In her summation the trial assistant urged that the jury should find the victim’s accusation of defendant credible by reason of defendant’s possession of a cigarette lighter resembling a pistol and toy handcuffs:
“How else do you know that [the victim] was credible and accurate. Well, you know that the defendant was armed with handcuffs. And [a] cigarette lighter as [defense counsel] keeps referring to . . . it. Although I submit to you if I pulled this [lighter] out, you’d be a little frightened.
“I’m not going to sit here and argue to you that he used these things during the robbery and therefore that makes this a more egregious case.
“What I’m going to say is if you could buy a home made kit to be a robber, right next to the ski-mask you would have the gun and handcuffs” (emphasis added).
To avoid precisely this use of the novelty items, defendant moved at the outset of the trial to preclude their introduction in evidence. Defendant’s attorney argued that the items had not been used during the charged offenses and were not relevant to their proof. Rather, the evidence of the items and the circumstances of their *1078recovery would only encourage a conviction impermissibly rooted in the supposition that robbery was defendant’s line— the sort of thing he did. The court nonetheless admitted the evidence, insisting that it was relevant as “part of the whole of the transaction.”
Notably, the People never advanced a theory of relevance to support the receipt of the items. While, subsequent to the court’s ruling, the trial assistant, in response to defense counsel’s argument that the items would not be probative of intent or identity since they were not used during charged crimes, argued that they might have been used if the victim had not resisted and were therefore relevant to proving intent, this fairly abstruse point was not subsequently pursued and, as noted, was not the basis of the court’s ruling. There was, moreover, no intent issue in the case. Defendant conceded that there had been at least an attempted robbery;1 the core disputed issue was whether defendant had been correctly identified as a perpetrator.
Defendant’s attorney requested an instruction forbidding the jury from making propensity-based inferences from the disputed evidence, reiterating his view that the evidence was admissible only if it was in some specified way directly probative of guilt. The request, however, was rebuffed. The court said, “I’m not going to parcel things out and break it down to minutiae. It doesn’t amount to a hill of beans anyway.” After defendant’s attorney made his record, the court reiterated, “I don’t have to designate [the relevance of the evidence] or place it under any particular rubric. We are just spinning wheels.”
The gist of the common-law rule with which People v Molineux (168 NY 264 [1901]) has become so closely identified is that evidence probative only of a propensity to conduct like that charged is not admissible. Although it may well seem that a person who has offended in some way would be a likely candidate for similar reoffense, the inference of propensity from prior conduct and from there to guilt is so fraught with opportunity for error that it is forbidden as a ground for a criminal conviction. Evidence of other, i.e., uncharged, crimes or bad *1079acts, however, may be admitted where it is in some way directly probative of the charged offense—that is to say, probative without the intermediate inference of criminal propensity. Molineux, of course, describes several ways in which evidence otherwise subject to exclusion as evidence of propensity, might be fairly and reliably used to establish in a reasonably direct and disciplined way an element of the crime charged (id. at 293-310). These theories of relevance are not exclusive, but prosecutors seeking to introduce proof demonstrative of bad propensity routinely attempt to bring such proof within one of the exceptions enumerated in Molineux. Here, however, it does not appear that the prosecutor offered any theory of relevance, much less one rooted in precedent. Rather, the evidence that defendant had on him at the time of his arrest, an imitation gun and keys to a pair of toy handcuffs found at the scene of his apprehension, was deemed admissible upon the ground, advanced by the court but not the prosecutor, that the disputed evidence was “part of the whole transaction.”
The court’s theory was obviously flawed since the novelty items simply were not part of the charged criminal transaction. If they had been, their admissibility would have been clear. It is precisely because they were not and invited speculation about defendant’s criminal inclinations that their relevance, if any, to the underlying transaction and the consequent criminal charges had to be spelled out.
It is true that in narrowly defined circumstances and with limiting instructions the People have been permitted to “complete the narrative” and, in the process, to place before the jury evidence of other crimes or bad acts (see People v Resek, 3 NY3d 385, 389 [2004]). The rationale for the exercise is to aid the jury in “trying to sort out ambiguous but material facts” (id. at 390). But here there was no showing that this jury needed a more complete context properly to evaluate the victim’s inculpatory identification. Indeed, situating that identification in a narrative in which it might well seem to acquire probative heft for reasons rooted in propensity-based speculation and having nothing to do with the identifier’s opportunity for observation and accuracy of recall, was merely a form of bolstering.
Although the People, in retrospect, have espoused the theory now embraced by the majority that the subject novelty items were probative of defendant’s intent to rob, that theory, not having figured in the trial court’s decision to admit the evidence, is not properly available to this Court as a ground for *1080affirmance (CPL 470.15 [1]; People v Concepcion, 17 NY3d 192, 195 [2011]). Moreover, as noted, intent simply was not an issue in this case.
And, while defendant’s identity as a perpetrator was closely disputed, the evidence in question was not permissively probative on that issue either. Under Molineux, it is only where the uncharged behavior demonstrates a signature—i.e., highly idiosyncratic—modus operandi or in some other very specific way tends to demonstrate that the perpetrator of the extraneous act and the charged crime were one and the same, that such proof is admissible to identify the accused as the person who committed the charged crime (168 NY at 313-314). While courts have, since Molineux, been somewhat less strict in applying the identity exception (see e.g. People v Beam, 57 NY2d 241, 252-253 [1982]), that nuance is of little moment here since there is no proof at all of any successive acts; there is no proof of a prior robbery, much less proof of one involving the novelty items at issue, and, even if there had been, its relevance to the charged crimes could not be made out under the Molineux identity exception since, as the prosecutor herself acknowledged, the crimes of which defendant was accused were not committed with those supposedly telltale props. For very good reason, then, the prosecutor never contended that the imitation gun and toy handcuffs were admissible under the Molineux identity exception. She argued instead that defendant’s possession of the novelty items showed that he was a robber and thus that the victim’s accusation that defendant had robbed him was creditworthy. But this was nothing more than an invitation to convict defendant of the charged robbery on the basis of his supposed tendency to rob. Leaving aside that the inference of such a tendency from the possession of novelties is far from ironclad, the further treacherous step of concluding that because defendant had a propensity to rob he must have robbed the victim on the occasion charged is precisely the sort of inference that the law has long forbidden as a ground for criminal liability.
Even if there had been some non-propensity-based rationale for admitting the disputed evidence, it was never explained to the jury. Nor, contrary to the suggestion of the Appellate Division, did the trial court ever undertake to exercise its discretion to determine whether the evidence, if relevant, would be more prejudicial than probative. The trial court, in default of performing these essential prophylactic functions specifically requested *1081by defendant, left the jury to speculate along the lines urged by the trial assistant—that defendant’s possession of the gun-like cigarette lighter and toy handcuffs was indicative of a career in robbery and that the victim’s accusation of defendant, even if doubtful on the basis of what he claimed to have seen and remembered of the incident, was in a general way believable and might well have hit a deserving mark. Obviously, this is not a permissible path to a guilty verdict.
Although the majority would deem any error in the admission of the novelty items harmless in light of the “overwhelming” testimony identifying defendant as one of the victim’s assailants (majority mem at 1076), the use of the term “overwhelming” to describe the evidentiary basis for defendant’s conviction is, on any actual examination of this record, a tad rich. This was in its essential aspect a one-witness identification case.2 The identifying witness, the victim, by his own account was held prone to the floor during the robbery, and observed his assailants only briefly. Although he thought that three men took part in the assault, he was completely unable to describe two of them, and as to the third, could say only that he was a “kind of white and Hispanic guy.” He was unable to recognize defendant when shown his photograph at trial,3 and conceded in his trial testimony that “even this guy [referring to defendant] I don’t remember.” It is true that, after his assailants fled, the victim, following an interval of uncertain duration, ran out of the building vestibule where the assault took place and, on scanning the busy rush hour street scene, thought he recognized defendant as one of his assailants, but his identification under those stressful circumstances based upon what could have been no more than a glimpse—the entire incident having, according to the victim, passed very quickly—was far from overwhelming in its proof of defendant’s participation in the robbery. Indeed, the use of the term “overwhelming” with reference to the inculpatory identification in this case is, on the present record, I believe, imprudent and discordant with what we have elsewhere acknowledged respecting the incidence of wrongful conviction attributable to mistaken eyewitness identification (see People v Santiago, 17 NY3d 661, 669 [2011]). To be clear, the issue is not *1082whether the properly admitted evidence was sufficient—it was. It is rather whether in addition to being sufficient it was overwhelming and, as such, on its own equal to the purpose of eliminating to a virtual certainty the hypothesis of defendant’s innocence (see People v Arafet, 13 NY3d 460, 467-468 [2009]). It would seem plain that such a robust predicate did not exist to insulate this verdict from the evidence introduced to show that defendant was by vocation a robber.
Judges Ciparick, Graffeo, Read, Smith, Pigott and Jones concur; Chief Judge Lippman dissents in an opinion.
Order affirmed in a memorandum.
. Defense counsel could not have been clearer. Before trial he stated on the record, “[my client]’s not disputing there was a robbery. He’s saying he wasn’t the person that did it.” He made the same concession as to what had happened—that there had been a robbery—to the jury. And, in summation, specifically addressing the adequacy of the proof of intent, defense counsel said, “I think it’s pretty obvious what the intent was. They were searching [the victim] to try and rob him. That’s what the intent was.”
. While there was testimony in addition to that of the victim about the assault and its aftermath—much of which was difficult to harmonize—only the victim identified defendant as one of his assailants.
. Defendant was not present at his trial, so the in-court identification was attempted using a photograph.