EASTERBROOK, Circuit Judge.

Jerry Montgomery, who is serving a 40-year term for murder, contends in this action under 28 U.S.C. § 2254 that Indiana violated the due process clause of the fourteenth amendment when it placed him in disciplinary segregation and reduced his credit-earning class. The district court denied his petition, and this appeal presents a series of questions: (1) is a collateral attack under § 2254 the right way to present a claim of constitutional error in a decision affecting the rate of earning good-time credits?; (2) does a lower rate of earning good-time credits deprive a prisoner of "liberty" or "property" within the meaning of the Constitution?; and, if we give affirmative answers to these questions, (3) did Indiana afford Montgomery all the process due him?

Section 2254 authorizes federal courts to grant collateral relief to state prisoners "in custody in violation of the Constitution or laws or treaties of the United States." Montgomery is in Indiana's custody, and the length of that custody depends on a resolution of his complaint about the award of credits. See Lackawanna County District Attorney v. Coss, 532 U.S. 394, 121 S.Ct. 1567, 1572-73, 149 L.Ed.2d 608 (2001). If the state had deprived him of good-time credits already earned, then § 2254 would provide the exclusive avenue for seeking federal relief. See Walker v. O'Brien, 216 F.3d 626 (7th Cir. 2000). A reduction in the rate of earning good-time credits should be treated identically. The stakes are the same: the length of incarceration. Montgomery was demoted from Class I (one day's good-time credit for each day served) to Class II (one day's credit for each two days served). See I.C. § 35-50-6-3(a), (b). This difference, applied from the start of a 40-year term, increases confinement by almost 7 years. (A prisoner sentenced to 40 years and in Class I throughout would be released after 20 years; the same prisoner in Class II would be released after 26 years and 8 months.) Any decision that determines the fact or duration of state custody may (and usually must) be challenged under § 2254 rather than 42 U.S.C. § 1983. See, e.g., Preiser v. Rodriguez, 411 U.S. 475, 93 S.Ct. 1827, 36 L.Ed.2d 439 (1973); Edwards v. Balisok, 520 U.S. 641, 117 S.Ct. 1584, 137 L.Ed.2d 906 (1997); Moran v. Sondalle, 218 F.3d 647 (7th Cir. 2000).

But only the change in credit-earning class may be challenged under § 2254. Disciplinary segregation affects the severity rather than duration of custody. More-restrictive custody must be challenged under § 1983, in the uncommon circumstances when it can be challenged at all. See Sandin v. Conner, 515 U.S. 472, 115 S.Ct. 2293, 132 L.Ed.2d 418 (1995); Meachum v. Fano, 427 U.S. 215, 96 S.Ct. 2532, 49 L.Ed.2d 451 (1976). Montgomery cannot use § 1983 to contest his segregation, for three reasons. First, the decision of a prison disciplinary board may not lead to an award of damages if it is open to contest under § 2254 yet remains in force. That's the holding of Edwards, an application of Heck v. Humphrey, 512 U.S. 477, 114 S.Ct. 2364, 129 L.Ed.2d 383 (1994). Cf. DeWalt v. Carter, 224 F.3d 607 (7th Cir. 2000). Montgomery can achieve review of the board's decision by concentrating on his credit-earning class, so Edwards blocks use of § 1983 unless Montgomery prevails in the § 2254 proceedings. Second, Montgomery has been barred from filing any § 1983 litigation as a result of his failure to pay sanctions imposed for his history of vexatious civil litigation. See Support Systems International, Inc. v. Mack, 45 F.3d 185 (7th Cir. 1995). The preclusion order has an exception for proceedings under § 2254 but forecloses other civil actions until the sanctions have been paid. Third, as Sandin holds, a prisoner has neither a "liberty" nor a "property" interest in remaining in a prison's general population, so Montgomery has nothing substantive to complain about. Only the reduction in credit-earning class requires further consideration.

Although § 2254 provides the right vehicle, the question remains whether a reduction in credit-earning class deprives a prisoner of "liberty" or "property" — for if not then the state is free to use any procedures it chooses, or no procedures at all. Any temptation to say that freedom from confinement (the maximum 6 year and 8 month difference between Class I and Class II) must be a form of "liberty" was squelched in opinions such as Greenholtz v. Nebraska Penal Inmates, 442 U.S. 1, 99 S.Ct. 2100, 60 L.Ed.2d 668 (1979), Connecticut Board of Pardons v. Dumschat, 452 U.S. 458, 101 S.Ct. 2460, 69 L.Ed.2d 158 (1981), and Ohio Adult Parole Authority v. Woodard, 523 U.S. 272, 118 S.Ct. 1244, 140 L.Ed.2d 387 (1998). These hold that a judgment of conviction extinguishes natural liberty for its full length — in Montgomery's case, 40 years. Opportunities for early release, such as parole or pardon, constitute either property interests or a form of synthetic liberty, and then only if the state has made a promise. Unilateral expectations and hopes for early release do not constitute property, which depends on a legitimate claim of entitlement. See Kentucky Department of Corrections v. Thompson, 490 U.S. 454, 109 S.Ct. 1904, 104 L.Ed.2d 506 (1989); Board of Pardons v. Allen, 482 U.S. 369, 107 S.Ct. 2415, 96 L.Ed.2d 303 (1987); Wallace v. Robinson, 940 F.2d 243 (7th Cir. 1991) (en banc). Sandin altered this positivest approach for sanctions such as segregation that do not affect the length of incarceration, but it did not modify the rules for determining when extension of custody affects a liberty or property interest, see 515 U.S. at 484, 115 S.Ct. 2293, so we apply to good-time credits the approach articulated by Greenholtz and its successors for parole release. In the federal system, and increasingly among the states, good-time credits are a prisoner's principal if not exclusive avenue of early release, so equating the constitutional treatment of the two systems is apt.

Good-time credits are statutory liberty interests once they have been awarded, just as parole is a form of statutory liberty once the prisoner has been released. That is the basis of Wolff v. McDonnell, 418 U.S. 539, 94 S.Ct. 2963, 41 L.Ed.2d 935 (1974), which holds that states must use appropriate procedures before revoking credits they have bestowed. See also Morrissey v. Brewer, 408 U.S. 471, 92 S.Ct. 2593, 33 L.Ed.2d 484 (1972) (same conclusion for parole revocation). What about the prisoner's quest for credits not yet awarded? A hope to be released before the expiration of one's term on good-time credits is no different in principle from a hope to be released on parole, and we know from Dumschat and, e.g., Heidelberg v. Illinois Prisoner Review Board, 163 F.3d 1025 (7th Cir. 1998), that a system making release entirely discretionary also means that the setting of a parole-release date does not entail "liberty" or "property." Similarly a state may, but need not, create a legitimate claim of entitlement to good-time credits. According to Malchi v. Thaler, 211 F.3d 953, 958-59 (5th Cir. 2000), the good-time system used by Texas does not create such an entitlement. What about Indiana's?

Indiana initially assigns each prisoner to Class I, see I.C. § 35-50-6-4(a), and provides that he may be reassigned to Class II or Class III (no credit) "if he violates any of the following . . .", id. at 4(b). The list includes rules of the prison system, the particular prison, and a community treatment program, but excludes rules of the parole and probation systems. This approach — an entitlement subject to defeasance for misconduct that must be specified by rule — curtails administrators' discretion and thus gives prisoners more than a subjective hope of receiving day-for-day credit in Class I. It is quite similar to the system that Greenholtz held to establish statutory liberty. Nebraska entitled each prisoner to release as soon as he became eligible for parole, unless the parole board came to one of four enumerated conclusions that would justify denial of parole. Although the statutory list afforded plenty of discretion — a prisoner could be kept in jail if, for example, "release would depreciate the seriousness of his crime or promote disrespect for law", Neb. Rev.Stat. § 83-1,114 (1)(b) (1976) — the Court concluded that a statute creating an entitlement to release unless findings are made suffices to convey more than a subjective hope. Just so with good-time credits in Indiana. Every prisoner begins in Class I and is entitled to stay there unless he violates a rule. Indiana's officials have less discretion than Nebraska's. It follows that Indiana must afford due process before reducing a prisoner's credit-earning class. So we assumed in Meeks v. McBride, 81 F.3d 717, 719 (7th Cir. 1996). Now the assumption is a holding.

What process is due? Both Montgomery and Indiana suppose that the answer is the list of procedures Wolff prescribed for the revocation of good-time credits. Yet for parole the Supreme Court has distinguished between the procedures required for revocation and the less formal requirements to deny release. See, e.g., Greenholtz, 442 U.S. at 13-15, 99 S.Ct. 2100. The Court noted that losing what you have is a more grievous loss than not getting what you want, an important distinction because the gravity of the loss matters to the due-process calculus under Mathews v. Eldridge, 424 U.S. 319, 96 S.Ct. 893, 47 L.Ed.2d 18 (1976). Although this approach jostles against a recognition of opportunity cost (a month in prison is a month in prison, whether caused by failure to award parole or by awarding and then revoking parole), the Court saw a difference at the level of the prisoner's expectations. In opportunity-cost terms, a chance of early release must be discounted by the probability that it will not occur, while release once granted has a probability of one. Moreover, the Court observed in Greenholtz, any one denial of parole is inconclusive; the state will offer another chance. Because any particular denial just defers release until the next hearing, the stakes and correspondingly the required procedures are diminished. That is equally true with credit-earning class assignments in Indiana. A person in Class II or III is entitled to a review every 6 months. I.C. § 35-50-6-4(d). Thus it is an overstatement to say that a person assigned to Class II soon after the commencement of a 40-year term spends an extra 6 years and 8 months in prison. The inmate may be returned to Class I at any of the biannual reviews. What follows is that the procedures required by due process likely are less elaborate than those prescribed by Wolff for revocation decisions. But we need not decide how much less, not only because the parties have not contested this issue but also because Indiana by statute gives prisoners more than Wolff requires. See I.C. § 35-50-6-4(c).

The prison rule that Montgomery violated requires inmates to obey state and federal criminal laws. According to the charge Montgomery violated I.C. § 35-45-10-5 by stalking a female guard. Indiana defines stalking as intentional harassment that would cause a reasonable person to feel frightened, intimidated, or threatened. See I.C. § 35-45-10-1. The prison concluded that Montgomery had become infatuated with officer Debbie Harrison and had decided that a course of harassment was the best way to induce her to feel the same way about him. The evidence at the disciplinary proceeding demonstrated that Montgomery had gone out of his way to be near Harrison, had threatened her, had refused to leave when she directed, and so on. The prison held a hearing at which Montgomery was represented by a lay advocate and was allowed to testify and present written and oral evidence. The disciplinary board found that the evidence supported the charge.

Montgomery's objections are niggling. One is that the board did not make adequate findings of fact. Whether it need make any findings ( Greenholtz held not) is open to question, but irrelevant here because the board explicitly relied on the conduct report and investigative report, both of which supplied details. See Pardo v. Hosier, 946 F.2d 1278, 1285 (7th Cir. 1991). Another is that the charge was defective because it identified the anti-stalking law as I.C. § 35-45-10.5 rather than I.C. § 35-45-10-5. How anyone could have been misled by the difference between a period and a dash is a mystery; there is no I.C. § 35-45-10.5, so access to a set of statute books (which the prison's library provides) would have led Montgomery to the right citation. He could have asked for clarification had he been in genuine doubt. Still another is a contention that he was interviewed without Miranda warnings, which may well be true but is irrelevant to prison discipline. See Baxter v. Palmigiano, 425 U.S. 308, 96 S.Ct. 1551, 47 L.Ed.2d 810 (1976). Montgomery has many more objections on his list, but all depend on the idea that prison discipline is a form of criminal prosecution, which as Baxter holds is not so. None of his other objections requires separate discussion.

AFFIRMED.