concurring.
I join the court’s opinion. I write separately to express my concern about the government’s last-minute change in theory as to which gun was used to commit the robbery.
Loggins’ theory of defense was that she did not know in advance of her co-defendants’ plan to rob the credit union and that she drove the car away from the credit union in a panic. Her odds of prevailing on this defense were not especially strong, given that participating in an escape from a robbery is considered participating in the robbery itself, see United States v. Smith, 415 F.3d 682, 689 (7th Cir.2005) (coll.cases), vacated & remanded on other grounds, - U.S.-, 126 S.Ct. 2859, 165 L.Ed.2d 892 (2006), and by Loggins’ own admission she realized what her co-defendants had done when they ran out of the credit union and jumped into the car with a bag full of money. R. 156-3 at 408, 471. Still, there is an intuitive difference between someone who knows about a robbery plan ahead of time and volunteers for the role of getaway driver and someone who remains in the dark until her co-defendants dash out of the bank, pile into the car, and scream at her to drive away. A sympathetic jury confronted with the second scenario might agree that the driver’s heat-of-the-moment decision to step on the gas and get the heck out of Dodge was not a deliberate and culpable decision to aid the robbery.1
Use of a real gun was not an element of the offense, as my colleagues point out, but as a practical matter it was quite relevant to Loggins’ defense. If it was Loggins’ Smith & Wesson that Johnson held in his hand during the robbery, then it was a more obvious inference that Loggins was in on the plan from the start; indeed, it would confirm Reynolds’ testimony that Loggins had participated in the discussion of the robbery the night before and had advised Johnson to use her gun in the robbery, see R. 156-1 at 108, 121. This would in turn rule out the possibility that Loggins found out about the robbery only after the fact and drove her cohorts away from the scene of the crime in a panic. Ante at 983.
*985I readily agree with my colleagues that the evidence before the jury supported the prosecutor’s contention that Loggins’ gun was used to commit the robbery. Ante at 983. Although none of the witnesses to the robbery were asked to identity the gun they saw in co-defendant Johnson’s hand, in addition to Reynolds’ testimony that Loggins had shown her revolver to Johnson and told him to carry it, there was testimony from the credit union tellers describing the gun Johnson was carrying as being big and silver. Having looked at the guns myself, I am satisfied that only Loggins’ Smith & Wesson truly meets that description; the plastic toy guns are smaller and have a distinctive blue-green cast to them (although they do have a somewhat metallic sheen).
What is troubling about the government’s contention — voiced for the first time in its rebuttal argument — that it was Loggins’ gun that was used, is that this was wholly contrary to the position that the government had taken — repeatedly, on the record, and in the presence of the district judge — right up to the eve of trial. When Love pleaded guilty to the May 28 robbery on May 3, 2005, four months before Loggins’ trial, the Assistant United States Attorney (“AUSA”) represented to the court that the government’s evidence would show that “[djuring the course of the robbery, [c]o-[d]efendant Johnson pointed a toy handgun at bank employees and customers while instructing them to put their hands up.” R. 170 at 13-14. The district judge took note of the prosecutor’s description of the gun as a toy and asked her what evidence she had to support that description. Id. at 17. The AUSA informed the court that two toy guns had been recovered from inside the getaway car, “[a]nd when compared to the bank surveillance photos and what [the] tellers described, those seemed, to match the guns that were used in both robberies. And we believe it was a toy gun used in both the May 22nd robbery and the May 28th robbery.” Id. at 18. The government made the same representation in its proffer of the evidence against Johnson (who pleaded guilty the same day as Love): the AUSA indicated that “[djuring the course of the robbery, ... [defendant Johnson ... pointed a toy handgun at bank employees and customers while instructing them to put their hands up.” Id. at 50.2 Love and Johnson themselves specifically represented to the court in their plea colloquies that the guns they had carried were toy guns (although Love could not recall what type of gun Johnson had carried). Id. at 21-22, 52. On September 1, 2005, just days before the start of Loggins’ trial, co-defendant Reynolds pleaded guilty. Again the government, in outlining its evidence, indicated that “during the course of the robbery, co-defendant Johnson pointed a toy handgun at bank employees and customers while instructing them to put their hands up.” R. 134 App. B at 4. Finally, on September 6, 2005, when the government asked for (and received) a delay in the start of Loggins’ trial because neither the toy guns nor Log-gins’ Smith & Wesson revolver could be located in the evidence locker, the AUSA specifically advised the court that the toy guns looked very much like the real gun (Loggins’ revolver). R. 173 at 5. “If you laid them side by side without picking them up, they would look almost identical *986in appearance.” Id. Again the court expressed interest in which of the guns was used to commit the robbery, and although on this occasion the government did not repeat its earlier refrain that Johnson had pointed a toy gun at the tellers and customers of the credit reunion, it did disavow any intent to show that Loggins’ gun had been used to commit the robbery:
THE COURT: Once this issue of the guns is resolved, then what inferences is the Court to draw with respect to the three guns, two apparently toy guns and one real gun?
AUSA: ... The inference regarding the guns would be that there were three guns used. We can’t tell the jury which guns were used inside of the May 28 robbery. All we can say is that three guns were found, and that the defendant had the only real gun on her person.
Id. at 25.
What the record does not reflect but what we have been told is that once the FBI tracked down the Smith & Wesson and the prosecutor apparently got her first side-by-side look at the three guns just a day or two before the trial commenced, she concluded that the toy guns were readily distinguishable from the real gun and that it likely was Loggins’ gun — the Smith & Wesson — that Johnson had used to carry out the May 28 robbery. We do not know whether the tellers themselves would have been able to identify the Smith & Wesson as the gun that Johnson had pointed at them, since none of them was asked to do so at trial. (And I would add that the one perspective from which it is difficult to distinguish the real gun from the toy guns is when one looks directly into the barrels of the guns — i.e., when they are pointed directly at an individual.) But in any event, the government changed its theory as to which gun Johnson had used, and as I have already acknowledged, the notion that Johnson had carried Log-gins’ gun into the bank was a plausible one based on the trial evidence, which included the guns themselves, the tellers’ testimony about what the gun that Johnson carried looked like, and Reynolds’ testimony that Loggins had urged Johnson to carry her gun with him. I agree with my colleagues that the government was entitled to make this argument to the jury. See ante at 983.
What I do not think the government was within its rights to do, given the position it had repeatedly taken on the guns prior to trial, was keep quiet about its change of theory until the evidence was closed and the defense counsel had given his closing argument. A defense by its nature is reactive; because it is the government that bears the burden of proof, the defense must respond to what evidence the prosecution puts on and what inferences it asks the jury to draw from that evidence. In the abstract, of course, the prosecutor may have had no obligation to share with the court and with the defense her thoughts as to which of the guns Johnson had used: as discussed, this was much more material to the defense case than it was to the government’s. But not only had the government previewed its theory regarding the guns before trial, it had in the course of three separate change-of-plea hearings repeatedly represented to the court as fact that Johnson had used a toy gun — not Loggins’ Smith & Wesson — in the course of the robbery, and as late as a few days before Loggins’ trial was representing to the court that it could not prove which of the guns was used inside the credit union. Nothing new or unexpected emerged at trial in regard to the guns: no witness identified the Smith & Wesson as the gun Johnson had pointed at the tellers and customers of the credit union, for example. Under these circumstances, the defense *987had no reason to anticipate the about-face that the government performed in its final closing argument.
The government suggests that the defense brought this on itself when Loggins’ counsel argued in his own closing statement that Johnson had used a toy gun rather than Loggins’ Smith & Wesson to commit the robbery, thus “mak[ing] ... an issue” of which gun was used. Gov’t Br. at 39; see also ante at 983. This is revisionist history. The government gift-wrapped this argument for the defense when its own counsel repeatedly represented to the court in advance of trial that Johnson had used a toy gun and that it could not prove otherwise. That Loggins’ attorney would take up that refrain could have surprised no one: it was part and parcel of the defense theory that Loggins took no part in planning the robbery and had no foreknowledge of it — a theory that was consistently advanced throughout the case. See, e.g., R. 173 at 10-11 (Loggins’ counsel outlines theory of defense at hearing conducted the week before trial commenced). The problem here was one of the government’s making, not Loggins’.
By keeping its change of mind about the guns under wraps until its final closing argument, the government effectively deprived Loggins of the chance to respond. Just as there was evidence to support the government’s eleventh-hour contention that Johnson had used Loggins’ gun, there was evidence to support Loggins’ contrary theory that he had not. During Johnson’s plea colloquy, for example, Johnson had read aloud (and swore to the truth of) a signed written statement concerning the May 28 robbery in which he stated that he “displayed a toy gun” during the robbery. R. 170 at 52. Moreover, an FBI forensic expert had prepared a report indicating that, on review of credit union surveillance photographs, he was unable to determine whether the weapon Johnson held in his hand during the robbery was Loggins’ revolver or one of the toy guns (because the photographs were insufficiently detailed). R. 134 App. D. Finally, the fact that the prosecutor did not ask any of the tellers whether they could identify which of the guns Johnson had used suggests it is at least possible that they too might be unable to do so. (Although the differences between the guns seemed rather obvious to me within the sedate confines of my chambers, I can readily appreciate how a witness who had only a moment’s glimpse of the gun under highly stressful conditions might be less able to state with confidence which of the guns he or she saw in Johnson’s hand.) Of course, Loggins could have sought to introduce this evidence during trial; nothing precluded her from doing that. But she had no reason to do so after the government effectively took that issue off the table by conceding that Johnson had used a toy gun and professing its inability to prove that Johnson had used Loggins’ gun. By the time the prosecutor finally put that issue back on the table, the evidence was closed and the defense’s opportunity to speak to the jury had come and gone.
As a former prosecutor and district judge, I know that in preparation for trial, opposing parties and judges alike routinely rely on the representations that the parties’ counsel make as to what issues and theories they will and will not pursue. This is not to say that a party is forever after bound by whatever pre-trial positions its counsel might take. Things rarely proceed according to plan: new evidence comes to light, witnesses give unexpected testimony, mistakes become apparent, strategies change. Although the government’s counsel can perhaps be faulted for not arranging to see the guns further in advance of trial than she did (we are told that Loggins’ counsel, by contrast, had *988viewed the guns months earlier), once she examined the guns and realized that the weapon witnesses had described as large and silver most likely was Loggins’ Smith & Wesson, I believe she was entitled to argue that theory. But given the government’s prior on-the-record assertions that Johnson had used a toy gun, I believe the AUSA was obligated to disclose her change of thought to both the court and the defense. Not to do so, and to reserve that announcement for the last moments of trial, was a serious error in judgment. Trial by ambush has absolutely nothing to recommend itself to the judicial process.
The question before us, however, is whether Loggins is entitled to be re-tried, and given the substantial deference we owe to the district court on this question, e.g., United States v. Woolfolk, 197 F.3d 900, 904-05 (7th Cir.1999), I cannot say that it was an abuse of discretion to deny Loggins’ Rule 33 motion for a new trial. The government’s rebuttal argument, although a surprise to the defense, was consistent with the appearance of the guns and with the tellers’ descriptions of the gun that Johnson had used. My sense is that the jury would have drawn the same inference on its own even if the government had not made the argument or if the defense, not having been lulled into a sense of false security by the government’s pre-trial statements on this subject, had introduced evidence suggesting that Johnson used (or may have used) a toy gun. Moreover, apart from Loggins’ gun, there was substantial evidence indicating that Loggins knew ahead of time that her co-defendants were planning to rob the credit union. Reynolds so testified, of course. Perhaps more importantly FBI Special Agent Gustavo Grodsinsky testified that following her arrest, Loggins admitted to him that she had overheard Johnson and Love discussing the robbery the night before (although she thought this was just idle chatter), R. 156-3 at 502, that on the morning of the robbery Johnson and Love had awakened her and told her “they were going to get the money,” id. at 503, and that she continued to think they were joking “except for when they told her to get dressed because she[ ][was] the only person that ha[d] a vehicle and they needed her to drive,” id. (emphasis mine). Loggins told Grodsinsky that after getting dressed, she drove her codefen-dants around town looking at different banks until they settled on the Uliana Financial Credit Union, which Johnson and Love said “looked like a good one.” Id. at 504. Moreover, Loggins’ own written statement (which Grodsinsky read aloud to the jury), although insisting that she had at first believed her co-defendants’ talk of a robbery to be a joke, acknowledged that once her codefendants got out of the car and proceeded into the bank, she realized they were not joking. R. 156-4 at 537. At the very least, this evidence suggests that Loggins did not remain in the dark about the robbery as long as her defense posited, and that her decision to drive the getaway car was not truly made in the panic of the chaotic moments after the robbery. Against that backdrop, I am not convinced that the government’s unexpected argument that Loggins’ gun was used to commit the robbery ultimately made a difference to the outcome of the trial or otherwise affected Loggins’ substantial rights. See generally United States v. Kuzniar, 881 F.2d 466, 470 (7th Cir.1989); United States v. Reed, 875 F.2d 107, 113 (7th Cir.1989) (quoting United States v. Martinez, 763 F.2d 1297, 1312-13 (11th Cir.1985)).
With these additional observations, I join the court’s opinion and judgment.
. Getaway drivers occasionally do succeed on such defenses. See, e.g., Briefs, Stuart: 20-year-old cleared of getaway charges, Stuart (Fla.) News, June 16, 2006, at Local News, 2006 WLNR 10585962; Madelaine Vitale, Split Verdict in Shootout Killing of Atlantic City Bar Manager, Press of Atlantic City (N.J.), Feb. 25, 2006, at Cl, 2006 WLNR 3340080; Sara Eaton, Alleged 01 getaway driver cleared; Wendy’s employee was shot during pair of robberies, Fort Wayne (Ind.) Journal Gazette, at 1, 2004 WLNR 15243884.
. The government made the same representation in the official version of the offense that it submitted to the probation office on May 19, 2005, for use in preparation of Johnson’s pre-sentence report. R. 134 App. C at 4 (“During the course of the robbery, defendant [Johnson] pointed a toy handgun at bank employees and customers while instructing them to put their hands up.”).