Robin Peoples is here for the third time contesting his convictions for bank robbery and associated offenses. Each time he has argued that he received ineffective assistance of counsel. Twice we resolved this claim on the merits; that is enough (if not once more than enough), and we decline to revisit the subject.

On his direct appeal Peoples's appellate lawyer challenged the competence and dedication of his trial lawyer. That was a risky tactic, given the circuit's rule, see United States v. Taglia, 922 F.2d 413, 418 (7th Cir.1991), that such a contention normally may be raised only once, and the fact that an ineffective-assistance claim is unlikely to succeed without additional evidence showing why trial counsel had acted as he did, and what the adverse consequences of a mistake would have been. But it may have been justified in this case, because by the time of the appeal the district judge had held a hearing on Peoples's request for a new trial, and evidence about his original attorney's conduct had entered the record. (He had a different attorney on appeal.) So we addressed the subject on the merits, though the decision was not to Peoples's liking. United States v. Peoples, 2001 U.S.App. LEXIS 7625 (7th Cir. Apr. 11, 2001) (unpublished order).

Meanwhile the new-trial request (Peoples had filed three motions) was still brewing in the district court. The district judge should have treated these as collateral attacks under 28 U.S.C. § 2255 but did not warn Peoples that his motions would use up the only collateral review allowed as of right, and as a result they did not count as an initial round of collateral review. See Castro v. United States, 540 U.S. 375, 124 S. Ct. 786, 157 L. Ed. 2d 778 (2003). The district judge denied these motions on the merits, and we affirmed, addressing the ineffective-assistance claim again, and with the same outcome. United States v. Peoples, 2002 U.S.App. LEXIS 14401 (7th Cir. July 16, 2002) (unpublished order).

His opportunities for direct review used up, Peoples filed a motion that was acknowledged to be under § 2255. Once again he complained about the legal assistance he had received. The district judge balked at evaluating the protest. He pointed out that he had done so once already, and we had done so twice. Circuit law is clear, the judge remarked, that a "defendant who complains on direct appeal about the quality of his lawyer can't try again on collateral attack unless there has been an intervening change of law." Ryan v. United States, 214 F.3d 877, 879 (7th Cir.2000). Discovery of previously withheld evidence also might reopen the subject, see Taglia, 922 F.2d at 418, but Peoples did not contend either that hidden evidence had at last surfaced or that the law governing his contentions had morphed since our second decision. All he did was add to the list of failings that he attributed to his former lawyers, but the district judge held that this is exactly the sort of thing that the law of the case cuts off: it blocks new theories as well as old ones. See, e.g., Ryan, 214 F.3d at 878-79; People Who Care v. Rockford Board of Education, 171 F.3d 1083, 1088 (7th Cir.1999); cf. Barrow v. Falck, 11 F.3d 729 (7th Cir.1993). The twin goals of this doctrine are to ensure that the parties marshal all of their facts and arguments so that a dispute may be resolved in one pass, and to conserve judicial resources. Treating new arguments as grounds for a second decision would contradict both rationales and in practical effect abandon the doctrine.

On this third appeal, and with the assistance of a third legal team, that is exactly what Peoples asks us to do. He contends that we should abrogate the doctrine of law of the case for all collateral proceedings and allow defendants freely to raise the same issue twice, once on direct appeal and again on collateral review, even if the law is unchanged and no new facts have come to light. He relies principally on two decisions: Massaro v. United States, 538 U.S. 500, 123 S. Ct. 1690, 155 L. Ed. 2d 714 (2003), and United States v. Galloway, 56 F.3d 1239 (10th Cir.1995) (en banc). Galloway is incompatible with this circuit's law (and with the Supreme Court's as well), and Massaro is irrelevant. But we begin with first principles.

Taglia did not invent the rule that a person who has raised an issue, and had it resolved by a federal court, cannot start from scratch on collateral review and ask the judiciary to proceed as if the first resolution had not occurred. This is a longstanding rule of federal practice. See, e.g., Himely v. Rose, 9 U.S. (5 Cranch) 313, 3 L. Ed. 111 (1809); Roberts v. Cooper, 61 U.S. (20 How.) 467, 481, 15 L. Ed. 969 (1857); Messinger v. Anderson, 225 U.S. 436, 444, 32 S. Ct. 739, 56 L. Ed. 1152 (1912). True enough, issue preclusion and claim preclusion (collateral estoppel and res judicata) do not apply on collateral review; if they did, there would be no collateral review. See Salinger v. Loisel, 265 U.S. 224, 44 S. Ct. 519, 68 L. Ed. 989 (1924); Wong Doo v. United States, 265 U.S. 239, 44 S. Ct. 524, 68 L. Ed. 999 (1924). But federal courts employ the more flexible doctrine of law of the case even when rules of preclusion do not govern. See Christianson v. Colt Industries Operating Corp., 486 U.S. 800, 815-18, 108 S. Ct. 2166, 100 L. Ed. 2d 811 (1988). This is true in collateral proceedings as well as in normal civil litigation.

Sanders v. United States, 373 U.S. 1, 83 S. Ct. 1068, 10 L. Ed. 2d 148 (1963), restated these common-law norms, which have been modified by the Antiterrorism and Effective Death Penalty Act for successive collateral reviews but left in place for the initial round. (Although Sanders dealt with successive rounds of collateral proceedings, the Court has held that its approach applies equally when the same issue is raised on direct appeal and again on an initial round of collateral review. See Davis v. United States, 417 U.S. 333, 342, 94 S. Ct. 2298, 41 L. Ed. 2d 109 (1974).) The Court held in Sanders that an initial federal determination controls in subsequent rounds of review if "(1) the same ground presented in the subsequent application was determined adversely to the applicant on the prior application, (2) the prior determination was on the merits, and (3) the ends of justice would not be served by reaching the merits of the subsequent application." 373 U.S. at 15, 83 S. Ct. 1068. What Davis, Taglia and other decisions, such as Ryan and United States v. Mazak, 789 F.2d 580 (7th Cir.1986), add to Sanders is that the "ends of justice" do support relitigation when the substantive law has changed or when evidence that could not have been discovered earlier despite diligent inquiries at last comes to light. (The AEDPA codifies a similar standard for relitigation. See 28 U.S.C. § 2244(b)(2).) See also Dobbs v. Zant, 506 U.S. 357, 113 S. Ct. 835, 122 L. Ed. 2d 103 (1993) (suggesting that the undeveloped-record exception to the common-law doctrine may be limited to capital cases).

Peoples presented ineffective-assistance arguments in his two prior appeals. That ground was resolved adversely to him on the merits, twice. Although Peoples now wants to present new instances of supposed shortcomings, ineffective assistance of counsel is a single ground for relief no matter how many failings the lawyer may have displayed. Counsel's work must be assessed as a whole; it is the overall deficient performance, rather than a specific failing, that constitutes the ground of relief. See Bell v. Cone, 535 U.S. 685, 697, 122 S. Ct. 1843, 152 L. Ed. 2d 914 (2002); Strickland v. Washington, 466 U.S. 668, 690, 104 S. Ct. 2052, 80 L. Ed. 2d 674 (1984); Holman v. Gilmore, 126 F.3d 876, 881-84 (7th Cir.1997).

That point was made with more generality in Sanders: "By `ground' we mean simply a sufficient legal basis for granting the relief sought by the applicant. For example, the contention that an involuntary confession was admitted in evidence against him is a distinct ground for federal collateral relief. But a claim of involuntary confession predicated on alleged psychological coercion does not raise a different `ground' than does one predicated on alleged physical coercion. In other words, identical grounds may often be proved by different factual allegations." 373 U.S. at 16, 83 S. Ct. 1068. See also Brannigan v. United States, 249 F.3d 584, 587 (7th Cir.2001), for a similar definition of a "claim" in the law of collateral review. Just as one who makes and loses a contention that a confession is involuntary because of physical coercion cannot start over by adding an allegation of psychological coercion, one who makes and loses a contention that counsel was ineffective for four reasons cannot start over by choosing four different (or four additional) failings to emphasize.

Let us turn, then, to the question whether Massaro supports a different understanding. The Court held in Massaro that a defendant never forfeits a claim of ineffective assistance by waiting until collateral attack. It disapproved our contrary decision in Guinan v. United States, 6 F.3d 468 (7th Cir.1993), which had held that a defendant who has new counsel on appeal, and who does not contend that any material outside the trial record is essential to evaluating the claim, must raise it at once or forever lose the opportunity to do so. But Massaro does not hold that the same ground of relief may be raised twice, once on direct appeal and again on collateral review. The question was left open. 538 U.S. at 508-09, 123 S. Ct. 1690. Since Massaro we have reiterated that a defendant who chooses to make an ineffective-assistance argument on direct appeal cannot present it again on collateral review. See, e.g., Fuller v. United States, 398 F.3d 644, 649 (7th Cir.2005); Harris v. United States, 366 F.3d 593, 595 (7th Cir.2004). Cf. White v. United States, 371 F.3d 900 (7th Cir.2004) (defendant who presents an argument on direct appeal, and therefore cannot present it in an initial collateral proceeding, cannot present the same ground on second or successive collateral proceeding either).

Now if the reason why Peoples made an ineffective-assistance claim on direct appeal was that our decision in Guinan compelled him to do so before essential support could be assembled and placed in the record, then Massaro would be a good reason to allow him to press the argument now. Peoples does not make such a contention, however. Guinan itself, and successors such as Duarte v. United States, 81 F.3d 75 (7th Cir.1996), allowed ineffective-assistance claims to be postponed when delay was needed to assemble evidence. Peoples does not contend that Guinan compelled him to present the argument prematurely; instead he was chomping at the bit to get this issue before us, which is why he made three motions for a new trial and filed a second appeal. Nothing in Guinan required that procedure. The earlier presentations were by choice rather than under compulsion, so Massaro does not open a third window of opportunity.

Galloway shows that the tenth circuit sees this issue differently. It did so because the judges thought it too labor-intensive to determine whether the initial decision is conclusive, when the defendant always can come up with reasons — newly discovered evidence, supposed changes of law, protests about the wisdom of appellate counsel's selection of issues — to avoid the effect of the initial decision. Instead of stewing about procedural doctrines, the tenth circuit concluded, judges should tackle the merits directly. Galloway did not attempt to reconcile this approach with Davis and Sanders, which the decision does not cite, and did not consider how permitting relitigation affects incentives in the initial appeal. When a court decides not to enforce rules that limit relitigation, it not only reduces the litigants' incentives to get things right the first time but ensures that judges will encounter the same claim repeatedly. Although it may be extra work to enforce rules against relitigation in a given case, doing so has valuable effects for many future cases in which litigation will be conducted more expeditiously and with better information. Galloway therefore does not persuade us to abandon the doctrine of law of the case in federal collateral review under § 2255. The tenth circuit stands alone on this subject.

This is not to say that Galloway misunderstood the extent of legal ingenuity. In the district court Peoples insisted that his prior appellate lawyer had furnished ineffective assistance by contending that his trial lawyer furnished ineffective assistance. The idea is that, if representation on direct appeal was indeed constitutionally deficient, then Peoples would be entitled to present his claims anew with the assistance of competent lawyers. This line of argument overlooks the fact that on his second direct appeal Peoples dismissed his lawyers and chose to represent himself. One who exercises the right of self-representation cannot contend that he received ineffective assistance of counsel. Faretta v. California, 422 U.S. 806, 834 n. 46, 95 S. Ct. 2525, 45 L. Ed. 2d 562 (1975). On that second appeal Peoples presented, and we resolved on the merits, an ineffective-assistance claim. On this third appeal Peoples has another team of lawyers, and they have dropped any challenge to the adequacy of prior appellate counsel. The only argument is that Peoples received ineffective assistance at trial. That argument has been considered and rejected before; it is no longer open.