Petitioner Dentrell Brown and his co-defendant Joshua Love were convicted of murder in a joint trial in an Indiana court. After exhausting state court remedies, Brown filed a federal habeas corpus petition under 28 U.S.C. § 2254. He claims he was denied effective assistance of counsel when his lawyer failed to insist that the judge give the limiting instruction required when evidence of a co-defendant’s out-of-court confession is introduced in a joint trial. See Bruton v. United States, 391 U.S. 123, 88 S.Ct. 1620, 20 L.Ed.2d 476 (1968) (protecting co-defendant from testimonial confessions of other co-defendants). The district court denied the habeas petition, finding that Brown had procedurally defaulted this claim for ineffective assistance of trial counsel by failing to assert it in state court so that federal review is barred. Brown has appealed.
On the issue of procedural default, we hold that the form of “cause” found in Martinez v. Ryan, 566 U.S. 1, 132 S.Ct. 1309, 182 L.Ed.2d 272 (2012), and expanded in Trevino v. Thaler, 569 U.S.-, 133 S.Ct. 1911, 185 L.Ed.2d 1044 (2013), is available to federal habeas corpus petitioners in Indiana who have substantial claims for ineffective assistance of trial counsel that have been procedurally defaulted in state post-conviction proceedings by lack of any counsel or lack of effective counsel. Brown is entitled to an opportunity to overcome procedural default of his claim for ineffective assistance of trial counsel for failure to request a limiting instruction if he can both demonstrate ineffective assistance of post-conviction counsel and assert a substantial claim of ineffective assistance of trial counsel. We conclude that he is entitled to an evidentiary hearing.
I. Factual and Procedural Background
On appeal we review de novo district court rulings on petitions for habe-as relief and review any findings of fact for clear error. See Lisle v. Pierce, 832 F.3d 778, 781 (7th Cir. 2016); Coleman v. Hardy, 690 F.3d 811, 814 (7th Cir. 2012). Those claims not adjudicated on the merits in the state court, like the one presented here, are also reviewed de novo. Cone v. Bell, 556 U.S. 449, 472, 129 S.Ct. 1769, 173 L.Ed.2d 701 (2009); Warren v. Baenen, 712 F.3d 1090,1096,1098 (7th Cir. 2013).
A. The Murder Trial of Joshua Love and Dentrell Brown
In the early morning hours of March 8, 2008, in Elkhart, Indiana, Gerald Wenger was murdered after trying to buy drugs. He was discovered lying dead in the street around 2:00 a.m., with a single nine-millim*507eter bullet wound to his head. Two bullet casings were found near Wenger’s body, one from a nine-millimeter handgun and a second from a .45 caliber handgun. No physical evidence was recovered beyond the shell casings.,
Following the murder, investigators relied on information. from interviews with community members. After interviews provided the names of Joshua Love and Dent-rell Brown, investigators began to rely on information from incarcerated individuals. On June 18, 2008, the State charged Brown with murder.
Brown was then just thirteen years old, and Love was nineteen years old. Brown was waived into adult felony court, and the two were tried together. At trial, the State’s key evidence tying Brown to the crime scene was the testimony of Mario Morris. Morris testified that, while Morris, Brown, and Love were all in the Elkhart County Jail, Brown and Love each confessed separately to involvement in the murder. Testifying first to his conversation with Love, Morris said that Love confessed to trying to sell fake drugs to Wenger the night of the murder, and then, after the sale went bad, shooting Wenger in the head with a nine-millimeter handgun.
Morris then testified that Brown had told him a similar story, but with some important differences. For example, Morris testified that Brown said he had struck Wenger with the butt of a .45 caliber handgun, discharging one unintentional shot. A critical feature of Morris’s testimony for Bruton purposes was that his account of Love’s confession included no mention of Brown or anyone else having been present at the shooting, and his account of Brown’s confession included no mention of Love or anyone else having been present when Brown hit Wenger in the head.
After Morris testified, Brown and Love both moved for a mistrial based on Bruton v. United States, 391 U.S. 123, 88 S.Ct. 1620, 20 L.Ed.2d 476 (1968). The trial judge denied both motions, emphasizing that at no time did Morris say Brown’s name when testifying against Love, nor did he say Love’s name when testifying against Brown. Both Love and Brown were convicted of murder, with Brown’s conviction based on a theory of accomplice liability. Brown was sentenced to 60 years in prison.
B. Direct & Collateral Review in the State Courts
On direct appeal, Brown’s counsel raised three claims, including that the trial court abused its discretion when it denied his Bruton motion for a mistrial. D.B. v. State (D.B. I), 916 N.E.2d 750, 2009 WL 3806084, at *1, 2-3 (Ind. App. 2009) (mem.). Brown’s appellate counsel argued that Morris’s testimony about Love’s statement violated Brown’s confrontation rights because Brown could neither compel Love to testify nor cross-examine him. Id. at *2. The appellate court was not persuaded. It found no Bruton violation because Morris’s account of Love’s confession to him never mentioned a third party present at the scene of the murder. Id. at *3.
Brown filed a petition for post-conviction relief in state court with the assistance of counsel. His post-conviction lawyer raised a single issue in the operative petition: ineffective assistance of trial counsel for having failed to move to sever Brown’s trial from Love’s. The argument relied on Bruton even though the appellate court on direct review had “specifically held” that there was no Bruton violation in Brown’s trial. The trial court denied relief, and the Indiana Court of Appeals affirmed, finding that the ineffective assistance of counsel claim was an attempt to revisit the Bruton issue decided against Brown on direct appeal and thus barred by res judicata. D.B. *508v. State (D.B. II), 976 N.E.2d 146, 2012 WL 4718965 at *2-3 (Ind. App. 2012) (mem.).
C. Brown’s Federal Habeas Petition
Brown’s habeas petition to the federal district court raised three issues, two of which have been dropped on appeal. The only claim before us is Brown’s claim that his “trial lawyer was ineffective for failing to request an instruction limiting the use of Love’s statement, offered through Morris, to Love.” Because it was not presented to the state courts, the claim for ineffective assistance of trial counsel would ordinarily be barred from federal review because of procedural default. In the district court, however, Brown argued that he should be given the opportunity to overcome that default under Martinez, 566 U.S. 1, 132 S.Ct. 1309, 182 L.Ed.2d 272, and Trevino, 569 U.S.-, 133 S.Ct. 1911, 185 L.Ed.2d 1044.
The district court held that Martinez and Trevino do not apply to § 2254 cases in Indiana, and thus Brown was not entitled to attempt to overcome procedural default on his claim for ineffective assistance of trial counsel. Brown v. Brown, No. 1:13-cv-1981-JMS-DKL, 2015 WL 1011371, at *2-3 (S.D. Ind. 2015). His request for an evidentiary hearing was denied and his petition dismissed. We granted Brown an expanded certificate of appealability that included this claim because Brown had “made a substantial showing of the denial of his right to effective assistance of trial counsel.”
II. Analysis
Brown’s claim for ineffective assistance of trial counsel requires a two-step analysis. We hold first that the Martinez-Trevino doctrine can apply to claims for ineffective assistance of counsel arising from the Indiana state courts. We next hold that Brown has offered some evidence of deficient performance by his post-conviction relief counsel and has asserted a substantial claim of ineffective assistance of trial counsel. We reverse and remand the case to the district court for an evidentiary hearing on both claims for ineffective assistance, first on the procedural default issue and then, if the default is excused, on the merits of the trial-based claim.
Before explaining our view on Martinez-Trevino, we pause to address the state’s assertion that petitioner’s argument on appeal has been forfeited. The state argues that in the federal district court, petitioner’s claims were based on the Confrontation Clause pursuant to Bruton rather than the Indiana Rules of Evidence. Under this theory, Brown’s claim of ineffective assistance of trial counsel for failure to request a limiting instruction based on the Indiana Rules of Evidence would be forfeited now on appeal. We disagree. Brown’s habeas petition claimed clearly that his “trial lawyer was ineffective for failing to request an instruction limiting the use of Love’s statement, offered through Morris, to Love.” The habeas petition discussed the failure to request the limiting instruction as something that should have occurred following the denial of the motion for a mistrial, and the federal district court evaluated Brown’s claim separately from the Confrontation Clause issue. Thus, petitioner’s specific claim— although not presented in the state courts — was not forfeited by any supposed failure to raise it in the federal district court. _
A. The Martinez-Trevino Doctrine Applies in Indiana
On appeal, petitioner argues that the rule established in Martinez and Trevino applies to § 2254 cases in Indiana so that he may try to overcome the procedural *509default of his claim for ineffective assistance of trial counsel. See Martinez, 132 S.Ct. 1309; Trevino, 133 S.Ct. 1911. We agree. We first explain the scope of the Martinez-Trevino doctrine. Against that backdrop, we then review the Indiana procedures for raising ineffective assistance of trial counsel, and we compare those procedures to those in other jurisdictions where the Martinez-Trevino doctrine applies. We find that Indiana procedures governing ineffective assistance of trial counsel claims fall into the category the Supreme Court addressed in Trevino.
1. The Martinez-Trevino Doctrine
A federal habeas petitioner’s claim is subject to the defense of procedural default if he does not fairly present his claim through a complete round of state-court review. Richardson v. Lemke, 745 F.3d 258, 268 (7th Cir. 2014). A . prisoner can overcome procedural default by showing cause for the default and resulting prejudice, or by showing he is actually innocent of the offense. Coleman v. Thompson, 501 U.S. 722, 750, 111 S.Ct. 2546, 115 L.Ed.2d 640 (1991). Until recently, a federal petitioner could not overcome the federal bar on procedurally defaulted claims by proving ineffective assistance of post-conviction counsel because there is no constitutional right to post-conviction counsel. See id. at 752-53, 111 S.Ct. 2546.
In 2012, however, the Supreme Court recognized a new form of cause for overcoming procedural default in Martinez: “Where, under state law, claims of ineffective assistance of trial counsel must be raised in an initial-review collateral proceeding, a procedural default will not bar a federal habeas court from hearing a substantial claim of ineffective assistance at trial if, in the initial-review collateral proceeding, there was no counsel or counsel in that proceeding was ineffective.” 132 S.Ct. at 1315, 1320. The Court explained that this route was needed to protect a prisoner with “a potentially legitimate claim of ineffective assistance of trial counsel” when state law required defendant to bring claim for ineffective assistance of trial counsel on collateral review. Id. If post-conviction counsel errs by failing, to raise a claim for ineffective assistance of trial counsel in the initial round of collateral review, it is unlikely that any state court at any level will hear the claim. Id. at 1316.
The next year, the Court expanded the Martinez form of “cause” in Trevino, holding that “a distinction between (1) a State that denies permission to raise the claim on direct appeal and (2) a State that in theory grants permission but, as a matter of procedural design and systemic operation, denies a meaningful opportunity to do so is a distinction without a difference.” Trevino, 133 S.Ct. at 1921. Examining Texas law, Trevino observed that even though Texas did not require a defendant to raise an ineffective assistance of trial counsel claim in state collateral review proceedings, the “structure and design of the Texas system in actual operation” worked effectively as a ban on claims on direct review. Id. at 1915. Like Texas, Indiana does not always require prisoners to bring claims for ineffective assistance of trial counsel on collateral review, so petitioner Brown must depend on the Trevino extension of Martinez to overcome procedural default.
In dissent in Trevino, Chief Justice Roberts predicted accurately á long process of state-by-state litigation on applying Trevino. Id. at 1923 (Roberts, C.J., dissenting). At least eight circuits, including this one, have decided whether Trevino applies to specific jurisdictions. This court applied the Martinez-Trevino doctrine to federal prisoners who bring motions for post-con*510viction relief under § 2255.1 Ramirez v. United States, 799 F.3d 845, 853 (7th Cir. 2015) (“[T]he federal courts have no established procedure ... to develop ineffective assistance claims for direct appeal,” so “the situation of a federal petitioner is the same as the one the Court described in Trevino.”); see also Coleman v. Goodwin, 833 F.3d 537, 543 (5th Cir. 2016) (Martinez-Trevino applies in Louisiana); Runningeagle v. Ryan, 825 F.3d 970, 981-82 (9th Cir. 2016) (Martinez-Trevino applies in Arizona); Woolbright v. Crews, 791 F.3d 628, 636 (6th Cir. 2015) (Martinez-Trevino applies in Kentucky); Fowler v. Joyner, 753 F.3d 446, 463 (4th Cir. 2014) (North Carolina procedures do “not fall neatly within Martinez or Trevino” and doctrine applies only in certain circumstances); Sutton v. Carpenter, 745 F.3d 787, 795-96 (6th Cir. 2014) (Martinez-Trevino applies in Tennessee); Sasser v. Hobbs, 735 F.3d 833, 852-53 (8th Cir. 2013) (Martinez-Trevino applies to capital defendants in Arkansas). Cf. Lee v. Corsini, 777 F.3d 46, 61 (1st Cir. 2015) (Martinez-Trevino does not apply in Massachusetts); Fairchild v. Trammell, 784 F.3d 702, 721 (10th Cir. 2015) (Martinez-Trevino does not apply in Oklahoma).
2. Claims for Ineffective Assistance of Trial Counsel in Indiana
With Trevino as our guide, two characteristics of Indiana practice — the “procédural design” and “systemic operation” — convince us that the Martinez-Trevino doctrine applies in Indiana. Trevino, 133 S.Ct. at 1921. First, while Indiana law does not always require claims for ineffective assistance of trial counsel to be brought on collateral review, the Indiana Supreme Court has adopted rules and doctrines that strongly discourage any other path. Second, in actual practice, the Indiana Supreme Court’s discouragement has worked to force almost all such claims to wait for collateral review.
a. Procedural Design
The Indiana Supreme Court acted to clear up the law governing claims for ineffective assistance of trial counsel in Woods v. State, 701 N.E.2d 1208 (Ind. 1998). After considering alternative approaches to procedural default, the court concluded “that the most satisfactory resolution of a variety of competing considerations is that ineffective assistance may be raised on direct appeal, but if it is not, it is available in postconviction proceedings irrespective of the nature of the issues claimed.” Id. at 1216. A claim for ineffective assistance of counsel is barred on collateral review if it was already raised on direct appeal. Id. at 1220. Critical for our purposes, presenting a claim for ineffective assistance is an all- or-nothing proposition in Indiana. A defendant may not present one specific ground on direct appeal and wait to present another on collateral review. Id.
The court in Woods explained that a claim for ineffective assistance of trial counsel will ordinarily require evidence beyond the record of the conviction and so *511should ordinarily be brought in a collateral post-conviction case where the defendant can offer new evidence. Id. at 1216. The complicating factor here — which shifts Indiana from Martinez to Trevino — is that Woods also recognized there may be an “exceptional case in which the defendant prefers to adjudicate a claim of ineffective assistance before direct appeal remedies have been exhausted.” Id. at 1219-20. Under these rare circumstances, Woods explained, a procedure known in Indiana as the Davis-Hatton procedure allows a convicted appellant to suspend or terminate his direct appeal to pursue a petition for post-conviction relief. Id. at 1219, citing Davis v. State, 267 Ind. 152, 368 N.E.2d 1149 (1977); Hatton v. State, 626 N.E.2d 442 (Ind. 1993); see also Ind. R. App. P. 37; Peaver v. State, 937 N.E.2d 896, 899 (Ind. App. 2010). The Davis-Hatton procedure might be appropriate if the trial record itself supports an indisputable claim of ineffective assistance of trial counsel that will result in the immediate release of a person who is in prison improperly.
If a trial court denies a Davis-Hatton petition, an appeal from that post-conviction denial and the original direct appeal will be consolidated but evaluated under separate standards of review. Peaver, 937 N.E.2d at 899-900; Slusher v. State, 823 N.E.2d 1219, 1222 (Ind. App. 2005); Dodd v. Knight, 533 F.Supp.2d 844, 852 (N.D. Ind. 2008). A defendant who uses the Davis-Hatton procedure will be barred from asserting any new claim for ineffective assistance on direct appeal, or in any of the consolidated proceedings or additional post-conviction proceedings that may follow. Peaver, 937 N.E.2d at 899. Most helpful for the issue we face here, the Indiana Supreme Court explained that the Davis-Hatton procedure is “not to be used as a routine matter in adjudicating the issue of trial counsel’s effectiveness.” Woods, 701 N.E.2d at 1220.
Like Texas in Trevino and Tennessee in Sutton, Indiana “permits defendants to raise the claims on direct appeal.” Compare Sutton, 745 F.3d at 791, quoting Trevino, 133 S.Ct. at 1918, with Woods, 701 N.E.2d at 1216, 1220. Because most claims for ineffective assistance of trial counsel cannot be shown within the four corners of the original trial court record, and because of the presumption of competence that applies in Indiana courts, claims for ineffective assistance of trial counsel brought on direct appeal “almost always fail.” Woods, 701 N.E.2d at 1216, quoting United States v. Taglia, 922 F.2d 413, 417-18 (7th Cir. 1991). Trevino made much the same point about the need to present evidence outside the original trial record. 133 S.Ct. at 1918.
Additional aspects of Indiana procedure align with other aspects of Trevino. As in our federal cases, a defendant who asserts a claim for ineffective assistance of trial counsel on direct appeal may not relitigate the claim on collateral review. Compare Ramirez, 799 F.3d at 853, ■ with Woods, 701 N.E.2d at 1220. Like the federal rules we reviewed in Ramirez, Indiana’s rule is even more restrictive than the Texas procedures in Trevino. See Ramirez, 799 F.3d at 853. Also, Indiana does not allow counsel on direct appeal from a conviction to use a motion to correct errors to supplement the record to assert a claim for ineffective assistance. Compare Woods, 701 N.E.2d at 1216, with Trevino,, 133 S.Ct. at 1918 (determining that a motion for new trial is an inadequate vehicle for ineffective assistance of trial counsel claims).
Moreover, because a Davis-Hatton petition in Indiana is a collateral attack on a conviction, it does not provide, in the Trevino Court’s words, “meaningful review” of an ineffective assistance counsel claim *512on direct review: it simply is not direct review. See Trevino, 133 S.Ct. at 1919. Perhaps most important, the Davis-Hatton procedure is neither “systematic” nor “typical.” It is, in the words of Trevino, “special, limited, ... [and] rarely used.” Compare Woods, 701 N.E.2d at 1220, with Trevino, 133 S.Ct. at 1919-21. Amicus Indiana Public Defender Council tells us that between 2008 and 2012, its attorneys filed approximately 2000 appeals and only four Davis-Hatton petitions.
b. Systemic Operation
Indiana rules work together to make it unlikely that an Indiana defendant will be able to raise adequately on direct appeal a claim for ineffective assistance of trial counsel. The Indiana Supreme Court said as much in Woods: “As a practical matter,” the confluence of these rules “will likely deter all but the most confident appellants from asserting any claim of ineffectiveness on direct appeal.” 701 N.E.2d at 1220. As in Trevino itself, “special, rarely used procedural possibilities” like the Davis-Hatton procedure cannot overcome the Indiana courts’ directives that the preferred forum for ineffective assistance of trial counsel claims is post-conviction review. See Trevino, 133 S.Ct. at 1920.
The Indiana courts, like the Texas courts in Trevino, routinely direct defendants to bring claims for ineffective assistance of trial counsel on collateral review and warn against bringing them on direct review. Compare Woods, 701 N.E.2d at 1219-20 (“[A] postconviction hearing is normally the preferred forum to adjudicate an ineffectiveness claim.”), with Trevino, 133 S.Ct. at 1919-20. The Indiana Supreme Court not only has reinforced the preference for collateral review but has gone so far as to decline addressing a defendant’s claim for ineffectiveness of trial counsel actually presented on direct appeal, believing it “preferable for the defendant to adjudicate his claim ... in a post-conviction relief proceeding.” McIntire v. State, 717 N.E.2d 96, 102 (Ind. 1999); see also Landis v. State, 749 N.E.2d 1130, 1132 (Ind. 2001).
The Indiana Court of Appeals has followed suit, routinely issuing non-prece-dential decisions that echo the lesson of Woods, especially when denying relief on direct appeal. E.g., Crockett v. State, 13 N.E.3d 556, 2014 WL 2202763, at *4 (Ind. App. 2014) (“[I]t is well-settled that a post-conviction proceeding is generally the preferred forum,” even if “a criminal defendant ... is at liberty to elect whether to present this claim on direct appeal or in post-conviction proceedings”); see also Johnson v. State, 46 N.E.3d 499, 2016 WL 327985, at *2 (Ind. App. 2016); Merriman v. State, 40 N.E.3d 1280, 2015 WL 5703912, at *5 n.3 (Ind. App. 2015); Beals v. State, 37 N.E.3d 977, 2015 WL 4105047, at *10 (Ind. App. 2015); Anderson v. State, 16 N.E.3d 488, 2014 WL 3511699, at *5 (Ind. App. 2014); Wine v. State, 9 N.E.3d 771, 2014 WL 1266285, at *3 (Ind. App. 2014); Reed v. State, 985 N.E.2d 1151, 2013 WL 1701879, at *3 (Ind. App. 2013).
Like the Texas bar in Trevino, the Indiana criminal defense bar “has taken this strong judicial advice seriously.” See Trevino, 133 S.Ct. at 1920. In its annual training, amicus Indiana Public Defender Council “consistently advises against appellate counsel presenting ineffective assistance claims on direct appeal.” When a public defender handling a direct appeal asked the Council if she should raise a claim for ineffective assistance of trial counsel in the direct appeal, the responses were best summarized by one that began, “NOOOOOOÜ!” Amicus Br. of Ind. Pub. Def. at 21a.
For these reasons, in the language of Trevino, “as a matter of its structure, de*513sign, and operation,” the Indiana procedural system “does not offer most defendants a meaningful opportunity to present a claim of ineffective assistance of trial counsel on direct appeal.” Trevino, 133 S.Ct. at 1921. The Martinez-Trevino form of cause to excuse procedural default is available to Indiana defendants who seek federal habe-as relief.2
B. Cause Under Martinez
In a state like Indiana where the Martinez-Trevino doctrine can apply, procedural default in the state courts will not bar federal habeas review when a petitioner can demonstrate cause for the default. See Trevino, 133 S.Ct. at 1918; Martinez, 132 S.Ct. at 1318; Coleman, 501 U.S. at 747-48, 111 S.Ct. 2546. To demonstrate cause under Martinez-Trevino, the petitioner must show deficient performance by counsel on collateral review as required under the first prong of the Strickland analysis. Martinez, 132 S.Ct. at 1318; see Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Actual resulting prejudice can be established with a substantial claim of ineffective assistance of trial counsel that would otherwise have been deemed defaulted. See Detrich v. Ryan, 740 F.3d 1237, 1245-46 (9th Cir. 2013) (observing that this reading is required to square the requirement with the structure of Martinez). Accordingly, to avoid procedural default, petitioner Brown must demonstrate that his collateral review counsel was deficient and must make a substantial claim of ineffective assistance of trial counsel. See Trevino, 133 S.Ct. at 1918, citing Martinez, 132 S.Ct. at 1318-19, 1320-21. Petitioner Brown has made a strong enough showing of each element to call for an evidentiary hearing in the district court.
1. Ineffective Assistance of Post-Conviction Relief Counsel
Brown claims that his lawyer in his post-conviction case was deficient because she did not raise a claim that his trial counsel was ineffective for failing to request a limiting instruction. To demonstrate that counsel’s performance was deficient, the petitioner “must show that counsel’s representation fell below an objective standard of reasonableness.” Strickland, 466 U.S. at 688, 104 S.Ct. 2052. Judicial review of counsel’s performance is “highly deferential,” with “every effort ... made to eliminate the distorting effects of hindsight.” Id. at 689,104 S.Ct. 2052.
The State has not directly addressed whether Brown’s collateral review lawyer was ineffective. On this record, and without having heard yet from the post-conviction attorney, we find that petitioner Brown has offered evidence that his post-conviction counsel’s representation fell below an objective standard of reasonableness so that he is entitled to an evidentiary hearing on the issue.
The amended complaint on collateral review made a single allegation of error: trial counsel had been ineffective “for failing to move for a severance of his trial from Petitioner’s codefendant” as a remedy for a Bruton violation. The problem with this claim, as the post-conviction courts held, was that on his direct appeal Brown had already argued the joint trial produced a Bruton violation. The state courts rejected that claim, squarely and definitively. Even if we account for the benefits of hindsight, a new claim built on the assumption of a *514Bruton violation would seem to have had little or no chance of success.
We recognize that the vast majority of claims for post-conviction relief are without merit, so an attorney’s failure to prevail, or even pursuit of an unpromising claim, does not show ineffective assistance. Also, the Supreme Court’s fundamental point in Strickland about avoiding the distorting effects of hindsight applies as much in the post-conviction process as in any other. See Strickland, 466 U.S. at 689, 104 S.Ct. 2052. A post-conviction attorney can and should use professional judgment in selecting which claims and issues to raise, just as we expect from attorneys in direct appeals. See Morris v. Bartow, 832 F.3d 705, 709-11 (7th Cir. 2016) (finding counsel’s performance competent despite mixed record indicating possible coerced plea); Vinyard v. United States, 804 F.3d 1218, 1225-27 (7th Cir. 2015) (Strickland not applied unreasonably; counsel advised client not to challenge guilty plea); Makiel v. Butler, 782 F.3d 882, 898-902 (7th Cir. 2015) {Strickland not applied unreasonably; counsel selected issues for appeal and did not include an additional obvious claim).
For purposes of applying Martinez and Trevino, the approach we take to claims of ineffective assistance of counsel on direct appeal provides the best available guide. Pursuit of unsuccessful arguments and claims does not show ineffective assistance of counsel. But we may compare the claims actually presented to those that might have been presented. Where counsel chose to pursue just one issue that was a virtually certain loser, as in Shaw v. Wilson, a petitioner may show deficient performance by showing that a much stronger claim or argument was available. 721 F.3d 908, 915 (7th Cir. 2013); see also Vinyard, 804 F.3d at 1228; Makiel, 782 F.3d at 898-99.
Even without relying on the benefits of hindsight, petitioner Brown makes a strong argument here that the one claim counsel pursued in the post-conviction petition was doomed from the beginning. The claim that counsel was ineffective by failing to move to sever Brown’s trial from Love’s appears to have been built on the assumption that the joint trial resulted in a Bruton violation. The state courts had already rejected that premise on direct appeal.
Petitioner argues that a viable ineffective assistance of trial counsel claim could have been premised on failure to seek a limiting instruction as to the hearsay Morris offered when testifying to his conversation with Love. See Ind. R. Evid. 801(c) (defining hearsay), 802 (making hearsay inadmissible), and 105 (providing a limiting instruction when evidence is presented that is “admissible against a party or for a purpose — but not against another party or for another purpose”). In contrast, the claim post-conviction review counsel presented instead was barred by res judicata. We do not mean to imply that we have reached a conclusion on the ultimate question of counsel’s performance. As noted, no court has heard testimony from Brown’s post-conviction counsel about the selection of issues and other factors that may affect the performance issue under Strickland. By showing that another, much stronger claim was available, however, petitioner has shown he is entitled to an evidentiary hearing on that issue.
2. Substantial Underlying Claim for Ineffective Assistance of Trial Counsel
Martinez also requires a petitioner to show “that the underlying ineffective-assistance-of-trial-counsel claim is a substantial one, which is to say that the prisoner must demonstrate that the claim has *515some merit.” Martinez, 132 S.Ct. at 1318— 19. Martinez offered little guidance as to what is a “substantial” claim for these purposes. It provided only a “cf.” citation to Miller-El v. Cockrell, 537 U.S. 322, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003), describing the standards for certificates of appealability. 132 S.Ct. at 1319. Miller-El held that a certificate of appealability should be granted when a substantial showing can be made “by demonstrating that jurists of reason could disagree with the district court’s resolution ... or that jurists could conclude the issues presented are adequate to deserve encouragement to proceed further.” Miller-El, 537 U.S. at 327, 123 S.Ct. 1029. “This threshold inquiry does not require full consideration.” Id. at 336, 123 S.Ct. 1029.
The Martinez dissent predicted the problem we face here: “to establish cause a prisoner must demonstrate that the ineffective-assistance-of-trial-counsel claim is ‘substantial,’ which apparently means the claim has at least some merit. ... The Court does not explain where this substan-tiality standard comes from.” Martinez, 132 S.Ct. at 1322 n.2 (Scalia, J., dissenting). Appellate opinions applying Martinez and Trevino thus far offer limited further guidance. See Ramirez, 799 F.3d at 854-56 (concluding that there was “some merit” to Ramirez’s argument without delving further into the standard). See also Running-eagle, 825 F.3d at 983 (deciding case on other grounds); Sexton v. Cozner, 679 F.3d 1150, 1161 (9th Cir. 2012) (same); Flores v. Stephens, 794 F.3d 494, 505 (5th Cir. 2015) (concluding “that reasonable jurists would not debate the district court’s decision ... because the claims are not ‘substantial’ within the meaning of Martinez”); Cox v. Horn, 757 F.3d 113, 119 (3d Cir. 2014) (applying standard for certificate of appealability); Detrich, 740 F.3d at 1245 (citing Miller-El standard); Cook v. Ryan, 688 F.3d 598, 610 n.13 (9th Cir. 2012) (observing that Martinez used Miller-El as “generally analogous support”).
In this case, petitioner argues that by granting a certificate of appealability, we have already determined that his defaulted ineffective assistance of trial counsel claim is substantial under Martinez. The State simply repeats that a “substantial claim is one that has ‘some merit,’ ” then argues that petitioner cannot satisfy cause and prejudice under Strickland. We conduct a separate and deeper review of the record, beyond our grant of a certificate of appeal-ability, and find a substantial ineffective assistance of trial counsel claim under Martinez.
We are guided by Strickland’s two-prong approach to claims of ineffective assistance of counsel. Brown must address whether his trial counsel’s performance was deficient, falling below an objective standard of reasonableness. He must also address whether the ineffective assistance caused actual prejudice. Strickland, 466 U.S. at 687-88, 691-92, 104 S.Ct. 2052. Substantiality is a threshold inquiry; full consideration of the merits is not required. Miller-El, 537 U.S. at 336, 123 S.Ct. 1029.
a. Substantial Showing of Trial Counsel’s Deficient Performance
At trial, the State relied heavily on the testimony of Mario Morris to place both Love and Brown at the scene of the murder. Morris testified to separate conversations he had in the Elkhart County Jail, one with Love and others with Brown. Without his testimony, only circumstantial evidence and one other reluctant witness implicated Brown.
Morris first testified to a conversation he had in Elkhart County Jail with Love. Like many of the witnesses called by the State, Morris suffered from credibility is*516sues. He claimed that over a card game in jail, Love admitted he was involved in the murder of Wenger. According to Morris, Love told'him that he had left a woman’s apartment at the Middlebury Apartments to sell a “gang pack” (something that appears to be crack cocaine but is not) to a “white guy, Mr. Wenger.” Love then told Morris he got in the back seat of a truck with Wenger in the driver’s seat, and drove around a few blocks. Once Wenger figured out the drugs were fake, an argument ensued. Both men got out of the truck. Love then shot Wenger with a nine-millimeter handgun. Afterwards, he got back into the truck and pulled off to park behind some houses. Love came back later to wipe down any fingerprints he might have left on the truck.
Immediately after describing Love’s tale for the jury, Morris testified that he had a separate conversation with Brown, also in the jail, who Morris said told a story very similar to Love’s. Morris testified that Brown told him that he left a woman’s apartment at the Middlebury Apartments on the night of the murder. He was going to try to sell some fake drugs. Morris was asked by the prosecutor, “And did they actually try to sell him those gang packs?” (emphasis added). Morris responded, ‘Tes, sir.”
Then, Morris testified, Brown told him that he had gotten out of the truck and hit Wenger on- the head with his .45 caliber handgun. The blow caused the gun to fire. Brown then got in the truck and drove to an alley behind some houses. During deliberations, the jury requested to review Morris’s testimony. It was read back to them in the courtroom.
Morris’s testimony as to his conversation with Love, although admissible against Love, was inadmissible hearsay as offered against Brown. As petitioner’s brief emphasizes, “like perhaps all jurisdictions, Indiana courts assume that jurors follow their instructions.” If Brown’s trial attorney had requested the limiting instruction to which Brown was probably entitled, it would have left the prosecution to rely on the arguably weak remainder of its case against Brown.
We are not convinced, on the limited record before us, that the decision not to seek a limiting instruction was objectively reasonable. Without the testimony of Morris’s conversation with Love, which mirrored so closely the testimony of Morris’s conversation with Brown, none of the evidence presented by the prosecution puts Brown at the murder scene with Love.
b. Substantial Showing of Prejudice
The additional evidence against Brown was not so strong that his claim of actual prejudice is not substantial for purposes of Martinez and Trevino. The State relies primarily on the testimony of Kendrick Lipkins, who at trial was treated as a witness hostile to the prosecution. He responded only reluctantly with a single word, “Correct,” to a leading question regarding ■ an overheard confession by Brown. Lipkins also testified that he had a separate conversation in a car with Love, in Brown’s presence, about the disposal of a .45 caliber handgun. But Lipkins, like most of the State’s witnesses, had serious credibility issues. He admitted to being interested in a reward offered for information on the case, and he was willing to cooperate with police in order to keep his brother, T.J. Lipkins, out of jail.
The remaining evidence against Brown was circumstantial and not conclusive. A witness testified that a few weeks before the shooting she saw Brown with what she thought was a gun. Another witness testified that she saw both Love and Brown around 10:30 p.m. the night of the shoot-*517mg. That was over three hours before Wenger was found, and she had a difficult time identifying Brown. A man testified that he saw two boys walking by Wenger’s truck the morning after the shooting, but he could neither identify Brown nor say what the two boys were doing. Two witnesses testified that Brown was trying to sell a nine-millimeter handgun in the weeks following the shooting. One of those witnesses testified that when Brown was asked whether he murdered someone with the gun, he laughed. Although at least sixteen fingerprints were pulled from the truck, they were all Wenger’s. No guns were recovered, but one .45 casing and one nine-millimeter casing were found at the crime scene. Bullet fragments found in Wenger’s body were from a single nine-millimeter bullet. The evidence was legally sufficient to permit a jury to convict Brown, but Brown has made a substantial claim of deficient trial counsel and resulting prejudice. His claim for ineffective assistance of trial counsel is not “wholly without factual support,” or lacking in all legal merit. Martinez, 132 S.Ct. at 1319. If Brown’s theory is proven at an evidentiary hearing, he will have made a successful ineffective assistance of trial counsel claim. On the record before us, reasonable jurists “could disagree ... or ... conclude the issues presented” in petitioner’s brief and borne out in the trial transcript “are adequate to deserve encouragement to proceed further.” Miller-El, 537 U.S. at 327, 123 S.Ct. 1029. Brown has presented a substantial claim of ineffective assistance of trial counsel, sufficient to avoid the procedural default because he has demonstrated that the claim has some merit. See Martinez, 132 S.Ct. at 1318-19.
[[Image here]]
In sum, the Martinez-Trevino doctrine applies to Indiana procedures governing ineffective assistance of trial counsel claims. Petitioner Brown has presented evidence of ineffective post-conviction counsel and made a substantial claim of ineffective assistance of trial counsel. Accordingly, we REVERSE the district court’s dismissal of Brown’s petition and REMAND to the district court for an evi-dentiary hearing on the issue of ineffective assistance of post-conviction counsel. If the district court finds deficient performance by post-conviction counsel, Brown’s default will be excused, and he will be entitled to an evidentiary hearing on the merits in the district court for the underlying claim of ineffective assistance of trial counsel for failure to request a limiting instruction.
. On two occasions, we have also observed that the Martinez-Trevino exception does not apply to the procedures that govern the typical ineffective assistance of trial counsel claim in Wisconsin courts. See Ramirez v. United States, 799 F.3d 845, 851 (7th Cir. 2015) ("Wisconsin law treats postconviction relief in an unusual way, insofar as it allows defendants to raise a claim of ineffectiveness of counsel simultaneously with a direct appeal.”); Nash v. Hepp, 740 F.3d 1075, 1079 (7th Cir. 2014) ("Wisconsin law expressly allows — indeed, in most cases requires — defendants to raise claims of ineffective assistance of trial counsel as part of a consolidated and counseled direct appeal, and provides an opportunity to develop an expanded record."). Our analysis here does not alter that analysis of Wisconsin law.
. On this issue; we respectfully disagree with both the district court here and the Northern District of Indiana in Brown v. Superintendent, 996 F.Supp.2d 704, 716-17 (N.D. Ind. 2014).