Appeals from the United States District Court for the Northern District of Illinois, Eastern Division. No. 95 C 4242--Ruben Castillo, Judge.

Before Coffey, Easterbrook, and Ripple, Circuit Judges.

Easterbrook, Circuit Judge.

Between 1976 and 1990 Richard Perlman worked at Equity Financial and Management Company, a real estate syndication and management group under the control of Samuel Zell and Robert Lurie. While at Equity, Perlman invested in several of the firm's real estate partnerships. Perlman also received "participations" in other partnerships--that is, fractional interests in Equity's share of these partnerships. Paying employees with participations put their self-interest to work; participation holders would profit only to the extent other investors did.

After leaving Equity, Perlman complained about the payments he received on account of his partnership and participation interests. Equity replied that payments were being offset against a $300,000 loan that Perlman had not repaid before he left; Perlman rejoined that the money was a bonus, not a loan. Equity also informed Perlman that distributions on participation interests were being deferred because some of the ventures were ongoing. According to Equity, a participation is measured by the value of all real estate partnerships established in the year it is awarded, and its final value cannot be calculated until all of these partnerships have been wrapped up or refinanced; Perlman contended, to the contrary, that participations created interests in each partnership and should be valued and paid out (if the value is positive) when each partnership concludes, rather than when all of the year's ventures come to an end. If some partnerships lose money, then the dispute about participations affects total payments as well as timing. Suppose Perlman had a 1% participation in Equity's share of three 1988 partnerships, two of which finished with a $1 million profit in 1992 and the third of which incurred a $1 million loss in 1995. Equity contends that nothing is payable until 1995, when all three ventures have been wound up, after which the loss would be offset against the profit and 1% of the net (or $10,000) paid to Perlman. On Perlman's view, however, Equity should have paid $20,000 for the two profitable ventures in 1992, and he would owe nothing to Equity in 1995 on account of the losing partnership.

Each side stuck to its guns and litigation ensued. Perlman sought relief under state law for breach of contract (the parties agree that Illinois law controls). Because most parties are citizens of Illinois, the diversity jurisdiction does not authorize litigation in federal court. To get around this problem--and to treble the stakes--Perlman contended that he is a victim of "racketeering" entitled to collect under the Racketeer Influenced and Corrupt Organizations Act (RICO), 18 U.S.C. sec.sec. 1962(c), (d), 1964. Recovery under RICO depends on proof that the defendants (Zell, Lurie, and partnerships they controlled) operated an "enterprise" (Equity) through a "pattern of racketeering activity" or at least conspired to do so. Section 1961(1) defines "racketeering activity"; mail fraud is the principal form alleged, although the possibility of securities fraud lurks in the background. Partnership interests may be securities, and the case began before Congress conditionally removed securities fraud from the list in sec.1961. See Fujisawa Pharmaceutical Co. v. Kapoor, 115 F.3d 1332, 1337-38 (7th Cir. 1997). State-law claims then came to federal court under the supplemental jurisdiction. 28 U.S.C. sec.1367. Perlman presented under sec.1367 not only breach of contract but also several flavors of consumer fraud.

Defendants performed their promises until Perlman's departure, which could have knocked out most of his fraud claims. Breach of contract is not fraud; only making a promise with the intent not to keep it deserves that epithet. Bower v. Jones, 978 F.2d 1004, 1012 (7th Cir. 1992); Restatement (2d) of Torts sec.530(1) (1977). When the substance of the promise is fairly debatable (and the participation program, at least, merits that label), and many promises are kept for extended periods before a dispute breaks out, it is hard to see how it could be said that the defendants never intended to keep their bargains. But Perlman had other theories of fraud, and the district judge concluded that all should be tried. 938 F. Supp. 1327 (N.D. Ill. 1996). The jury awarded him approximately $700,000 for the value of his participations and another $700,000 for his partnerships, and it ruled in Perlman's favor under RICO. The jury also found that the $300,000 was a loan rather than a bonus (vindicating defendants' setoffs) and rejected Perlman's state-law fraud claims.

On post-judgment motions, the district court concluded that the RICO judgment could not be sustained, even viewing the evidence in the light most favorable to the verdict, because Perlman failed to establish a "pattern" of racketeering. The court held that Perlman is entitled to prejudgment interest on the $1.4 million and that defendants are not entitled to prejudgment interest on the $300,000. The net judgment in Perlman's favor is approximately $1.8 million. Both sides have appealed, raising many issues, most with multiple subparts. This opinion addresses only the more substantial of the disputes; the judgment is affirmed with respect to the rest substantially for the district court's reasons.

* Breach of contract is not fraud, and a series of broken promises therefore is not a pattern of fraud. It is correspondingly difficult to recast a dispute about broken promises into a claim of racketeering under RICO. Uniroyal Goodrich Tire Co. v. Mutual Trading Corp., 63 F.3d 516, 522-23 (7th Cir. 1995); J.D. Marshall International, Inc. v. Redstart, Inc., 935 F.2d 815, 821 (7th Cir. 1991). Difficult is not impossible. Sometimes the evidence shows outright lies and a plan not to keep one's promises-- enough of them to meet RICO's continuity-plus-relationship formula for a "pattern." H.J. Inc. v. Northwestern Bell Telephone Co., 492 U.S. 229 (1989); see also Sedima, S.P.R.L. v. Imrex Co., 473 U.S. 479 (1985). But the burden of proving this is on the plaintiff, and like the district court we conclude that Perlman did not adduce evidence from which a trier of fact properly could find a pattern of predicate acts. Indeed, the jury itself seems to have been of this view; it rejected each of the explicit fraud claims Perlman leveled against the defendants. This leaves its RICO verdict as something of a puzzle, but one best resolved as the district court did. In an effort to avoid the conclusion that this case is more than just a dispute about the meaning of the partnership and participation contracts, Perlman has advanced so many different theories of fraudulent schemes that it would be tedious to discuss them all. A sample will show the flavor of the claims, and why they fail.

Zell and Lurie were (through an intermediary) the general partners of Four Lakes Village Associates, in which Perlman Marketplace Investors owned a 4% interest as a limited partner. Perlman maintains that the documents concerning Four Lakes misrepresented some facts about that venture, and that Zell and Lurie fraudulently appropriated the partnership's value by shifting its assets to a real estate investment trust (REIT) that Zell and Lurie controlled. Perlman contends that Zell and Lurie deceived other investors about both the purpose and the consequences of the transfer to the REIT, obtaining their approval for the transaction under false pretenses that amounted to mail fraud. A separate claim in the complaint demanded compensation for the value of the 4% interest, and misrepresentations concerning the Four Lakes transaction also formed the centerpiece of a contention that defendants committed consumer fraud in violation of Illinois law.

Defendants responded to these contentions by noting, first, that the supposedly fraudulent transaction was set up by Perlman himself. If deceit occurred, defendants observed, then Perlman was the author of the fraud--but it would be strange to find fraud when Perlman, who must have known the truth, bought 4% of the units. Was Perlman such a good liar that he deceived himself?, defendants inquire. They concede that the general partners caused the restructuring of Four Lakes, and that the REIT ended up with its assets, but they say that the partnership received fair value and that the limited partners were told exactly what had been done and why. Zell lent the Four Lakes partnership approximately $6 million to finance its operations, but it was unable to make a profit and defaulted on the loan, which Zell then foreclosed. By using the foreclosure to transfer the partnership's assets to the REIT, Zell and Lurie were able to produce tax benefits for the limited partners that compensated them for the financial reverse. The general partners sent the limited partners a description of the transaction that revealed both the partnership's financial distress and the plan of restructuring. This notice was sufficiently ominous that a limited partner other than Perlman had his accountant investigate whether anything was amiss--and the accountant concluded that everything was on the up-and-up.

After hearing these conflicting versions of events, the jury decided that defendants did not owe Perlman Marketplace Investors any money on account of the 4% interest, a verdict that necessarily rejects any claim that the restructuring fraudulently deprived the limited partners of their entitlements. The jury also brought back a verdict in defendants' favor on the state consumer fraud theory, which necessarily rejects any claim that the sale of interests in Four Lakes can be attributed to mail, wire, or securities fraud. Perlman has four theories about why these verdicts, which supply the principal underpinning for the district judge's Rule 50 judgment, are not dispositive.

First, he contends that these deals also entailed tax fraud, a theory that the district judge did not allow to be presented to the jury. But how tax fraud (of which we see no evidence) could have injured the limited partners, as opposed to the United States Treasury, Perlman does not explain.

Second, Perlman contends that a RICO conspiracy may be actionable without proof of injury attributable to racketeering acts. Our opinion in Schiffels v. Kemper Services, Inc., 978 F.2d 344, 348 (7th Cir. 1992), offers support for this view. Next Term the Supreme Court may decide whether this aspect of Schiffels is correct. See Beck v. Prupis, 162 F.3d 1090, 1098-99 & n.15 (11th Cir. 1998) (disagreeing with Schiffels), cert. granted, 119 S. Ct. 2046 (1999). Perlman's problem lies deeper, however: it is not simply a failure to show that the injury flowed directly from the racketeering acts as opposed to some other conduct in furtherance of the conspiracy, but inability to show either injury (from any source) or a pattern of racketeering.

Third, Perlman believes that it is possible to establish mail fraud without establishing deceit; if this is so, then the jury's verdicts do not knock out his claims. But how can one have mail fraud without fraud? The word "fraud" in the mail-fraud statute means deliberate, material misrepresentations. See Neder v. United States, 119 S. Ct. 1827, 1839-41 (1999). No fraud, no mail fraud. And the jury found that the Four Lakes transactions did not entail fraud.

Fourth and finally, Perlman points to the damages the jury awarded for his participation interests. Perlman's assessment of the value of his participation interests at the time of trial was $633,288. The jury actually awarded him $733,288. He continues: "defendants admit they cannot otherwise explain at least $100,000 in damages awarded under Counts V, VIII and IX. Those damages could easily be based on Four Lakes." Defendants treat this unexplained $100,000 as an error in addition; the jury's figures for both partnership and participation damages were otherwise exactly what Perlman sought. The jury asked for a calculator but returned its verdict before the court supplied one. Could the $100,000 have been a disguised verdict in Perlman's favor on the Four Lakes partnership? If it was, the disguise is a good one--for the jury was given a separate verdict form concerning Four Lakes and brought back a verdict for defendants. Why would a jury that expressly ruled against Perlman on his theory that Four Lakes was operated fraudulently have smuggled a contrary decision into the damages award on the participation claims? It is best to take the jury's verdicts at face value, just as the district judge did; these verdicts mean that the events concerning the Four Lakes partnership are not predicate acts of fraud.

None of Perlman's other arguments persuades us that the district judge erred. Perlman contends, for example, that the defendants "fraudulently converted" partnership distributions to which he was entitled. But the "fraudulent conversion" was the offset of partnership distributions against the $300,000 loan. No "fraud" was involved because the defendants told Perlman exactly what they were doing; and the jury found that there was no "conversion" either, because the $300,000 was indeed a loan rather than a bonus. A setoff cannot be treated as mail fraud (or any other kind of fraud). Perlman's other efforts to establish a pattern of predicate acts have similar shortcomings; no more need be said on the RICO issues.

Perlman has two other challenges to the verdict: first, he believes that the district judge erred in failing to submit a punitive- damages instruction to the jury; second, he believes that the judge should have allowed the jury to consider a claim that the defendants converted the value of his stock in Rosenberg, Perlman & Associates, P.C. (now known as Rosenberg & Liebentritt, P.C.), a professional corporation that served as in-house counsel to Equity, its only client. Whatever could be said for punitive damages earlier, the jury's conclusion that the defendants did not commit fraud, and the district judge's conclusion that they did not commit any RICO predicate acts, removes any foundation for punitive damages. As for the stock in the captive law firm: Perlman received this without payment as an accounting measure when the firm was converted from partnership to professional-corporation form of organization; he never carried it at any value on his statements of net worth. The stock was transferred to Perlman's replacement in the firm as a formality when Perlman moved from in-house lawyer to business executive at Equity. Perlman did not protest or seek compensation until many years later. None of the other participants in this firm has ever paid for the stock or received anything for its surrender. The district judge sensibly concluded that a reasonable jury could not find that Perlman had been deprived of a valuable interest.

For their part, defendants advance in a conditional cross-appeal several arguments in favor of a new trial on issues concerning compensatory damages. But because they concede owing Perlman something in the vicinity of $1.4 million--from defendants' perspective, the payment is too soon, not necessarily too high-- they do not want a new trial to argue about a few thousand dollars here or there (or even about $100,000) unless the RICO claim is going to be retried. Because we have agreed with the district judge about Perlman's RICO claim, the condition on which the cross-appeal is based has not been satisfied, and we need not address defendants' arguments about the conduct of the trial.

On one liability issue, however, the cross- appeal is not conditional. Nancy Kresek was held liable on a single claim: that she failed to act as Perlman's fiduciary in the administration of the participation program. For this supposed failing she was ordered to pay Perlman more than $700,000 in damages and $544,000 in prejudgment interest. Kresek contends that she was a subordinate at Equity--essentially a glorified bookkeeper in the participation program's administration, an employee too junior to participate in the program herself and therefore hardly its fiduciary. Challenged to show how Kresek could be treated as a fiduciary, Perlman's appellate brief is silent. Instead Perlman responds with the generality that Kresek "participated in" other defendants' acts. That may be so, but why is she liable for them? If a bank's president wrongfully instructs a teller not to pay a check, the bank and its president may be liable, but the teller will not be, for the teller does not exercise discretion in the matter. Perlman relies on Allabastro v. Cummins, 90 Ill. App. 3d 394, 413 N.E.2d 86, 89 (1st Dist. 1980), and National Acceptance Co. of America v. Pintura Corp., 94 Ill. App. 3d 703, 418 N.E.2d 1114, 1116-17 (2d Dist. 1981), for the proposition that Illinois holds "participants" personally liable even when they exercise no discretion, but neither case supports that proposition. National Acceptance did not concern fiduciary obligations, and in Allabastro both persons held liable apparently had the discretion to satisfy the plaintiff's demands. Kresek therefore is entitled to judgment in her favor. We doubt that this will matter in the end; Kresek's liability was joint and several with other defendants who are good for the judgment (indeed, have posted a supersedeas bond securing its payment). She is entitled, however, to the assurance that her personal wealth is secure.

II

Perlman contended that he had been effectively frozen out of the partnerships and participation program, and he asked the jury to award as damages the full value of his interests as of the day of trial. The jury obliged. Defendants then asked the district judge to cancel all of these interests as part of the judgment, observing that, once they compensated Perlman for the market value of his interests, Perlman could not keep the interests themselves; the judgment amounted to a sale of his investments back to the defendants for a price fixed by the jury. The district judge agreed with this proposition in principle, and the final judgment cancels many of Perlman's interests. But it does not cancel them all. The court treated the judgment as a sale of only those partnerships that had been named as defendants and listed in that role on the verdict forms submitted to the jury. (Not all of the defendants were listed in the verdict forms. The reason for the discrepancy is mysterious.)

Whether a given partnership was formally a defendant (or included in the verdict forms) is not material to the question whether the judgment has compelled the persons who are defendants to repurchase particular interests. Every interest that has been cashed out in the judgment must be canceled. Perlman's main damages exhibit listed all of his partnership and participation interests, with a dollar value attached to each; the jury gave him what he asked for (plus $100,000). It follows that all of his partnership and participation interests have been liquidated and must be canceled.

The law of preclusion (res judicata) leads to the same result. Perlman put all of these interests in issue with his complaint. The final decision in a suit covers the whole case, which here includes the whole portfolio of partnership and participation interests. According to Perlman, this is not true. His brief asserts that the district court struck from the complaint all claims relating to two particular partnerships, Galesburg Venture and Sandburg Investors, and that his interests in these two partnerships, at least, must pass through the case unaffected. Actually, however, the district court's ruling was that Perlman could not amend his complaint to make extra claims concerning these two partnerships. Perlman's valuation expert included Galesburg Venture and Sandburg Investors in the calculation of damages--and, though Perlman's interests in these two partnerships were valued at $0, this appears to reflect the fact that both are under water, rather than their exclusion from the case. Because the district court neither entered an order preventing Perlman from seeking compensation for the value of his interests in these two partnerships nor blocked his expert from including them in the calculation of his loss, we conclude that Perlman's legal interest in these partnerships too must be canceled.

III

Prejudgment interest depends on the law that supplies the substantive rule of decision--in this case, Illinois law. In re Oil Spill by Amoco Cadiz, 954 F.3d 1279, 1333 (7th Cir. 1992). Illinois authorizes prejudgment interest when the claim depends on written instruments or "on money withheld by an unreasonable and vexatious delay of payment." 815 ILCS 205/2. The partnership and participation agreements are written, and the district judge added that in his view defendants unreasonably withheld payments as they came due. Section 205/2 specifies a rate of 5% per annum. Nonetheless, the district judge awarded prejudgment interest at the rate of prime plus 2%. Perlman does not attempt to reconcile this with the terms of the governing law. Instead he contends that the defendants "agreed" to pay interest at the higher rate.

What Perlman calls an "agreement" turns out to be no more than defendants' effort to show that delay in payment of participation units did not injure Perlman in the first place. Recall the example in the second paragraph of this opinion: two partnerships wind up with profits in 1992, and a third concludes with a loss in 1995. Defendants deferred payment on the profitable ventures until the net for all partnerships established in a given year could be computed. To reflect the time value of money in the interim, Equity credited participation holders with interest on the profitable partnerships at a rate of prime plus 2% while waiting for final settlement. This is why, in defendants' view, delay in distribution did not harm Perlman or other participation holders. Nothing in this internal accounting convention supersedes sec.205/2, which specifies what rate must be used if litigation ensues.

Using the wrong rate is not the only problem with the award of prejudgment interest. The district judge treated the entire $1.4 million as if it had been due on the date suit was filed, and he added interest from that time. Yet the bulk of the jury's verdict does not represent sums that should have been paid in 1995 or before. At least $880,000 of the damages represents the capital value of the partnership and participation interests as of the trial. Payment of these items was not delayed; none of this money was "due" until the jury decided to liquidate Perlman's investments.

Valuing these investments as of the trial automatically compensated Perlman for the time value of money through trial, so prejudgment interest is double counting. His investments received the same economic return as the other partners did, and, because Perlman voluntarily chose to become a partner, he cannot demand a higher rate than they received in the same investments. Suppose a plaintiff contends that in 1995 the defendant converted 100 shares of General Motors stock, and in 1998 the jury agrees and directs defendant to pay plaintiff the value of these shares as of the date of verdict. No prejudgment interest could be added to that award; the change in the shares' market price between 1995 and 1998 would compensate the plaintiff for the time value of money. (Put otherwise, valuing the shares as of the trial ensures that the plaintiff rather than the defendant receives the economic benefit of the investment for the 1995-98 period.) Prejudgment interest would be due only on dividends the defendant retained between 1995 and 1998. The same principle applies when the investments are traded over the counter (as interests in these partnerships were) rather than on a stock market. Provided the valuation as of trial is accurate, the plaintiff automatically receives compensation for the time value of money.

Only sums in the nature of dividends or distributions, which should have been paid before trial, are eligible for prejudgment interest. Because the jury found that the $300,000 was a loan rather than a bonus, the first $300,000 of distributions should be set off against this loan before calculation of interest. Although the jury determined that the loan was not an interest- bearing advance, and the district judge therefore properly denied defendants' motion for prejudgment interest on the $300,000, the setoff of this loan against partnership distributions was proper, and this setoff annulled the debt to that extent. Because sums that were set off were not wrongfully withheld from Perlman, he can't collect interest on them; he received full value from the reduction of his indebtedness to the defendants. For amounts withheld in excess of the $300,000, however, prejudgment interest is applicable--but from the date payment should have been made, not from the date the complaint was filed.

Some of the payments were not due until well after the case began; an award of prejudgment interest from an earlier date is improper. Perlman believes that other overdue payments predated the complaint and contends that running interest from the complaint for the whole sum is a good compromise. "Close enough for government work" is not a good refrain, however. The payment due dates are ascertainable, and a district court can do as well as a bank in tracking these dates and adding interest. Because the jury followed Perlman's damages calculation so closely, we need not be derailed by the possibility that the jury's award represents an unknowable combination of deferred payments and capital value. See Williamson v. Handy Button Machine Co., 817 F.2d 1290, 1298 (7th Cir. 1987). We therefore vacate the award of prejudgment interest for the capital value of Perlman's investments as of trial, and we remand for an award (at 5% annually) on unpaid distributions (to the extent they exceed $300,000) from the dates they should have been paid through the date of judgment. This is a mechanical task, on which the parties' damages experts should be able to reach quick agreement.

IV

After concluding that Perlman is the prevailing party, the district court awarded him approximately $90,000 in costs under Fed. R. Civ. P. 54(d). Defendants contend that this award is an abuse of discretion, if not downright impermissible, for several reasons:

* Four defendants prevailed on every claim in the district court, and a fifth (Kresek) has won on appeal. These five defendants are prevailing parties entitled to recover their own costs; they may not be required to pay Perlman's.

* Those defendants who filed the counterclaim concerning the $300,000 loan prevailed on that issue.

* Perlman lost his RICO claim outright, and he failed on the state-law claims relating to Four Lakes, fraud, and punitive damages.

* Perlman ultimately recovered only about 7% of the sum demanded in his complaint. Even if all RICO and punitive damages theories are put to one side, Perlman received only 26% of the amount he initially sought under state law.

Just the other day we stressed that Rule 54(b) gives the prevailing party a strong presumptive entitlement to recover costs. Luckey v. Baxter Healthcare Corp., No. 98-2043 (7th Cir. July 12, 1999). Five defendants have prevailed outright; the district court did not explain why it ordered these prevailing defendants to chip in toward Perlman's costs. They are entitled to recover their own costs of defense.

Whether Perlman is himself a "prevailing party" is a tougher issue. He lost on the $300,000 loan issue, but he nonetheless received a substantial net judgment. Still, two features of this case support a conclusion that Perlman did not "prevail"; and though we cannot say that an award of costs in his favor is forbidden we think that the subject deserves a fresh look by the district judge.

First, Perlman's RICO claim is why this case is in federal court. Yet he lost on RICO and recovered only on state-law claims that would not have supported a federal suit. So Perlman is a loser on the question whether this was a federal case. Congress provided in 28 U.S.C. sec.1332(b) that, when a claim based on diversity of citizenship does not produce a recovery high enough to justify its presence in federal court--that is, when the recovery falls below the $75,000 jurisdictional amount--"the district court may deny costs to the plaintiff and, in addition, may impose costs on the plaintiff." Section 1332(b) does not apply directly to our situation, but the principle for which it stands--that if the outcome shows that the case did not belong in federal court, then costs may be denied or shifted--is generalizable. Drawing a parallel to sec.1332(b) does not forbid an award of costs; but it shows that the district court has discretion to treat defendants as the prevailing parties because they prevailed on the only claim that justified the presence of the case in federal court. Following the approach of sec.1332(b) enables federal courts to protect themselves from suits in which federal claims have been trumped up only to impose upon federal jurisdiction.

Second, Perlman's modest recovery (modest in relation to his original demand, that is) implies that the defendants won more of the dispute than they lost. The term "prevailing party" appears in many statutes and rules other than Rule 54(d). Many statutes provide, for example, that a prevailing party recovers attorneys' fees. We have held in a series of recent cases that a litigant who wins less than 10% of his initial demand either is not a prevailing party for purposes of fee-shifting statutes or should be treated as if he had not prevailed. E.g., Cole v. Wodziak, 169 F.3d 486 (7th Cir. 1999); Fletcher v. Ft. Wayne, 162 F.3d 975 (7th Cir. 1998); Simpson v. Sheahan, 104 F.3d 998 (7th Cir. 1997). Recovery of less than 10% of the claim shows that, even making allowances for puffing in the complaint, the defendant has won on the bulk of the seriously disputed items. Just so here. Perlman recovered far less than he sought, losing the two big-ticket items (RICO and punitive damages) and most of the major skirmishes (Four Lakes, the $300,000 loan, and state-law consumer fraud).

What Perlman recovered is in the end substantially less than 7% of his claim, for he "won" only on the partnership and participation claims with respect to which his interest was conceded by the defendants. The jury valued his interests at $1.4 million as of 1998, but even had every issue in the suit been decided in the defendants' favor, Perlman still would have received this amount (less the mystery $100,000) in a few years. The partnership and participation claims are about the timing of payments, rather than entitlement to them; and if defendants are computing interest on deferrals at prime plus 2%, then Perlman shot himself in the foot by "winning" on these claims! He would have been better off had he waited for defendants to pay. (Defendants' valuation expert calculated the worth of Perlman's participation interests at more than Perlman's expert did, precisely because defendants' expert included interest at the higher rate.) The nominal award of $1.4 million therefore substantially overstates the value of Perlman's victory; his real success is measured by the difference between the $1.4 million now and the amount he would have received had he waited, discounted to present value. Perhaps that figure is negative, but even if positive it is much less than $1.4 million. The real stakes of this case always have been treble damages under RICO and punitive damages under state law, and on these Perlman lost.

None of what we have written means that the district court is forbidden to award costs in Perlman's favor (except with regard to the five defendants who prevailed on all claims). District judges have substantial discretion to allocate costs under Rule 54. But that discretion must be exercised with full understanding of the governing legal principles, and in this case the district judge's short treatment of costs, which did not acknowledge the principles we have elucidated, is inadequate to sustain its award.

V

The judgment on damages is affirmed, except to the extent it concerns Kresek. We remand for entry of judgment in Kresek's favor.

The portion of the judgment canceling plaintiffs' partnership and participation interests is vacated, and the case is remanded for entry of a revised judgment canceling all of the interests involved in the litigation.

The award of prejudgment interest is vacated, and on remand the district judge must exclude interest on the capital value of plaintiffs' investments as of the trial, and recompute prejudgment interest at 5% on payments (in excess of $300,000) that should have been made before trial, from the date these payments were due.

The award of costs is vacated, and on remand the district court must reconsider this subject in light of Part IV of this opinion. Defendants recover their costs of these appeals.