*488According to Joann Matouk Romain's estate, two local police departments and many officers covered up Joann's murder. The alleged plot required duping other officers, the U.S. Coast Guard, and Canadian authorities. The motive? To help a friend who sold the officers alcohol at prices cheaper than Costco. What at first sounds fanciful is moored by some odd facts. For example, Joann's daughter swears that an unidentified officer questioned her about Joann before the police suspected Joann was missing. Another officer had Joann's spare key that allegedly went missing a month before her disappearance. That said, a reasonable jury could not return a verdict for the estate. As a result, we affirm the district court's decision to grant summary judgment to the defendants.

I.

In January 2010, police found Joann Matouk Romain's car alone in a church driveway. Two months later, a fisherman found her body in the Detroit River. This case is about the investigation of Joann's disappearance.

According to the police, the investigation started when an officer spotted Joann's car around 9 p.m. After running her license plate, the officer took no further action because the car was on private property. About an hour later, a different officer noticed the car. That officer was more concerned because the car was alone in the driveway, nobody was around, and it was a cold night. The officer approached the car and looked inside. Nobody was there.

People often park at the church, cross the street, and go down to Lake St. Clair. Suspecting that the car's driver may have done the same, the officer looked around. Nearby, he saw footprints in the snow heading towards an embankment. The officer followed the footprints. He then saw imprints suggesting that someone sat on a breakwall near the lake, pushed off to a second breakwall, and then sat on that second breakwall above the water. No footprints led back from the lake.

Based on those observations, the officer thought the person from the car might be in the water. He notified his supervisor, who then came to the scene. The supervisor agreed, so he activated the police department's dive team and contacted the U.S. Coast Guard. The search continued until the next afternoon. They did not find a body.

While some officers searched the water, others investigated the car. That investigation included sending an officer to Joann's house. There, the officer spoke with Joann's daughter. The daughter reported that Joann had attended an evening Mass and that she was not answering her phone. In the weeks that followed, police questioned people who knew Joann, who saw her at Mass, and who were near the lake the night she disappeared. Nothing concrete materialized.

Two months after the disappearance, a fisherman found a body on the Canadian side of the Detroit River. Canadian authorities responded and ultimately identified *489Joann with help from their American counterparts. Before the Canadians released the body, a Canadian coroner performed an autopsy. He concluded that Joann drowned, but he could not determine the manner of death.

The body then went to a county medical examiner's office in Michigan. The county coroner performed another autopsy and drew the same conclusions. The coroner noted, however, that homicide was "less likely" than suicide because Joann had no significant injuries. He also opined that an accident seemed "quite unlikely" because Joann had no reason to be near the water.

Joann's daughter requested a third autopsy. Like the coroners, a doctor at the University of Michigan reported drowning as the cause of death with an undetermined manner of death.

Ultimately, the police could not find any answers. They have stopped actively investigating Joann's death, but the case is still open.

Joann's estate is unsatisfied with the investigation. It believes that an unknown person murdered Joann. Even worse, the estate claims that the police knew about the murder before it happened and botched the investigation to protect the killer. In support of that theory, the estate points to evidence it says shows that the police communicated with the killer and then improperly investigated the disappearance.

Starting with the communications with the killer, the estate says the evidence shows that the police knew about Joann's disappearance before finding her abandoned car. In support, the estate relies largely on an affidavit from Joann's daughter. The affidavit states that an unidentified officer arrived at Joann's house about thirty minutes before the police reported following the footprints to the water. That officer, the affidavit says, was not the officer that the police later claimed they sent. What is more, the officer asked about Joann. But Joann's car was registered in her daughters' names, so the officer should not have known that Joann was missing based only on a license plate check. The estate asserts that a jury could infer from those facts that the officer knew about Joann's disappearance before it happened. That would make sense, the estate says, only if the officer knew about the murder plot.

As further proof, the estate says the police had Joann's spare keys that went missing a month before her disappearance. An officer claims he picked up the keys from Joann's house the morning after she disappeared. The estate seems to insinuate that the officer is lying and that the police were in contact with the person who took the keys or that they took the keys themselves.

Turning to the investigation, the estate claims that the police concealed key evidence. The estate, for example, challenges the account that a single set of footprints led to the lake. In support, there are pictures showing many footprints by the water. The police explain that those prints are from the search party that was scrambling to find Joann. The family maintains, however, that the photos show either that there were more than one set of prints originally or that the police intentionally failed to preserve the original set of prints.

The estate also suggests that the police ignored a struggle. Joann's purse had a rip along the seam of a decorative flap. Her daughter says Joann carried her purse on her left shoulder, and that it was "in pristine condition" the day of her disappearance. And an autopsy report said that Joann had a contusion on her upper left arm. The estate says that these facts are *490proof of a struggle that ended in murder. The police disagreed.

Finally, the estate criticizes the police's failure to investigate certain witness accounts. For example, a witness said she told the police that she saw an underdressed man wearing a scarf running near the water on the night Joann disappeared. Consistent with that observation, the police found a black scarf. But the witness's account is not in the police report, and the police disposed of the scarf after some time. A different witness says on the night Joann disappeared, he saw a motionless woman slumped over on the breakwall and two men standing nearby.1 The police did not pursue that tip because they thought the witness was not credible.

Based on the alleged evidence of a cover up, the estate sued the two investigating police departments and many officers. Relevant to this appeal, the estate brought a Monell claim and a § 1983 claim alleging that there was a state-created danger because the officers made "it known to Killer John Doe that they would immediately cover up the murder and rule it a suicide." The district court granted summary judgment to the defendants. The estate now appeals that order.

II.

We review de novo a summary judgment ruling. Smith v. Perkins Bd. of Educ. , 708 F.3d 821, 825 (6th Cir. 2013). Summary judgment is proper "if the movant shows that there is no genuine dispute as to any material fact and the movant is entitled to judgment as a matter of law." Fed. R. Civ. P. 56(a). A fact is material if it would establish or refute an "essential element[ ] of a cause of action or defense asserted by the parties[.]" Kendall v. Hoover Co. , 751 F.2d 171, 174 (6th Cir. 1984) (quotation omitted). And a factual dispute is genuine if it is based on evidence that a reasonable jury could use to return a verdict for the nonmoving party. Anderson v. Liberty Lobby, Inc. , 477 U.S. 242, 248, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). When reviewing the record, we view the evidence in the light most favorable to the nonmoving party and draw all reasonable inferences in the nonmoving party's favor. Smith , 708 F.3d at 825.

III.

The estate makes grave factual allegations. But to prevail, it needs to turn those factual allegations into cognizable legal theories. The estate has struggled with that challenge. After failing below, the estate changed lawyers and legal theories. On appeal, it says that the police, as co-conspirators, are directly responsible for the murder because it occurred during a conspiracy.

A clever move, but not the one made below. There, the estate sought relief for a cover up-not a murder. Alleging a "state-created danger," the estate's complaint claimed that the defendants increased the risk of harm to Joann by telling the unknown killer that they would cover up the crime. And in its summary judgment briefing, the estate identified an "overall conspiracy to not investigate [Joann's] disappearance as a crime."

Putting those arguments aside, the premise of a state-created danger theory belies the estate's expanded appellate argument. If the estate intended to hold the defendants liable for murder, then the police did not create the danger-they were *491the danger. See Jones v. Reynolds , 438 F.3d 685, 695 (6th Cir. 2006) ("Had the officers organized or participated in [the crime], the issue would cease to turn on whether they were responsible for harm caused by a private actor and would turn instead on whether they had caused the harm themselves.").

As the district court explained, a state-created danger theory provides a narrow path for recovery. Now that the case is on appeal, the estate cannot seek a more forgiving path. Bowles v. Marx Hide & Tallow Co. , 153 F.2d 146, 149 (6th Cir. 1946) ("Appellant, on appeal, cannot change the theory of his case made by his pleadings and his conduct of the case.") We therefore consider the claim as the estate presented it below.

The estate cannot recover under a state-created danger theory. The state-created danger doctrine allows plaintiffs to bring due process claims under § 1983 for harms caused by private actors-an anomaly because neither the Fourteenth Amendment nor § 1983 regulates private actors. Hunt v. Sycamore Cmty. Sch. Dist. Bd. of Educ. , 542 F.3d 529, 534 (6th Cir. 2008) ; Barber v. Overton , 496 F.3d 449, 458 (6th Cir. 2007). Although the doctrine stems from a Supreme Court case, it does not have a strong foundation. See Hunt , 542 F.3d at 534. In DeShaney v. Winnebago County Department of Social Services , county authorities returned a child to his abusive father who then beat the child again. 489 U.S. 189, 192-93, 109 S.Ct. 998, 103 L.Ed.2d 249 (1989). The child and his mother sued the county and its employees, alleging a violation of the child's substantive due process rights. Id. at 193-95, 109 S.Ct. 998. The Supreme Court rejected the claim because "nothing in the language of the Due Process Clause itself requires the State to protect the life, liberty, and property of its citizens against invasion by private actors." Id. at 195, 109 S.Ct. 998.

The Court also addressed an argument that a state may have a duty, arising out of "special relationships," to protect certain individuals. Id. at 197, 109 S.Ct. 998. When addressing that argument, the Court acknowledged precedent requiring states to protect inmates, to provide medical care to inmates, and to ensure the reasonable safety of involuntarily committed mental patients. Id. at 198-99, 109 S.Ct. 998. But the Court held that the precedent was limited to situations where a state holds a citizen against his will because then the state has affirmatively restrained the citizen's ability to care for himself. Id. at 199-200, 109 S.Ct. 998. That logic did not apply in DeShaney , Chief Justice Rehnquist wrote, because "[w]hile the State may have been aware of the dangers that [the child] faced in the free world, it played no part in their creation, nor did it do anything to render him any more vulnerable to them." Id. at 201, 109 S.Ct. 998.

That passage could be read as simply explaining that the county in DeShaney had not restrained the child's ability to protect himself and thus the cited precedent did not apply. But our circuit and others have read it more broadly to imply a new theory of constitutional liability-the state-created danger doctrine. McQueen v. Beecher Cmty. Sch. , 433 F.3d 460, 464 (6th Cir. 2006) ; Erwin Chemerinsky, The State-Created Danger Doctrine , 23 Touro L. Rev. 1 (2007). Under that doctrine, plaintiffs can recover for private harms if they show: "(1) an affirmative act by the state which either created or increased the risk that the plaintiff would be exposed to an act of violence by a third party; (2) a special danger to the plaintiff wherein the state's actions placed the plaintiff specifically at risk, as distinguished from a risk that affects the public *492at large; and (3) the state knew or should have known that its actions specifically endangered the plaintiff." Cartwright v. City of Marine City , 336 F.3d 487, 493 (6th Cir. 2003) (citation omitted).

That is a "demanding standard." Jones , 438 F.3d at 691 (quoting Sargi v. Kent City Bd. of Educ. , 70 F.3d 907, 913 (6th Cir. 1995) ). Indeed, some of our cases have recognized an "additional element"-that the government's conduct shocks the conscience-to ensure "that the standards for proving a state-created-danger violation based on governmental inaction correspond to the standards necessary for proving a substantive due process violation based on governmental action." Schroder v. City of Fort Thomas , 412 F.3d 724, 730 (6th Cir. 2005). We therefore often reject allegations of a state-created danger. McQueen , 433 F.3d at 465 (collecting cases). This case is no different.

The estate cannot make the first showing, that the defendants' affirmative actions increased the risk of harm to Joann. When looking for an increased risk, we focus on "whether [the victim] was safer before the state action than he was after it." Koulta v. Merciez , 477 F.3d 442, 445-46 (6th Cir. 2007) (quoting Cartwright , 336 F.3d at 493 ) (brackets in original). At first blush, the estate seems to satisfy that requirement. If the alleged murderer knew that the police would cover up the crime, the reduced risk of penalty may have encouraged the murderer to proceed with his plot.2 Other circuits have accepted similar theories. See Dwares v. City of New York , 985 F.2d 94, 99 (2d Cir. 1993) recognized as overruled on other grounds by Montero v. City of Yonkers , 890 F.3d 386, 403 (2d Cir. 2018) ; Freeman v. Ferguson , 911 F.2d 52, 54-55 (8th Cir. 1990).

But even if we accepted that theory, the estate has failed to establish a genuine issue regarding the officers' alleged assurance that they would conceal the murder. The estate cites limited circumstantial evidence and asks this court to infer not only that the police knew beforehand of the plan to murder Romain, but also that the police told the killer that they would conceal the murder. No direct evidence supports this conclusion; and no reasonable jury could make the speculative inferences required to reach it.

Two final points. First, the estate asks us to revive its Monell claim. Because the estate's theories against the individual officers fail, we decline that request. See Fox v. DeSoto , 489 F.3d 227, 238 (6th Cir. 2007) (recognizing that there can be no municipal liability if the municipality's agent or employee did not violate the Constitution). Second, the estate argues that the alleged botched investigation denied it access to the courts. The estate did not include that claim in its operative complaint, and the district court denied the estate's request to add it later. Rather than challenging the denial of leave to amend, the estate acts as if its request for leave is enough for us to consider the claim's merits on appeal. We decline the invitation.

IV.

In sum, no reasonable jury could return a verdict for the estate. We thus affirm the district court's decision to grant summary judgment to the defendants.

CONCURRENCE

Our holding that this suit must end should not be taken to minimize the case's *493tragic facts-Joann Matouk Romain lost her life and Michelle Marie Romain lost her mother. Joann's estate and daughter make serious accusations against the police who investigated Joann's disappearance, claiming that the police intentionally masked Joann's murder and refused to bring the alleged murderer to justice. But, as the court correctly holds, the plaintiffs did not factually support these charges of misconduct with evidence that would allow a reasonable jury to credit their substantive-due-process claim, which is the only claim preserved on appeal. While not critical to the outcome here, I write to make two additional legal points: Our cases may have overread one statement from DeShaney v. Winnebago County Department of Social Services , 489 U.S. 189, 109 S.Ct. 998, 103 L.Ed.2d 249 (1989), while overlooking another statement in that opinion.

First , our court and many others have used a single sentence in DeShaney to adopt different versions of a "state-created-danger" theory of substantive due process compelling the state at times to protect an individual from a private actor's violence. Kallstrom v. City of Columbus , 136 F.3d 1055, 1066 (6th Cir. 1998) ; see Irish v. Maine , 849 F.3d 521, 525-26 (1st Cir. 2017). DeShaney held that the state did not violate due process by failing to protect a child from his father's abuse even though the state knew of the abuse and did not remove the child from harm's way. 489 U.S. at 191, 109 S.Ct. 998. The relevant sentence noted that, "[w]hile the State may have been aware of the dangers that [the child] faced in the free world, it played no part in their creation, nor did it do anything to render him any more vulnerable to them." Id. at 201, 109 S.Ct. 998. The point of this sentence was to distinguish cases holding that a state must protect individuals (like prisoners) who are in custody and whose liberty the state has, in fact, restricted. Id. at 198-201, 109 S.Ct. 998. Yet from this line, our court has developed a three-factored test that allows parties who are not under state control to pursue substantive-due-process claims challenging a state's failure to protect them. We allow these claims if a plaintiff can show that: (1) a state actor's "affirmative act" created or increased the risk that the plaintiff would suffer private violence; (2) this risk qualifies as a "special danger" that exceeds the general risk of private violence faced by the public at large; and (3) the state "knew or should have known" that its action would trigger this danger. See, e.g. , Engler v. Arnold , 862 F.3d 571, 575 (6th Cir. 2017) ; Nelson v. City of Madison Heights , 845 F.3d 695, 700 (6th Cir. 2017) ; Cartwright v. City of Marine City , 336 F.3d 487, 493 (6th Cir. 2003).

I am not sure DeShaney supports our test. In many respects, DeShaney is a surprising source for this right. More than any other case, that decision teaches that the Due Process Clause's text-focused on state deprivations of life, liberty, or property without adequate process-makes the clause a poor fit for claims that the state refused to protect a person from private deprivations of life, liberty, or property. 489 U.S. at 194-97, 109 S.Ct. 998. Plaintiffs making allegations like these do not seek notice and a hearing ("process") before the state refuses to protect them; they seek substantive protection from a private actor's crime or tort. Id. at 194-95, 109 S.Ct. 998. And even substantive due process safeguards fundamental rights only against state interference; it does not compel the state to protect those rights against private invasion. Id. at 196-97, 109 S.Ct. 998. DeShaney thus holds that a "State's failure to protect an individual against private violence simply does not constitute a violation of the Due Process Clause." Id. at 197, 109 S.Ct. 998. Whether or not our test can be defended *494based on the one sentence in DeShaney , it surely runs counter to the opinion's general thrust-that the Due Process Clause is ill-suited for claims seeking state protection from private violence.

Our test also encounters difficulties when assessed against other substantive-due-process principles. As a general matter, "because the 'guideposts for responsible decisionmaking in this unchartered area are scarce and open-ended ...,' '[t]he doctrine of judicial self-restraint requires us to exercise the utmost care whenever we are asked to break new ground in this field.' " Kerry v. Din , --- U.S. ----, 135 S. Ct. 2128, 2134, 192 L.Ed.2d 183 (2015) (plurality op.) (quoting Collins v. Harker Heights , 503 U.S. 115, 125, 112 S.Ct. 1061, 117 L.Ed.2d 261 (1992) ). This note of caution should inform our reading of DeShaney . The opinion could just as well be read (consistent with the Due Process Clause's use of the word "deprive") to trigger a right to state aid only for custodial-like situations in which there has been a state deprivation of liberty. See Pinder v. Johnson , 54 F.3d 1169, 1175 (4th Cir. 1995) (en banc) (opinion of Wilkinson, J.); cf. Rios v. City of Del Rio , 444 F.3d 417, 422 (5th Cir. 2006). After all, another sentence says that "[t]he affirmative duty to protect arises not from the State's knowledge of the individual's predicament or from its expressions of intent to help him, but from the limitation which it has imposed on his freedom to act on his own behalf." DeShaney , 489 U.S. at 200, 109 S.Ct. 998. And while our test's "affirmative act" factor also responds to DeShaney 's holding that the Due Process Clause limits state (not private) conduct, it does so with an act/omission "distinction" that "blur[s] when stared at too long." Trinity Lutheran Church of Columbia, Inc. v. Comer , --- U.S. ----, 137 S. Ct. 2012, 2025-26, 198 L.Ed.2d 551 (2012) (Gorsuch, J., concurring in part). A good lawyer often can recast an omission as an act, Pinder , 54 F.3d at 1175-76 (opinion of Wilkinson, J.), which makes "an end run around DeShaney 's core holding," Rivera v. Rhode Island , 402 F.3d 27, 38 (1st Cir. 2005).

As a specific matter, our test could be read as adopting a substantive-due-process framework for private-actor harm that is easier to meet than the framework that governs state-actor harm. For one thing, none of our three factors mentions the Supreme Court's central substantive-due-process teaching in this context: that a state action must be "so egregious, so outrageous, that it may fairly be said to shock the contemporary conscience." County of Sacramento v. Lewis , 523 U.S. 833, 847 n.8, 118 S.Ct. 1708, 140 L.Ed.2d 1043 (1998). And in the context of privately inflicted harm, this test must focus not on what will typically be egregious misconduct by a private party , but on the state action that increased the risk of that misconduct. Barber v. Overton , 496 F.3d 449, 458 (6th Cir. 2007) (Cook, J., concurring). While our cases have occasionally invoked this demanding standard, e.g. , Schroder v. City of Fort Thomas , 412 F.3d 724, 730-31 (6th Cir. 2005), it should be the critical question in all of these cases.

For another thing, our third factor's "should-have-known" language suggests a classic negligence formulation. E.g. , Global-Tech Appliances, Inc. v. SEB S.A. , 563 U.S. 754, 769-70, 131 S.Ct. 2060, 179 L.Ed.2d 1167 (2011). But the Supreme Court's cases explain that negligence can never establish a substantive-due-process violation. Davidson v. Cannon , 474 U.S. 344, 347-48, 106 S.Ct. 668, 88 L.Ed.2d 677 (1986). The Court has instead chosen between two more culpable mental states. It has recognized that state action undertaken with a purposeful intent to harm is "most likely" to be "conscience-shocking." Lewis , 523 U.S. at 849, 118 S.Ct. 1708. And *495it has debated when, if ever, state action undertaken with a mere deliberate indifference to a risk of harm can prove a substantive-due-process violation in non-custodial settings. Compare id. at 848-54, 118 S.Ct. 1708, with id. at 863-64, 118 S.Ct. 1708 (Scalia, J., concurring in the judgment). We should not permit state liability in this private-harm context to rest on a less culpable mental state. Indeed, despite many cases reciting the should-have-known test, e.g. , Schneider v. Franklin County , 288 F. App'x 247, 252-53 (6th Cir. 2008), others have applied a deliberate-indifference rule, see, e.g. , Schroder , 412 F.3d at 729-31 ; Ewolski v. City of Brunswick , 287 F.3d 492, 510-12 (6th Cir. 2002). That deliberate-indifference element, too, should be the floor in all of these cases.

In sum, the court in this case is quite correct, in my view, to say that our three-factored substantive-due-process test for state-created dangers lacks a "strong foundation."

That brings me to my second point: a different disclaimer in DeShaney deserves more attention than the sentence our cases have quoted. The Supreme Court has said that if a different constitutional provision "provides an explicit textual source of constitutional protection, a court must assess a plaintiff's claims under that explicit provision and 'not the more generalized notion of substantive due process." Conn v. Gabbert , 526 U.S. 286, 293, 119 S.Ct. 1292, 143 L.Ed.2d 399 (1999) (internal quotation marks omitted). In that regard, DeShaney elsewhere noted that the plaintiff had not asserted an equal-protection claim and that the "State may not, of course, selectively deny its protective services to certain disfavored minorities without violating the Equal Protection Clause." 489 U.S. at 197 n.3, 109 S.Ct. 998.

At first blush, the Equal Protection Clause's text-barring the State from "deny[ing] to any person ... the equal protection of the laws," U.S. Const. amend. XIV, § 1 -may provide a more plausible textual hook than the Due Process Clause for claims that the police intentionally denied a specific person (a so-called "class of one") the protection of the criminal laws that everyone else enjoys. Indeed, while it is now well-established that the Equal Protection Clause provides a general antidiscrimination mandate for all state acts, Village of Willowbrook v. Olech , 528 U.S. 562, 564, 120 S.Ct. 1073, 145 L.Ed.2d 1060 (2000) (per curiam), some commentators have cited historical materials suggesting that remedial laws (like the laws implicated here) fall within the center of the "protection of the laws," see John Harrison, Reconstructing the Privileges or Immunities Clause , 101 Yale L.J. 1385, 1435-36 (1992) (citing 1 W. Blackstone, Commentaries *55-56). These protective laws stand in contrast to, say, public-employment decisions in which "the government acts in a more proprietorial and less regulatory capacity." SECSYS, LLC v. Vigil , 666 F.3d 678, 690 (10th Cir. 2012) (opinion of Gorsuch, J.) (discussing Engquist v. Ore. Dep't of Ag. , 553 U.S. 591, 128 S.Ct. 2146, 170 L.Ed.2d 975 (2008) ).

History supports this instinct too. "[T]he suppression of private violence [was] the core concern of the Equal Protection Clause." Christopher R. Green, The Original Sense of the (Equal) Protection Clause: Subsequent Interpretation and Application , 19 Geo. Mason U. C.R. L.J. 219, 254 (2009). "The unwillingness of the law enforcement authorities in southern states to protect the newly freed blacks from white vigilante groups such as the Ku Klux Klan was an important motive for the enactment of the equal protection clause." Del Marcelle v. Brown Cty. Corp. , 680 F.3d 887, 889 (7th Cir. 2012) (en banc) (opinion of Posner, J.). For example, Congress enacted the 1866 Civil Rights Act to give "nonwhites 'full and equal benefit of *496all laws and proceedings for the security of person and property, as is enjoyed by white citizens.' " David P. Currie, The Constitution in the Supreme Court: The First Hundred Years, 1789-1888 , at 349 (1985) (quoting Act of Apr. 9, 1866, § 1, 14 Stat. 27, 27). And the framers enacted the Fourteenth Amendment to ensure this act's constitutionality. Id. at 347-49. To be sure, the Equal Protection Clause's primary target may have been racially discriminatory refusals to protect persons from private violence. Id. at 349. But, whatever its purpose , the Equal Protection Clause's text is not limited to race-based denials of the protection of the laws. Cf. District of Columbia v. Heller , 554 U.S. 570, 577-78, 128 S.Ct. 2783, 171 L.Ed.2d 637 (2008) ; Crawford v. Washington , 541 U.S. 36, 61-62, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004).

The Equal Protection Clause's text and history suggest that the right question to ask is: When, if ever, do equal-protection principles give a specific individual the right to challenge a state officer's intentional refusal to provide the protection of the laws that keep the public safe from private violence? Figuring out the right question is the easy part; determining the appropriate answer is much harder. Thoughtful jurists have suggested a variety of approaches to this class-of-one problem. Perhaps the "den[ial]" of the "equal protection of the laws" in the law-enforcement context simply does not occur unless there has been class-based discrimination like the racial discrimination that motivated the clause. Cf. Del Marcelle , 680 F.3d at 901 (Easterbrook, C.J., concurring in the judgment); Flowers v. City of Minneapolis , 558 F.3d 794, 799-800 (8th Cir. 2009). Or perhaps the clause should reach claims alleging a one-off refusal to provide police protection only if the refusal flows from personal reasons (such as animosity toward the plaintiff) that are unrelated to public duties. Cf. Del Marcelle , 680 F.3d at 889 (opinion of Posner, J.). Or maybe the traditional equal-protection test works just fine here, requiring a party to show that the police have intentionally discriminated against the plaintiff and that the discrimination lacked a rational basis. Cf. id. at 913 (Wood, J., dissenting); SECSYS , 666 F.3d at 689-90 (opinion of Gorsuch, J.). Any approach also must account for the tradition of prosecutorial and law-enforcement discretion. See Town of Castle Rock v. Gonzales , 545 U.S. 748, 760-61, 125 S.Ct. 2796, 162 L.Ed.2d 658 (2005).

I leave the scope of equal protection for other cases. For present purposes, I note only that the equal-protection question strikes me as a more appropriate question to ask for the types of failure-to-protect allegations that are presented in this appeal. That, in turn, indicates that the due-process question is the wrong question to ask. See Conn , 526 U.S. at 293, 119 S.Ct. 1292. So I am inclined to think this area worthy of reexamination in a suitable future case.