Hamilton, Circuit Judge.

Carmen G. Iguina Gonzalez, Attorney, American Civil Liberties Union, Washington, DC, Spencer Elijah Wittmann Amdur, Attorney, American Civil Liberties Union, San Francisco, CA, for Petitioner - Appellant.

Peter Blackett, Bob Wood, Attorneys, Office of the United States Attorney, Indianapolis, IN, for Respondent - Appellee.

Before Rovner, Hamilton, and Scudder, Circuit Judges.

Hamilton, Circuit Judge.

This appeal raises fundamental questions about what is expected of defense counsel in the federal criminal justice system, where almost all defendants plead guilty. Counsel must negotiate guilty pleas and argue for more lenient sentences, both of which require expert knowledge of the federal Sentencing Guidelines. This knowledge is a core competency for federal criminal defense. The issue here is whether a lawyer's failure to raise an important and, in this case, ultimately meritorious guideline argument may constitute ineffective assistance of counsel even where there was no directly on-point precedent within the circuit at the relevant time. We find that it may in this case.

Now in his sixties, petitioner Jeffery Bridges has been in and out of prison since he was a teenager and has been battling drug addiction even longer. After staying out of trouble for eight years, Bridges got involved in drugs again and committed four robberies in two days in March 2017. He netted scarcely $700 in total and was easily caught by the police. A federal grand jury indicted Bridges for four counts of robbery in violation of the Hobbs Act, 18 U.S.C. § 1951.

Bridges agreed to a guilty plea stipulating that he was subject to the guideline career offender enhancement, U.S.S.G. § 4B1.1, which could apply only if his crimes of conviction were "crimes of violence" as defined by the Guidelines. This enhancement more than doubled his advisory guideline sentencing range. The district court imposed a below-guideline sentence of 140 months. Bridges did not appeal. He had waived that right in his plea deal.

Bridges now seeks postconviction relief under 28 U.S.C. § 2255, alleging he was denied effective assistance of counsel in pleading guilty. He argues that his lawyer failed to realize and argue that Hobbs Act robbery did not then qualify as a "crime of violence" under the Guidelines, so he should not have been categorized as a career offender. When Bridges pleaded guilty and was sentenced, there was no binding precedent in this circuit on this issue. Bridges argues that competent counsel still would have recognized the issue or at least known to investigate it. The district court denied relief without holding a hearing, reasoning that counsel's failure to anticipate arguments that we have not yet accepted cannot be constitutionally deficient.

We reverse for an evidentiary hearing on defense counsel's performance under 28 U.S.C. § 2255(b). First, we join the other circuits that have concluded that Hobbs Act robbery is not a "crime of violence" as that phrase is currently defined in the Guidelines. Although we had not so ruled when Bridges pleaded guilty, the building blocks for a successful legal argument were already in place. Effective counsel would have considered this question that was so important in this case. At that time, minimal research would have uncovered a Tenth Circuit decision squarely holding that Hobbs Act robbery was no longer a crime of violence under a 2016 amendment to the guideline definition of a crime of violence.

We realize how counterintuitive it is to argue or hold that Hobbs Act robbery is not a crime of violence—and that counsel could be deficient for failing to argue for that unexpected result. Yet defense lawyers, prosecutors, and judges in the federal system all appreciate that both statutory and guideline sentencing enhancements for recidivism and crimes of violence have produced many counterintuitive results over the last several decades. During those years, both federal statutes and the Sentencing Guidelines have used the "categorical method" to classify prior convictions and current offenses. The Sentencing Commission proposed guideline amendments in 2018 to reduce reliance on the categorical method. 83 Fed. Reg. 65400, 65407–65412 (Dec. 20, 2018). The Commission has been unable to act on those proposed amendments, though, because it has lacked a quorum for years. Bridges may be a beneficiary of that odd circumstance.

I. Factual and Procedural Background

Over two days in March 2017, Bridges robbed four retail stores in Indianapolis. In three, he held his hand in a pocket in the shape of a gun. In the fourth, he threatened to "hurt" the cashier if his demands were not met. Bridges netted a total of $719.35.

Police quickly identified Bridges as a suspect. Five days after the last robbery, Bridges was stopped by police and admitted that he had committed the robberies. A federal grand jury indicted him on four counts of robbery in violation of the Hobbs Act, 18 U.S.C. § 1951(a).

Bridges pleaded guilty to all four robberies. His plea agreement stipulated to certain guideline calculations, including application of the career offender enhancement in § 4B1.1. This enhancement applies when a defendant with a sufficient history of violent or drug convictions is sentenced for a new "controlled substances offense" or "crime of violence." Bridges' plea stipulated that this enhancement applied based on two prior convictions for robbery in state court and assumed that his new Hobbs Act robbery convictions also counted as crimes of violence. The plea agreement was finalized on April 24, 2018. On July 31, 2018, he pleaded guilty in court and was sentenced on the same day.

At sentencing and after accounting for Bridges' acceptance of responsibility, his guideline offense level was 29, with criminal history VI as a career offender, for a range of 151 to 188 months. Without the career offender guideline, Bridges' advisory range would likely have been 57 to 71 months, based on offense level 21 and criminal history category IV.

Defense counsel did not challenge the guideline calculations. Instead, counsel focused on mitigation arguments. Bridges, he told the court, had had a difficult life and was trying to improve himself to help his family. Bridges was more or less abandoned by his parents and subjected to physical, emotional, and sexual abuse by family members. His drug problems, which he blames for the 2017 robberies, began before he was even thirteen years old. Defense counsel argued that despite these challenges, Bridges continued to try to better himself through treatment and education, and suggested that his age made him unlikely to commit future offenses. Bridges asked for a below-guideline 108-month sentence. The government asked for a 160-month sentence.

The district court accepted the agreed guideline calculations putting the advisory range at 151 to 188 months. The court dismissed out of hand the aging-out-of-crime argument. After all, Bridges had committed these robberies in his sixties. Bridges also had a long history of violent crime, including a now 45-year-old conviction for participating in a home invasion and rape and robbery, and he had shown dangerous tendencies and disrespect for the law and others. Other mitigation arguments had some weight. The judge explained that the final sentence was below the guideline range because of Bridges' experiences as a child and his continuing efforts at rehabilitation. The court sentenced Bridges a little below the low end of the guideline range, to 140 months in prison.

Pursuant to his plea deal, Bridges did not appeal directly. Instead, he brings this suit under the federal habeas corpus statute, 28 U.S.C. § 2255. His only path to challenge his sentence and avoid his waiver of his right to appeal is to argue that his lawyer was ineffective in negotiating the plea agreement with the career offender enhancement. Bridges argues that Hobbs Act robbery is not a crime of violence as the Guidelines now define that phrase and that his counsel should have recognized as much. Without holding an evidentiary hearing, the district court denied relief, reasoning that even if Hobbs Act robbery is not a crime of violence, there was no binding circuit precedent on that issue and that the Sixth Amendment does not require defense counsel to anticipate future developments in case law.

II. Legal Standard

The federal habeas statute permits a prisoner to petition for relief "upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States." 28 U.S.C. § 2255(a). The trial court must "grant a prompt hearing" on the petition unless the motion and record "conclusively show that the prisoner is entitled to no relief." § 2255(b). When reviewing the denial of a federal prisoner's § 2255 petition, we review the district court's legal conclusions de novo, its factual findings for clear error, and its decision to deny an evidentiary hearing for abuse of discretion. Martin v. United States , 789 F.3d 703, 705 (7th Cir. 2015). A decision that rests on an error of law is always an abuse of discretion. Id. at 706.

III. Hobbs Act Robbery Is Not a Guidelines "Crime of Violence"

The career offender guideline applies here only if an offense of a conviction is a "crime of violence" as defined in the Guidelines. See U.S.S.G. §§ 4B1.1 ; 4B1.2. Before 2015, there would have been no doubt that Hobbs Act robbery was a crime of violence. That year, however, the Supreme Court decided Johnson v. United States , 576 U.S. 591, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015), which held that the so-called residual clause in the statutory definition of a "violent felony" in 18 U.S.C. § 924(e)(2)(B) was unconstitutionally vague. In response to uncertainty about whether Johnson would apply to the similar—but advisory—definition in the Sentencing Guidelines, the U.S. Sentencing Commission amended the guideline definition of "crime of violence" to remove the similar residual clause and otherwise narrow its scope. See Guidelines Amendment 798; U.S.S.G. § 4B1.2 (as amended August 1, 2016). The new definition, which applied to Bridges, provides:

The term "crime of violence" means any offense under federal or state law, punishable by imprisonment for a term exceeding one year, that—

(1) has as an element the use, attempted use, or threatened use of physical force against the person of another, or

(2) is murder, voluntary manslaughter, kidnapping, aggravated assault, a forcible sex offense, robbery, arson, extortion, or the use or unlawful possession of a firearm described in 26 U.S.C. § 5845(a) or explosive material as defined in 18 U.S.C. § 841(c).

Even though Hobbs Act robbery is one of the two most common federal robbery crimes (the other is bank robbery under 18 U.S.C. § 2113 ), it does not qualify as "robbery" within the meaning of the new guideline definition of crimes of violence. The heart of the problem is that a person may commit Hobbs Act robbery by threatening physical violence against any "person or property ." 18 U.S.C. § 1951(b)(1). The guideline definition of robbery is narrower, as it is understood to require a threat of physical violence against a person. A threat to property does not count. Under the categorical method used to decide such questions, the broader definition of Hobbs Act robbery means that it does not count as "robbery" under Guideline Amendment 798.

Following Amendment 798, every federal appellate court to address the issue squarely has concluded that Hobbs Act robbery does not fall within the narrow definition of "crime of violence" in Amendment 798 and codified in § 4B1.2. United States v. O'Connor , 874 F.3d 1147 (10th Cir. 2017) ; United States v. Camp , 903 F.3d 594 (6th Cir. 2018) ; United States v. Eason , 953 F.3d 1184 (11th Cir. 2020) ; see also United States v. Rodriguez , 770 F. App'x 18 (3d Cir. 2019) (non-precedential); United States v. Edling , 895 F.3d 1153 (9th Cir. 2018) (recognizing impact of Amendment 798 on similar Nevada robbery statute); but see United States v. Andrews , 768 F. App'x 189 (5th Cir. 2019) (non-precedential) (declining to revisit prior circuit precedent regarding a similar California robbery statute after Amendment 798). We may summarize our colleagues' careful and persuasive analysis of the issue.

In determining whether an offense of conviction, here Hobbs Act robbery, qualifies as a crime of violence under the Guidelines, we use the so-called categorical approach. United States v. Campbell , 865 F.3d 853, 855 (7th Cir. 2017) ; see generally Camp , 903 F.3d at 599. Under the categorical approach, which also applies to the Armed Career Criminal Act, see 18 U.S.C. § 924(e), sentencing courts may look only to the statutory definitions—i.e., the elements—of a defendant's offenses, and not to the actual facts underlying those convictions. Descamps v. United States , 570 U.S. 254, 261, 133 S.Ct. 2276, 186 L.Ed.2d 438 (2013). The categorical method presumes that a conviction rests on the least serious acts that would satisfy the statute, regardless of the offender's actual conduct. Campbell , 865 F.3d at 856, citing Johnson v. United States , 559 U.S. 133, 137, 130 S.Ct. 1265, 176 L.Ed.2d 1 (2010). If Hobbs Act robbery can be committed in a way that falls outside the definition of a "crime of violence" in U.S.S.G. § 4B1.2, then it is not a categorical fit, regardless of how the defendant actually committed his robberies. The Hobbs Act criminalizes a broad range of criminal behavior under the "robbery" banner:

The term ‘robbery’ means the unlawful taking or obtaining of personal property from the person or in the presence of another, against his will, by means of actual or threatened force, or violence, or fear of injury, immediate or future, to his person or property , or property in his custody or possession, or the person or property of a relative or member of his family or of anyone in his company at the time of the taking or obtaining.

18 U.S.C. § 1951(b)(1) (emphases added).

The phrase "crime of violence" is defined in § 4B1.2 in two ways. The first is where the offense "has as an element the use, attempted use, or threatened use of physical force against the person of another." Hobbs Act robbery does not satisfy this first alternative under the guideline definition because a use of force or a threat of force against property is sufficient.

The second part of the definition lists a number of crimes: "murder, voluntary manslaughter, kidnapping, aggravated assault, a forcible sex offense, robbery, arson, extortion, or the use or unlawful possession" of a machine gun or explosive. § 4B1.2(a)(2). We focus here on whether Hobbs Act robbery is a categorical fit for guideline robbery or extortion.

The application notes to § 4B1.2 do not define robbery, so we turn to the "generic" definition of the crime. Klikno v. United States , 928 F.3d 539, 542 (7th Cir. 2019), quoting Descamps , 570 U.S. at 257, 133 S.Ct. 2276. The generic definition of robbery does not include threats of force against property. O'Connor , 874 F.3d at 1154–55 (collecting cases, discussing the approach of "a significant majority of states," and reviewing criminal law treatises); accord Camp , 903 F.3d at 601–02 (generic robbery includes "the requirement that the danger be both immediate and to a person ") (emphasis added); Eason , 953 F.3d at 1193–94 ; see also Klikno , 928 F.3d at 544–47, discussing Stokeling v. United States , ––– U.S. ––––, 139 S. Ct. 544, 552, 202 L.Ed.2d 512 (2019) (recognizing that generic robbery is distinct from larceny in that it requires "force that overcomes a victim's resistance"); Shabazz v. United States , 912 F.3d 73, 77 (2d Cir. 2019) ("Scholars of the criminal law underline the inherent potential for physical harm to the victim as the explanation why robbery developed as, and continues to be treated as, an aggravated felony, generally carrying harsher punishments than other forms of larceny.") (emphasis added). Hobbs Act robbery is not a categorical match for generic robbery because it can be committed through threats against property.

The government has suggested that a Hobbs Act robbery that threatens only property might still qualify as extortion in the Guideline's list of enumerated crimes of violence, triggering the career offender enhancement. The application notes define extortion as "obtaining something of value from another by the wrongful use of (A) force, (B) fear of physical injury, or (C) threat of physical injury." U.S.S.G. § 4B1.2, cmt. 1. The central issue here is whether fear or threat of "physical injury" could fit the Hobbs Act robbery branch allowing conviction for harm or threats to property. That depends on how broadly "physical injury" is defined. The government argues that "threat of physical injury" means any threat to harm a physical object, be it animal, vegetable, or mineral. Bridges counters that "physical injury" simply means injury to a person.

We agree with our colleagues in other circuits who have rejected the government's argument. Extortion does not fill the categorical gaps for Hobbs Act robbery because the Guidelines use "physical injury" to mean injury to a person. As the Tenth Circuit explained in O'Connor : "The Guidelines contain several examples where ‘physical injury’ is distinguished from property damage or where the phrase clearly refers to injury to a person." 874 F.3d at 1157 & n.17, discussing U.S.S.G. §§ 5K2.2, 2C1.1(c)(3), 2J1.2(b)(1)(B), and 5K2.12. Moreover, the common meaning of the phrase "physical injury" is "bodily injury." Injury , Black's Law Dictionary (11th ed. 2019) (defining "physical injury" as "bodily injury," that is, "physical damage to a person's body"); O'Connor , 874 F.3d at 1157 (quoting 10th edition of Black's); accord Eason , 953 F.3d at 1194 ; Camp , 903 F.3d at 603. Hobbs Act robbery thus extends more widely than guideline extortion.

The government argues next that this definition of "extortion" must be wrong because it would be redundant, overlapping completely with generic robbery. We are not persuaded. The two are not identical. Generic robbery requires that the threat of violence be imminent; guideline extortion does not. See Camp , 903 F.3d at 601 (discussing immediacy requirement of generic robbery). In addition, a textual argument based on avoiding redundancy is not necessarily a show-stopper, as "the presence of some redundance is rarely fatal on its own to a statutory reading." White v. United Airlines , 987 F.3d 616, 622 (7th Cir. 2021). We have accordingly cautioned that drafters of legal documents may "intentionally err on the side of redundancy to ‘capture the universe.’ " See Sterling National Bank v. Block , 984 F.3d 1210, 1218 (7th Cir. 2021), quoting Abbe R. Gluck & Lisa Schultz Bressman, Statutory Interpretation from the Inside—an Empirical Study of Congressional Drafting, Delegation, and the Canons: Part I , 65 Stan. L. Rev. 901, 934 (2013).

In sum, Hobbs Act robbery criminalizes threats against property, and both generic robbery and guideline extortion reach only threats against persons. Hobbs Act robbery is not a categorical fit, so Bridges was not convicted of a crime of violence as the Guidelines define the phrase. The career offender guideline should not have been applied.

IV. Effective Assistance of Counsel

For Bridges, it is not enough to show that a challenge to the career offender enhancement should have prevailed. The issue is whether his defense lawyer failed to provide effective assistance of counsel in 2018 by stipulating to the enhancement as part of the plea agreement rather than objecting to it.

The government argues that counsel's performance did not fall below the constitutional threshold because all this hair-splitting under the categorical approach is so convoluted and counterintuitive. Competent counsel, according to the government, need not ask and answer legal koans, such as when is a robbery not a robbery? Indeed, federal defenders in this circuit did not regularly raise this argument until well after Bridges was sentenced. Bridges counters that the categorical approach is a well-known tool in applying the Guidelines, and as categorical approach cases go, this analysis is straightforward. He also asserts that counsel should have been on the lookout for categorical approach problems given the recent Amendment 798, the dramatic impact of this factor on his guideline range, and cases around the country where counsel did recognize and raise these arguments. We conclude that under these facts Bridges is at least entitled to a hearing.

The Sixth Amendment right to counsel applies during plea negotiations. E.g., Lafler v. Cooper , 566 U.S. 156, 162, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012). We use the familiar two-part Strickland v. Washington test to determine whether counsel's performance fell below the constitutional minimum. 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Under this rubric, counsel was ineffective if both "counsel's performance was deficient" and "the deficient performance prejudiced the defense." Id. at 687, 104 S.Ct. 2052. The performance issue is whether "counsel made errors so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment." Id.

A. Performance

Bridges is entitled to a hearing unless the record and motion "conclusively show that [he] is entitled to no relief." 28 U.S.C. § 2255(b). The district court reasoned that the lack of controlling precedent in this circuit defeated Bridges' Sixth Amendment argument. We respectfully disagree. In some circumstances, defense counsel may be required to anticipate arguments foreshadowed but not yet adopted by existing case law. Based on the timeline and allegations in the petition, we cannot say that this claim is conclusively foreclosed. Bridges is entitled to try to prove his allegations at a hearing.

We apply an objective standard of reasonableness considering all the circumstances. Hinton v. Alabama , 571 U.S. 263, 272–73, 134 S.Ct. 1081, 188 L.Ed.2d 1 (2014). Deference must be given to counsel's strategic decisions, but "strategic choices made after less than complete investigation [of law and facts] are reasonable precisely to the extent that reasonable professional judgments support the limitations on investigation. In other words, counsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary." Id. at 274, 134 S.Ct. 1081, quoting Strickland , 466 U.S. at 690–91, 104 S.Ct. 2052. "In the plea bargaining context, reasonably competent counsel will attempt to learn all of the facts of the case, make an estimate of a likely sentence, and communicate the results of that analysis before allowing his client to plead guilty." Gaylord v. United States , 829 F.3d 500, 506 (7th Cir. 2016) (quotation omitted); see also Anderson v. United States , 981 F.3d 565, 575–76 (7th Cir. 2020) (remanding habeas petition for evidentiary hearing where counsel's failure to investigate one available theory of defense before negotiating plea deal could constitute deficient performance).

"An attorney's ignorance of a point of law that is fundamental to his case combined with his failure to perform basic research on that point is a quintessential example of unreasonable performance under Strickland ." Hinton , 571 U.S. at 274, 134 S.Ct. 1081 (describing examples); see Osagiede v. United States , 543 F.3d 399, 409 (7th Cir. 2008) ("All lawyers that represent criminal defendants are expected to know the laws applicable to their client's defense," and where "simple computer research" would turn up those laws, counsel is ineffective for failing to rely on them absent a strategic justification) (cleaned up); see also Cates v. United States , 882 F.3d 731, 736 (7th Cir. 2018) ("[A] mistake of law is deficient performance.").

Defense attorneys, it is true, are generally not obliged to anticipate changes in the law. E.g., Valenzuela v. United States , 261 F.3d 694, 700 (7th Cir. 2001). Yet there are some circumstances where they may be obliged to make, or at least to evaluate, an argument that is sufficiently foreshadowed in existing case law. Shaw v. Wilson , 721 F.3d 908, 917 (7th Cir. 2013) ; see also United States v. Carthorne , 878 F.3d 458, 465–66 (4th Cir. 2017) ("[T]he ineffective assistance standard may require that counsel raise material issues even in the absence of decisive precedent."), citing Ramirez v. United States , 799 F.3d 845, 855 (7th Cir. 2015).

If proved, counsel's apparent failure to investigate or raise a challenge to the career offender enhancement here could be deemed to have been deficient performance. There was no binding in-circuit precedent on this exact issue, but case law sufficiently foreshadowed this argument, which had been brought to the forefront by both a recent amendment to the Guidelines and several court of appeals decisions interpreting Amendment 798. Bridges is entitled to a "prompt" hearing to prove his allegations. 28 U.S.C. § 2255(b).

When Bridges was pleading guilty and being sentenced, the categorical approach was of course well-established. See generally Taylor v. United States , 495 U.S. 575, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990). So was its applicability to the Guidelines. E.g., United States v. Woods , 576 F.3d 400, 403 (7th Cir. 2009) ("In deciding whether a crime fits [ U.S.S.G. § 4B1.2's definition of crime of violence], the Supreme Court has instructed lower courts to use a categorical approach."). The overwhelming majority of federal defendants plead guilty. Administrative Office of U.S. Courts, Statistical Tables for the Federal Judiciary , Table D-4 (Dec. 31, 2018) (89.9% of federal criminal cases end in guilty plea). And the Guidelines remain the "lodestar" of federal sentencing. Molina-Martinez v. United States , ––– U.S. ––––, 136 S. Ct. 1338, 1346, 194 L.Ed.2d 444 (2016). Sentencing advocacy, both during plea negotiations and before a district court, is in most cases the most important task of a federal criminal defender. Understanding how to interpret the Guidelines and making guideline arguments are core competencies of federal defense counsel.

Competent counsel would also know that the categorical approach frequently produces counterintuitive results and has been the subject of much judicial handwringing. See Amit Jain & Phillip Dane Warren, An Ode to the Categorical Approach , 67 U.C.L.A. L. Rev. Disc. 132, 134 (2019) (collecting colorful criticisms before defending categorical approach on the merits). Many of these criticisms acknowledge that the categorical approach frequently benefits defendants. For example, Judge Wilkinson has decried it as "a protracted ruse for paradoxically finding even the worst and most violent offenses not to constitute crimes of violence." United States v. Doctor , 842 F.3d 306, 313 (4th Cir. 2016) (Wilkinson, J., concurring). The need to investigate whether Bridges had a viable categorical approach argument was especially pronounced in this case because the Sentencing Commission had narrowed its definition of "crime of violence" months before Bridges committed these robberies, and the enhancement nearly tripled the advisory sentencing range. That effect made this a crucial issue to investigate. United States v. Winstead , 890 F.3d 1082, 1089 (D.C. Cir. 2018) ; see also Wilson v. Gaetz , 608 F.3d 347, 352 (7th Cir. 2010) ("Especially because of the severity of the penalty that [defendant] was facing, [counsel] should have done more."). Reasonable counsel would investigate such a new and timely issue, known to benefit criminal defendants, especially considering the dramatic effect on the guideline range.

We have previously found that counsel's failure to compare statutory definitions in resolving a guideline question during plea negotiations would, if proven, constitute deficient performance. In Brock-Miller v. United States , 887 F.3d 298, 309–10 (7th Cir. 2018), we remanded for a hearing on allegations of deficient performance. We rejected the government's argument that examining the "statute to determine if it met the federal definition of felony drug offense" was too "sophisticated" and "convoluted" to expect from defense counsel. Id. at 310–11. So too here. The categorical approach may be counterintuitive and artificial, but, like other complex factual and legal inquiries that regularly arise in criminal defense, competent counsel need to understand it and be able to use it. See Hinton , 571 U.S. at 274, 134 S.Ct. 1081 ; see also, e.g., Dunn v. Jess , 981 F.3d 582, 592–94 (7th Cir. 2020) (deficient performance where counsel failed to assimilate conflicting medical reports); United States v. Sepling , 944 F.3d 138, 146–50 (3d Cir. 2019) (deficient performance where counsel failed both to investigate whether the probation officer had correctly listed 3,4-Methylenediox-ymethamphetamine as the appropriate drug analog for methylone and to debunk the medical research underlying the Sentencing Commission's recommendation of a severe sentence for trafficking either drug); Shaw , 721 F.3d at 917 (deficient performance where counsel failed to argue that state appellate courts had misunderstood an Indiana Supreme Court holding about a fine point of state procedural rules). As in Brock-Miller , if Bridges demonstrates on remand that he entered into the plea agreement based on advice of counsel that fell below constitutional standards, the district court should not hold him to his pleas. 887 F.3d at 310, citing Hurlow v. United States , 726 F.3d 958, 966–67 (7th Cir. 2013).

With modern methods of legal research, it would not have taken long in 2018 for counsel to have found the Tenth Circuit decision holding that Hobbs Act robbery is not a crime of violence under the 2016 guideline amendment. The Tenth Circuit published O'Connor months before Bridges signed his plea agreement. We cannot say that the record "conclusively shows" that reasonable counsel would not have found this noteworthy published opinion, one of the few at the time of Bridges' plea that considered the effects of Amendment 798. See Osagiede , 543 F.3d at 409.

O'Connor was not the only federal appellate opinion discussing the new guideline definition of crime of violence or the Hobbs Act. In Edling , the Ninth Circuit noted O'Connor in concluding that guideline extortion does not categorically match a state robbery statute. 895 F.3d at 1158. Counsel checking citations to O'Connor , the guideline definition of "crime of violence," or the phrase "Hobbs Act" would have also found the Fourth Circuit's opinion in United States v. McCollum , 885 F.3d 300 (4th Cir. 2018), issued just one month before Bridges signed his plea agreement. That opinion noted that O'Connor "analyzed whether Hobbs Act robbery is a crime of violence under § 4B1.2(a)(2)." Id. at 306 n.6. On the merits, McCollum applied the categorical method to the new guideline definition of crime of violence to hold that conspiring to commit murder in aid of racketeering is not a crime of violence under § 4B1.2 as amended. Id. at 303. That prompted a noteworthy dissent calling for divine guidance and intervention, id. at 310, 314 (Wilkinson, J., dissenting), and a concurrence that would have been satisfied with help from Congress, the Supreme Court, or perhaps the Sentencing Commission. See id . at 309 (Traxler, J., concurring).

The case for ineffective assistance of counsel here appears to be stronger than other similar cases concluding that failures to raise new but well-supported arguments were deficient performance. The most directly analogous case is Cuthbertson v. United States , 833 F. App'x 727 (10th Cir. 2020) (non-precedential), which concerns the same issue—whether counsel was ineffective for failing to recognize that Hobbs Act robbery is not a crime of violence under the amended § 4B1.2 definition. In Cuthbertson , though, the defendant was sentenced even before O'Connor was issued. The Tenth Circuit still found counsel's performance wanting. 833 F. App'x at 733–34. We need not reach that far here because O'Connor had already laid out that path when Bridges pleaded guilty and was sentenced.

In that respect, this case is also more straightforward than United States v. Winstead , in which the D.C. Circuit found counsel ineffective for failing to raise an administrative law argument against applying the career offender guideline. 890 F.3d at 1090. Unlike this case, however, there was no out-of-circuit precedent supporting the criminal defendant's position in Winstead . In fact, all the circuits that had addressed Winstead's issue at the time of his sentencing had rejected his argument. See id. at 1091 ; see also United States v. Adams , 934 F.3d 720, 728–29 (7th Cir. 2019) (declining to follow Winstead because of binding circuit precedent). The D.C. Circuit relied, as we do here, on the fact that the career offender guideline dramatically increased the advisory sentence, so that the issue deserved more attention than counsel had apparently given it. Winstead , 890 F.3d at 1090.

The government argues here that defense counsel in this circuit were not raising this issue in 2018, so that Bridges' lawyer's omission of this guideline argument was not objectively unreasonable. That argument may carry some weight, but we do not view it as conclusive. Even in the context of "doubly deferential" review under the Antiterrorism and Effective Death Penalty Act, see Burt v. Titlow , 571 U.S. 12, 15, 134 S.Ct. 10, 187 L.Ed.2d 348 (2013) ; 28 U.S.C. § 2254(d)(1), we have explained that a lawyer's "fail[ure] to take advantage of a clear avenue of relief for her client is no less concerning because many others made the same error." Jones v. Zatecky , 917 F.3d 578, 580 (7th Cir. 2019) ; see generally Shaw , 721 F.3d at 917 (despite double deference, appellate counsel's failure to raise issue under largely moribund state statute was deficient performance where some counsel and trial courts had recognized the issue and counsel offered no comparably strong alternative arguments on appeal). We see no reason to hold federal defenders in Indianapolis, Chicago, and Milwaukee to a lower standard than those in Kansas City, Las Vegas, Ann Arbor, and Brooklyn, who, after all, must apply the same Supreme Court precedents to the same Guidelines Manual. See Shaw , 721 F.3d at 917.

In this case, the erroneous application of the career offender enhancement almost tripled the low end of the guideline range from 57 months to 151 months. Bridges has shown that counsel's performance may have been deficient in not raising this issue, following the recent Amendment 798, which substantially narrowed the definition of "crime of violence," and the well-known counterintuitive and defendant-friendly results produced by the categorical approach.

Accordingly, the district court's decision not to hold a hearing was an abuse of discretion because it rested on the erroneous legal conclusion that counsel's failure to anticipate this career offender argument could not constitute ineffective assistance of counsel in the absence of controlling circuit precedent. See Brock-Miller , 887 F.3d at 304 ; Martin , 789 F.3d at 706. To be clear: we are not holding as a matter of law that counsel was ineffective. It is possible that defense counsel had a reasonable sentencing strategy that deserves deference. See Anderson , 981 F.3d at 578 (remanding for hearing); Brock-Miller , 887 F.3d at 311 (same); Osagiede , 543 F.3d at 411 (same).

We must emphasize that "[j]udicial scrutiny of counsel's performance must be highly deferential," as Strickland and its countless progeny have made clear. 466 U.S. at 689, 104 S.Ct. 2052. Strategic choices made after a thorough investigation are "virtually unchallengeable." Id. at 690, 104 S.Ct. 2052. Yet deferential review does not mean no review: a "court adjudicating a Strickland claim can't just label a decision ‘strategic’ and thereby immunize it from constitutional scrutiny." Jones v. Calloway , 842 F.3d 454, 464 (7th Cir. 2016). In any case, one must evaluate a defense counsel's performance by considering the other options available. Compare, e.g., Brock-Miller , 887 F.3d at 302, 311 (remanding § 2255 action for hearing where defense counsel had raised frivolous objection to sentencing enhancement under 21 U.S.C. § 851 but overlooked meritorious objection), with Harris v. United States , 366 F.3d 593, 596 (7th Cir. 2004) (counsel not deficient for strategically declining to argue for "safety valve" reduction in guideline range where its applicability was an open question and pushing that theory would detract from defendant's mitigation arguments); see also Vinyard v. United States , 804 F.3d 1218, 1225 (7th Cir. 2015) (discussing similarities between deference afforded trial and appellate counsel's strategic decisions; "winnowing down possible claims to those an attorney believes are strongest is a classic example of a strategic decision").

In one Hobbs Act robbery case, other available theories for defense or mitigation may seem solid enough that a lawyer might reasonably decide, after a full investigation, not to pursue a technical and counterintuitive guideline argument asking the court to break new legal ground. Such a strategy might be particularly appropriate before a judge noted for minimizing the role of guideline calculations vis-à-vis the more general considerations in 18 U.S.C. § 3553(a), where such a guideline argument might also seem unlikely to gain much traction. Other judges may have well-known antipathy to the categorical approach and regularly deviate from the Guidelines' recommendations based on this policy disagreement. Effective defense counsel might reasonably choose not to pursue this guideline issue.

But in other cases, where other defenses and mitigation arguments offer little traction, a failure to raise this available argument that would cut the guideline sentencing range by more than half could be deemed ineffective. Moreover, a theory that counsel sought to prioritize mitigation arguments over guideline arcana cannot be considered in a vacuum. The defense here proposed a sentence of 108 months, which was 50% higher than the high end of the proper guideline range. And even discounting for the benefits of hindsight, counsel's argument that his 62-year-old client with a 45-year criminal history, fresh off this robbery spree, would age out of crime at age 68 never could have looked promising. Likewise, refusing to do needed research because it "would have been extremely difficult to do and time consuming" does not necessarily excuse deficient performance, Ramirez , 799 F.3d at 855, and the research needed here does not seem to have been particularly difficult or onerous in any event. See also Washington v. Smith , 219 F.3d 620, 631 (7th Cir. 2000) (performance was ineffective where counsel failed to investigate because of "busy" schedule).

B. Prejudice

The government has rightly conceded that the district court's denial of habeas relief without a hearing cannot be affirmed based solely on lack of prejudice. We briefly explain why. To prove ineffective assistance of counsel, the defendant must show "that there is a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different." Strickland , 466 U.S. at 694, 104 S.Ct. 2052. In the context of federal sentencing, "the Guidelines are not only the starting point for most federal sentencing proceedings but also the lodestar." Molina-Martinez , 136 S. Ct. at 1346 (reviewing caselaw and statistical reports). "In most cases a defendant who has shown that the district court mistakenly deemed applicable an incorrect, higher Guidelines range has demonstrated a reasonable probability of a different outcome." Id.

Here, the record shows that the district court and defense counsel treated the erroneous range of 151 to 188 months as the starting point. No evidence suggests that the district court would have imposed a 140-month sentence even if the advisory guideline had a high end of 71 months. See United States v. Marks , 864 F.3d 575, 576 (7th Cir. 2017) ; see also Molina-Martinez , 136 S. Ct. at 1346 (listing factors that may rebut presumption of reasonable probability of prejudice).

In this case, the district judge looked beyond the Guidelines, as she should have under § 3553(a). See also Gall v. United States , 552 U.S. 38, 128 S.Ct. 586, 169 L.Ed.2d 445 (2007) ; United States v. Booker , 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005). "Sentencing convicted offenders is generally recognized as the most difficult part of the job of a United States District Judge," United States v. Mzembe , 979 F.3d 1169, 1176–77 (7th Cir. 2020), and this case was especially challenging. Bridges was older than roughly 98% of all other federally sentenced robbers in the year he was sentenced. U.S.S.C., Sourcebook of Federal Sentencing Statistics Table 7 (Fiscal Year 2018). He had a long and disturbing criminal history dating back to his involvement in a home invasion and rape and robbery as a teenager. Since those crimes, he had been convicted of a bevy of other crimes, including criminal confinement and several other robberies. Without downplaying his robbery spree in this case, a judge might still recognize that his Hobbs Act robberies did not display the violence and cruelty of previous crimes for which Bridges has already been punished. His case also features significant mitigating circumstances, including demonstrated progress in combatting his addictions and learning disabilities. In other words, the statutory factors that the sentencing court must consider in imposing a sentence point strongly in different directions. See 18 U.S.C. § 3553(a). The district court thoughtfully engaged with these facts in a nuanced way.

We have encouraged district courts facing seemingly arbitrary or difficult guideline issues to ask, "Why should anyone care?" Marks , 864 F.3d at 582. The district court seemingly did so here in attempting to consider where the 140-month sentence would fall if the career offender guideline did not apply. If the court had computed the counterfactual guideline range properly, counsel's possibly deficient performance might well have withstood scrutiny. See Molina-Martinez , 136 S. Ct. at 1346. Unfortunately, however, the district court's calculation was erroneous, dropping the criminal history category from VI to IV, but without also accounting for the large drop in the offense level, from 29 to 21. The court thus mistakenly thought that the 140-month sentence it imposed would have been within the guideline range if the career offender enhancement had not applied.

Cases such as this illustrate the importance of interrogating the Guidelines when unintuitive distinctions have a significant impact on the advisory guideline range. The Guidelines are a comprehensive and detailed system that accounts for the seriousness of criminal conduct and the defendant's criminal record. Such a system necessarily includes some arbitrary rules and cutoffs. Marks , 864 F.3d at 582. Sentencing courts are required to consider that advice, but they must also consider the other factors listed in 18 U.S.C. § 3553(a) in crafting a "sentence sufficient, but not greater than necessary, to comply with" the statutorily enumerated purposes of sentencing: punishment, deterrence, incapacitation, and rehabilitation. District judges recognize that the career offender guideline frequently represents one such arbitrary cutoff. See U.S.S.C., The Influence of the Guidelines on Federal Sentencing at *55–56 (2020) (the career offender guideline has "been the subject of longstanding concern and has among the lowest within-guideline rates each year"; the average sentence for someone who has had the career offender guideline applied is five and a half years shorter than the advisory guideline minimum). The categorical approach grafts another arbitrary cutoff onto this already problematic guideline.

That should not be a problem if sentencing courts follow the process the district court took in this case: recognize that the career offender guideline can significantly, and perhaps arbitrarily, ratchet up an advisory range and attempt to craft a sentence accounting for that discontinuity. See Gall , 552 U.S. at 49–50, 128 S.Ct. 586. Where appropriate, such a move could be justified by reference to either the § 3553(a) factors as they apply in the particular case or a broader but reasonable policy disagreement with the Guidelines. See United States v. Corner , 598 F.3d 411, 416 (7th Cir. 2010) (en banc), discussing, among other cases, Spears v. United States , 555 U.S. 261, 129 S.Ct. 840, 172 L.Ed.2d 596 (2009) (per curiam), Kimbrough v. United States , 552 U.S. 85, 128 S.Ct. 558, 169 L.Ed.2d 481 (2007), and Gall , 552 U.S. 38, 128 S.Ct. 586, 169 L.Ed.2d 445.

We recognize that this decision further elevates the abstract and artificial categorical approach over the known facts of the case. But this method ensures that dramatic sentencing enhancements are applied to only those defendants who clearly fall within their intended scope. See United States v. Sonnenberg , 628 F.3d 361, 367 (7th Cir. 2010). A district court finding that a guideline definition is abstract, arbitrary, or arduous should ask itself, with the § 3553(a) factors in mind, why anyone should care how the answer comes out. "We are ill-equipped to speculate" as to how the district court will answer that question at a hypothetical future resentencing. United States v. McGee , 985 F.3d 559, 564 (7th Cir. 2021).

The judgment in favor of respondent is REVERSED and this case is REMANDED for further proceedings consistent with this opinion.