After leading police on a high-speed chase, Le'Ardrus Burris rammed his vehicle into a police officer, "throwing him backwards for 15 feet." R. 496, Pg. ID 3684. As a result of this action, an Ohio jury convicted Burris of felonious assault. A casual reader of today's decision might struggle to understand why we are even debating if ramming a vehicle into a police officer is a crime of violence. The reader's struggle would be understandable. The time has come to dispose of the long-baffling categorical approach. My colleagues have repeatedly said as much, and I urge any frustrated readers of today's decision-particularly Congress or the Sentencing Commission-to read and heed their words. See, e.g. , Mathis v. United States , --- U.S. ----, 136 S.Ct. 2243, 2266-71, 195 L.Ed.2d 604 (2016) (Alito, J., dissenting); Cradler v. United States , 891 F.3d 659, 672 (6th Cir. 2018) (Kethledge, J., concurring); United States v. Valdivia-Flores , 876 F.3d 1201, 1210-11 (9th Cir. 2017) (O'Scannlain, J., specially concurring); United States v. Davis , 875 F.3d 592, 595, 604 (11th Cir. 2017) (Ed Carnes, C.J.); United States v. Chapman , 866 F.3d 129, 136-39 (3d Cir. 2017) (Jordan, J., concurring); United States v. Faust , 853 F.3d 39, 60-61 (1st Cir. 2017) (Lynch, J., concurring); United States v. Tavares , 843 F.3d 1, 19 (1st Cir. 2016) (Kayatta, J.); United States v. Doctor , 842 F.3d 306, 312-19 (4th Cir. 2016) (Wilkinson, J., concurring). I write to join them and to suggest a possible path forward.
As the jurists before me have explained, there are two big problems with the categorical approach. First, the categorical approach is an elements lottery that leads to arbitrary results. Second, it is very difficult to administer. Both problems are on display here.
*408The elements lottery. Under the categorical approach, whether a prior state conviction qualifies as a crime of violence is a matter of chance that depends on how a state legislature and its judiciary have defined and interpreted the crime. This leads to some absurd results. For example, a North Carolina conviction for "knowingly discharging a firearm into an occupied building" or for raping a "mentally disabled person" is not a crime of violence. Doctor , 842 F.3d at 315 (Wilkinson, J., concurring). In Florida, a conviction for attempted second degree murder is not a crime of violence, and in Oklahoma, a conviction for kidnapping is not. Id.
The absurd results are even more apparent in our circuit. As the majority opinion correctly notes, if you are in Cincinnati, Ohio, and you "cause serious physical harm to another," it is not a crime of violence. But if you drive one mile across the Ohio River and commit the very same crime in Kentucky, it is a crime of violence-all because the analogous statute in Kentucky defined serious physical harm a little differently. See Ky. Rev. Stat. Ann. § 508.040 ; United States v. Maynard , 894 F.3d 773, 775 (6th Cir. 2018) ; United States v. Colbert , 525 F. App'x 364, 369-70 (6th Cir. 2013). So an increased sentence in this circuit turns not on whether a defendant's aggravated assault was actually violent, but on where the assault took place. In the future, those who have prior convictions for aggravated assault under subsection (A)(1) in Ohio will not be labeled a career offender, but those who have an identical conviction in Kentucky for the exact same conduct will be.
But the arbitrariness does not stop there. Each circuit decides for itself whether a crime qualifies as a crime of violence. For example, subsection (A)(1) of the Ohio felonious assault statute is no longer a crime of violence in this circuit, but it may still be a crime of violence in other circuits. United States v. Clark , 131 F. App'x 29, 30 (4th Cir. 2005) (per curiam). In other words, whether a defendant commits felonious assault in Ohio could matter less than where he commits a later crime. Id. ; see also, e.g. , United States v. Harris , 844 F.3d 1260, 1262 (10th Cir. 2017) (observing that of eleven circuits to have considered whether robbery is a "violent felony," five say no, and six say yes), cert. denied , --- U.S. ----, 138 S.Ct. 1438, 200 L.Ed.2d 716 (2018) ; compare United States v. Fluker , 891 F.3d 541, 549 (4th Cir. 2018) (holding that Georgia's robbery statute is indivisible and not a crime of violence), with United States v. Cooper , 689 F. App'x 901, 906-07 (11th Cir. 2017) (holding that Georgia's robbery statute is divisible and thus that robbery by sudden snatching is a crime of violence). So here again, geography matters more than actual violence under the categorical approach.
And yet arbitrariness abounds even within single states. In Ohio, subsection (A)(1) of the felonious and aggravated assault statutes is no longer a crime of violence, but subsection (A)(2) is. As the majority opinion properly explains, these subsections set forth separate but similar crimes. Each prohibits knowingly causing physical harm to another. The difference is that subsection (A)(1) applies when the defendant caused serious physical harm, and subsection (A)(2), requiring only physical harm, applies when the defendant used a deadly weapon or dangerous ordnance. Compare Ohio Rev. Code Ann. § 2903.11(A)(1), with id. § 2903.11 (A)(2). After today's decision, a person's prior conviction for felonious assault would not be a violent felony if he had caused the victim to have a seizure and a serious brain injury by violently striking her dozens of times with his fists. See State v. Perez , No. WD-17-017, 2018 WL 2278169, at *1 (Ohio Ct. App. May 18, 2018). But if *409he had instead caused a bruise on the victim's leg by striking her with a crowbar, it likely would be a violent felony. See State v. Smith , No. 14 CA 83, 2015 WL 1884222, at *2 (Ohio Ct. App. Apr. 24, 2015). One must accept some level of arbitrariness with any law, but not the amount here, which a fact-based approach would avoid.
Administrative problems. The arbitrariness of the categorical approach ties in with its second flaw: the costs of administering the approach outweigh its benefits. Some initially thought that avoiding an inquiry into the facts of a prior conviction would be easier for judges. Taylor v. United States , 495 U.S. 575, 601, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990). And, to be sure, looking into the facts behind a prior conviction could be difficult in some cases where state records are unclear. This was particularly so in the pre-digital age in which the categorical approach emerged, when electronic filing was not commonplace. But whatever administrative difficulty a factual approach still hazards, the categorical approach ups the ante. By "simply swapp[ing] factual inquiries for an endless gauntlet of abstract legal questions," the categorical approach requires judges to throw away common sense. Doctor , 842 F.3d at 313 (Wilkinson, J., concurring); see Davis , 875 F.3d at 595. Each categorical-approach case (and there is no shortage of them) instead requires the judge to (1) mull through any number of hypothetical ways to commit a crime that have nothing to do with the facts of the prior conviction; (2) mine electronic databases for state court cases (precedential or not) depicting non-violent ways of commission; and (3) scrutinize those state court cases, some of which are old and predate the categorical approach, to determine their import. See Doctor , 842 F.3d at 313 (Wilkinson, J., concurring). Here, this meant that our court (1) considered hypothetical ways to feloniously assault someone without actually striking them, such as by exposing a child to his mother's dead body; (2) poured over countless Ohio Court of Appeals decisions, some decades-old, e.g. , State v. Elliott , 104 Ohio App.3d 812, 663 N.E.2d 412 (1995) ; and (3) tried to decide if one or two of those decisions, almost all involving violence, nevertheless demonstrate that Ohio felonious assault can be committed nonviolently. No judge who has engaged in this "Rube Goldberg jurisprudence" or traveled to this "pretend place" leaves thinking that life has been made easier. Davis , 875 F.3d at 595 ; Tavares , 843 F.3d at 19. And any judge who could end the matter upon finding that a defendant actually committed a violent act would find that task more administrable. This is particularly so when-as here-that question is undisputed.
In light of these problems, I join others in proposing an alternative approach. That approach would permit judges to deem a prior conviction a crime of violence if the underlying criminal conduct was actually violent. If the government can prove that the state court record establishes violent conduct, end the inquiry there. See Mathis , 136 S.Ct. at 2269-70 (Alito, J., dissenting). If the government cannot, or if the record is insufficient to sustain the government's burden of proof, then the prior conviction does not count. See id. (Alito, J., dissenting).
I recognize that a fact-based approach is not without its own potential problems. Any difficulty in proving the facts underlying a past conviction would inure to the benefit of the defendant, meaning a defendant's career-offender status could turn not on the violence of his conduct, but on record-keeping. But it is incumbent on the government to keep good records whenever the government asks a court to put a defendant in prison, and it should be no *410different here. United States v. Dupree , 323 F.3d 480, 491 (6th Cir. 2003) (observing that the government must prove whether a sentencing enhancement applies). And in the new world of electronic filing, many of the records will be readily available going forward. In addition, requiring sentencing courts to probe the facts underlying a conviction could pose Sixth Amendment problems, at least in the ACCA context. See Mathis , 136 S.Ct. at 2259 (Thomas, J., concurring). But whatever the difficulty of those problems under the ACCA, the Sixth Amendment presents no obstacle under the advisory Sentencing Guidelines. United States v. Sanders , 406 F. App'x 995, 997-98 (6th Cir. 2011) ; cf. Beckles v. United States , --- U.S. ----, 137 S.Ct. 886, 892, 197 L.Ed.2d 145 (2017) ("Unlike the ACCA, however, the advisory Guidelines do not fix the permissible range of sentences."). Therefore, Congress and the Sentencing Commission are well-positioned to make a change for the better. I urge them to consider such a change so courts no longer have to "delve into pointless abstract questions" when determining whether striking a police officer with a car is a crime of violence. See Mathis , 136 S.Ct. at 2268 (Alito, J., dissenting).
I respectfully concur.
CONCURRING IN PART AND IN THE JUDGMENT
ROGERS, Circuit Judge.
I concur in the result, and in the following parts of the lead opinion: Parts I and II, all but the last sentence of Part III.B, and Part III.C. For the reasons given by Judge Kethledge, we ought not address in this case whether United States v. Anderson , 695 F.3d 390 (6th Cir. 2012), was rightly decided. The problematic part of Anderson poses no problem for the defendant before the court. To resolve this appeal, we need only conclude that Ohio Rev. Code § 2903.11(A) is divisible and that § 2903.11(A)(2) describes a crime of violence under the Armed Career Criminal Act and the Guidelines. The lead opinion rightly resolves these issues, and I join the portions of the lead opinion that so hold. Burris was properly sentenced as a career offender.
CONCURRING IN THE JUDGMENT
KETHLEDGE, Circuit Judge, concurring in the judgment.
The idea of judicial restraint encompasses not only the content of our decisions, but also whether we should make them in the first place. Advisory opinions, everyone agrees, are out of bounds. See, e.g. , Steel Co. v. Citizens for a Better Env't , 523 U.S. 83, 101, 118 S.Ct. 1003, 140 L.Ed.2d 210 (1998) ; Hayburn's Case , 2 U.S. 408, 2 Dall. 409, 1 L.Ed. 436 (1792). And dictum is usually a bad idea, because judges think differently-more carefully, more focused, more likely to think things through-when our words bring real consequences to the parties before us. (The same criticism goes for alternative holdings.) Yet when judges do indulge in dictum, at least, it typically amounts to only a stray comment or two on matters incidental to the real issue in the case.
But here, respectfully, dictum is the raison d'être for the plurality's opinion. We reheard this case en banc to revisit our decision in United States v. Anderson , 695 F.3d 390 (6th Cir. 2012), which most of the en banc court agrees was mistaken to the extent it held that Ohio Rev. Code § 2903.11(A)(1) describes a "crime of violence" under the Armed Career Criminal Act. That part of Anderson 's holding, of course, would have real consequences in a case where the defendant's sentence was enhanced as a result of a prior conviction under § 2903.11(A)(1). But Burris instead *411has a prior conviction under § 2903.11(A)(2), which does describe a crime of violence under the ACCA. Thus we could decide this case simply by holding (as the plurality correctly holds) that § 2903.11(A) is divisible and that the particular crime of which Burris was previously convicted-namely § 2903.11(A)(2) -is a crime of violence. That is reason enough to deny him relief. The question whether § 2903.11(A)(1) describes a crime of violence, therefore, is not a matter in controversy in this case. Hence the court's discussion of that issue is dictum-indeed dictum so central to the court's opinion as to render it nearly advisory. See, e.g., Chafin v. Chafin , 568 U.S. 165, 172, 133 S.Ct. 1017, 185 L.Ed.2d 1 (2013).
I therefore concur only in the judgment.
CONCURRING IN PART AND DISSENTING IN PART
COLE, Chief Judge, concurring in part, dissenting in part.
Despite growing concerns regarding mass incarceration, federal law imposes longer prison sentences on certain individuals deemed "career criminals." Although a seemingly simple concept, courts across the country have struggled to determine which crimes qualify as violent offenses under both the ACCA and the Guidelines. This is one such case.
Here, the court asks whether a conviction under Ohio's felonious assault statute qualifies as a violent-felony predicate under the Guidelines and the ACCA. As the lead opinion notes, in 2012, a panel of this court held that both Ohio felonious assault and aggravated assault qualify as violent-felony predicates under the ACCA. See United States v. Anderson , 695 F.3d 390, 399-402 (6th Cir. 2012). But in the time since Anderson was decided, the Supreme Court has clarified the appropriate analysis for reviewing whether certain convictions qualify as violent offenses for sentencing enhancements. See Mathis v. United States , --- U.S. ----, 136 S.Ct. 2243, 2248-49, 195 L.Ed.2d 604 (2016). In light of these developments, we granted en banc review to revisit Anderson . While I concur in the lead opinion's holding that Anderson was wrongly decided and should be overruled, my agreement ends there.
I.
The lead opinion states that two analyses must be done in this case: "One is an overbreadth analysis to determine whether the statute in question is too broad to categorically qualify as a violent-felony predicate because it criminalizes more conduct than is described in the ACCA and Guidelines.... The other is a divisibility analysis...." Lead Op. 393. In a footnote, the lead opinion notes that it expresses "no opinion as to the order in which these two issues should be addressed in other cases, noting only that in this case, addressing overbreadth first is the more useful approach." Lead Op. 396, n.6. I disagree. The Supreme Court has explicitly stated that the divisibility analysis should be conducted first. See Mathis , 136 S.Ct. at 2256 ("The first task for a sentencing court faced with an alternatively phrased statute is [ ] to determine whether its listed items are elements or means."). This makes sense, as the so-called "overbreadth analysis" is necessarily dependent on the divisibility analysis. I thus struggle to see how analyzing the statute in its entirety (as if the statute were indivisible) is the more useful approach. The practical reality of doing so-as Judge Rogers and Judge Kethledge recognize-is that an opinion can begin to look advisory. With that said, in performing the "overbreadth analysis," the lead opinion correctly overrules Anderson and I concur in the stated reasons for doing so. As a matter of practicality *412and common-sense, however, and for the reasons described below, I would conduct the divisibility analysis first, and then turn to the analysis of the statute at issue.1
A.
The Guidelines' career-offender provision applies a sentencing enhancement if, as relevant here, the defendant has at least two prior felony convictions of either a "crime of violence or a controlled substance offense." U.S.S.G. § 4B1.1(a). At the time of Burris's sentencing, the Guidelines defined a "crime of violence" as any crime punishable by over one-year imprisonment that:
(1) Has as an element the use, attempted use, or threatened use of physical force against the person of another ["the elements clause"]; or
(2) is burglary of a dwelling, arson, or extortion, involves use of explosives [the "enumerated clause"], or otherwise involves conduct that presents a serious potential risk of physical injury to another ["the residual clause"].
U.S.S.G. § 4B1.2(a)(2015). Two weeks after Burris was sentenced, the Guidelines were amended and the residual clause was removed. See U.S.S.G. § 4B1.2(a) ; see also United States v. Smith , 881 F.3d 954, 959 n.1 (6th Cir. 2018) ("In 2016, the U.S. Sentencing Commission removed the residual clause and replaced it with a fuller list of enumerated offenses.").
In reviewing potential predicate crimes, courts apply "a 'categorical' approach to determine the nature of a prior conviction," and whether it qualifies as a crime of violence. United States v. Denson , 728 F.3d 603, 607 (6th Cir. 2013). "Under the 'categorical approach,' courts must compare the elements of the statute of conviction to the elements of the generic definition of the offense."
*413Cradler v. United States , 891 F.3d 659, 667 (6th Cir. 2018) (citing Taylor v. United States , 495 U.S. 575, 599-600, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990) ). Put another way, if the statute of conviction criminalizes more conduct than a "crime of violence" as defined by the Guidelines, then the statute is deemed "overbroad" and the conviction does not qualify as a "crime of violence" for sentencing enhancement purposes. See id. ("If the statute of conviction criminalizes more conduct than the generic definition [in the ACCA], then that conviction is not the same offense enumerated in the ACCA's 'violent felony' definition" and "is not a violent felony for ACCA purposes."). In applying the categorical approach, courts are prohibited from looking to the particular facts underlying the conviction. See Cradler , 891 F.3d at 667. In other words, how a defendant actually perpetrated the crime "makes no difference." Mathis , 136 S.Ct. at 2251.
The comparison of elements under the categorical approach involves a straightforward analysis "when a statute sets out a single (or 'indivisible') set of elements to define a single crime." Id. at 2248. If a statute is indivisible, a court need only line up the crime's statutory elements alongside those of the Guidelines' elements and see if they match. See id. But, as the Supreme Court has recognized, some statutes' structures are more complicated, making the comparison of elements more difficult. Id. at 2249. At times, a "defendant's prior conviction is based on a statute that sets out one or more elements of the offense in the alternative, thereby defin[ing] multiple crimes." Cradler , 891 F.3d at 668 (citations and quotations omitted). These statutes, referred to as "divisible statutes," add an additional level of analysis for courts when determining "which set of statutory elements the defendant violated." Id.
Mindful of such situations, the Supreme Court has "suggested that courts may need to 'go beyond the mere fact of conviction' and look to the facts underlying a conviction in order to determine which element or set of elements was the basis of the defendant's conviction." Id. (citing Taylor v. United States , 495 U.S. 575, 602, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990) ). In these circumstances-where a statute contains multiple alternative elements-the "modified categorical approach" should be utilized. Mathis , 136 S.Ct. at 2249. Under this approach, a court may look to a limited class of documents, such as the indictment, jury instructions, or plea agreement, "to determine what crime, with what elements, a defendant was convicted of." Id. ; see also Descamps v. United States , 570 U.S. 254, 257, 133 S.Ct. 2276, 186 L.Ed.2d 438 (2013) ("[T]he modified categorical approach permits sentencing courts to consult a limited class of documents, such as indictments and jury instructions, to determine which alternative formed the basis of the defendant's prior conviction.").
This court recently explained the interplay of the modified categorical approach and the categorical approach:
When we speak of the "modified approach," we simply refer to this additional analytical step in cases involving divisible statutes. After utilizing this additional step to determine which elements in the statute formed the basis of the defendant's prior conviction, courts resume their application of the categorical approach as they would in any other case. "[T]he modified approach merely helps implement the categorical approach when a defendant was convicted of violating a divisible statute. The modified approach thus acts not as an exception, but instead as a tool. It retains the categorical approach's central feature: a focus on the elements, rather than the *414facts, of a crime." [ Descamps , 570 U.S.] at 263, 133 S.Ct. 2276.
Cradler , 891 F.3d at 668. Put simply, the modified categorical approach is to be used only when a statute is deemed divisible. See id.
Thus, despite the lead opinion's contention to the contrary, whether a statute is divisible or indivisible is a threshold issue. Id. (quoting Descamps , 570 U.S. at 263, 133 S.Ct. 2276 ).
B.
As the lead opinion recognizes, this court has never held definitively in a published opinion whether Ohio's felonious assault statute is divisible or indivisible. Lead Op. 402-03. I agree with the lead opinion that the time has come to provide a clear answer. Regrettably, I disagree with the answer provided.
To determine a statute's divisibility, we look first to state law, and then to the statute itself-namely the structure and whether statutory alternatives carry different punishments. United States v. Ritchey , 840 F.3d 310, 318 (6th Cir. 2016). If these sources fail to provide a definitive answer, we turn to the record of conviction. Id. (citing Mathis , 136 S.Ct. at 2256-57 ). Here, the question is whether Ohio's felonious assault statute lists multiple elements disjunctively (divisible), or enumerates various factual means of committing a single offense (indivisible).
The lead opinion claims that each factor supports a finding of divisibility. I disagree. The lead opinion acknowledges only the factors supporting divisibility, while conveniently ignoring other relevant considerations. Importantly, after performing a divisibility analysis, if a court "still cannot discern whether a statute presents elements or means, the statute is indivisible ." United States v. Stitt , 860 F.3d 854, 862 (6th Cir. 2017), rev'd on other grounds , --- U.S. ----, 139 S.Ct. 399, --- L.Ed.2d ---- (2018) (citing Mathis , 136 S.Ct. at 2257 ) (emphasis added); see also Shepard v. United States , 544 U.S. 13, 21, 125 S.Ct. 1254, 161 L.Ed.2d 205 (2005) (noting the "demand for certainty" in the analysis as required by the Supreme Court in Taylor ). The lead opinion acknowledges this demand for certainty in principle, Lead Op. 402-03, but fails to apply it in practice.
The first consideration under Mathis is seemingly dispositive: whether "a state court decision definitively answers the question." 136 S.Ct. at 2256. Mathis explained that if a "ruling of that kind exists, a sentencing judge need only follow what it says." Id. In State v. Brown , the Ohio Supreme Court analyzed Ohio's aggravated assault statute2 and stated that the two subdivisions "set forth two means of committing the same offense[.]" 119 Ohio St.3d 447, 895 N.E.2d 149, 150 (2008). The opinion contains additional language suggesting that the statute sets out two means to commit the same crime. See Brown , 895 N.E.2d at 155 ("these two alternate theories of aggravated assault"); id. ("these two forms of aggravated assault"); id. at 156 ("subdivisions (1) and (2) set forth two means of committing the offense").
But the ultimate holding of the Ohio Supreme Court was as follows: "the General Assembly did not intend violations of [O.]R.C. 2903.11(A)(1) and (A)(2) to be separately punishable when the offenses result *415from a single act undertaken with a single animus. Thus, aggravated assault in violation of [O.]R.C. 2903.12(A)(1) and (A)(2) are allied offenses of similar import." Id. at 156. The opinion makes clear that the purpose of the "allied offenses of similar import" inquiry is to prevent a defendant from receiving cumulative sentences for a single instance of conduct in violation of the Double Jeopardy Clause. Id. at 152 ; see also O.R.C. § 2941.25(A) ("Where the same conduct by defendant can be construed to constitute two or more allied offenses of similar import, the indictment or information may contain counts for all such offenses, but the defendant may be convicted of only one."). Thus, while the language in Brown at least seems to suggest that Ohio's felonious assault and aggravated assault statutes contain two means of committing the same offense-making the statutes indivisible-I recognize the lead opinion's notion that the context of the case casts some doubt as to whether the Ohio Supreme Court definitively answered the question as Mathis requires. To say this factor supports divisibility, though, is misleading, at best.
A survey of recent Ohio appellate cases similarly fails to provide certainty. It is true that one could cite-as the lead opinion does-numerous Ohio appellate court decisions that show that defendants are regularly charged under only one of the two felonious assault subsections, or, alternatively, under both subsections of the statute separately. See Lead Op. 404. And these cases certainly lend support to divisibility. But it is equally true that just as many recent Ohio appellate court cases charge felonious assault as a single crime, without differentiating between subsections, therefore supporting indivisibility. See, e.g. , State v. Perry , No. 17CO0009, --- N.E.3d ----, ----, 2018 WL 4677661, at *1 (Ohio Ct. App. September 24, 2018) (defendant indicted on "one count of felonious assault, a felony of the second degree in violation of [O.]R.C. 2903.11(A)"); State v. Kennedy , No. 8-18-01, 2018 WL 4961487, at *1 (Ohio Ct. App. Oct. 15, 2018) (defendant indicted for felonious assault in violation of O.R.C. § 2903.11(A), without specifying a subsection); State v. Betters , No. 17AP-901, 2018 WL 4896813, at *1 (Ohio Ct. App. Oct. 9, 2018) (defendant indicted on "one count of felonious assault in violation of [O.]R.C. 2903.11"). Indeed, a case currently before this en banc court illuminates this very principle. See Williams v. United States , 875 F.3d 803 (6th Cir. 2017), reh'g en banc granted , vacated by 882 F.3d 1169 (6th Cir. 2018) (A grand jury indicted Williams for one count of felonious assault under § 2903.11, charging that he "knowingly cause[d] serious physical harm to [the victim] and/or did knowingly cause or attempt to cause physical harm to [victim] by means of a deadly weapon or dangerous ordnance[.]") (emphasis added); (R. 39-3, PageID 227). The conflicting cases demonstrate that this factor cannot "support" divisibility. Any assertion to the contrary is disingenuous.
Because I believe no conclusive state court decisions exist, I would next look to the additional factors as directed by Mathis . One such factor is the punishment delineated in the statute: "[i]f statutory alternatives carry different punishments, then under Apprendi [v. New Jersey , 530 U.S. 466, 489, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000) ] they must be elements[,]" as opposed to means. Mathis , 136 S.Ct. at 2256. The lead opinion notes that under O.R.C. § 2903.11(D)(1)(b), a mandatory prison term is imposed if a defendant commits felonious assault against a "peace officer" who "suffer[s] serious physical harm." Lead Op. 404-05. The lead opinion finds that because the term "serious physical harm" is found only in subsection (A)(1), *416the two subsections carry different punishments, making the statute divisible.
Yet the two subsections of Ohio felonious assault do not carry different punishments. It is true that felonious assault on a peace officer carries a mandatory prison term if the officer suffers "serious physical harm," but Ohio courts have treated peace-officer felonious assault as an entirely different crime than ordinary felonious assault. See, e.g. , State v. Mundy , No. 05CA0025-M, 2005 WL 3416186, at *1 (Ohio Ct. App. Dec. 14, 2005) (affirming conviction that included two counts of peace-officer felonious assault and three counts of ordinary felonious assault); O.R.C. § 2903.11(D)(1)(a) (specifying that "a violation of division (A)" constitutes a felony in the first degree if the victim is a peace officer, as opposed to a felony in the second degree if the victim is not a peace officer). Ironically, the lead opinion seemingly recognizes that felonious assault against a peace officer is a separate crime (and felony class), Lead Op. 393-94, but fails to recognize this distinction in conducting the divisibility analysis. At base, that peace-officer felonious assault carries a potentially different punishment in different circumstances does not mean that punishment for ordinary felonious assault varies based on whether a defendant is convicted under subsection (A)(1) or subsection (A)(2). A defendant who commits felonious assault on an average citizen does not face a longer (or a shorter) sentence if the crime resulted in "serious physical harm" to the victim. Accordingly, I do not see how the punishment factor supports divisibility.
Mathis also suggests that "a statute may itself identify which things must be charged (and so are elements) and which need not be (and so are means)." Mathis , 136 S.Ct. at 2256. The lead opinion contends that "[e]ach of these statutes on its face appears to be divisible" because "[e]ach appears to set forth one or more elements of the offense in the alternative, thereby defining multiple crimes." Lead Op. 404. The lead opinion fails to provide, however, any explanation, or cite any case, supporting the idea that the statute is divisible "on its face." To the contrary, the statute does not explicitly identify whether the statute lists means or elements, meaning this factor does not provide a clear answer to the divisibility question.
With no definitive answer, courts are permitted to "peek at the record documents," such as an indictment and correlative jury instructions, for "the sole and limited purpose of determining whether the listed items are elements of the offense." Mathis , 136 S.Ct. at 2256-57 (internal brackets and quotations omitted). Here, Burris was specifically charged with O.R.C. § 2903.11(A)(2) and not subsection (A)(1), which admittedly could support divisibility. But in looking at Ohio's pattern jury instructions, the divisibility analysis is once again muddied.3 As the government notes, Ohio's jury instructions provide alternative instructions for felonious assault under the (A)(1) and (A)(2) subsections and direct the court to select "the appropriate alternative." Ohio Jury Instruc., Part II Crim. Instruc., 503.11(A). Ohio courts, however, do not require a jury to unanimously choose between the two subsections when charged in a single count. See Jackson , 2004 WL 1045402, at *3; Chappell , 2002 WL 337726, at *4. Often courts have not considered which subsection formed the basis of the juries' verdicts and instead affirmed defendants' convictions because the evidence was sufficient to convict under either subsection. Id. This treatment *417of the subsections as alternative means of committing the same crime is a hallmark of an indivisible statute. See Mathis , 136 S.Ct. at 2249, 2256 (finding state court decision "definitively answers" the elements or means question when it holds that a jury need not unanimously decide between two methods of committing the crime).
As a final aside, I note the parties' own inconsistencies. The government argued before the original panel that Ohio's felonious assault statute was indivisible . (See Gov't Br. 38-39 (noting that Anderson treated section 2903.11(A) as indivisible and arguing that "Ohio law confirms that doing so was correct"); id. at 39 (arguing that "under Ohio law, subsections (A)(1) and (A)(2) of Ohio's felonious assault statute provide two different means for committing one crime"); id. at 41 (arguing that "even if Section 2903.11(A) were divisible (which it is not)[,]" Burris's claims would still fail).) In its en banc brief, however, the government changed its tune: "Having conducted a more thorough review of this sentencing concept, the structure of Ohio's statute, and the other factors identified in Mathis , the government now agrees with Burris's original position that subsection (A)(1) and (A)(2) require proof of different elements and are divisible." (Gov't Supp. Br. 25.)4
Burris's position on divisibility has also evolved. In his opening brief before the original panel he argued that "[b]ecause there are multiple ways to commit felonious assault, this is a divisible statute." (Pl. Br. 37.) Then, in his supplemental brief before the en banc court, Burris stated that his "conviction does not qualify as a crime of violence whether the statute is indivisible or divisible." (Pl. Supp. Br. 23.) The parties' inconsistent positions further demonstrate that the divisibility analysis is far from clear.
According to Mathis , employing the above tools should make answering the divisibility question "easy." Mathis , 136 S.Ct. at 2256. In a case such as this one, though, where the answer remains uncertain, " Mathis makes [ ] clear what we must do-or, more precisely, what we must not do: If state law fails to provide clear answers and the record materials [don't] speak plainly, then we won't be able to satisfy Taylor 's demand for certainty." United States v. Degeare , 884 F.3d 1241, 1246 (10th Cir. 2018) (internal quotations omitted) (citing Mathis , 136 S.Ct. at 2257 ). In other words, we are to presume indivisibility without a clear answer. Stitt , 860 F.3d at 862 (citing Mathis , 136 S.Ct. at 2257 ). The lead opinion ignores this presumption. In the absence of certainty, I believe it would be a fundamental error-not to mention a miscarriage of justice-to treat a defendant's prior conviction as a violent-felony predicate. I would, therefore, hold that the statute is indivisible.
II.
Having analyzed divisibility, I would then look to whether Burris's statute of conviction qualifies as a "crime of violence" under the elements, enumerated, or residual clauses of the Guidelines. As mentioned, *418the lead opinion performs this analysis first. For the reasons articulated, that approach makes little sense, but I agree that Ohio felonious assault criminalizes more conduct than is described in both the elements and enumerated clauses of the Guidelines.5 I disagree, however, with the lead opinion's conclusion that Ohio felonious assault is too broad to categorically qualify under the Guidelines' residual clause. Because the lead opinion reviews the residual clause under an inaccurate standard, it is worth elaborating on the proper analysis.
The residual clause states that crimes of violence encompass those state convictions that "otherwise involve[ ] conduct that presents a serious potential risk of physical injury to another." U.S.S.G. § 4B1.2 (2015). The government relies on this court's precedent in United States v. Perry , which held that Ohio's aggravated assault statute was a crime of violence under the residual clause of the ACCA. 703 F.3d 906, 910 (6th Cir. 2013). The Perry court held, albeit with little explanation, that "[b]y its plain terms, [the Ohio aggravated assault statute] proscribes conduct that presents a serious potential risk of physical injury to another." Id. at 910 (internal quotations omitted). However, the Perry court failed (as did the court in Anderson and Rodriguez ) to recognize that "serious physical harm" is defined in Ohio to include not only physical harm, but also mental harm. This flaw, however, is not fatal.
When analyzing a statute under the residual clause, our inquiry differs from that of the elements and enumerated clause analyses. The inquiry is not-as the lead opinion assumes without citation, Lead Op. ----- whether it is possible for a defendant to commit a crime in any way that would not be a violent felony. See United States v. Smith , 881 F.3d 954, 958 (6th Cir. 2018), cert. denied , No. 17-8857, --- U.S. ----, 139 S.Ct. 85, 202 L.Ed.2d 56, 2018 WL 2163626 (U.S. Oct. 1, 2018). Rather, we ask "whether the conduct encompassed by the elements of the offense, in the ordinary case , presents a serious potential risk of [physical] injury to another." James v. United States , 550 U.S. 192, 208, 127 S.Ct. 1586, 167 L.Ed.2d 532 (2007), overruled on other grounds by Johnson v. United States , --- U.S. ----, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015) (emphasis added). Thus, we look at how Ohio courts ordinarily apply the statute in practice. See Smith , 881 F.3d at 958.
Of course, we are still left with the essential question: how do we define the "ordinary" case? The Supreme Court has articulated previously the difficulty of this analysis, albeit in the context of the ACCA:
[T]he residual clause leaves grave uncertainty about how to estimate the risk posed by a crime. It ties the judicial assessment of risk to a judicially imagined "ordinary case" of a crime, not to real-world facts or statutory elements. How does one go about deciding what kind of conduct the "ordinary case" of a crime involves? "A statistical analysis of the state reporter? A survey? Expert evidence? Google? Gut instinct?"
*419United States v. Mayer , 560 F.3d 948, 952 (C.A.9 2009) (Kozinski, C.J., dissenting from denial of rehearing en banc). To take an example, does the ordinary instance of witness tampering involve offering a witness a bribe? Or threatening a witness with violence? Critically, picturing the criminal's behavior is not enough; as we have already discussed, assessing "potential risk" seemingly requires the judge to imagine how the idealized ordinary case of the crime subsequently plays out.
Johnson , 135 S.Ct. at 2557-58. Due to these deficiencies, the Supreme Court ultimately held that the residual clause of the ACCA violated the Constitution's guarantee of due process and overruled James . Id. at 2563. But shortly thereafter, the Supreme Court held that the residual clause of the Guidelines was still valid, due to the Guidelines' advisory nature. Beckles v. United States , --- U.S. ----, 137 S.Ct. 886, 197 L.Ed.2d 145 (2017).
In light of Beckles , this court has held that the ordinary-case standard articulated in James "remains good law for purposes of the Guidelines." Smith , 881 F.3d at 957 (brackets omitted) (quoting United States v. Goodson , 700 F. App'x 417, 423 (6th Cir. 2017) ); United States v. Morris , 885 F.3d 405, 412 (6th Cir. 2018) ("Although Johnson abrogated [ James ] for purposes of ACCA, the case remains good law for purposes of analyzing the residual clause of the Guidelines."). While we question the prudence of this standard (for all the reasons articulated in Johnson ), it is the law of the land. Thus, we must determine whether Ohio felonious assault involves a "serious potential risk of physical injury" in an "idealized ordinary case of the crime." Welch v. United States , --- U.S. ----, 136 S.Ct. 1257, 1262, 194 L.Ed.2d 387 (2016) (citation omitted); see also Beckles , 137 S.Ct. at 892.
Under James , "[t]he mere possibility that a person could, conceivably, commit the offense without creating a serious risk of physical injury to another is not enough to exclude the offense from the ambit of the Guidelines' residual clause." Goodson , 700 F. App'x at 424 ; see also James , 550 U.S. at 208, 127 S.Ct. 1586 ("One can always hypothesize unusual cases in which even a prototypically violent crime might not present a genuine risk of injury[.]"). As the lead opinion discusses, however, individuals committing felonious assault in Ohio without creating a serious risk of physical injury to another, is not a mere possibility or a hypothetical. It is a certainty. Nevertheless, it appears that "ordinary" felonious assault in Ohio likely poses serious risk of physical injury, sufficient to conclude that O.R.C. § 2903.11 is a crime of violence under the residual clause of the Guidelines. I therefore disagree with the lead opinion's conclusion to the contrary.
III.
Everyone agrees that we review the district court's classification of Ohio felonious assault as a "crime of violence" under the Guidelines for plain error. See Lead Op. 396. But, because I believe Burris would not be subject to the career-offender enhancement in the absence of the residual clause-and his sentencing guideline would be significantly lower without the enhancement-the district court should be given an opportunity to consider Burris's case in light of the amendment that eliminated the residual clause just weeks after his sentencing. This court, as well as courts across the country, have previously exercised the discretion to remand under similarly unique circumstances. See United States v. Atkinson , 354 F. App'x 250, 254 (6th Cir. 2009) (per curiam); see also United States v. McMillan , 863 F.3d 1053, 1058-59 (8th Cir. 2017) ;
*420United States v. Frates , 896 F.3d 93, 102-04 (1st Cir. 2018). I believe we should do so again.
To be clear, I would not hold that the district court erred-plainly or otherwise. See U.S. v. Taylor , 648 F.3d 417, 428-29 (6th Cir. 2011). But the decision to remand does not hinge solely on a showing of plain error. United States v. Ahrendt , 560 F.3d 69, 80 (1st Cir. 2009) ("We d[o] not dwell on the standard of review in our decision to remand for resentencing[.]"). And a remand would serve to give the district court the opportunity to consider the amendment eliminating the residual clause-a task that was previously futile under Anderson . Under these circumstances, prudence dictates that the district court be given that opportunity. I respectfully dissent from the affirmance of Burris's sentence.
Some might find it strange to engage in a divisibility analysis when Burris was convicted of an identified subsection of the felonious assault statute. Why not simply ask whether that subsection satisfies the elements clause? We must first ask the divisibility question because it tells us whether the identification of the specific subsection is significant. As explained in Mathis, if the aggravated assault and felonious assault statutes are divisible, Burris's conviction of subsection (A)(2) tells us that the jury necessarily found each of the elements of that subsection. But if the offenses are not divisible, Burris's conviction of a specific subsection simply tells us the "brute facts" of his offense, not the elements. Mathis , 136 S.Ct. at 2248 (internal quotation marks omitted). In the latter case, the situation is analogous to the hypotheticals discussed by the dissenting justices in Mathis -Shepard documents that unequivocally establish that the defendant was convicted of breaking into a building, not a boat or tent. Notwithstanding these arguments, the Mathis majority held that if the offense is not divisible, the facts, or means, are irrelevant. Because the divisibility analysis (see infra pp. 414-18) reveals that Ohio sometimes treats the aggravated assault and felonious assault offenses as divisible and sometimes treats them as setting forth alternative means of committing a single offense, the concerns animating the Court's decision in Mathis are applicable. Indeed, in Burris's direct appeal of his felonious assault conviction, the Ohio Court of Appeals erroneously stated, "Felonious assault under [O.]R.C. 2903.11(A)(2) provides that no person shall knowingly cause serious physical harm to another," State v. Burris , No. 24088, 2008 WL 4151321, at *3 (Ohio Ct. App. Sept. 10, 2008), thus supporting that the Ohio courts sometimes merge or conflate the two subsections and do not consistently treat them as setting forth distinct offenses. See, e.g ., State v. Jackson , No. 82724, 2004 WL 1045402, at *3 (Ohio Ct. App. May 6, 2004) (holding that "trial court did not err by failing to issue separate instructions" for each subsection because "the evidence presented in this case was sufficient ... to find the [defendant] guilty of felonious assault under either applicable code section"); State v. Chappell , No. 79589, 2002 WL 337726, at *4 (Ohio Ct. App. Feb. 21, 2002) (same).
Because the aggravated assault statute is nearly identical to the felonious assault statute, the reasoning and holding in cases analyzing one statute have been found to apply equally to both statutes. See Person v. Sheets , 527 F. App'x 419, 424, n.4 (6th Cir. 2013) (applying Brown 's interpretation of Ohio's aggravated assault statute to Ohio's felonious assault statute because the relevant subdivisions "are identical under both statutes").
The jury instructions in Burris's felonious assault case are not part of the record before this Court. Consequently, we review Ohio's jury instructions generally.
One could make a strong argument that the government has forfeited, or maybe even waived, its argument that Ohio's felonious statute is divisible. See, e.g. , Miller v. Texas Tech Univ. Health Sci. Ctr. , 421 F.3d 342, 348-49 (5th Cir. 2005) ("The maxim is well established in this circuit that a party who fails to make an argument before either the district court or the original panel waives it for purposes of en banc consideration."); Ryan v. United States , 688 F.3d 845, 848 (7th Cir. 2012) (holding that the reason a litigant comes to its decision to not assert an argument "is irrelevant, and a mistake in reaching a decision to withhold a known defense does not make that decision less a waiver").
In analyzing the enumerated clause, the lead opinion notes that the panel in United States v. Rodriguez , 664 F.3d 1032 (6th Cir. 2011), "did not refer to or analyze Ohio's statutory definition of 'serious bodily harm,' nor did it perform the divisibility analysis now required by Descamps and Mathis ." Lead Op. 400, n. 15. Confusingly, the lead opinion ended its discussion there. To the extent the panel's decision in Rodriguez can be read as finding that "serious physical harm" as used in Ohio's aggravated assault and felonious assault statutes requires violence, rather than note the error as the lead opinion does, I would overrule it.