POSNER, Circuit Judge.

It is a curiosity of class action litigation that often there is greater ferocity in combat among the class lawyers over the allocation of attorneys' fees than there is between the class lawyers and the defendants. The contest among the lawyers is a zero-sum game. But the contest between them and the defendants is a positive-sum game because the class lawyers are naturally very interested in the fee component of any settlement, while the defendants care only about the size of the settlement, including fees. So the lawyers may be willing to settle for less for the class if the defendants will help them obtain a generous fee award, and the defendants will be happy to help them if the sum of the fee award and the relief granted to the class is smaller than it would be if the class lawyers pressed for more generous relief for the class. E.g., Thorogood v. Sears, Roebuck & Co., 624 F.3d 842, 848-49 (7th Cir.2010); Vollmer v. Selden, 350 F.3d 656, 660 (7th Cir.2003); Reynolds v. Beneficial National Bank, 288 F.3d 277, 282 (7th Cir.2002); In re General Motors Corp. Pick-Up Truck Fuel Tank Products Liability Litigation, 55 F.3d 768, 778, 802 (3d Cir.1995); Weinberger v. Great Northern Nekoosa Corp., 925 F.2d 518, 524 (1st Cir.1991).

Indeed, class lawyers may try to fend off interlopers who oppose a proposed settlement as insufficiently generous to the class; and given the role of such interlopers in preventing cozy deals that favor class lawyers and defendants at the expense of class members, their requests for fees must not be slighted. Mirfasihi v. Fleet Mortgage Corp., 356 F.3d 781, 782-83 (7th Cir.2006); Vollmer v. Selden, supra, 350 F.3d at 659-60; Crawford v. Equifax Payment Services, Inc., 201 F.3d 877, 880-82 (7th Cir.2000); In re General Motors Corp. Pick-Up Truck Fuel Tank Products Liability Litigation, supra, 55 F.3d at 803.

One such objecting lawyer in the present case, the appellant, Dawn Wheelahan, was awarded $2.7 million in attorneys' fees by the district court for her contribution to the generous settlement of the Trans Union Corporation Privacy Litigation. She contends that she's entitled to three times as much as she was awarded, both because the total attorneys' fees awarded to class counsel were too little and because her share of them was too small. She is not opposed by the class action defendant, Trans Union, because this is a "common fund" suit; attorneys' fees come out of the amount of damages awarded the class, and so Trans Union has no stake in the dispute over fees. Wheelahan's only opponents are some of the other class lawyers, who fear that an increase in the amount of fees awarded to her would come at their expense. They don't object to an increase in the total fees awarded, or indeed to an increase in the share awarded to her that would not reduce their fees. With the class members unrepresented and the defendant indifferent to the overall award of attorneys' fees, we must decide the appeal with limited assistance from an adversary presentation. But this is a standard dilemma in class action adjudication, as we noted at the outset, and may be unavoidable without elongating the litigation disproportionately to the stakes in the fee dispute.

The other lawyers for the class (with one exception) were prepared to settle the case for $40 million, consisting of $20 million in cash plus relief in kind valued at $20 million. Wheelahan opposed the settlement, and as a result solely of her opposition (she argues) the case was eventually settled for a little under $110 million (we'll ignore the little under), of which $75 million was in cash and the other $35 million was the estimated value of the relief in kind included in the settlement. At the oral argument Wheelahan told us that she should receive at least 20 percent of the added value of $70 million as her fee. That would be $14 million, rather than the $2.7 million that she was awarded. Her brief, however, asks for only $8.4 million. As that amount is, as we'll see, excessive, the $14 million is pie in the sky and can be ignored.

The dispute over fees grows out of a litigation that began in 1998 with the filing of a number of class actions charging Trans Union, a large credit-reporting agency, with violating the Fair Credit Reporting Act, 15 U.S.C. §§ 1681 et seq., by selling information in its consumer credit files to advertisers without the consumers' authorization. The actions were consolidated for pretrial proceedings in the Northern District of Illinois. In 2006 lawyers referred to by the parties to the appeal as "MDL Counsel," who had filed the earliest cases, agreed with Trans Union to a $40 million settlement. Wheelahan, who had filed her own class action in 2005, opposed the proposed settlement, and her opposition, together with that of class counsel in a case filed against Trans Union in Texas, persuaded the district judge to rescind his preliminary approval.

He approved the final settlement of $110 million in 2008. The settlement placed a ceiling of $18.75 million on attorneys' fees for the class lawyers (remember that the fees come out of the common fund created by the settlement). The lawyers urged the district court to award this amount—17 percent of the estimated value of the settlement. The district judge, however, thought $18.75 million excessive and reduced it to $10.83 million, but later reconsidered and referred the issue of fees to a special master, who recommended an award of $13 million, of which about two-thirds ($7.8 million) would go to the MDL counsel and most of the rest to Wheelahan and to the Texas counsel. The district judge accepted the recommendation.

The special master arrived at these figures by first determining the total amount of attorneys' fees that would be reasonable to award and then allocating that amount among the lawyers. He placed little weight on the contingent fee agreements between the lawyers and the "clients" (the named plaintiffs in the class actions), recognizing that named plaintiffs are usually cat's paws of the class lawyers. In re Cavanaugh, 306 F.3d 726, 737-38 (9th Cir. 2002); Goldberger v. Integrated Resources, Inc., 209 F.3d 43, 53 (2d Cir.2000); John C. Coffee, Jr., "The Regulation of Entrepreneurial Litigation: Balancing Fairness and Efficiency in the Large Class Action," 54 U. Chi. L.Rev. 877, 885-86 (1987). That's especially true when, as in this case, the prospective relief for an individual class member is minuscule: the class has 190 million members! —though most have not filed claims and the deadline for filing is past. The special master recognized that his task was to estimate the contingent fee that the class would have negotiated with the class counsel at the outset had negotiations with clients having a real stake been feasible. In re Synthroid Marketing Litigation, 264 F.3d 712, 718-20 (7th Cir.2001); see also Missouri v. Jenkins, 491 U.S. 274, 285-86, 109 S.Ct. 2463, 105 L.Ed.2d 229 (1989); but see Goldberger v. Integrated Resources, Inc., supra, 209 F.3d at 51-52.

Such estimation is inherently conjectural, and what the special master mainly did was examine data on awards of attorneys' fees in other class actions. He relied heavily on an academic study which had found that between 1993 and 2002 the average awards of attorneys' fees in common fund consumer class actions had been either 16.2 or 24.3 percent of the amount of the settlement, depending on which of two datasets of such awards were consulted. Theodore Eisenberg & Geoffrey P. Miller, "Attorney Fees in Class Action Settlements: An Empirical Study," 1 J. Empirical Legal Stud. 27, 51 (tab.1) (2004). The average award declined in percentage terms as the size of the settlement increased, in recognition of the fixed-cost component of a lawyer's activity in a case—there is an irreducible minimum of lawyer activity that must be undertaken if the client is to have a reasonable chance of prevailing, no matter how small the stakes in the case. In the case of settlements of between $79 million and $190 million (the range within which the settlement in this case fell), the study found that the average attorneys' fee awards were 17.6 or 19.5 percent of the settlement, again depending on the dataset. Id. at 73 (tab.7). The award that the parties to the settlement in the present case recommended—$18.75 million—was just below that range: 17 percent of $110 million, as we said.

The special master then looked at a group of securities cases and from the fee awards made in them estimated that if the settlement in such a case had as in this case been $110 million the fee award would have been between 3.8 percent and 23.2 percent. He thought this range relevant because plaintiffs' discovery costs tend to be higher in securities class actions than in consumer class actions. But he did not indicate how one might adjust the fee award in this case to reflect the presumed lower cost of discovery.

He then considered the noncash portion of the settlement: a choice, valued as we said at $35 million, between two "free" credit monitoring services that Trans Union would offer each class member who filed a complaint:

In re Trans Union Privacy Litigation, "Detailed Notice," p. 5, www.listclass action.com/content/Detailed_Notice.pdf (visited Jan. 5, 2011) (emphasis in original). The special master thought this in-kind relief worth less to class members than $35 million in cash would have been worth and that the attorneys' fee awards should therefore be lower than the awards for cash relief—only 5 percent of the $35 million. In contrast, he awarded 15 percent of the cash portion of the settlement.

The $35 million figure is the retail price of the offered services multiplied by the number of class members who asked for the services in lieu of cash. The value of the services to a member of the class could be less than the retail price. The members of the class were presented with a choice between cash relief—not knowing the actual amount they'd receive—and inkind relief. (This is a standard feature of settlements in common fund class actions.) The amount of cash to which each class member asking for cash rather than services would be entitled would be $75 million divided by the number of such claimants. If they are very numerous, the cash that each would receive would be less than the value of the in-kind relief even if a class member valued that relief at less than its retail price. Suppose the class member can obtain cash relief of only $30, and the retail price of the service he could obtain instead is $60. Then even if he values that service at only $31, he is better off choosing it. (On the assumption that the amount of fees requested was proper and that one percent of the 190 million eligible claimants requested cash relief, their individual recovery would be only $29.60: $75 million, minus $18.75 million in attorneys' fees, divided by 1.9 million.)

But this possibility need not justify a lower attorneys' fee award (let alone two-thirds lower) for the in-kind component of the settlement. The fact that the amount of cash that a class member can expect to receive is likely to be small makes the in-kind option very attractive—maybe not $35 million attractive but the special master did not try to estimate a lower value.

Next he considered the risk of losing that the lawyers for the class faced when they embarked on the litigation. If they lost they would receive no fees at all, and the higher the risk of failure the higher the contingent fee that a client would have to pay in an arm's length negotiation with the lawyer in advance of the suit. As we've noted in previous cases, the logic of scaling the fee to the risk leads to absurdity if pressed too hard: it would justify an astronomical fee in a frivolous suit in which the plaintiff prevailed by a fluke. Kirchoff v. Flynn, 786 F.2d 320, 326 (7th Cir.1986); McKinnon v. City of Berwyn, 750 F.2d 1383, 1392-93 (7th Cir.1984); see also Pennsylvania v. Delaware Valley Citizens' Council for Clean Air, 483 U.S. 711, 719-23 and n. 6, 107 S.Ct. 3078, 97 L.Ed.2d 585 (1987); Laffey v. Northwest Airlines, Inc., 746 F.2d 4, 26-27 (D.C.Cir.1984), overruled on other grounds by Save Our Cumberland Mountains, Inc. v. Hodel, 857 F.2d 1516 (D.C.Cir.1988); John Leubsdorf, "The Contingency Factor in Attorney Fee Awards," 90 Yale L.J. 473, 474 (1981). But within the set of colorable legal claims, a higher risk of loss does argue for a higher fee. Risk aversion to one side (for a lawyer with a diversified portfolio of cases should not be risk averse, Rand v. Monsanto Co., 926 F.2d 596, 599 (7th Cir. 1991)), if the market-determined fee for a sure winner were $1 million the market-determined fee for handling a similar suit with only a 50 percent chance of a favorable outcome should be $2 million. E.g., In re Continental Illinois Securities Litigation, 962 F.2d 566, 569, 573 (7th Cir. 1992); Staton v. Boeing Co., 327 F.3d 938, 967-68 (9th Cir.2003).

The special master thought the risk of loss in the present litigation had been low because in 2001, when many of the separate class actions against Trans Union that were eventually consolidated for pretrial proceedings had not yet been filed and those that had been filed were still in their early stages, and so presumably the costs that the lawyers would have borne in vain if the suits collapsed were small, a decision by the Federal Trade Commission holding that Trans Union's practices attacked in the class actions indeed violated the Fair Credit Reporting Act was affirmed. Trans Union Corp. v. FTC, 245 F.3d 809 (D.C.Cir.2001). The special master thought that outcome made the risk of loss to the class counsel in this case less than the risk in In re Synthroid Marketing Litigation, 325 F.3d 974, 980 (7th Cir. 2003), where we had upheld a 22 percent attorneys' fee award. Without trying to quantify the difference in risk ("no one can know"), the special master concluded that the attorneys' fees in this case should be limited to 12 percent of the settlement. (Fifteen percent of the $75 million in cash plus 5 percent of the $35 million of in-kind relief equals 12 percent of $110 million.)

The 12 percent figure was plucked out of a hat, and a hat with three holes in it: the unresolved comparison with securities class actions, the arbitrary reduction in attorneys' fees for the nonpecuniary relief, and the perfunctory (less than a page) consideration, also left unresolved, of the relative risk of loss in the present case and in Synthroid.

Having lowered the attorneys' fees ceiling to 12 percent of the settlement from the 17 percent that the lawyers had recommended to him, thus setting an overall limit on attorneys' fees of $13 million, the special master turned to its allocation among counsel. He thought Wheelahan and counsel in the Texas suit piggish to be requesting, between them, $10.8 million of the $13 million. The request was based on their claim to be responsible for the entire $70 million increase in the final settlement over the settlement to which MDL counsel had agreed earlier. So far as we can judge, had it not been for the efforts of Wheelahan and the Texas lawyer the class would indeed be poorer by $70 million. The special master did not suggest otherwise. But then, remarking that the premise of Wheelahan's and the Texas counsel's claims was that the initial, rejected settlement had not reflected the true settlement value of the case—which was obviously so—he asked: "But who `created' that extra value [the $70 million]? It was principally the efforts of MDL Counsel." The others had "helped unlock the true enhanced value in this case, but they did not solely create it," and "MDL Counsel remain entitled to share in the value of the case above and beyond the original $20 million in cash and $20 million of in-kind value embodied in the rejected settlement." He decided they should get full credit for having created the first $20 million of the cash component of the final settlement, for that was the amount that had been negotiated in the first settlement agreement, and 50 percent of the credit for the additional $55 million in cash in the final settlement, for a total of 63 percent ($20 million plus .5 × $55 million = $47.50 million ÷ $75 million = 63 percent) of the cash component of the final settlement. He used the same percentages in determining fees for the in-kind relief, and so ended up awarding MDL counsel 63 percent of the $13 million total of attorneys' fees awarded. He gave Wheelahan 22 percent of that total (hence the $2.7 million that he awarded her) and Texas counsel 15 percent.

In making these adjustments the special master was wrestling with a problem of joint causation. The final settlement was the result of the combined efforts of MDL counsel and of the other two class lawyers. The fact that these efforts were successive rather than simultaneous has no significance. The MDL counsel created an asset—the expected gain from the lawsuits— the value of which they did not realize. The efforts of the other lawyers enabled the full value to be obtained.

Suppose the value of an uncut diamond owned by A is $1,000, he hires B to cut it for $5,000, and after it is cut it is worth $10,000. Does that mean that B should receive not $5,000 but $9,000 because he created additional value in that amount? That isn't how prices are set in a competitive market, or, since we're talking about attorneys' fees, wages either. The suggestion that they are echoes the "comparable worth" theory of fair compensation. See Paul Weiler, "The Wages of Sex: The Uses and Limits of Comparable Worth," 99 Harv. L.Rev. 1728, 1756-58 (1986). B in our example received $5,000 because that was the competitive price for cutting the diamond. He was not entitled to the surplus that his service created, because he wasn't the diamond's owner. MDL counsel correspond to A in our example and Wheelahan and the Texas counsel to B.

The special master may have had an inkling that his percentage allocations were arbitrary, because he also looked at the relative time productively invested by the different lawyers, multiplied by a reasonable hourly rate, and came up with an estimate of the relative costs incurred by the different lawyers. This computation, which accords with the evidence, yielded a ratio of MDL counsel's costs to Wheelahan's costs of 4.3 to 1, which is considerably higher than the 63 to 22 percent ratio of fees that the special master awarded to the two counsel. In effect he penalized MDL counsel for having proposed an inadequate settlement. Those counsel aren't complaining, however.

The special master did fine in his allocation of fees among counsel, but, as we explained earlier, failed to justify the $13 million ceiling on fees, and one result—critical to this appeal—is that because a $110 million settlement is large and Wheelahan's contribution major though not so great as she contends, the $2.7 million award for her contribution—a mere 2.5 percent—was too small. But as we have upheld the allocation among counsel, her only valid complaint is about the aggregate award.

Remember that the special master ruled that as far as the cash relief was concerned, an award of fees equal to 15 percent of the cash award was proper, and that he failed to justify a lower percentage of the value of the in-kind relief. An award of attorneys' fees equal to 15 percent of the entire $110 million award, which would correct the special master's error, would be $16.5 million. But considering that he erred consistently in favor of reducing fees, we think he did not make a case that the $18.75 million requested by the lawyers for the class was excessive.

Consider against that background what Wheelahan is asking for. She wants an additional $5.7 million, which on top of the $2.7 million awarded to her on the recommendation of the special master would give her the $8.4 million requested in her brief. If we are correct that she is entitled to only 22 percent of the total attorneys' fees awarded, the implication of her request is that the special master should have recommended a total fee award to all counsel of $38.2 million, because $8.4 million is 22 percent of that amount. That would be 35 percent of the settlement. She does not defend that percentage, or complain about the total of $18.75 million in attorneys' fees agreed to in the final settlement negotiations, but on the contrary that is the starting point of the calculation that mysteriously generates her claim to an $8.4 million award.

We conclude that she's entitled to 22 percent of the $18.75 million in total fees that should have been awarded, which is $4.125 million. Since she's already been awarded $2.7 million, she's entitled to an additional award of $1.425 million. No other lawyers have appealed from the fee awards, so they are not entitled to additional fees.

The additional award to Wheelahan does only rough justice, but a remand would produce only speculative refinements, given the inherent uncertainties of estimation in such a case, and would do so at a heavy cost in judicial and party resources unlikely to be offset by any benefit in greater precision, which would in any event be illusory. We are mindful that the $1.425 million will come out of the relief awarded to the class members, who are not represented in this appeal, but we do not think that this litigation should be protracted further by our appointing a lawyer to represent the class in place of the existing class counsel, who at this stage are concerned only with their fees. That lawyer's fees would come out of the common fund and so reduce the amount of relief obtained by the members of the class. We therefore order the judgment of the district court to be modified as explained above, and we remand for the entry of a corrected judgment.

One issue remains to be addressed. The district court had directed the special master not only to recommend fee awards but also to investigate complaints that Wheelahan had engaged in unethical behavior in the litigation. The special master concluded that although Wheelahan had behaved unprofessionally in various respects, she had not behaved unethically, and so he recommended against the imposition of any sanction and the district judge accepted the recommendation. Nevertheless her appeal challenges the judgment on the further ground that the special master and the district judge have made unwarranted criticisms of her. Whether this is true or not, it is not a basis for an appeal. A criticism is not an appealable order.

MODIFIED AND REMANDED.