dissenting, with whom FAGG, BOWMAN, MAGILL and BEAM, Circuit Judges, join.
I respectfully dissent.
The court today ignores the Supreme Court’s instruction that in reviewing the performance of Chambers’ lawyer, Hager, “every effort be made to eliminate the distorting effects of hindsight, to reconstruct the circumstances of counsel’s challenged conduct, and to evaluate the conduct from counsel’s perspective at the time.” Strickland v. Washington, 466 U.S. 668, 689, 104 S.Ct. 2052, 2065, 80 L.Ed.2d 674 (1984). Like the district court and the state trial court that reviewed this claim, I conclude that Hager properly made a strategic judgment that Jones should not be called to testify and need not be interviewed because Jones’ testimony would have been more harmful than helpful to Chambers. Furthermore, after studying Jones’ testimony during the first trial, I believe that neither element of the Strickland test has been satisfied, and I would affirm the district court judgment denying the writ.
Although the decision of the Missouri Supreme Court, reviewing the direct appeal from Chambers’ first trial, clearly states that Jones’ testimony would have supported a self-defense instruction,1 Jones was not called to testify at the second trial. *834However, the state trial court, in considering the collateral attack under Missouri’s Rule 27.26, concluded that Jones’ testimony, on balance, was more damaging than helpful to Chambers. After observing that Hager “could cho[o]se between a weak self-defense theory that carried with it a strengthening of the State’s case,” or try the case as he did, the state court concluded that the decision to not call Jones was a reasonable one.2 The Missouri Court of Appeals affirmed the conviction, Chambers v. State, 745 S.W.2d 718 (Mo.Ct.App.1987), and Chambers’ application for transfer to the Missouri Supreme Court was denied.
The district court, in this habeas corpus action, concluded that Jones’ testimony would have supported the State’s theory of the case. It also concluded that, because Hager’s failure to interview Jones resulted from a strategic decision, his performance was not deficient. Because it decided that Hager rendered effective assistance, the district court did not reach the question of prejudice.3
I.
The effectiveness component of the Strickland test asks whether the defendant received “reasonably effective assistance.” 466 U.S. at 687, 104 S.Ct. at 2064. Moreover, Strickland teaches us that judicial scrutiny of counsel’s performance must *835be “highly deferential,” id. at 689, 104 S.Ct. at 2065, and should eliminate the “distorting effects of hindsight,” id.
In performing the first part of the Strickland analysis, courts distinguish between actions that result from inadequate pretrial preparation and those that are the product of trial strategy decisions. See Burger v. Kemp, 483 U.S. 776, 788-95, 107 S.Ct. 3114, 3122-26, 97 L.Ed.2d 638 (1987); Kimmelman v. Morrison, 477 U.S. 365, 384-87, 106 S.Ct. 2574, 2587-89, 91 L.Ed.2d 305 (1986); Darden v. Wainwright, 477 U.S. 168, 184-87, 106 S.Ct. 2464, 2473-74, 91 L.Ed.2d 144 (1986); Strickland, 466 U.S. at 687-91, 104 S.Ct. at 2064-66; United States v. Gray, 878 F.2d 702, 711 (3d Cir.1989); Laws v. Armontrout, 863 F.2d 1377, 1382-86 (8th Cir.1988) (en banc), cert. denied, — U.S. -, 109 S.Ct. 1944, 104 L.Ed.2d 415, reh’g denied, — U.S. -, 109 S.Ct. 3179, 104 L.Ed.2d 1041 (1989). “[C]ounsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary. In any ineffectiveness case, a particular decision not to investigate must be directly assessed for reasonableness in all the circumstances, applying a heavy measure of deference to counsel’s judgments.” Strickland, 466 U.S. at 691, 104 S.Ct. at 2066. As the Third Circuit recently stated, “Ineffectiveness is generally clear in the context of complete failure to investigate because counsel can hardly be said to have made a strategic choice against pursuing a certain line of investigation when s/he has not yet obtained the facts on which such a decision could be made.” Gray, 878 F.2d at 711.
In contrast to the close scrutiny which courts give to an attorney’s preparatory activities, greater deference is given to an attorney’s informed strategic choices. Indeed, it has been clear since Strickland that “strategic choices made after thorough investigation of law and facts relevant to plausible options are virtually unchallengeable.” Strickland, 466 U.S. at 690, 104 S.Ct. at 2066. Also, in reviewing the performance of counsel, “courts must resist the temptation to second-guess a lawyer’s trial strategy.” Laws, 863 F.2d at 1393 (quoting Blackmon v. White, 825 F.2d 1263, 1265 (8th Cir.1987)), because even a losing strategy “may have been reasonable in the face of an unfavorable case.” Id. at 1394 (emphasis removed) (quoting Blackmon, 825 F.2d at 1265).
Chambers attempts to formulate arguments based upon Hager’s allegedly inadequate investigation. However, as the Seventh Circuit has observed:
When the allegation of the ineffectiveness of counsel centers on a supposed failure to investigate, we cannot see how, especially in the context of a habeas proceeding that collaterally attacks the state court conviction, the petitioner’s obligation can be met without a comprehensive showing as to what the investigation would have produced. The focus of the inquiry must be on what information would have been obtained from such an investigation and whether such information, assuming its admissibility in court, would have produced a different result.
United States ex rel. Cross v. DeRobertis, 811 F.2d 1008, 1016 (7th Cir.1987).
Hager, trial counsel in Chamber’s second trial, read the transcript of Jones’ testimony in the first trial and concluded that the testimony was more damaging than helpful. While the court today states that Hager’s decision to not interview Jones reflects inadequate preparation for trial rather than a conscious trial strategy, that conclusion ignores the fact that Hager carefully studied and analyzed Jones’ testimony from the first trial and knew that any departures from it would open up a strong credibility attack. His decision to not interview Jones does not demonstrate inadequate preparation for trial, but rather a careful analysis of known testimony. The only reason for Hager to have interviewed Jones would have been to see if Jones would change his story. Hager decided not to interview Jones because even any substantial, beneficial changes in his story would have created an excessive danger of devastating impeachment. Chambers v. State, 745 S.W.2d at 720. Furthermore, as is clear from Jones’ testimony at the 27.26 hearing, the only new information that an *836interview would have produced was Jones’ rather lame explanation that when he said that Chambers had held the gun “back here” and against his leg, he did not mean that it was hidden. See ante at 827 n. 2, 829-830 n. 5. Such an embellishment is precisely the sort of a change that could have led to damaging cross-examination of Jones, particularly because Jones had explicitly stated at the first trial that Chambers was not displaying the weapon. (Chambers I Tr. at 740-41).
My study of the record convinces me that both the state trial court and the district court properly assessed Jones’ testimony. At the first trial, Jones testified that Chambers arrived in a car which was turned to face the exit from a bar’s parking lot. (Chambers I Tr. at 748). As Chambers entered the building, the car was left running and was still occupied by the driver. (Id. at 748-49). Jones testified that he saw Chambers come out the door, get about half the distance of an automobile or truck, and turn half-way toward the door. (Id. at 738). Oestricker followed Chambers out of the door and struck Chambers hard enough to knock him down. (Id.). Chambers then got up, took a step forward, and shot Oes-tricker. (Id.).
However, on cross-examination, Jones revealed that when Chambers walked out the door and turned around half-way, he already had a pistol in his hands, (Id. at 740), with the gun against his leg and positioned behind him, (Id. at 741). Oestricker was just emerging through the door when Chambers stopped, turned around with the gun in hand, and waited for Oestricker to come out. (Id. at 741-42). Jones also testified that he had not seen Oestricker attempt to strike Chambers before Chambers initially took the gun out. (Id. at 742). According to Jones, after Chambers shot Oestricker, Chambers said either “[tjake that, tough guy,” or “[t]ake that.” (Id.). After being shot, Oestricker made a grunting sound and backed up three or four steps. (Id.). Chambers then walked toward him and slapped him in the head with the pistol “[o]ver and over and over again.” (Id. at 742-43). Furthermore, Oestricker was standing about six feet away from Chambers at the time of the shot and was not moving toward Chambers. (Id. at 747). Jones also said that, after shooting Oes-tricker, Chambers walked into the building and asked “if anybody else wanted any of this.” (Id. at 746). As he left the building, Chambers said to Oestricker, “Lay there and die.” (Id. at 747).
Based upon this testimony, I cannot conclude that Hager acted in an unreasonably ineffective manner by deciding to not call Jones. Even if Jones’ testimony supported a self-defense instruction, as the Supreme Court of Missouri held, the testimony also indicated that Chambers, with a pistol concealed against his leg, both waited for Oes-tricker to come out of the door and, after being struck, fired the fatal shot while Oestricker was six feet away and was not moving toward him. After threatening the crowd in the bar, Chambers ran to the car which had waited for him, with its motor running, during the entire incident.
While the question of whether there was enough evidence to support a self-defense instruction involves considering the evidence in the light most favorable to Chambers, a professional evaluation of the testimony’s trial impact involves considering it in the light that the jury would consider it. This is a far broader analysis, and I cannot conclude that Hager was unreasonably ineffective in his assessment of the impact of Jones’ testimony on the jury. The Supreme Court has refused to find ineffective assistance where a lawyer did not introduce helpful evidence which, in turn, could have led to the introduction of other more harmful testimony. See Burger, 483 U.S. at 788-95, 107 S.Ct. at 3122; Darden, 477 U.S. at 184-87, 106 S.Ct. at 2473-74. The testimony by Jones presented just such a dilemma for Hager, and we should follow the teaching of the Supreme Court by refusing to hold that there was ineffective assistance in this respect.
The court today has only one answer to the damaging aspects of Jones’ testimony: it states that “any damaging testimony that Jones gave at the first trial was cumulative.” Ante at 829-830. An appellate court often categorizes testimony as cumu*837lative in deciding evidence questions, but this is no answer at all in the context of evaluating Hager’s decision. While the court establishes conclusively that testimony by James Fowler, Fred Ieppert, Dennis Simmons and several other witnesses overlapped with testimony by Jones, ante at 829-830 n. 5, it fails to establish that Jones’ testimony would have had only a negligible impact on the jury, thus underscoring the wisdom of Hager’s decision to not call Jones to testify. Jones was the only witness who saw the whole incident outside the bar. Hager, when evaluating the probable impact of Jones’ testimony on the jury, could have reasonably concluded tht Jones’ testimony would drive the damaging points home to the jury. That strategic decision is one that must be viewed from the testimony’s impact on the jury, because we are here deciding how the jury’s verdict would have been affected. An appellate court engages in a far different exercise when it concludes that evidence is cumulative in deciding whether evidence either should have been admitted or excluded, or whether error was harmless or prejudicial. It was the jury impact, however, that Hager analyzed.
Moreover, the court’s assertion that all of Jones’ harmful testimony was already before the jury in the second trial4 is patently incorrect. Had Jones testified at the second trial, he would have introduced an important piece of information that would have helped establish an element of capital murder, and he would have hurt Chambers by directly contradicting the testimony of another witness.
At the first trial, Jones testified Chambers arrived in a car that made a U-turn to face the street, that someone stayed in the car while Chambers went inside, that Chambers was inside the bar for only two or three minutes, and that the car's engine was left running during the entire episode. (Chambers I Tr. at 748-49). At the second trial, the jury was instructed that it could convict Chambers of capital murder only if it found that he “considered taking the life of Jerry Lee Oestricker and reflected upon this matter coolly and fully before doing so.” (Chambers II Tr. at 681). The most damning evidence of cool planning was Jones’ testimony that the car was left running, because that testimony undercuts Chambers’ theory that he innocently went into the bar to drink with Oes-tricker, but it squarely supports the State’s theory that Chambers planned the shooting even before entering the bar. Because Jones was not called at the second trial, the jury at that trial was unaware of this damaging information.
Jones also would have hurt Chambers by directly contradicting the testimony of Fred Ieppert, the only witness besides Jones who testified concerning the events that immediately preceded the shooting. At the first trial, both Jones and Ieppert testified that Chambers shot Oestricker while Oestricker was standing still, approximately six feet away from Chambers. (Chambers I Tr. at 464, 746-47, 750-51). After Hager confronted Ieppert with statements that Ieppert had made at a pre-trial hearing, however, Ieppert testified at the second trial that Oestricker was moving toward Chambers when the shot was fired. (Chambers II Tr. at 446, 451). Iep-pert’s testimony at the second trial helps support a self-defense theory. Had Jones testified at the second trial, he would have directly contradicted Ieppert. Thus, it is difficult to understand how the court can call this part of Jones’ testimony “cumulative.”
It is also important to consider the fact that, before the second trial, Chambers signed a statement in which he agreed with the decision to not call Jones. The Supreme Court stated in Strickland that “[t]he reasonableness of counsel’s actions may be determined or substantially influenced by the defendant’s own statements or actions,” 466 U.S. at 691, 104 S.Ct. at 2066, and that those statements are critical to a proper assessment of litigation decisions, id. When Chambers’ signed state*838ment is considered in combination with the content of Jones’ testimony at the first trial, I am convinced that the decision to not call Jones was reasonable under Strickland.
The court, purporting to answer arguments made by the State, engages in a substantial discussion of Jones’ credibility.5 The court’s affirmation of Jones’ credibility hardly supports its position today, however, because Jones’ believable testimony simply hammered home the State’s case.
The court views Hager’s actions in a myopic sense when it concludes that Jones was not called as a witness because he was not interviewed. Hager had the full benefit of the trial transcript of Jones’ earlier testimony and made a careful strategic determination that the testimony was more harmful than helpful and that any changes in the testimony would open Jones up for vigorous cross-examination that would hurt Chambers. We should not second-guess that decision or place it in a light contrary to that required by Strickland.
In holding that Hager’s assistance was unreasonably ineffective, the court today reaches a result contrary to that reached in earlier decisions in which we recognized that Strickland is not violated when a counsel, in the exercise of professional judgment, decides not to produce mitigating evidence that could reasonably be considered more damaging than helpful. In Smith v. Armontrout, 888 F.2d 530 (8th Cir.1989), we held that certain medical records would have hurt the defendant at least as much as they helped, and we refused to flyspeck the decision of a lawyer long after the fact. Id. at 534-35. Similarly, in Swindler v. Lockhart, 885 F.2d 1342 (8th Cir.1989), cert. denied, — U.S. -, 110 S.Ct. 1938, 109 L.Ed.2d 301 (1990), we held that it was not unreasonable for counsel to refrain from offering into evidence medical reports, concerning the defendant’s mental condition, that he felt to be more damaging than helpful. Id. at 1352-53. See also Laws, 863 F.2d at 1387-91.
II.
Even if Hager should have called Jones, the Strickland test is not satisfied unless Chambers can also demonstrate “that the deficient performance prejudiced the defense.” Strickland, 466 U.S. at 687, 104 S.Ct. at 2064. In order to prove prejudice, “[t]he 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.” Id. at 694, 104 S.Ct. at 2068. A reasonable probability is defined as one which is “sufficient to undermine confidence in the outcome.” Id. After a thorough examination of the record, I conclude that there is not a reasonable probability that the introduction of Jones’ testimony would have changed the outcome of the second trial.6
Accordingly, I would affirm the judgment of the district court denying the writ.
. The Missouri Supreme Court, in reversing the Chambers I judgment, observed that there was conflicting evidence as to the incident and stat*834ed that ”[i]n examining the record for evidence of self-defense, we must consider the evidence in [the] light most favorable to appellant Chambers.” State v. Chambers, 671 S.W.2d 781, 783 (Mo.1984) (en banc). After reviewing the evidence in that manner, the court concluded that "[w]hile the evidence of self-defense is not so unequivocal as to mandate a directed verdict of acquittal, the evidence is sufficient to justify submission of self-defense to the jury.” Id. at 784.
. The detailed reasoning of the state trial judge is as follows:
During this proceeding, Donald Hager testified that the decision not to call Jones was [a] deliberate one, based upon strategic concerns. That, having the benefit of Jones' testimony on cross-examination adduced at the first trial, in his professional opinion, the disadvantages of Jones' testimony outweighed the advantages. The State’s cross-examination ... was highly damaging in that it supported the State's theory of the case under a capital murder submission. Mr. Hager knew that although Jones' testimony would have supported a self-defense instruction, it corroborated the State’s main witness — Fred Ieppert— and conflicted with his defense strategy. His strategy at trial was to: 1) attack the credibility of the State's witnesses; 2) suggest that Oestricker had a pair of pliers in his hands; and 3) attempt to negate the element of Chambers reflecting "cooly” upon ... taking the life of Oestricker. The fact that Jones was in a position to observe the condition of the getaway car with running engine and the distance between the victim and petitioner at the time of the fatal shot would have made this trial strategy almost impossible from a practical standpoint.
******
Without Jones’ testimony a jury might believe, as at least one [Missouri] Supreme Court Judge did, that the whole matter was just "an ordinary barroom altercation” thus negating the cool reflection that might not exist under those circumstances.
******
In light of the foregoing, the Court finds that petitioner’s trial counsel’s decision not to call Jim Jones was a reasonable one based on his professional judgment in consideration of the evidence and the circumstances in the first trial.
Chambers v. Missouri, No. CV186-4580-CC-J3, slip op. at 12-13 (23d Cir.Ct. March 11, 1987). The court also rejected Chambers' claim that he had not read the signed statement in which he agreed with the decision to not call Jones. Id. at 14 n. 2.
. The district court’s reasoning is of interest:
The Court finds reasonable counsel’s conclusion that Jones’ testimony would have tended to support the state’s theory of the case and thus his decision not to call Jones as a witness. This is especially true in view of petitioner’s written and signed pretrial statement that he agreed with counsel’s decision in this regard. As the United States Supreme Court noted, ”[t]he reasonableness of counsel’s actions may be determined or substantially influenced by the petitioner’s own statements or actions.” Strickland, supra, 466 U.S. at 691 [104 S.Ct. at 2066]. Furthermore, counsel reasonably assessed the affect [sic] of Jones' earlier testimony on both the state’s theory of the case and Jones' credibility as a witness.
Chambers v. Armontrout, No. 88-0567C(3), slip op. at 12 (E.D.Mo. July 19, 1988).
. Although the court today relies upon testimony from both trials, it does not explain why testimony at the first trial, by witnesses other than Jones, is relevant to the question of whether Jones should have been called at the second trial.
. The statement that the Missouri Supreme Court found Jones' testimony credible is simply not based upon its opinion. See ante at 827 n. 1. The fact that the State did not attempt to impeach Jones’ credibility is completely understandable in light of the support that Jones gave to the State's version of the case.
. We need not discuss in detail the statement and assumption made by the court today that the two trials were essentially the same except for the fact that Jones was not called at the second trial. There were significant differences which point to the strength of the defense waged by Hager. For example, under cross-examination at the second trial, but not at the first, Kenneth Vaughn stated that Oestricker was drunk and "wanting to fight like crazy— wanting to fight anybody.” (Chambers II Tr. at 385). Hager also significantly impeached several of the other witnesses. {See, e.g., Chambers II Tr. at 581-615) (Testimony of James Fowler).