Walter Thomas was convicted in an Illinois state court in 1987 of murder, burglary, and aggravated arson, and was sentenced to death.. After exhausting his state remedies, see People v. Thomas, 137 Ill.2d 500, 148 Ill.Dec. 751, 561 N.E.2d 57 (1990); People v. Thomas, 164 Ill.2d 410, 207 Ill.Dec. 490, 647 N.E.2d 983 (1995), he sought federal habeas corpus; lost, 951 F.Supp. 1338 (N.D.Ill.1996); and appeals. His principal challenge is to the adequacy of his lawyer at the sentencing hearing, and we begin there.
The victim discovered Thomas, a 31-year-old employee of a cleaning service that she used, stealing perfume from her garage, which he had broken into, and in which she had stored a large quantity of perfume. Thomas (who admitted to his boss previous burglaries from the garage) stabbed the victim 12 times, killing her, then poured perfume over her body, the floor of the garage, and the car that was parked in the garage, and lit the perfume in an effort to obliterate the traces of the crime. The garage was in a condominium complex that might have gone up in flames had the car exploded. After the murder, Thomas resumed his normal routine, acting as if nothing had happened. He discarded the murder weapon and cleaned his blood-stained sweater, and, when arrested, at first denied his guilt, though eventually he confessed.
Because the murder was committed in the course of committing another felony, Thomas was eligible under Illinois law for the death penalty. 720 ILCS 5/9-1(a)(6), formerly Ill.Rev.Stat.1985 ch. 38, para. 9-1. At his sentencing hearing the jury heard evidence of both aggravating and mitigating factors. 720 ILCS 5/9—1(c), (g). The prosecution emphasized Thomas’s criminal history. Thomas had been convicted of six prior assaults on women, primarily involving attempts at knifepoint to rob and sexually assault them. In one of these cases he raped the woman three times and also stabbed her. (The rape victim testified at the sentencing hearing.) In another he carjacked the woman and at knifepoint tried to get her to drive to a secluded area; she deliberately collided with another car and, eluding his knife thrust, fled. (She also testified.) Thomas had struck one of his teachers with a table leg. He had committed this and three other of his offenses when he was a juvenile. When questioned in connection with the murder, he dismissed his previous offenses as a joke. He exhibited no remorse either for those offenses, which he attributed to his desire to prove his manhood, or for the murder.
In preparing the case in mitigation—a preparation begun a year before the sentencing hearing—Thomas’s lawyer interviewed Thomas, his mother, his former live-in girlfriend, and other friends and relatives. The lawyer asked them whether he had any history of psychological problems, and all of them *515including Thomas denied that he did. She hired a psychologist who prepared a narrative of Thomas’s life but did not (so far as anyone can recall) recommend that he undergo a psychiatric examination. The lawyer did not subpoena Thomas’s prison or school records. That was not done until, in the postconviction proceedings, new counsel for Thomas hired another mitigation expert. This time the records were searched, and they showed that Thomas was of low average intelligence (his I.Q. was in the range of 81 to 85), that he was emotionally disturbed, that he had “schizoid tendencies” and a “paranoid personality,” and that he had a learning disability—and might have organic brain damage—as a result of having had scarlet fever and two falls as a youth. Armed with these records, Thomas’s new counsel reinterviewed Thomas’s family and friends and learned that his mother had beaten him and his siblings with an extension cord when they were children, that his father had been alcoholic, and that his parents had fought with each other in front of the children.
The evidence presented in mitigation at the sentencing hearing consisted of testimony by Thomas’s friends and relatives that he was a good person and that the murder was an inexplicable deviation from his normal behavior. His former girlfriend testified that she had lived with him for four or five years and he had never been violent to her or her daughters, and that the Walter Thomas whom she knew (and was continuing to visit in prison) could not have committed the murder or other crimes that Thomas had committed. The other defense witnesses testified in similar vein. His mother testified that a police officer had told her that, if Thomas (who denied committing the murder) would confess he would not be sentenced to death, and this was the part of the mitigation testimony that the lawyer emphasized in her closing argument.
Thomas argues that his trial lawyer’s failure to subpoena his prison and school records, which would have revealed serious psychological problems, demonstrated that the lawyer failed to provide him with even minimally competent representation. In a death case, the defendant’s lawyer is required to conduct a reasonable investigation into the possibilities for proving mitigating factors, Strickland v. Washington, 466 U.S. 668, 691, 104 S.Ct. 2052, 2066-67, 80 L.Ed.2d 674 (1984); Kubat v. Thieret, 867 F.2d 351, 369 (7th Cir.1989); Cargill v. Turpin, 120 F.3d 1366, 1385 (11th Cir.1997); Porter v. Singletary, 14 F.3d 554, 557 (11th Cir.1994), factors that might persuade the jury not to impose the death penalty. A reasonable investigation is not, however, the investigation that the best criminal defense lawyer in the world, blessed not only with unlimited time and resources but also with the inestimable benefit of hindsight, would conduct. Kokoraleis v. Gilmore, 131 F.3d 692, 696 (7th Cir.1997); Stewart v. Gramley, 74 F.3d 132, 135 (7th Cir.1996); Waters v. Thomas, 46 F.3d 1506, 1514 (11th Cir.1995); White v. Singletary, 972 F.2d 1218, 1220 (11th Cir.1992).
It is reasonable for a lawyer to place a certain reliance on his client, so that if the client and his family and friends throw the lawyer off the scent, here by denying the existence of psychological problems that might have provided an alternative theory of mitigation, the lawyer cannot be faulted for failing to go down the path thus closed off. Strickland v. Washington, supra, 466 U.S. at 691, 699, 104 S.Ct. at 2066-67, 2070-71; Kokoraleis v. Gilmore, supra, 131 F.3d at 697; Stewart v. Gramley, supra, 74 F.3d at 135; LaRette v. Delo, 44 F.3d 681, 685 (8th Cir.1995); Whitmore v. Lockhart, 8 F.3d 614, 621 (8th Cir.1993). Thomas’s lawyer was led to believe that Thomas was a normal person from a warm close-knit family, a man who held a'steady job, was not violent, and whose crimes were therefore aberrant and did not mark him as irredeemably evil. His criminal history, however, drew the profile of a sexual predator, and sex criminals are often emotionally disturbed people. It would have been easy for Thomas’s lawyer to subpoena his prison and school records and to submit them for evaluation by a psychiatrist. Since the “nice guy” evidence was not powerful— the lawyer herself placed little emphasis on it in her closing argument, instead emphasizing the state’s alleged promise to Thomas’s mother that if he confessed he would be *516spared—the felt incentive to find an alternative line of defense should have been great.
Thomas’s brief might be read as asking us to lay down a per se rule that capital defendants’ lawyers must always, or at least in all cases in which sex may have been an element in the murder (may have, for there is no evidence, even after the second mitigation investigation, of a sexual element in the murder, unlike the case of Thomas’s previous crimes), subpoena all available institutional records. Such a rule would be a long step down the path that leads to -the courts’ laying down a rigid protocol that criminal defense lawyers must follow if they are to escape being later deemed ineffective. It would also be a new rule, for while many cases hold that defense counsel need not (depending on the circumstances) present any mitigation evidence at all, e.g., Preston v. Delo, 100 F.3d 596, 603 (8th Cir.1996); Marek v. Singletary, 62 F.3d 1295, 1300 (11th Cir.1995); Porter v. Singletary, supra, 14 F.3d at 557, we cannot find any case that holds that counsel must follow a particular line in investigating the possibilities for obtaining such evidence. The usual view is that an assessment of ineffective assistance of counsel depends, barring extreme cases, on the particular circumstances of the case. E.g., Payne v. United States, 78 F.3d 343, 348 (8th Cir.1996). We are not supposed to grant relief in a habeas corpus proceeding on the basis of a new rule. Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989); Gacy v. Page, 24 F.3d 887 (7th Cir.1994); Curtis v. Duval, 124 F.3d 1, 4-5 (1st Cir.1997); Miller v. Lockhart, 65 F.3d 676, 685 (8th Cir.1995). The state does not argue Teague, but this is excusable neglect, because Thomas denies wanting a new rule, though that is one construal of his submission. Anyway a court can invoke Teague even if the state has waived it. Caspari v. Bohlen, 510 U.S. 383, 389, 114 S.Ct. 948, 952-53, 127 L.Ed.2d 236 (1994); Eaglin v. Welborn, 57 F.3d 496, 499 (7th Cir.1995) (en banc); Curtis v. Duval, supra, 124 F.3d at 5.
So Thomas cannot prevail by asking for a rule requiring the subpoenaing of institutional records in all capital cases; but that leaves open the possibility that his lawyer failed to come up to minimum professional standards by not subpoenaing the records in the particular circumstances of this case. Even if she did fail to come up to the minimum—and a strong argument can be made that, given the absence of other good mitigating evidence, a reasonable lawyer in her position would indeed have subpoenaed the institutional records—Thomas has not succeeded in proving that he was prejudiced by the lawyer’s failure. The right to effective assistance of counsel presupposes that if counsel had been competent, it would have made, to some substantial likelihood, a difference in the outcome. Therefore “the. defendant must show that there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland v. Washington, supra, 466 U.S. at 694, 104 S.Ct. at 2068. See also Liegakos v. Cooke, 106 F.3d 1381, 1387 (7th Cir.1997); David v. United States, 134 F.3d 470, 477-78 (1st Cir.1998). Counsel cannot confidently be deemed ineffectual if it is unlikely that a better performance would actually have helped his client.
This is such a ease. To begin with, the two possible strategies, the “nice guy” strategy that failed and the “psycho” strategy which Thomas’s current counsel tells us any competent lawyer would have preferred, could not have been combined; it was either/or. At argument Thomas’s current lawyer, equating “schizoid” to having a split personality—when what the term actually means is having difficulty expressing emotions and forming relationships, American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders 301.20, p. 638 (4th ed.1994)—suggested that the two strategies could have been combined in a theory that Thomas was nice most of the time but went crazy under particular stimuli. This argument is neither made in Thomas’s brief nor supported by the psychiatric evidence obtained after the sentencing hearing.
All that the extensive research of the post-conviction counsel and mitigation expert turned up was that Thomas was of dull normal intelligence, that his mother may have beaten him when he was growing up and fought with his alcoholic father, that he is *517suspicious (“mildly paranoid”), that he has difficulty relating to other people (“schizoid tendencies”), and that- he may have some brain damage, though not enough to push him below the average level of intelligence. This is unimpressive evidence in comparison with the evidence in cases in which the failure to obtain psychological evidence for presentation at the sentencing hearing was the basis for ordering a new hearing. Emerson v. Gramley, 91 F.3d 898, 907 (7th Cir.1996); Brewer v. Aiken, 935 F.2d 850, 858-59 (7th Cir.1991); Bloom v. Calderon, 132 F.3d 1267, 1273-77 (9th Cir.1997); Clabourne v. Lewis, 64 F.3d 1373, 1385-86 (9th Cir.1995); Antwine v. Delo, 54 F.3d 1357, 1365-68 (8th Cir.1995); Baxter v. Thomas, 45 F.3d 1501, 1512 and n. 30 (11th Cir.1995); Hendricks v. Calderon, 864 F.Supp. 929, 933-35 (N.D.Cal. 1994), aff'd, 70 F.3d 1032, 1044-45 (9th Cir. 1995). It is actually weaker than evidence in eases in which courts have upheld the denial of such relief. Eddmonds v. Peters, 93 F.3d 1307, 1319-22 (7th Cir.1996); Green v. Johnson, 116 F.3d 1115, 1123 (5th Cir.1997). And, as in those cases, mixed in with the psychological evidence in mitigation is evidence that, had a mitigation defense based on psychological evidence been' attempted, would have been used by the prosecutor as further evidence in aggravation. Eddmonds v. Peters, supra, 93 F.3d 1307, 1321; Green v. Johnson, supra, 116 F.3d at 1123. This included evidence that Thomas was belligerent and hostile, and had been so in the classroom; that he had poor impulse control; that he is “quite intelligent” and his speech is “fluent and grammatical”; that when irritated and frustrated he could “act out in explosive ways”; that one judge had said that “it would be dangerous to allow this person to walk the street”; that a clinical evaluation had said that “in his present- condition, [Thomas] must be regarded as quite dangerous, specifically from a sexual point of view.” This evidence in aggravation, secreted as it were in the interstices of the psychological defense, further demonstrates the impossibility of combining that defense with a “nice guy” defense. One of the documents unearthed by the postconviction investigation, incidentally, is a social worker’s report confirming “the relative stability of the home situation.” Another such report states, “this appears to be a good family unit.”
Even viewed in isolation, the mitigation evidence was unimpressive. The principal document is the report of a psychiatrist who examined Thomas on death row. The report attributes the murder to “an unfortunate synchronicity of events,” Thomas having been “in the process of leaving the victim’s garage when she screamed and tried to restrain him from behind. Mr. Thomas was conditioned to act reflexively in such a circumstance.” The only evidence that the murder victim screamed and tried to “restrain” Thomas, or that he was already in the process of leaving the garage when she surprised him, came from Thomas’s -confession. The psychiatrist’s uncritical dependence on the murderer’s uncorroborated self-exculpatory statement, along with the psychiatrist’s effort to blame Thomas’s behavior on an “oedipal situation,” would have exposed the psychiatrist to lacerating cross-examination had he testified at the sentencing hearing.
Men who repeatedly assault women, finally murdering one, are unlikely to have a normal psychological profile. Cf. Kokoraleis v. Gilmore, supra, 131 F.3d at 697. They are unlikely to be the highly intelligent products of a warm and loving home. It is difficult to believe that a jury otherwise minded to sentence Thomas to death would have been moved by the evidence, evaluated as a whole, which his current lawyers not unreasonably fault his trial lawyer for having failed to uncover. Even if she had' dug up all this evidence, she might reasonably have decided that on balance it was better to keep it away from the jury and instead stake her all on convincing the jury (or at least one member of the jury, all that it takes in Illinois to prevent the imposition of the death penalty, People v. Ramey, 152 Ill.2d 41, 178 Ill.Dec. 19, 36, 604 N.E.2d 275, 292 (1992)) that Thomas was a generally decent person who yielded at times to an irresistible impulse to violence; that he had some redeeming qualities; and that it would be unfair to sentence a man to death when the police had extracted a confession from him by promising his mother that if he confessed he would not be executed. The fact that in the end a compe*518tent lawyer might have decided to do just what Thomas’s lawyer did, fearing that psychological evidence would simply have made her client look weird and dangerous, suggests that her failure to investigate was harmless. But more plainly it shows that Thomas has failed to establish a substantial likelihood that had his trial lawyer conducted the same thorough investigation that his postconviction counsel conducted Thomas would have been spared.
This is not to say that his trial lawyer in fact chose the better defense. By trying to portray Thomas as a “nice guy,” she opened herself to the withering scorn of the prosecutor, who in rebuttal to the closing argument of Thomas’s lawyer remarked, “He never rapes his family. He never attacks his friends. And he is a nice guy. He is nice enough to leave his neighborhood and go to other people’s neighborhoods to terrorize them.” He added, “There is no evidence that Walter came from a bad home environment. In fact, the evidence is the opposite. There is no evidence that he had any psychological problems.” But the prosecutor’s main emphasis was on Walter’s being cruel and pitiless, devoid of rehabilitative potential, remorseless, sadistic; and he could thus have used the psychological evidence that Thomas’s lawyer did not present to as good purpose as he used the “nice guy” evidence. (Toward the end of his rebuttal, he said, “Walter Thomas is every woman’s nightmare”—a theme to which the psychological evidence would have added resonance.) Indeed, it is predictable that if Thomas’s lawyer had taken the path that his current lawyer says she should have taken, we would now be faced with the argument that the lawyer made the wrong choice—that the psychological evidence was too thin to be worth putting before the jury—and that therefore a new sentencing hearing should be ordered.
His best case is Emerson v. Gramley, supra, where we held that the failure to investigate the defendant’s history, in a case in which no mitigation evidence at all was introduced at the sentencing hearing, entitled him to a new hearing. Emerson’s history, however, was garish and shocking, including his being shot and seriously wounded at the age of eight and the death of his child possibly strangled by her mother; we referred to “a life that one juror in twelve might find so bleak, so deprived, so harrowing, so full of horrors ... as to reduce Emerson’s moral responsibility” for the murder. 91 F.3d at 907. There is nothing comparable here. Thomas’s history is that of a soeiopath-a compulsive, and eventually a homicidal, criminal. Moreover, because in Emerson there was no alternative theory of mitigation, it was not possible there as it is here to conclude that while the alternative approach that was taken may not have been very good, it may have been as good as the psychohistory approach that was not taken.
His only other argument for reversal is that the prosecutor, in his opening argument at the sentencing hearing, appealed to racial prejudice and by doing so fatally tainted the hearing and the resulting sentence. The prosecutor remarked in passing that a detective would testify that one or both of Thomas’s prior sexual offenses involved “young white women and both of those cases involved knives.” Thomas’s counsel did not object to the remark, and the Supreme Court of Illinois therefore considered the issue waived; and though it added that there was no plain error, People v. Thomas, supra, 148 Ill.Dec. at 769, 561 N.E.2d at 75, review for plain error does not cure a procedural default. Neal v. Gramley, 99 F.3d 841, 843-44 (7th Cir.1996).
This may not seem, and in other cases might not be, an entirely satisfactory resolution, since if the prosecutor was appealing to racial prejudice an objection would only have magnified the appeal. But in context it appears that the only purpose of the remark was to establish that Thomas was following a pattern in his assaultive behavior; and his lawyer testified later that she had not objected because she had not considered it an appeal to racial prejudice. The prosecution did not elsewhere suggest a possible racial motive for Thomas’s crimes; the remark was too fleeting and isolated to have planted such a suggestion in the jury’s mind; and the detective never did testify about the race of any of Thomas’s victims. The remark did not prevent Thomas from having a fair trial, *519and so it did not deny Mm due process of law. Donnelly v. DeChristoforo, 416 U.S. 637, 642-43, 94 S.Ct. 1868, 1871-72, 40 L.Ed.2d 431 (1974).
No error of constitutional magnitude was committed in the sentencing hearing. The judgment of the district court is therefore
Affirmed.