POSNER, Circuit Judge.

A commodity pool is an investment fund made up of contributions by a number of different investors. The contributions are commingled and used by the commodity pool operator to buy and sell futures contracts. Because commodity pools are common vehicles for fraud, including Ponzi schemes, the Sentencing Commission has ordained a 4-level guidelines sentencing enhancement for fraud committed by a commodity pool operator. U.S.S.G. § 2B1.1(b)(18)(B)(iii). Brant Rushton, who operated a commodity pool that he used as the vehicle for a Ponzi scheme, pleaded guilty to one count each of mail fraud and money laundering. 18 U.S.C. §§ 1341, 1956(a)(1)(B)(i). The statutory maximum prison sentence for each of these crimes is 20 years. The probation service calculated Rushton's guidelines sentencing range by adding to the base offense level for the mail fraud the 4-level enhancement for commodity pool operator fraud and a 2-level enhancement for abuse of a position of trust. U.S.S.G. § 3B1.3. Other adjustments brought the total offense level in the presentence report to 28 and the guidelines sentencing range to 78 to 97 months. Neither side objected to the presentence report at the sentencing hearing. Rushton's lawyer did argue that enhancements for operating a commodity pool and for abuse of trust overlap and therefore that including both in calculating a sentencing range overestimated the appropriate sentence for his client. But he was appealing to the sentencing judge's discretion rather than challenging the probation service's calculation of the guidelines range.

The judge sentenced Rushton to 96 months in prison and ordered him to make restitution to his victims of $1.62 million. The appeal challenges just the prison sentence.

The judge was indignant that Rushton's victims had, in the judge's words at the sentencing hearing, "include[d] your parents [defrauded of $116,000], relatives [including an uncle defrauded of $30,000 that he had intended for the care of his mentally disabled son—Rushton's cousin], friends, senior citizens, and disabled children." As is typical of such schemes, much of the money that Rushton stole he spent on luxury items, including $150,000 on horses alone.

The judge dwelled particularly on the plight of Dorris Dunn, who "was 85 years old when she invested [in Rushton's commodity pool]. If there's one thing we all know, [it's that] the main thing that senior citizens worry about is that they won't have enough money to live on and will have to ask their children or others for help. And they're generally too proud to do that. Your [Rushton's] actions made sure that her concerns came true." Despite the judge's strong language about exploiting an elderly victim, the government did not seek, nor the judge impose or even mention, the 2-level "vulnerable victim" enhancement authorized by U.S.S.G. § 3A1.1(b)(1).

On appeal Rushton argues not that it was merely inappropriate for the judge to add an abuse of trust enhancement on top of the enhancement for his being a commodity pool operator, as he argued in the district court, but that it violated the guidelines. And the government now agrees, noting that the parties had overlooked U.S.S.G. § 2B1.1, Application Note 14(c), which bars the abuse of trust enhancement in a fraud case if the enhancement for being a commodity pool operator applies. The guidelines commentary explains that because commodity pool operators "are subject to heightened fiduciary duties imposed by securities law or commodities law" and therefore the sentencing court "is not required to determine specifically whether the defendant abused a position of trust," Application Note 14(c) "provides that, in cases in which the new, four level enhancement [for commodity pool operator fraud] applies, the existing two level enhancement for abuse of position of trust... shall not apply." U.S.S.G. App. C, vol. II, p. 367 (Amendment 653, Nov. 1, 2003).

While conceding the error in the calculation of the guidelines range, the government argues that it is not a plain error, as it must be for the appellant to prevail, because he didn't argue in the district court that it was an error. E.g., United States v. Garrett, 528 F.3d 525, 527 (7th Cir.2008). (Recall that in objecting to the overlap Rushton was merely appealing to the judge's exercise of sentencing discretion, conferred on sentencing judges by 18 U.S.C. § 3553(a), which lists factors that the judge must consider in deciding, after calculating the defendant's guidelines sentencing range, what sentence to give.) A plain error is an error that is not only indisputable but also prejudicial—that is, that had an adverse effect on the party complaining of it. United States v. Marcus, 560 U.S. 258, 130 S.Ct. 2159, 176 L.Ed.2d 1012 (2010); Johnson v. United States, 520 U.S. 461, 466-69, 117 S.Ct. 1544, 137 L.Ed.2d 718 (1997); United States v. Olano, 507 U.S. 725, 734, 113 S.Ct. 1770, 123 L.Ed.2d 508 (1993); United States v. Paladino, 401 F.3d 471, 481-82 (7th Cir.2005). The government argues that the error was not plain—but for reasons that have changed during the course of this appeal.

In its appeal brief the government argued that any error in adding the 2-level abuse of trust enhancement was offset by the judge's failure to include a 2-level vulnerable-victim enhancement; for had the judge been apprised of this omission he would surely (in light of what he said about Dorris Dunn) have added—and would have been required by the guidelines to add—a vulnerable-victim enhancement in calculating the defendant's guidelines range. And since a 2-level vulnerable victim enhancement would be identical to the enhancement that the judge mistakenly imposed for abuse of trust, the applicable guidelines range would have been exactly what the judge calculated, albeit erroneously. Moreover, if he thought there were many vulnerable victims of Rushton's fraud—as he may well have thought—he was required to impose an additional 2-level enhancement, on top of the 2-level enhancement for one vulnerable victim. § 3A1.1(b)(2).

But the government withdrew its argument after making it, in acknowledgment of decisions of ours that forbid the government to seek additional sentencing enhancements on remand from an unrelated sentencing appeal. United States v. Love, 706 F.3d 832, 842 n. 4 (7th Cir.2013); United States v. Tello, 687 F.3d 785, 798-800 (7th Cir.2012); United States v. Sutton, 582 F.3d 781, 786 (7th Cir.2009).

Our court appears to be alone in refusing to allow the government to seek a sentencing enhancement that had not been rejected at the original sentencing, as in United States v. Wyss, 147 F.3d 631, 633 (7th Cir.1998), but merely had not been advocated, so that the government was not trying to take a second bite from the same apple. In other circuits the government is allowed on appeal to ask the appellate court to order or authorize the district judge to add a previously unmentioned enhancement if the court is remanding the case for a complete rather than limited resentencing, provided that the government is not seeking to punish the defendant for his temerity in having challenged his original sentence by appealing. (That is, provided the government's purpose in seeking a higher sentence on remand is not "vindictive." Alabama v. Smith, 490 U.S. 794, 798-99, 109 S.Ct. 2201, 104 L.Ed.2d 865 (1989); Waring v. Delo, 7 F.3d 753, 758 (8th Cir.1993).)

Illustrative cases from the other circuits are United States v. Matthews, 278 F.3d 880, 885-87 (9th Cir.2002) (en banc) ("as a general matter, if a district court errs in sentencing, we will remand for resentencing on an open record—that is, without limitation on the evidence that the district court may consider"); United States v. Ynfante, 78 F.3d 677, 679-80 (D.C.Cir. 1996) ("the government relied on `the existing record' and did not introduce any new evidence at the resentencing. The remand was occasioned not by the government's failure to meet its burdens of production and persuasion at the original sentencing, but by the district court's legal error in construing the guidelines. At resentencing, the district court's task was to apply a proper construction of the guidelines to the record already before it"); United States v. Cornelius, 968 F.2d 703, 705 (8th Cir.1992) ("once a sentence has been vacated or a finding related to sentencing has been reversed and the case has been remanded for resentencing, the district court can hear any relevant evidence on that issue that it could have heard at the first hearing," though of course "the sentencing court must ... adhere to any limitations imposed on its function at resentencing by the appellate court"); United States v. Sanchez Solis, 882 F.2d 693, 699 (2d Cir.1989) ("in the interests of truth and fair sentencing a court should be able on a sentence remand to take new matter into account on behalf of either the Government or the defendant [provided that] ... both parties have had an opportunity to be heard").

Thus, according to the rule in other circuits, when the evidence that would justify an enhancement was before the district court at the original sentencing (namely the evidence that Dorris Dunn was indeed a vulnerable victim), and was merely overlooked, the government could seek its correction on remand even though it had failed to object in the district court to the error. But that is not our rule—in defense of which we note that it prevents the government from holding in reserve an objection at the original sentencing, so that if need be it can spring it on the defendant and the court should the defendant succeed in getting his original sentence overturned. Fear of such a tactic might dissuade some defendants from appealing a sentence. But this is not the case in which to reexamine our rule and by rejecting it eliminate a conflict with the other circuits; for the government has not challenged it.

Later still the government told us, agreeing with a point that we'd raised at the oral argument, that the sentencing judge had committed a further error, which if corrected would make up for the erroneous imposition of the enhancement for abuse of trust. Remember that Rushton had pleaded guilty to money laundering as well as to fraud, but that the presentence report had ignored his money laundering plea in calculating his guidelines range. That was error. The report should have calculated offense levels for both counts, fraud and money laundering, and selected the higher of the two as the basis for calculating the defendant's guidelines sentencing range. U.S.S.G. § 3D1.3(a), Application Note 2. And the higher level was the offense level for the money laundering count, not for the fraud count.

The offense level for money laundering in violation of 18 U.S.C. § 1956 (as in the present case) levitates from the underlying offense (the crime that produced the money that the defendant laundered) by adding 2 levels to the total offense level for that offense. U.S.S.G. §§ 2S1.1(a)(1), (b)(2)(B); see, e.g., United States v. Hodge, 558 F.3d 630, 636-37 (7th Cir. 2009); United States v. Anderson, 526 F.3d 319, 328 (6th Cir.2008); United States v. Cruzado-Laureano, 440 F.3d 44, 48 (1st Cir.2006). The result in this case would be a total offense level of 28—the same as the offense level that the judge calculated by erroneously adding the 2-level abuse of trust enhancement to the total offense level for fraud. The effect of the offsetting errors is shown in detail in the following table:

The ranges would not have been the same had the 2-level enhancement for abuse of trust been permissible. It is permissible in a money laundering case— but only when the abuse of trust relates to the money laundering itself rather than to the underlying offense (the offense that generated the money that the defendant laundered). See U.S.S.G. § 2S1.1, Application Note 2(C); United States v. Keck, 643 F.3d 789, 799-801 (10th Cir.2011); United States v. Byors, 586 F.3d 222, 226-28 (2d Cir.2009); United States v. Anderson, 526 F.3d 319, 328 (6th Cir.2008). It is impermissible when, as stated in the abuse of trust guideline, U.S.S.G. § 3B1.3, "an abuse of trust ... is included in the base offense level or specific offense characteristic." See, e.g., United States v. Cruzado-Laureano, supra, 440 F.3d at 48. True, a sentencing judge is not always or even often forbidden to double count. E.g., United States v. Vizcarra, 668 F.3d 516, 519 (7th Cir.2012). But the passage we just quoted from the abuse of trust guideline shows that the double counting in this case was impermissible; Rushton's 4-level enhancement for operating a commodity pool fraud already took account of abuse of trust, as explained in the guidelines commentary that we quoted earlier.

Here the government's argument stops. As the government sees it, because Rushton would face the same offense level (28) on remand, there can be no plain error— and resentencing would be futile to boot because the sentencing range would be unchanged (not so, as we'll soon see). But this overlooks the fact that the failure to sentence under the money laundering guideline was not the sentencing judge's only error. Given what he said about Dorris Dunn, he should have added at least a 2-level vulnerable-victim enhancement, raising the defendant's total offense level to 30 and the guidelines sentencing range to 97 to 121 months. That would make the 96-month sentence that the judge imposed a below-guidelines sentence, doubtless contrary to what he intended.

The defendant insists that Dorris Dunn was not a vulnerable victim, because her money was in trust (the "Dorris Dunn Trust") and presumably the trustee, unlike the beneficiary, was not old, infirm, afflicted with a mental illness (Dunn has bipolar disorder), financially inexperienced, or otherwise especially vulnerable to a Ponzi schemer. The defendant is wrong. His mistake may seem academic now that the government has disclaimed seeking a vulnerable-victim enhancement (and would in any event be barred from doing so by our forfeiture rule) should we remand for resentencing. Not so; it is not academic, because the judge can impose the enhancement on his own, as part of a resentencing proceeding, and should do so.

For there is no basis for doubting that Dorris Dunn was a vulnerable victim. The presentence report—to which, remember, the defendant made no objection—describes Dunn as having been introduced to Rushton by her son, another victim of Rushton's Ponzi scheme, so that she might invest in his commodity pool. In the son's words, quoted in the presentence report, Dunn was swayed by "Brant's way of words to have her trust him and his abilities in investing her life savings with him." And the money lost as a result of the scheme was her money; she was the beneficial owner. The loss was $190,000, and though we don't know her overall financial situation—only that the $190,000 loss was her entire investment in the commodity pool—it left her feeling "betrayed." She was a vulnerable victim of the Ponzi scheme. United States v. Goldberg, 406 F.3d 891, 892-93 (7th Cir.2005); United States v. Sims, 329 F.3d 937, 944 (7th Cir.2003); United States v. Pol-Flores, 644 F.3d 1, 4 (1st Cir.2011); United States v. Hawes, 523 F.3d 245, 255 (3d Cir.2008). The judge should also have considered whether there were other vulnerable victims, maybe enough others (recall his reference to the defrauded uncle, senior citizens, and disabled children) to require a further 2-level enhancement.

So what we have is a thoroughly botched sentencing in which the parties, the probation service, and the sentencing judge are all implicated. We can't blame any of them too harshly, because the sentencing guidelines are absurdly complex. But botched the sentencing was, and the remedy is not speculation about what the judge would have done had he calculated the sentencing range accurately; it is a resentencing from scratch, see United States v. Tovar-Pina, 713 F.3d 1143, 1147-48 (7th Cir.2013); United States v. Langford, 516 F.3d 205, 216-19 (3d Cir.2008); United States v. Lozano, 514 F.3d 1130, 1135-36 (10th Cir.2008), beginning with a revision by the probation service of the presentence report.

The alternative to ordering resentencing would be to pronounce the errors not plain because they were offsetting: the enhancement for abuse of trust was wrong, but so was the judge's failure to sentence under the money laundering guideline. But that ignores the judge's failure to impose a further enhancement, or enhancements, for the presence of a vulnerable victim, or vulnerable victims. That was another error. Properly computed, Rushton's total offense level would have been either 30 and his guidelines sentencing range 97 to 121 months, or even 32 (sentencing range 121 to 151 months) if there were, as there may well have been, many vulnerable victims, not just Dorris Dunn.

It can be argued that since the government is not asking for a longer sentence than the judge gave, we should let it stand even though it is based on a miscalculation of the sentencing range. But while the government was precluded by our forfeiture rule from arguing in this court for a vulnerable-victim enhancement, a defendant who appeals from a sentence takes a risk that if the case is remanded for resentencing, as the defendant in this case urges be done, he will receive a longer sentence should the court of appeals notice an error in his favor committed in the sentencing proceeding that he has appealed. As long as the higher sentence on remand is not vindictive, its exceeding the defendant's original sentence does not invalidate it. See, e.g., United States v. Warda, 285 F.3d 573, 580-81 (7th Cir. 2002); United States v. Goldberg, supra, 406 F.3d at 894; United States v. Johnson, 715 F.3d 179, 181-82 (6th Cir.2013); United States v. Garcia-Guizar, 234 F.3d 483, 487-90, (9th Cir.2000) ("Garcia's higher sentence resulted solely from the district court's correction of an error in Garcia's first presentence report, an error the district court was obligated to correct," id. at 489).

We cannot predict what sentence the district judge will impose on remand; it is unlikely to be shorter but uncertain whether it will be longer. The correct guidelines range is a range, not a point. There thus is sentencing discretion even when the judge imposes a sentence within the range. And the calculation of the defendant's guidelines sentencing range is merely the first step in sentencing, though an essential step because it has what psychologists call an "anchoring" effect. The calculation is complicated, mandatory, and done first; thus it is likely to exert a not wholly conscious tug on the judge when, after having determined the guidelines range, he is deciding what sentence to give, guided by the sentencing factors in 18 U.S.C. § 3553(a). See, e.g., Stephanos Bibas, "Plea Bargaining Outside the Shadow of Trial," 117 Harv. L.Rev. 2463, 2515-19 (2004); Birte Englich & Thomas Mussweiler, "Sentencing Under Uncertainty: Anchoring Effects in the Courtroom," 31 J. Applied Soc. Psychol. 1535 (2001). A mistake—in this case a cascade of mistakes—in calculating the guidelines range can affect the sentence even if a correct calculation would result in the same range. But in this case it would not.

Having calculated the range correctly, the judge must then decide, as a matter of discretion, whether to sentence within it; and that decision may be influenced by the factors that determined the range, since he is free to reject those factors. The judge in this case may believe, for example, that money laundering is not a more serious offense than fraud, and so he might sentence at the low end of the new guidelines range (for his disagreement with the weight that the guidelines give particular conduct does not alter the range, though it may influence his exercise of his sentencing discretion) despite the 2-level enhancement for the money laundering count to which Rushton pleaded guilty. In that event the judge might even decide to reimpose his original sentence, which was just a month short of the bottom of the correct guideline. Alternatively he might decide that since Rushton's total offense level is at least 30 (because there should have been a vulnerable-victim enhancement, and maybe a double such enhancement), his previous sentence—a below-guidelines sentence because the correctly calculated total offense level is at least 30—was too lenient. And finally there is the unresolved question whether there were enough additional vulnerable victims to warrant an additional 2-level enhancement.

REVERSED AND REMANDED.