WOOD, Circuit Judge.

William and Joan Kerr had been married almost a decade when Joan struck up a romantic relationship with Anthony Graff. (We will refer to William Kerr by his last name and to Joan Kerr by her first to keep the actors clear.) The affair between Joan and Graff started in the summer of 1999, but it did not last long. On August 8th of that year, the Kerrs spent the day drinking with friends, first at the Wisconsin State Fair and later at their home. During the day, Joan met with Graff in secret at the fair. That evening, Graff decided to call the Kerrs' house to speak with Joan. Kerr picked up the phone, however, and the conversation was hostile. Ultimately, Graff decided to pay Kerr a visit.

Upon learning that Graff was on his way over, Kerr loaded a 12-gauge shotgun and placed it within reach in the kitchen. Less than an hour later, Graff appeared at the Kerrs' back door and told Kerr that they needed to talk. Graffs arrival was followed almost immediately by the appearance of officers from the Wauwatosa (Wisconsin) Police Department. A neighbor had summoned the police after spotting Graff as he sneaked up to the Kerrs' house. When the officers arrived, Kerr told them that Graff — who was by then sitting at the Kerrs' kitchen table — was a friend and sent them on their way. Kerr returned to the kitchen and, according to his version of events, had the following brief exchange with Graff:

Kerr: What's the problem?

Graff: I'm your fucking problem. I'm going to take your wife and you're not going to stop me.

Kerr: Over my dead body.

Graff: That will be easy enough.

At that point, Graff started to stand up. Kerr picked up his shotgun and shot Graff once in the abdomen, killing him instantly. Kerr called the police back to the house and turned himself in.

A Wisconsin jury convicted Kerr of first-degree intentional homicide in December 1999, and the state court sentenced him to life imprisonment with eligibility for parole after 21 years. After losing his direct appeal, Kerr filed a section 974.06 motion pro se in the state trial court. (A section 974.06 motion is the statutory equivalent of a petition for a writ of habeas corpus in Wisconsin. See Wis. STAT. § 974.06; Morales v. Boatwright, 580 F.3d 653, 656-57 (7th Cir. 2009) (discussing Wisconsin's post-conviction procedures).) The trial court denied Kerr's motion, the Wisconsin Court of Appeals summarily affirmed, and the Supreme Court of Wisconsin denied review.

Kerr then turned to federal court to challenge his conviction. Again acting pro se, he filed a petition for a writ of habeas corpus under 28 U.S.C. § 2254, in which he asserted that constitutional errors in his state-court proceedings justified relief. The district court denied the petition on the merits without an evidentiary hearing, but it issued a certificate of appealability on a number of Kerr's claims. Kerr is now before this court, where he is arguing that he is entitled to relief because his defense attorneys provided ineffective assistance of counsel. He advances two different theories in support of that position.

Kerr first contends that his lawyers were ineffective because they gave him incorrect information about the sentence that he would have faced if he had accepted a plea deal offered by the state before trial. According to Kerr, the prosecutors were prepared to allow him to plead guilty to first-degree reckless homicide, but his lawyers convinced him not to accept the deal when they incorrectly informed him that he would face 13 years to life for that offense. In fact, reckless homicide in Wisconsin is punishable by 0-40 years. Kerr asserts that he would readily have accepted the state's offer if he had known of the actual penalties associated with the different crimes.

The second theory that Kerr asserts relates to his lawyers' failure to present the defense of adequate provocation at the trial. He points to the events in the days and hours leading up to the shooting and argues that Graff did more than enough to provoke the attack that killed him. Had the jury been persuaded that provocation existed, this would have reduced his crime from first- to second-degree intentional homicide.

We review the district court's decision to deny Kerr's petition for a writ of habeas corpus de novo and any of its factual findings for clear error. Ebert v. Gaetz, 610 F.3d 404, 411 (7th Cir. 2010). Ken-would like us either to reinstate the plea deal that he says was offered by the state or to grant a new trial. In the alternative, Kerr asks for an evidentiary hearing in the district court so that he can develop his claims. We conclude that Kerr is not entitled to relief on his adequate provocation theory, but we find that a hearing is necessary before his plea bargain claim can be resolved.

I

We begin with Kerr's argument that his trial lawyers were ineffective for failing to present an adequate provocation defense at trial. We agree with both parties and the district court that the state trial court — the Circuit Court of Milwaukee County — ruled on the merits of this argument. Accordingly, we evaluate the state court's decision under the Antiterrorism and Effective Death Penalty Act of 1996 ("AEDPA"). See 28 U.S.C. § 2254(d); Harrington v. Rickter, ___ U.S. ___, 131 S.Ct. 770, 780, 178 L.Ed.2d 624 (2011); Ylst v. Nunnemaker, 501 U.S. 797, 802-04, 111 S.Ct. 2590, 115 L.Ed.2d 706 (1991); George v. Smith, 586 F.3d 479, 484-485 (7th Cir. 2009). The statute provides that we may grant relief to Kerr only if the state court's decision was contrary to or an un-reasonable application of clearly established Supreme Court precedent, or if it was based on an unreasonable determination of the facts in light of the evidence. 28 U.S.C. § 2254(d)(1) (2); Williams v. Taylor, 529 U.S. 362, 404-06, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). In this case, the applicable Supreme Court precedent is Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). To succeed on his claim that his lawyers provided constitutionally ineffective assistance, Kerr had to show deficient performance on the part of his attorneys and prejudice from that deficiency, meaning a substantial likelihood that the result would have been different. Id. at 687, 104 S.Ct. 2052; see also Richter, 131 S.Ct. at 791-92. Our responsibility is not to decide ourselves whether Strickland's standards were met; it is instead to decide whether the state courts were unreasonable when they concluded that Kerr's ineffective assistance claim could not prevail. Richter, 131 S.Ct. at 785. Here, since the state courts correctly identified Strickland as the governing law, we have only to decide "whether the state court's application of clearly established federal law was objectively unreasonable." Williams, 529 U.S. at 409, 120 S.Ct. 1495; see also Raygoza v. Hulick, 474 F.3d 958, 963 (7th Cir. 2007).

A few additional legal principles inform our resolution of that question. First, we do not as a rule second-guess counsel's strategy. Instead, judicial scrutiny of a lawyer's performance must be highly deferential, and courts assessing counsel's performance generally presume that decisions at trial fall within "the `wide range' of reasonable professional assistance." Richter, 131 S.Ct. at 787 (quoting Strickland, 466 U.S. at 689, 104 S.Ct. 2052). "The law does not require counsel to raise every available nonfrivolous defense." Knowles v. Mirzayance, 556 U.S. Ill, 129 S.Ct. 1411, 1422, 173 L.Ed.2d 251 (2009). Moreover, even if a lawyer's decision to omit a defense falls below this forgiving performance standard, Strickland's requirement that prejudice be shown means that the defendant must make a case that there is a reasonable probability — one sufficient to undermine our confidence in the outcome — that the result of the proceeding would have been different if the defense had been presented. 466 U.S. at 694, 104 S.Ct. 2052; see also Long v. Krenke, 138 F.3d 1160, 1164 (7th Cir. 1998) ("[T]he mere possibility of success based on a defense for which there existed little or no evidentiary support is not enough to establish constitutionally in-adequate counsel.").

Whether Kerr's lawyers should have put on the defense and whether it might have succeeded also depend in part on Wisconsin law. As we have mentioned, a first-degree intentional homicide in Wisconsin is mitigated to second-degree intentional homicide when death is caused under the influence of adequate provocation. WIS. STAT. §§ 940.01(2)(a) and 939.44. "Provocation" is defined as "something which the defendant reasonably believes the intended victim has done which causes the defendant to lack self-control completely at the time of causing death," id. § 939.44(1)(b); and provocation is "adequate" when it is "sufficient to cause complete lack of self-control in an ordinarily constituted person," id. § 939.44(1)(a). The adequacy requirement is the objective component of the defense; it requires proof of "such mental disturbance, caused by a reasonable, adequate provocation as would ordinarily so overcome and dominate or suspend the exercise of the judgment of an ordinary man, as to render his mind for the time being deaf to the voice of reason: make him incapable of forming and executing that distinct intent to take human life. . . ." Johnson v. State, 129 Wis. 146, 108 N.W. 55, 60-61 (1906). Provocation, on the other hand, is the subjective part of the defense. A defendant must show that the events in question did in fact produce the required mental disturbance at the time of the homicide. State v. Williford, 103 Wis.2d 98, 307 N.W.2d 277, 284 (1981).

The state trial court addressed the question of adequate provocation when it ruled on Kerr's section 974.06 motion. In that motion, Kerr explained that he shot Graff because of the turmoil caused by Graffs ongoing affair with Joan, the threats of violence Graff conveyed in the phone calls leading up to the killing, and Graffs threat physically to remove Joan from the Kerrs' house once he had arrived there. Kerr wrote, "Certainly, one would expect a man to defend himself, family, and home by any means necessary." The state court was unimpressed. It concluded that there was no evidence that would have justified a jury instruction on the adequate provocation defense, and it implied that this was the reason why Kerr's assertion that his lawyers were ineffective had little force. The court wrote:

No reasonable interpretation of the evidence would lead to the conclusion that Mr. Kerr completely lost self-control. To the contrary, he loaded and placed his weapon in a convenient place before the victim entered his home, he permitted the victim to enter and he dealt calmly with the police only minutes before the shooting. And it appears that all of these actions came after potentially provocative revelations that his wife and the victim were having an affair.

The court thought that nothing in the brief exchange we have described between Kerr and Graff after the police had left the house amounted to objectively adequate provocation. The district court agreed with this view of the evidence, concluding that there was nothing to "show a high degree of exasperation, rage, or anger" and that "Kerr was calm just moments before shooting Graff, and their verbal exchange was not such that it could reasonably have been considered adequate provocation for Kerr's action."

We agree with the district court that the state court's resolution of this issue represents a reasonable application of Strickland. A fairminded jurist could readily conclude that Kerr satisfied neither the objective nor the subjective component of the adequate provocation defense, and that as a result counsel's decision not to put on the defense fell within the wide range of professionally competent assistance. The state court's decision is thus consistent with the Supreme Court's decisions in this area and Kerr is not entitled to relief. Richter, 131 S.Ct. at 786. In particular, Kerr's decision to inform the police that Graff was his friend and to send them away when they came to help severely undermines any argument that he was laboring under a mental disturbance that interfered with rational thought. Kennow argues that he had no time to "cool down" between the time that he began to argue with Graff on the telephone and the time that he killed Graff in his kitchen. But the visit by police provided an ideal opportunity to defuse the situation. Moreover, it is a stretch to say that Graff, who was unarmed, adequately provoked Kerr during the brief exchange after the police had left. The state suggests that Kerr's lawyers made a strategic decision to omit the adequate provocation defense so that they would not undermine other defenses that they presented at trial. Perhaps so. Or perhaps they merely came to the conclusion that the state judge reached later: the facts simply did not support the defense.

Kerr's efforts to avoid this conclusion focus on what he sees as the broader context of the killing. Under Wisconsin law, the question whether provocation is adequate — whether the objectively reasonable person would have lost control — is typically limited to those events that immediately precede the killing. Kerr urges us not to be too literal about the concept of immediacy. Instead, he wants us to look beyond the "final provocatory incident" to the days and months leading up to the crime. Graffs affair with Joan tormented him over a period of a few months, he says, and then Graff made several violent threats during the phone calls earlier in the evening. He points to a series of Wisconsin cases that stand for the proposition that "actions over a long period of time can have . . . cumulative effect upon any ordinary person so that the provocation just before the shooting would be greatly magnified." State v. Lucynski, 48 Wis.2d 232, 179 N.W.2d 889, 890 (1970) (internal quotation marks omitted); see also State v. Felton, 110 Wis.2d 485, 329 N.W.2d 161, 172-73 (1983); State v. Hoyt, 21 Wis.2d 284, 128 N.W.2d 645, 649 (1964). These cases represent a well-defined exception to the normal rules of heat-of-passion defenses: a court may consider a pattern of prior conduct in circumstances where the defendant is a battered spouse who suffers persistent abuse that culminates in a killing. Kerr, however, wants an extension of that principle — one that the state court was unwilling to adopt. (Kerr did not even know Graffs name until after Graff was dead.) Federal law did not compel the state court to take that step.

Before turning to Kerr's other argument, it is worth mentioning that Kerr's trial lawyers argued to the jury that Kerr acted in self-defense and that his actions were reckless rather than intentional. The trial court instructed the jury not only on first-degree intentional homicide, but also on second-degree intentional homicide, first-degree reckless homicide, second-degree reckless homicide, and self-defense. The jury's decision to convict Kerr of first-degree intentional homicide rather than any of these other offenses reinforces the state court's conclusion that Kerr could not have benefitted from the adequate provocation defense. We conclude that Kerr's argument based on the lawyers' decision not to raise the adequate provocation defense falls far short of clearing the hurdle imposed by § 2254(d)(1).

II A

That leaves Kerr's claim that his lawyers provided ineffective assistance of counsel when they gave him bad advice that caused him to forego a favorable plea deal offered by the state and instead proceed with a trial. Before we consider whether Kerr's claim warrants relief, we must decide whether he may now present the argument, given the litigation that has occurred thus far in both state and federal court.

The state advances two alternative arguments to respond to Kerr's claim. First, it says that Kerr has procedurally defaulted this point. see, e.g., Ward v. Jenkins, 63 F.3d 692, 696-97 (7th Cir. 2010). Federal habeas corpus relief from a state conviction is not available when the decision rests on a state procedural ground independent of any federal issue and adequate to support the judgment. Coleman v. Thompson, 501 U.S. 722, 729-30, 111 S.Ct. 2546, 115 L.Ed.2d 640 (1991). As the state sees it, the state trial court rejected Kerr's argument because it was fatally underdeveloped, and this, the state argues, constitutes an adequate and independent state ground. In the alternative, the state continues, one could conclude that the state trial court denied Kerr's plea bargain claim on the merits. If so, then our review of this claim, like the adequate provocation claim discussed above, is governed by AEDPA. This alternative argument concludes that Kerr has not shown that the state court's rejection of his argument was an unreasonable application of Strickland. The state's two arguments are mutually exclusive: either the state court refused to decide Kerr's claim because of his failure properly to present it to the court, or it did decide the claim adversely to Kerr on the merits. See Ylst, 501 U.S. at 802-04, 111 S.Ct. 2590. We must decide which characterization better describes the court's rationale.

It is helpful to begin by examining in more detail the motion for post-conviction relief that Kerr filed in the state trial court. Kerr's motion says that his lawyers were ineffective because they gave him "inappropriate advise [ sic] on not excepting [ sic] the prosecutions [ sic] plea bar-gain just before trial. . . ." The nature of the inappropriate advice is unclear from Kerr's pro se pleading; as far as we can tell, he suggests that his lawyers were incorrect in their belief that they could defend the case, and so they encouraged Kerr not to plead guilty because of their faulty evaluation of feasible defenses. Kerr's papers leave no doubt, however, that he asserted in his state post-conviction motion that he would have negotiated a plea rather than going to trial if his lawyers had given him accurate information and that his lawyers' mistake led to a violation of his Sixth Amendment right to effective assistance of counsel. (We note that we see no problem here, in light of the developed argument in Kerr's state post-conviction motion, with fair presentment to the state court, Baldwin v. Reese, 541 U.S. 27, 30-31, 124 S.Ct. 1347, 158 L.Ed.2d 64 (2004); the record satisfies us that Kerr is essentially presenting to us the same "factual and legal bases" that he advanced in the state courts, Curtis v. Montgomery, 552 F.3d 578, 583 (7th Cir. 2009). As we observed recently, "a petitioner may reformulate his claims somewhat, so long as the substance of his argument remains the same." Ward, 613 F.3d at 697 (internal quotation marks omitted).)

At the conclusion of its opinion denying Kerr relief, the state court called his argument "a mish-mash of allegations" and rejected it, saying:

Mr. Kerr's allegations are just too underdeveloped for me to rule on. His allegations beg a number of questions that I cannot answer on the record before me, and which must be answered before I could determine whether there is a reasonable probability that Mr. Kerr would have decided to plead guilty had he been advised differently. [After listing examples of such questions, the court continued:] Even if the State offered to resolve the case on a lesser charge, such as first degree reckless homicide, Mr. Kerr might still be serving a sentence which would make him ineligible for parole until 2021, in which case Mr. Kerr would not be able to demonstrate that he was prejudiced by going to trial.

When, in order to decide issues, a court must first develop them, the court may decline to rule on them. State v. Pettit, 171 Wis.2d 627, 646-47 [ 492 N.W.2d 633] ([Wis.]Ct.App. 1992). Accordingly, I reject the final claim Mr. Kerr makes in his motion.

(Emphasis added.) At oral argument in this court, the state took the position that the language we have emphasized shows that the state court resolved the merits of Kerr's plea bargain theory.

We would be doing state courts in general a disservice if we began picking individual sentences or phrases out of a longer explanation and assuming that these isolated phrases represent the court's view. And even if that tactic were available here, we do not read the language the state has singled out as a ruling on the merits. The statement is perfunctory and uses the conditional mood. It reads as if it were another example of a question (How much time might Kerr have actually gotten?) that the state court thought had to be answered before it could decide the merits. In addition, as we discuss in a moment, taken literally the state court's remark incorrectly describes state law. That indicates to us that it was nothing but a casual comment, rather than a considered ruling on the merits of Kerr's argument. "[W]e hesitate to apply a stricter standard of review [that of § 2254(d)] without a clearer indication that Wisconsin fully considered [the petitioner's] claim on the merits." George, 586 F.3d at 485.

We conclude that the brief comment on which the state is now relying does not reflect an alternative ruling on the merits. Instead, the state court's decision to reject Kerr's claim because he failed to develop it in a manner that complied with state procedural rules "`fairly appears' to rest primarily on state law." Coleman, 501 U.S. at 740, 111 S.Ct. 2546. This conclusion is buttressed by the fact that the only legal authority mentioned by the state court was State v. Pettit, 171 Wis.2d 627, 492 N.W.2d 633 (App. 1992). Pettit describes Wisconsin statutes that permit state courts to ignore a claim that has not properly been developed. Id. at 646-47, 492 N.W.2d 633. We recently ob-served that "Wisconsin waiver law . . . constitutes an adequate and independent state law ground" barring federal habeas corpus, Promoter v. Pollard, 628 F.3d 878, 886-S7 (7th Cir. 2010); see also Gray v. Hardy, 598 F.3d 324, 329 (7th Cir. 2010) (discussing Illinois waiver rules and federal habeas corpus review), and we see no reason to treat Wisconsin's rules about proper development of legal claims differently. Just as in federal court, a party's failure adequately to develop an argument has consistently been a reason to reject claims advanced by litigants in the Wisconsin state courts. That the state court might have reached the merits of Kerr's claim and chose not to do so makes no difference in the adequate-and-independent ground analysis. Walker v. Martin, ___ U.S. ___, 131 S.Ct. 1120, 1125, 179 L.Ed.2d 62 (2011) (citing Beard v. Kindler, ___ U.S. ___, 130 S.Ct. 612, 175 L.Ed.2d 417 (2009)). The record leaves no doubt that the state court's decision rested on this independent and adequate procedural ground.

Nothing in the Supreme Court's decision in Richter changes this aspect of habeas corpus law. As we have mentioned, the Wisconsin Court of Appeals summarily affirmed the trial court's decision without further discussion of Kerr's claim that his lawyers caused him to forego a plea in favor of trial (and the Supreme Court of Wisconsin later denied review). The important question after Richter is whether the state appellate court's summary decision signals that Kerr's claim was "adjudicated on the merits" by the state court. See 131 S.Ct. at 783-S5. In our view, it does not. Richter did not say that all summary affirmances in the course of state-court proceedings are rulings on the merits. To the contrary, the Richter Court wrote, "When a federal claim has been presented to a state court and the state court has denied relief, it may be presumed that the state court adjudicated the claim on the merits in the absence of any indication or state-law procedural principles to the contrary." Id. at 784-35 (emphasis added). That presumption leaves intact the Court's prior instructions in Ylst, which established rules for cases in which a federal court is confronted with a series of state-court judgments issued over the course of a prisoner's collateral proceedings in more than one tribunal:

Where there has been one reasoned state judgment rejecting a federal claim, later unexplained orders upholding that judgment or rejecting the same claim rest upon the same ground. If an earlier opinion "fairly appear[s] to rest primarily upon federal law," Coleman, 501 U.S. at 740 [ 111 S.Ct. 2546] we will presume that no procedural default has been invoked by a subsequent unexplained order that leaves the judgment or its consequences in place. Similarly where, as here, the last reasoned opinion on the claim explicitly imposes a procedural default, we will presume that a later decision rejecting the claim did not silently disregard that bar and consider the merits.

501 U.S. at 803, 111 S.Ct. 2590 (alteration in original); see also Richter, 131 S.Ct. at 785 (citing Ylst, 501 U.S. at 803, 111 S.Ct. 2590).

The Court in Richter had before it a summary order of the California Supreme Court denying without comment a state prisoner's request for a writ of habeas corpus. 131 S.Ct. at 783. California's system of collateral review is unique among the states in that all California courts have original jurisdiction in habeas corpus proceedings. See Martin, 131 S.Ct. at 1125-26. A prisoner petitioning for a writ of habeas corpus in a particular California court — whether trial, appellate, or the California Supreme Court — deals only with the court in which the petition is filed; in California, "`no appeal lies from the denial of a petition for writ of habeas corpus.'" Id. at 1125 (quoting In re Clark, 5 Cal.4th 750, 21 Cal.Rptr.2d 509, 855 P.2d 729, 740 n. 7 (1993)). Thus, the order issued by the California Supreme Court and considered in Richter did not affirm (or decline to review) a lower court's judgment; it denied a petition filed by Richter directly with the state supreme court. The summary affirmance entered by the Wisconsin Court of Appeals in Kerr's case is different. It affirmed a lower court's decision denying Kerr relief on procedural grounds. In Kerr's situation, unlike Richter's, the state court's decision plainly states that state-law procedural principles are at play, and Ylst instructs that we should read the appellate court's summary decision as an affirmation of the trial court's procedural disposition. That is not to say that Richter's rule is reserved for California. Of course not. If, for example, a Wisconsin trial court summarily rejects a prisoner's claim without explanation, after which the state appeals court similarly denies relief in a summary order, a situation comparable to the one in Richter would exist. There, in the absence of any other clue, Richter and Ylst indicate that the default rule is to treat it as a decision on the merits. That is not our case, however. We can and must proceed in light of our conclusion that the Wisconsin courts rejected Kerr's claim on procedural grounds.

B

Whether Kerr's procedural default bars federal review is complicated by the fact that the state first presented its procedural default argument in this court, and then only in a single paragraph of its brief. The application of the independent and adequate state grounds doctrine does not affect federal jurisdiction; it exists instead to promote federalism and judicial economy. Kaczmarek v. Rednour, 627 F.3d 586, 591-92 (7th Cir. 2010) (citing Trest v. Cain, 522 U.S. 87, 89, 118 S.Ct. 478, 139 L.Ed.2d 444 (1997)). It is thus possible that procedural default may be equitably excused, if the petitioner demonstrates cause for the default and resulting prejudice, or shows that the merits of the case must be discussed in order to avoid a miscarriage of justice. Perruquet v. Briley, 390 F.3d 505, 514 (7th Cir. 2004). Kerr argues that any procedural default he has committed should be forgiven under the exception for cause and prejudice, but we do not think it is necessary to reach that argument. What is important for the purposes of this case is that Trest holds that "procedural default is normally a defense that the State is obligated to raise and preserve if it is not to lose the right to assert the defense thereafter." 522 U.S. at 89, 118 S.Ct. 478 (citations, alterations, and internal quotation marks omitted).

As a general matter, a state must assert the defense in its answer to the federal petition. Kaczmarek, 627 F.3d at 591-92. We have recognized, however, that a state might successfully preserve a procedural default argument even if it does not raise the issue in its first response in the district court, see id. at 594 (entertaining the argument after it was raised and fully briefed for the first time in a motion for reconsideration in the district court); and there are limited circumstances where we may consider procedural default even though the state presents the defense for the first time in this court, see Perruquet, 390 F.3d at 516-19 (reaching the state's argument because the petitioner had failed to present his constitutional claim to any state court and because the state had not engaged the merits of that argument in the district court); see also Day v. McDonough, 547 U.S. 198, 206-07, 126 S.Ct. 1675, 164 L.Ed.2d 376 (2006). We must therefore decide whether the normal rule requiring the state to raise the defense prevails here, or if this is one of the unusual cases in which the state's own forfeiture should be overlooked.

Here, the state ignored Kerr's default until past the eleventh hour. It addressed Kerr's plea bargaining argument on the merits in both its answer to Kerr's federal petition and in a brief later filed in the district court. Although the state asserted that the district court should reject some of Kerr's arguments as procedurally defaulted, Kerr's plea bargain theory was not on the list. Only after the district court said in its opinion that it had decided to reach the merits of Kerr's plea bargaining issue because the state failed to raise its defense, and the case was before this court, did the state consider procedural default. Even then, the state hardly touched on the point. It devoted a scant half-page to the issue in its responsive brief and referred to no legal authority in support of its argument. Federal courts, just like their Wisconsin counterparts, consider such poorly developed arguments forfeited on appeal, see, e.g., Judge v. Quinn, 612 F.3d 537, 557 (7th Cir. 2010), and there is no exception that applies to the state's argument here. The state trial court flagged Kerr's plea bargain theory as underdeveloped, and so it would have been easy for the state to have argued procedural default. It did not do so in the district court, and its argument in this court was skeletal. Under the circumstances, we conclude that the default was forfeited. Whether because this is the result that the state wanted at some point during this litigation or because of the normal operation of the rules of forfeiture, the state's forfeiture means that we may move on to the merits of Kerr's plea bargain theory.

III A

Our conclusion that the state courts never spoke to the merits of Kerr's plea bargaining claim has consequences for the standard of review we apply. As Richter reiterated, section 2254(d) bars re-litigation of any "claim" that the state court "adjudicated on the merits," subject to the exceptions found in § 2254(d)(1) and § 2254(d)(2). 131 S.Ct. at 784. That is the reason that we reviewed under the terms of 28 U.S.C. § 2254(d)(1) the state trial court's conclusion that Kerr's trial attorneys provided constitutionally acceptable representation even though they did not raise an adequate provocation defense. The state trial court found meritless Kerr's claim that the adequate provocation defense should have been presented, and we asked whether that decision was an unreasonable application of Supreme Court precedent. But the deferential standard of section 2254(d) applies only when a state court has adjudicated the federal claim "on the merits." Canaan v. McBride, 395 F.3d 376, 382-83 (7th Cir. 2005). As we have already noted, Richter recognizes a default rule that a state court decision rejecting a federal claim without comment is a decision on the merits. 131 S.Ct. at 783-85. But if the state court issues a summary order that denies a petition for post-conviction relief and in the order it furnishes a procedural reason for its decision (whether in a lengthy discussion or through a brief reference to the governing law), then we must take the state court at its word and treat the decision as procedural, not merits-based. See Martin, 131 S.Ct. at 1124. We have said that "[a]n adjudication on the merits is perhaps best understood by stating what it is not: it is not the resolution of a claim on procedural grounds." Muth v. Frank, 412 F.3d 808, 815 (7th Cir. 2005); see also Sellan v. Kuhlman, 261 F.3d 303, 311 (2d Cir. 2001) ("`Adjudicated on the merits' has a well settled meaning: a decision finally resolving the parties' claims, with res judicata effect, that is based on the substance of the claim advanced, rather than on a procedural, or other, ground."). Richter leaves this dichotomy in place, while at the same time it elaborates on the proper treatment of silence from the state court.

As Kerr's plea bargain claim reaches us — by means of a confusing presentation in the state court that was deemed a procedural default, followed by the state's forfeiture of whatever default occurred — the one thing that is clear is that no state court has squarely addressed the merits. In these circumstances, we review Kerr's plea bargain claim under the pre-AEDPA standard of review set out in 28 U.S.C. § 2243. see, e.g., Choker v. Crogan, 428 F.3d 1215, 1220-21 (9th Cir. 2005); cf. Johnson v. Thurmer, 624 F.3d 786, 789-91 (7th Cir. 2010); George, 586 F.3d at 484-85; Braun v. Powell, 227 F.3d 908, 916-17 (7th Cir. 2000). Under that standard, we are to "dispose of the matter as law and justice require," 28 U.S.C. § 2243, which we have interpreted to mean that we review the petitioner's constitutional claim with deference to the state court, but ultimately de novo. Richter, 131 S.Ct. at 788 ("Even under de novo review, the standard for judging counsel's representation is a most deferential one."); McGee v. Bartow, 593 F.3d 556, 572 n. 10 (7th Cir. 2010).

As we assess Kerr's claim, we first consider his argument as if there were no dispute about the factual allegations he makes; later we return to the evidentiary problems. According to Kerr, the district attorney offered him a deal before jury selection began that would have allowed him to plead guilty to the lesser crime of first-degree reckless homicide. Kerr's lawyers, he asserts, told him that reckless homicide carried a maximum sentence of life imprisonment and a 13-year minimum term. This was wrong. In fact, the law in effect at the time made reckless homicide a Class B felony, punishable by a prison term between zero and 40 years. Wis. STAT. §§ 940.02(1) and 939.50(3)(b) (1997-1998). A person convicted of that crime would have been eligible for parole after serving six months or 25% of a sentence, whichever was greater, id. § 304.06(1)(b), and for "presumptive mandatory release" — release that the state parole commission can deny only in enumerated circumstances — after two-thirds of the sentence, id. § 302.11(1) and (lg). First-degree intentional homicide, on the other hand, carried a mandatory term of life imprisonment, id. §§ 940.01(1) and 939.50(3)(a), with eligibility for parole after a minimum of 20 years, id. § 304.06(1)(b), and no presumptive mandatory release at all, id. § 302.11(1m). Kerr says that if he had known the truth about how much time he would have to serve after pleading guilty, he would have entered a plea on the spot rather than facing a jury trial on charges of intentional homicide.

As things unfolded, Kerr did go to trial, and he received a life sentence in January 2000, with eligibility for parole after 21 years. The state court assumed that Ken was not prejudiced by a trial, because he "might still be serving a sentence which would make him ineligible for parole until 2021" even if he had pleaded guilty to first-degree reckless homicide. But the statutes we have just described demonstrate that this assumption was based on a faulty premise. If Kerr had pleaded guilty to first-degree reckless homicide and been sentenced to the statutory maximum of 40 years, he would have been eligible for pa-role after 10 years — in 2010, not 2021. The district court made a related mistake. It thought that Kerr had not suffered the prejudice required by Strickland, even if his lawyers had performed deficiently. The court reasoned that because Kerr was eligible for parole after 21 years and because a guilty plea would have resulted in a term of 27 years, it was not clear that "the difference in penalty would have been the determinative factor in accepting or rejecting the plea offer." This was a comparison of apples and oranges: the district court contrasted Kerr's actual parole eligibility date pursuant to the conviction for intentional homicide (the year 2021) with an estimate of what his presumptive release date would have been after a guilty plea for reckless homicide (2026, if Kerr had pleaded guilty and received the maximum term of imprisonment). After his trial, Kerr received a life sentence as required by statute, and he was eligible for parole after 21 years; under his conviction for intentional homicide, he was not eligible for presumptive mandatory release under any circumstance. Had he pleaded guilty, the worst-case scenario was a 40-year term, eligible for parole after 10 years, with presumptive release in just under 27 years. The state court's error of state law, of course, is not grounds for granting federal relief. Swarthout v. Cooke, ___ U.S. ___, 131 S.Ct. 859, 861, 178 L.Ed.2d 732 (2011). We describe the miscalculations in detail only to point out that Kerr has identified how his alleged loss of an opportunity to plead guilty (if it occurred) almost certainly prejudiced him.

B

Hill v. Lockhart holds that Strickland's two-step framework applies "to challenges to guilty pleas based on ineffective assistance of counsel." 474 U.S. 52, 58, 106 S.Ct. 366, 88 L.Ed.2d 203 (1985); see also Premo v. Moore, ___ U.S. ___, 131 S.Ct. 733, 743, 178 L.Ed.2d 649 (2011); Padilla v. Kentucky, ___ U.S. ___, 130 S.Ct. 1473, 1480-81, 176 L.Ed.2d 284 (2010) ("Before deciding whether to plead guilty, a defendant is entitled to `the effective assistance of competent counsel.'") (quoting McMann v. Richardson, 397 U.S. 759, 771, 90 S.Ct. 1441, 25 L.Ed.2d 763 (1970)). In Hill, the Court considered a Strickland claim based on allegations that the petitioner's lawyer had given bad advice that caused him to plead guilty instead of proceeding to trial. While we have encountered many cases analogous to Hill, we have also understood that decision to establish a rule applicable to other circumstances where lawyers advise their clients at the plea-bargaining stage of the case. Cf. Padilla, 130 S.Ct. at 1485 n. 12. For instance, we have applied Hill to a case where a lawyer provides ineffective assistance of counsel when the defendant rejects a plea deal and proceeds to trial in the face of overwhelming evidence of guilt and lacking any viable defense. Tow v. Fairman, 940 F.2d 1065, 1068 (7th Cir. 1991). In addition, we have suggested (without deciding) that a defendant might demonstrate ineffective assistance in a case where bad advice caused him to take a plea agreement less favorable than a different deal he otherwise might have made. United States v. Parker, 609 F.3d 891, 894-95 (7th Cir. 2010). We have also decided a number of cases involving allegations that an attorney performed deficiently by giving erroneous advice that caused a defendant to reject a favorable plea deal and proceed instead with a trial. see, e.g., Paters v. United States, 159 F.3d 1043 (7th Cir. 1998); United States v. Golden, 102 F.3d 936 (7th Cir. 1996); Johnson v. Duckworth, 793 F.2d 898 (7th Cir. 1986).

But the question in the final analysis is not how we think Hill applies to differing factual settings; the constitutional standard in all of the separate phases of a criminal trial where the Sixth Amendment applies, including the point at which a defendant decides whether to plead guilty to a crime, requires only "reasonable competence in representing the accused." Moore, 131 S.Ct. at 742. The Supreme Court has recently granted certiorari to consider two cases that will provide additional guidance in this area. Missouri v. Frye, ___ U.S. ___, 131 S.Ct. 856, 178 L.Ed.2d 622 (2011) (10-444), presents the question whether a defendant who enters a valid plea may later claim that but for an error made by his lawyer he would have pleaded guilty on more favorable terms (the issue we touched upon in Parker, supra). In Lafler v. Cooper, ___ U.S. ___, 131 S.Ct. 856, 178 L.Ed.2d 622 (2011) (10-209), the Court will consider an issue that appears to affect Kerr's case directly: whether a person seeking habeas corpus relief from a state conviction may prevail when his lawyer deficiently advises him to reject a favorable plea bargain but the defendant is later convicted and sentenced pursuant to a fair trial. In both cases, the Court has instructed the parties to address what remedy, if any, a court should provide when a lawyer is constitutionally ineffective at the plea-bargaining phase but the defendant is later convicted and sentenced pursuant to constitutionally adequate procedures. See id. We have considered whether to hold this case pending the decision in Lafler, but in the interest of efficiency, we have decided not to do so. The record is not sufficiently developed to permit a decision in Kerr's case, if the Court rules such a case may proceed at all. We think it best to move forward now, recognizing that if the Court rules that the later trial erases any possible claim relating to potential plea bargains, then it is likely that Kerr's case will have to be dismissed at that time.

To show that his lawyers provided ineffective assistance at the plea-bargaining phase in violation of the Sixth Amendment, we turn again to Strickland's dual requirements of deficient performance and prejudice. In the plea-bargain context, Strickland requires us to assess whether counsel performed effectively, based strictly on what was known at the time. Moore, 131 S.Ct. at 741-42. When a plea bargain is involved, the prejudice inquiry turns not on the effect of the counsel's performance on the trial, but instead on the question "whether the deficient information was the decisive factor in a defendant's decision to plead guilty or to proceed to trial." Julian v. Bartley, 495 F.3d 487, 498 (7th Cir. 2007); see also Moore, 131 S.Ct. at 743-45 (describing the inquiry when the petitioner did plead guilty and is arguing that he would have insisted on a trial). The certainty brought to the criminal justice system by the plea bargain "make[s] strict adherence to the Strickland standard all the more essential when reviewing the choices an attorney made at the plea bargain stage." Moore, 131 S.Ct. at 741; id. at 745-46.

The district court thought that Kerr's lawyers had not performed deficiently, even if they had provided Kerr with incorrect information, because there was evidence that Kerr discussed the plea deal with his lawyers, considered the strength of his defense, and participated in the decision to plead guilty. But this version of events takes account of only part of the story. The content of those discussions is critical. If Kerr's lawyers told him (as he alleges) that he would face 13 years to life in prison as a result of pleading guilty to reckless homicide, they overstated the minimum sentence by 13 years and the maximum sentence by much more than that (depending on how long Kerr lives). This is a basic legal error that easily falls "`below an objective standard of reasonableness.'" Richter, 131 S.Ct. at 787 (quoting Strickland, 466 U.S. at 688, 104 S.Ct. 2052). If Kerr's allegations are correct — a big "if," to which we return shortly — we would have little trouble concluding that the deficient-performance part of Strickland's test has been established.

That brings us to the evidence that Kenhas presented in support of his plea bargain theory. There is no dispute that Kerr was offered a plea deal; nor is there any question that right before sentencing Kerr's attorneys said decisively that if he pleaded guilty he would face life imprisonment with a 13-year statutory minimum. All of that is reflected in a discussion among Kerr, the state trial judge, the district attorney, and Kerr's lawyers, in the judge's chambers right before sentencing in January 2000. The decisive nature of the attorney's prediction of a petitioner's sentence is relevant to our assessment of how influential that advice was in the petitioner's decision whether to plead guilty. See Julian, 495 F.3d at 498 ("The strength of an attorney's prediction . . . goes toward determining whether the attorney's statements were the decisive factor in the defendant's decision. . . .").

In his affidavit, Kerr asserts that he would have accepted the state's plea offer rather than going to trial if his attorneys had accurately conveyed the correct sentencing range carried by first-degree reckless homicide. Kerr acknowledges that his own sworn statement is not enough to show prejudice without some objective corroborating information. see, e.g., Bethel v. United States, 458 F.3d 711, 718 (7th Cir. 2006). But he has pointed to corroborating evidence. The district court found that one of the lawyers (Gerald Boyle) told Kerr that he would have to plead to life imprisonment with a 13-year statutory minimum. This is at least some evidence that Kerr's accusation that he was given erroneous advice is well founded. We have already described how Kerr's evaluation of the plea would have been distorted by an error of such proportions. Such advice easily could have removed any incentive to take the plea, because it would have made it seem as if he had little to lose by going to trial. See Julian, 495 F.3d at 498-99 (petitioner rejected a plea because his lawyer told him he would face a maximum 30 years after trial and 23 if he pleaded guilty, when in fact he faced 60 years if convicted); Moore v. Bryant, 348 F.3d 238, 240 (7th Cir. 2003) (petitioner accepted a plea after his lawyer said petitioner would face 22-27 years if convicted and 10 if he pleaded guilty, when actually he only risked 12-15 years at trial).

We think that the combination of Kerr's affidavit, the district court's finding that apparently incorrect advice was actually given, and the large disparity between the sentence his lawyers thought would be required under the plea and the true sentence is enough to demonstrate a serious question on the issue of prejudice. We are not prepared, however, to say that this is enough to carry the day for Kerr. Apart from the parties' allegations, the only information that we have about the plea deliberations in this case comes from letters between Kerr and his attorneys. Kerr wrote to his former defense lawyers and asked three questions about what he referred to as the district attorney's offer of a first-degree reckless homicide plea. He asked whether he could have a copy of the offer in writing; he asked whether the district attorney had planned to recommend a particular sentence to the jury if he had accepted the plea; and he asked why he had not been offered the plea further in advance of trial, so that he might have had the opportunity to discuss it with his family. Kerr's lawyers responded without mentioning exactly what plea the state had offered:

At no time did Mr. Williams [the district attorney] offer any type of deal that was sufficient until the morning of trial. That was not in writing and he does not have to put the offer in writing. Because your matter is complete, he will not put a letter in writing to memorialize an offer that was given on the day of trial.

* * *

You are well aware that both my father and I had a lengthy conversation with you after receiving that offer. You determined that you did not want that offer and proceeded to trial. At no time did you ask for time to discuss it with you [ sic] family in order to assist you in making a decision.

In his affidavit, Kerr says that he had no idea that he could confer with his family before making a decision about the plea offer. An evidentiary hearing is necessary so that both Kerr and the state can have an opportunity to develop evidence about how the plea negotiations proceeded.

An evidentiary hearing will also give the state an opportunity to present its evidence on exactly what crime (if any) it was prepared to permit Kerr to plead guilty to. Kerr alleges that it was first-degree reckless homicide. We know that during the judge's in-chambers conference one of Kerr's lawyers said to him, "[Y]ou now understand that we really had no offer other than the possibility of your pleading guilty to life imprisonment with a 13 — with a statutory minimum." But the discussion ends there, without any hint about what charge was the basis of the plea. In this court, the state denies that Kerr was ever offered the chance to plead guilty to reckless homicide, but the state court's handling of the issue meant that the prosecutor never had the chance to support that position with affidavits or otherwise to present any evidence of what the state did offer to Kerr. Without more, we cannot tell whether Kerr's allegation is true. For all we know, the district attorney offered Kerr a plea agreement for some other crime that would have resulted in a sentence between 13 years and life (which would mean that Kerr's attorneys provided correct advice). Without evidence of the crime that was the basis of the plea agreement, we cannot assess whether Kerr's lawyers performed deficiently or whether Kerr suffered prejudice.

When a petitioner fails to develop the factual basis of a claim in state court, AEDPA normally prevents federal courts from holding an evidentiary hearing to supplement the record. See 28 U.S.C. § 2254(e)(2); Williams v. Taylor, 529 U.S. 420, 430, 120 S.Ct. 1479, 146 L.Ed.2d 435 (2000); Boyko v. Parke, 259 F.3d 781, 789-92 (7th Cir. 2001). But this bar is lifted if the applicant can show "a factual predicate that could not have been previously discovered through the exercise of due diligence." 28 U.S.C. § 2254(e)(2)(A)(ii). In his motion for state post-conviction relief, Kerr asserted that his lawyers gave him incorrect advice about the plea offered by the state, and he requested a Machner hearing — a hearing provided by Wisconsin state courts so that prisoners can preserve the testimony of their trial lawyers for ineffective assistance of counsel claims. See generally Ward, 613 F.3d at 695-96. But the state trial court denied his request for a hearing, in part because Kerr's "claims against his attorneys are merely conclusory or under-developed. . . ." In fact, no state or federal court has ever held an evidentiary hearing, and so the factual basis of Kerr's allegation that the district attorney offered to let him plead guilty to first-degree reckless homicide is entirely undeveloped. It is difficult in these circumstances to say that Kerr was not diligent enough in developing his claim. See Ward, 613 F.3d at 697-99; Davis v. Lambert, 388 F.3d 1052, 1060-62 (7th Cir. 2004). Kerr similarly cannot be accused of being less than diligent when the missing factual information critical to this case — the basis of the plea agreement — is missing because Kerr's lawyers failed to record it in any way and failed even to inform Kerr what deal was offered in their later letter, which has made it into the record.

Even if AEDPA does not bar an evidentiary hearing, Kerr must still show (1) that he has alleged facts that, if proved, would entitle him to relief, and (2) that the state courts, for reasons beyond his control, never considered the claim in a full and fair hearing. Matheney v. Anderson, 253 F.3d 1025, 1039 (7th Cir. 2001); Toumsend v. Sain, 372 U.S. 293, 312, 83 S.Ct. 745, 9 L.Ed.2d 770 (1963), overruled on other grounds, Keeney v. Tamayo-Reyes, 504 U.S. 1, 5-6, 112 S.Ct. 1715, 118 L.Ed.2d 318 (1992). For all of the reasons we discussed above, subject to our earlier caution about the pending Lafler case in the Supreme Court, we are satisfied that Kerr meets both requirements. He is thus entitled to a hearing in the district court on his plea bargain theory, which will give both parties the opportunity to present evidence to address the factually undeveloped aspects of the case that we have identified.

We VACATE the decision of the district.