After a two-day trial, Wanda Figueroa was convicted by a jury of embezzling almost $65,000 from her employer, the First National Bank of Chicago. At trial, her appointed attorney did not try to prevent — indeed, he invited — testimony revealing that she had failed a lie detector test administered by federal agents investigating the disappearance of monies from the bank. On appeal, Figueroa argues (as she may, see United States v. Castillo, 965 F.2d 238, 243 (7th Cir.1992); United States v. Taglia, 922 F.2d 413, 417 (7th Cir.1991)) that the pursuit of this tactic constituted ineffective assistance of counsel mandating a reversal of her conviction. Our task is “to reconstruct the circumstances of counsel’s challenged conduct, and to evaluate the conduct from counsel’s perspective at the trial” while indulging the “strong presumption that ... the challenged action might be considered sound trial strategy.” Strickland v. Washington, 466 U.S. 668, 689, 104 S.Ct. 2052, 2065, 80 L.Ed.2d 674 (1984) (internal quotation omitted). After carefully reviewing the record (to which, of course, our inquiry is confined), we conclude that the decisions made by Figueroa’s counsel were not, under the circumstances of this case as they unfolded, outside “the wide range of reasonable professional assistance,” id., which we may not second-guess.
Wanda Figueroa was a teller at the Evans-ton, Illinois branch of First National. On July 12 and September 12, 1991, large sums of cash disappeared from the bank’s main vault. Evidence linking Figueroa to the *707missing funds initially was not strong. Yet certain aspects of her behavior did arouse suspicions. For example, Figueroa left work early on July 12, claiming that she needed to go to the hospital because her baby was sick; and a large check which Figueroa supposedly (according to bank records) received and cashed on September 12, while assisting a trainee at the drive-up teller position, could not be (and never was) found. (Four days after the latter incident, Figueroa was fired.) However, no one ever saw Figueroa take any money, investigators could not trace the missing funds to her, and other employees also had access to the vault on the days in question.
The FBI agent investigating the July incident, Special Agent Kevin Deery, interviewed Figueroa in September. She described her activities on the day in question and denied wrongdoing. Not long after this encounter, Agent Deery tried to recontact Figueroa to discuss the later episode at the bank but was unsuccessful, apparently because she had recently left town on her way to California and Mexico. By January, 1992, however, Figueroa had returned to Chicago and learned from her parents that the agent was looking for her. Figueroa called the agent and, at his suggestion, agreed to submit to a polygraph examination.
As arranged, Figueroa appeared at the FBI office on January 17. After she was advised of her Miranda rights and her right not to take the polygraph test and after she signed separate consents to be interviewed and to be administered the exam, Figueroa was interviewed and tested by Special Agent Michael Hanna, a polygraph specialist. In the course of the examination, Agent Hanna queried Figueroa about the July and September shortfalls at the bank and concluded that her disavowals of involvement were untruthful. According to Agent Hanna, when he then confronted Figueroa with her poor performance on the test and urged her to come clean, she confessed in detail to taking money from the bank on each occasion. Agent Hanna then called Agent Deery into the room, and, according to both men, Figueroa repeated her confession. The only recordation of her confessions was a contemporaneous written report prepared by Agent Hanna and not signed by Figueroa.1 In September, 1992, Figueroa was indicted on two counts of embezzling money entrusted to a federally insured bank. See 18 U.S.C. § 656.
Enter Sheldon Nagelberg, appointed by the district-court on October 8 to represent Figueroa. Faced with a trial date originally set for November 23 and a government case relying mainly on Agents Deery’s and Hanna’s accounts of Figueroa’s confessions, Na-gelberg needed to develop relatively rapidly a strategy for defusing the evidentiary impact of the confessions at trial.2 His options were rather limited. He could deny that they ever took place, effectively accusing two FBI agents of deliberate falsification; he could claim they were garbled and inconclusive, despite the clear acknowledgment of guilt and detailed descriptions of the incidents recorded and sure to be recounted at trial by the agents; or, he could argue that they were involuntary and thus unreliable. The alternatives dubious, Nagelberg chose to pursue a voluntariness/coercion theory.
Nagelberg apparently planned to rely on Figueroa’s testimony at trial and a perceived *708inconsistency between Figueroa’s confession — in which she may have implied that the claimed hospitalization of her son on July 12 was a ruse and untrue, see infra note 2 — and hospital records — which show her child was admitted, treated and released to her on that day — to support the contention that she was coerced by the FBI into saying untrue things.3 Hoping that an independent assessment of the circumstances surrounding the polygraph exam and subsequent confession would buttress this theory, on October 19 Nagelberg requested that the court appoint a polygraph expert to examine the procedures employed at the January 17 session. To that end, Nagelberg also made a motion to preserve the agents’ handwritten notes from that day.
After his request for polygraph expert assistance was denied, Nagelberg evidently felt that the defense would be best served if no references to the failed exam were made at trial and coercion were established through his client’s version of the confession, the purported conflict between the confession and Figueroa’s actions as evidenced by the hospital records,4 and cross-examination of the agents. Thus, he made a motion requesting that all mention of the polygraph examination and its results be barred at trial. The district court granted the motion with respect to the government’s case-in-chief but, significantly, reserved ruling on admissibility of the polygraph evidence in the government’s rebuttal case, citing United States v. Kampiles, 609 F.2d 1233 (7th Cir.), cert. denied, 446 U.S. 954, 100 S.Ct. 2923, 64 L.Ed.2d 812 (1979). In Kampiles, this court held that when a confession is preceded by a failed polygraph examination and the confession is challenged as coerced, the polygraph evidence may be admitted to show the circumstances surrounding the confession and rebut the charge of coercion. See id at 1244. (More precisely, evidence that the defendant was informed of the failure before confessing is admissible.)
This ruling, made four weeks before the commencement of trial, set the stage for Nagelberg’s next (and, for present purposes, critical) adjustment in trial strategy. Presented with the inevitability that Figueroa’s polygraph failure would be brought to the jury’s attention,5 Nagelberg decided to incor*709porate the polygraph evidence into his casein-chief and somehow try to use it to bolster the claim that Figueroa’s confessions were unreliable. The government, actually fearing that he would do so,6 filed a motion in limine on the eve of trial requesting that Figueroa be barred from using the polygraph evidence in his case-in-chief. When the district court intimated that if Nagelberg planned to refer to Figueroa’s awareness of having failed the polygraph the government could discuss the same in its main case, Nagelberg stuck with his revised plan to use the polygraph evidence offensively and orally withdrew his original motion to the extent it barred government reference to Figueroa’s knowledge that she failed the exam.7 As a result, the government in its opening statement disclosed that Figueroa was asked to take a polygraph exam and confessed after the examiner said “I believe you are lying to me.” In his opening, after suggesting that the FBI agents had preconceived beliefs as to Figueroa’s guilt, Nagelberg told the jury *710that Figueroa confessed only after she was told that she failed a polygraph test, to which she had originally agreed to submit in the expectation of clearing her name. He went on to indicate that Figueroa would testify as to how she felt further pressure to confess by unfair and intimidating verbal tactics used by the interviewing agent, including a suggestion that she could avoid separation from her child by admitting guilt.
Looking at Nagelberg’s gambit and the circumstances under which it was made, we can only conclude that it was a valid tactical maneuver. Mounting an assault on the government’s prime piece of evidence meant the jury was eventually going to find out, one way or another, that before she confessed, Figueroa was told that she failed a polygraph examination. The district court made this much quite clear by its citation to Kampiles and comments in open court. It undoubtedly was reasonable lawyering for Nagelberg to attempt to deflate the impact of this information by putting his spin on it as early as possible in the course of the trial. Expecting his client to testify about the coercive nature of the confession-inducing interview, Nagel-berg plausibly sought to portray her as initially weakened and made susceptible to the government’s pressuring practices by her belief that the government and its imposing machine were already disposed against her. Nagelberg, previously denied expert assistance to challenge polygraph reliability, cab-ined the collateral prejudice caused by the unavoidable polygraph references in the only way he could. He made sure that witnesses would only testify that Figueroa was told that she failed and not directly that she did in fact fail, and he obtained a special instruction specifically charging the jury not to consider the reported polygraph failure in determining the truthfulness of Figueroa’s confessions.8 Admittedly, it is especially hard to put this type of genie back into the bottle, but from Nagelberg’s perspective, the genie was going to make her way out anyway. By trying to have some say in the terms of her departure, Nagelberg did not act unreasonably.
That events at trial did not unfold as favorably as hoped does not retroactively transmute legitimate strategy into ineffective assistance. After the agents testified about the events of January 17, including how Figueroa was told that she failed the polygraph test, Nagelberg cross-examined them vigorously but without spectacular success. Although he managed to elicit some helpful testimony during the trial — for example, that practices at the bank for keeping track of cash were lax — , Nagelberg was unable to make good on his assurance to the jury that they would hear testimony showing that the confession was made under coercive circumstances. This blow to the defense was largely a result of an unexpected decision midway through trial that Figueroa would not testify in her own defense as originally planned. Apparently experiencing trepidations about taking the witness stand, Figueroa made up her mind not to testify.9 The defense had to rest *711without presenting Figueroa’s first-hand claim of coercion to the jury.10
The upshot of this was that the jury heard how Figueroa was administered a polygraph test and that she failed without ever being told — apart from never-substantiated allegations in the defense’s opening and insinuating questions on cross-examination— a coherent story about coercion. Thus, in hindsight, Nagelberg’s strategy completely backfired. But a retrospective vista is precisely the wrong perspective from which to determine whether a criminal defendant received effective assistance. We evaluate an attorney’s actions as if we were sitting alongside him at counsel’s table at each moment a challenged decision was made. And, as the above summary of the litigation makes clear, a contemporaneous assessment of Nagelberg’s conduct of Figueroa’s defense based solely on the trial record (and thus assuming reasonableness whenever possible, see Bond v. United States, 1 F.3d 631, 635 (7th Cir.1993)), indicates that it met the constitutional minimum for competency.11
In sum; to attack the confession, Nagel-berg acquiesced to some polygraph testimony; to minimize the damage, he planned to use the testimony offensively; to receive court approval for the plan, he assented to earlier government use of the testimony; and to quell his Ghent’s unanticipated agitation, he discarded the plan. On the record before us, each decision was, though not indisputable, objectively reasonable at the time it was made.
AFFIRMED