 918 (4th Cir.1997); United States v. Spriggs, 102 F.3d 1245, 1255 (D.C.Cir.1997); United States v. Stewart, 65 F.3d 918, 926 (11th Cir.1995); United States v. Alvarado, 951 F.2d 22, 25 (2d Cir.1991); United States v. Lance, 853 F.2d 1177, 1181 (5th Cir. 1988); United States v. McCoy, 848 F.2d 743, 745 (6th Cir.1988); United States v. Lewis, 837 F.2d 415, 417 n. 5 (9th Cir.1988).
. Batson, 476 U.S. at 98 n. 21, 106 S.Ct. 1712.
. This case is very different from Jones v. Ryan, 987 F.2d 960 (1993). There, exercising plenary review in the absence of any findings of fact by a state court, we held that Batson was violated where the prosecutor excluded a black juror who had a child approximately the same age as the defendant, while retaining a white juror who was similarly situated. Jones, 987 F.2d at 973. In the present case, we are limited to deciding whether the state court finding is fairly supported by the evidence.
. The state’s brief stated that it "emphatically denie[d] that the prosecutor exercised any of his challenges solely on the assumption that the juror's race, in the context of the facts of this case, indicated a verdict position adverse to the prosecution.” App. 896. The majority seizes on the word "solely” in this sentence as a tacit admission that race played a part in the decision to exercise peremplo-ries. In my view, it is wholly unreasonable to read that much into the word "solely.” The Supreme Court in Batson itself used this same word in the same context. See 476 U.S. at 89, 106 S.Ct. 1712 (emphasis added)("[T]he Equal Protection Clause forbids the prosecutor to challenge potential jurors solely on account of their race.").
. According to Riley's statistics, five of those struck by the state were white, and the race of two is not provided.
. The majority's statement that "the procedural posture of the case” provided "no opportunity” for Riley to offer an expert analysis of his statistics (Maj. Op. at 292) is difficult to understand. What stopped Riley from offering the evidence of a statistician as to the significance of the scant statistics that Riley provided?
. See "Forget Left-Wing. Say Hello to Left-Handed Politics,” New York Times, Jan. 23, 2000.
. Romano v. Oklahoma, 512 U.S. at 9, 114 S.Ct. 2004 (quoting Dugger v. Adams, 489 U.S. at 407, 109 S.Ct. 1211).
. Romano v. Oklahoma, 512 U.S. at 9, 114 S.Ct. 2004 (quoting Darden v. Wainwright, 477 U.S. at 184, n. 15, 106 S.Ct. 2464).