delivered the opinion of the Court.
We granted certiorari to decide whether the retroactive application of a Georgia law permitting the extension of intervals between parole considerations violates the Ex Post Fasto Clause. The Court of Appeals found that retroactive application of the change in the law was necessarily an ex post facto violation. In disagreement with that determination, we reverse its judgment and remand for further proceedings.
*247t — I
In 1974 respondent Robert L. Jones began serving a life sentence after Ms conviction for murder in the State of Georgia. He escaped from prison some five years later and, after being a fugitive for over two years, committed another murder. He was apprehended, convicted, and in 1982 sentenced to a second life term.
Under Georgia law, at all times relevant here, the State’s Board of Pardons and Paroles (Board or Parole Board) has been required to consider inmates serving life sentences for parole after seven years. Ga. Code Ann. §42-9-45(b) (1982). The issue in this case concerns the interval between proceedings to reconsider those inmates for parole after its initial denial. At the time respondent committed his second offense, the Board’s Rules required reconsiderations to take place every three years. Ga. Rules & Regs., Rule 475-3-.05(2) (1979). In 1985, after respondent had begun serving his second life sentence, the Parole Board, acting under its authority to “set forth ... the times at which periodic reconsideration [for parole] shall take place,” Ga. Code . Ann. §42-9-45(a) (1982), amended its Rules to provide that “Reconsideration of those inmates serving life sentences who have been denied parole shall take place at least every eight years,” Ga. Rules & Regs., Rule 475-3-.05(2) (1985).
The Parole Board considered respondent for parole in 1989, seven years after the 1982 conviction. It denied release and, consistent with the 1985 amendment to Rule 475-3-05(2), reconsideration was set for 1997, eight years later. In 1991, however, the United States Court of Appeals for the Eleventh Circuit held that retroactive application of the amended Rule violated the Ex Post Facto Clause. Akins v. Snow, 922 F. 2d 1558, cert. denied, 501 U. S. 1260 (1991). In compliance with that decision, in effect reinstating its earlier 3-year Rule, the Parole Board reconsidered respondent’s case in 1992 and in 1995. Both times parole was denied, the Board citing for its action respondent’s “multiple offenses” *248and the “circumstances and nature of” the second offense. App. 53-54.
In 1995 the Parole Board determined that our decision in California Dept. of Corrections v. Morales, 514 U. S. 499 (1995), had rejected the rationale underlying the Eleventh Circuit’s decision in Akins. The Board resumed scheduling parole reconsiderations at least every eight years, and so at respondent’s 1995 review it set the next consideration for 2003. Had the Board wished to do so, it could have shortened the interval, but the 8~year period was selected based on respondent’s “multiple offenses” and the “circumstances and nature of” his second offense. App. 54. Respondent, acting pro se, brought this action under Rev. Stat. § 1979, 42 U. S. C. § 1983, claiming, inter alia, the amendment to Rule 475-3-.05(2) violated the Ex Post Facto Clause. The suit was filed against individual members of the Parole Board, petitioners in this Court. Respondent requested leave to conduct discovery to support his claim, but the District Court denied the motion and entered summary judgment for petitioners. The court determined the amendment to Rule 475-3- 05(2) “change[dj only the timing between reconsideration hearings” for inmates sentenced to life in prison, thereby “relieving the Board of the necessity of holding parole hearings for prisoners who have no reasonable chance of being released.” App. to Pet. for Cert. 27a. Because the Parole Board’s policies permit inmates, upon a showing of “a change in their circumstance or where the Board receives new information,” App. 56, to receive expedited reconsideration for parole, the court further concluded the amendment created “ ‘only the most speculative and attenuated possibility’ ” of increasing a prisoner’s measure of punishment, App. to Pet. for Cert. 27a (quoting Morales, supra, at 509).
The Court of Appeals reversed, finding the amended Georgia Rule distinguishable in material respects from the California law sustained in Morales. 164 F. 3d 589 (CA11 1999). In finding the Georgia law violative of the Ex Post Facto *249Clause, the court posited that the set of inmates affected by the retroactive change — all prisoners serving life sentences — is “bound to be far more sizeable than the set . . . at issue in Morales” — inmates convicted of more than one homicide. Id., at 594. The Georgia law sweeps within its coverage, the court continued, “many inmates who can expect at some point to be paroled,” ibid., and thus “seems certain to ensure that some number of inmates will find the length of their incarceration extended in violation of the Ex Post Facto Clause of the Constitution,” id., at 595. “Eight years is a long time,” the court emphasized, and “[m]uch can happen in the course of eight years to affect the determination that an inmate would be suitable for parole.” Ibid. The Court of Appeals recognized that the Parole Board would set a new parole review date three years or more into the future (up to eight years) only where it concludes that “ ‘it is not reasonable to expect that parole would be granted’ ” sooner. Ibid, (quoting policy statement of Parole Board). The court thought this policy insufficient, however, because, unlike the statute in Morales, it does not require the Board “to make any particularized findings” and is not “carefully tailored.” 164 F. 8d, at 594-595. The court also recognized that the Board’s policy permitted it to reconsider any parole denials upon a showing of a “change in cir-cumstanee[s]” or upon the Board’s receipt of “new information.” The court deemed the policy insufficient, however, stating that “[p]olicy statements, unlike regulations are unenforceable and easily changed, and adherence to them is a matter of the Board’s discretion.” Id., at 595.
We granted certiorari, 527 U. S. 1068 (1999), and we now reverse.
II
The States are prohibited from enacting an ex post facto law. U. S. Const., Art. I, § 10, cl. 1. One function of the Ex Post Facto Clause is to bar enactments which, by retroactive operation, increase the punishment for a crime after its com*250mission. Collins v. Youngblood, 497 U. S. 37, 42 (1990) (citing Beazell v. Ohio, 269 U. S. 167, 169-170 (1925)). Retroactive changes in laws governing parole of prisoners, in some instances, may be violative of this precept. See Lynce v. Mathis, 519 U. S. 433, 445-446 (1997) (citing Weaver v. Graham, 450 U. S. 24, 32 (1981)); Morales, 514 U. S., at 508-509. Whether retroactive application of a particular change in parole law respeets the prohibition on ex post facto legislation is often a question of particular difficulty when the discretion vested in a parole board is taken into account.
Our recent decision in Morales is an appropriate beginning point. There a California statute changed the frequency of reconsideration for parole from every year to up to every three years for prisoners convicted of more than one homicide. Id., at 503. We found no ex post facto violation, emphasizing that not every retroactive procedural change creating a risk of affecting an inmate’s terms or conditions of confinement is prohibited. Id., at 508-509. The question is “a matter of ‘degree.’ ” Id., at 509 (quoting Beazell, supra, at 171). The controlling inquiry, we determined, was whether retroactive application of the change in California law created “a sufficient risk of increasing the measure of punishment attached to the covered crimes.” 514 U. S., at 509.
The amended California law did not violate this standard. It did not modify the Statutory punishment imposed for any particular offenses. Nor did the amendment alter the standards for determining either the initial date for parole eligibility or an inmate’s suitability for parole. Id., at 507. The amendment did not change the basic structure of California’s parole law. It vested the California parole board with discretion to decrease the frequency with which it reconsidered parole for a limited class, consisting of prisoners convicted of more than one homicide. Id., at 507, 510. If the board determined a low likelihood of release existed for a member within that class, it could set the prisoner’s next consider*251ation date three years hence. The change in California law did not, however, prohibit requests for earlier reconsideration based on a change of circumstances. Id., at 512-513. Historical practices within the California penal system indicated “about 90% of all prisoners are found unsuitable for parole at the initial hearing, while 85% are found unsuitable at the second and subsequent hearings.” Id., at 510-511 (citing In re Jackson, 39 Cal. 3d 464, 473, 703 P. 2d 100, 105 (1985)). On these facts we determined the Ex Post Facto Clause did not prohibit California from conserving and reallocating the resources that would otherwise be expended to conduct annual parole hearings for inmates with little chance of release. 514 U. S., at 511-512. The sum of these factors illustrated that the decrease in the frequency of parole suitability proceedings “ereate[d] only the most speculative and attenuated possibility of producing the prohibited effect of increasing the measure of punishment for covered crimes.” Id., at 509.
Consistent with the Court of Appeals’ analysis, respondent stresses certain differences between Georgia’s amended parole law and the California statute reviewed in Morales. The amendment to Rule 475-3-.05(2), respondent urges, permits the extension of parole reconsiderations by five years (not just by two years); covers all prisoners serving life sentences (not just multiple murderers); and affords inmates fewer procedural safeguards (in particular, no formal hearings in which counsel can be present). These differences are not dispositive. The question is whether the amended Georgia Rule creates a significant risk of prolonging respondent’s incarceration. See ibid. The requisite risk is not inherent in the framework of amended Rule 475-3-.05(2), and it has not otherwise been demonstrated on the record.
Our decision in Morales did not suggest all States must model their procedures governing consideration for parole after those of California to avoid offending the Ex Post Facto Clause. The analysis undertaken in Morales did identify *252factors which convinced us the amendment to California law created an insignificant risk of increased punishment for covered inmates. Our opinion was careful, however, not to adopt a single formula for identifying which legislative adjustments, in matters bearing on parole, would survive an ex post facto challenge. Ibid. We also observed that the Ex Post Facto Clause should not be employed for “the micromanagement of an endless array of legislative adjustments to parole and sentencing procedures.” Id., at 508. These remain important concerns. The States must have due flexibility in formulating parole procedures and addressing problems associated with confinement and release.
The case turns on the operation of the amendment to Rule 475-3- 05(2) within the whole context of Georgia’s parole system. Georgia law charges the Parole Board with determining which prisoners “may be released on pardon or parole and [with] fixing the time and conditions thereof.” Ga. Code Ann. §42-9-20 (1997). In making release decisions, the same law, in relevant part, provides:
“Good conduct, achievement of a fifth-grade level or higher on standardized reading tests, and efficient performance of duties by an inmate shall be considered by the board in his favor and shall merit consideration of an application for pardon or parole. No inmate shall be placed on parole until and unless the board shall find that there is reasonable probability that, if he is so released, he will live and conduct himself as a respectable and law-abiding person and that his release will be compatible with his own welfare and the welfare of society. Furthermore, no person shall be released on pardon or placed on parole unless and until the board is satisfied that he will be suitably employed in self-sustaining employment or that he will not become a public charge.” §42-9-42(c).
*253See also § 42-9-48 (listing information the Board should consider, including wardens’ reports, results of physical and mental examinations, and reports regarding prisoners’ performance in educational programs). These provisions illustrate the broad discretion the Parole Board possesses in determining whether an inmate should receive early release. Accord, Sultenfuss v. Snow, 35 F. 3d 1494, 1501-1502 (CA11 1994) (en banc) (describing the discretion Georgia law vests with Parole Board). Only upon a showing that the Board engaged in a “gross abuse of discretion” can a prisoner challenge a parole denial in the Georgia courts. Lewis v. Griffin, 258 Ga. 887, 888, n. 3, 376 S. E. 2d 364, 366, n. 3 (1989).
The presence of discretion does not displace the protections of the Ex Post Facto Clause, however. Cf. Weaver, 450 U. S., at 30-31. The danger that legislatures might disfavor certain persons after the fact is present even in the parole context, and the Court has stated that the Ex Post Facto Clause guards against such abuse. See Miller v. Florida, 482 U. S. 423, 429 (1987) (citing Calder v. Bull, 3 Dall. 386, 389 (1798) (Chase, J.)). On the other hand, to the extent there inheres in ex post facto doctrine some idea of actual or constructive notice to the criminal before commission of the offense of the penalty for the transgression, see Weaver, supra, at 28-29, we can say with some assurance that where parole is concerned discretion, by its very definition, is subject to changes in the manner in which it is informed and then exercised. The idea of discretion is that it has the capacity, and the obligation, to change and adapt based on experience. New insights into the accuracy of predictions about the offense and the risk of recidivism consequent upon the offender’s release, along with a complex of other factors, will inform parole decisions. See, e.g., Justice v. State Board of Pardons and Paroles, 234 Ga. 749, 751-752, 218 S. E. 2d 45, 46-47 (1975) (explaining, by illustration to one prisoner’s circumstances, that parole decisions rest upon the *254Board's consideration of numerous factors specific to an inmate’s offense, rehabilitative efforts, and ability to live a responsible, productive life). The essence of respondent’s case, as we see it, is not that discretion has been changed in its exercise but that, in the period between parole reviews, it will not be exercised at all. The statutory structure, its implementing regulations, and the Parole Board’s unrefuted representations regarding its operations do not lead to this conclusion.
The law changing the frequency of parole reviews is qualified in two important respects. First, the law vests the Parole Board with discretion as to how often to set an inmate’s date for reconsideration, with eight years for the maximum. See Ga. Rules & Regs., Rule 475-3-05(2) (1985) ("Reconsideration . . . shall take place at least every eight years”). Second, the Board’s policies permit “expedited parole reviews in the event of a change in their circumstance or where the Board receives new information that would warrant a sooner review.” App. 56. These qualifications permit a more careful and accurate exercise of the discretion the Board has had from the outset. Rather than being required to review cases pro forma, the Board may set reconsideration dates according to the likelihood that a review will result in meaningful considerations as to whether an inmate is suitable for release. The Board’s stated policy is to provide for reconsideration at 8-year intervals “when, in the Board's determination, it is not reasonable to expect that parole would be granted during the intervening years.” Ibid. The policy enables the Board to put its resources to better use, to ensure that those prisoners who should receive parole come to its attention. By concentrating its efforts on those cases identified as having a good possibility of early release, the Board’s Rules might result in the release of some prisoners earlier than would have been the case otherwise.
The particular case of respondent well illustrates that the Board’s Rule changes are designed for the better exercise of *255the discretion it had from the outset. Given respondent’s criminal history, including his escape from prison and the commission of a second murder, it is difficult to see how the Board increased the risk of his serving a longer time when it decided that its parole review should be exercised after an 8-year, not a 3-year, interval. Yet if such a risk develops, respondent may, upon a showing of either “a change in [his] circumstance^]” or the Board’s receipt of “new information,” seek an earlier review before the 8-year interval runs its course.
We do not accept the Court of Appeals’ supposition that Rule 475-3-.05(2) “seems certain” to result in some prisoners serving extended periods of incarceration. 164 F. 3d, at 596. The standard announced in Morales requires a more rigorous analysis of the level of risk created by the change in law. Cf. 514 U. S., at 506-507, n. 3 (“After Collins, the focus of the ex post facto inquiry is not on whether a legislative change produces some ambiguous sort of ‘disadvantage’ . .. but on whether any such change . .. increases the penalty by which a crime is punishable”). When the rule does not by its own terms show a significant risk, the respondent must demonstrate, by evidence drawn from the rule’s practical implementation by the agency charged with exercising discretion, that its retroactive application will result in a longer period of incarceration than under the earlier rule. The litigation in Morales concerned a statute covering inmates convicted of more than one homicide and proceeded on the assumption that there were no relevant differences between inmates for purposes of discerning whether retroactive application of the amended California law violated the Ex Post Facto Clause. In the case before us, respondent must show that as applied to his own sentence the law created a significant risk of increasing his punishment. This remains the issue in the ease, though the general operation of the Georgia parole system may produce relevant evidence and inform further analysis on the point.
*256The record before the Court of Appeals contained little information bearing on the level of risk created by the change in law. Without knowledge of whether retroactive application of the amendment to Rule 475-3- 05(2) increases, to a significant degree, the likelihood or probability of prolonging respondent’s incarceration, his claim rests upon speculation.
On the record in this case, we cannot conclude the change in Georgia law lengthened respondent’s time of actual imprisonment. Georgia law vests broad discretion with the Board, and our analysis rests upon the premise that the Board exercises its discretion in accordance with its assessment of each inmate’s likelihood of release between reconsideration dates. If the assessment later turns out not to hold true for particular inmates, they may invoke the policy the Parole Board has adopted to permit expedited consideration in the event of a change in circumstances. App. 56.
The Court of Appeals erred in not considering the Board’s internal policy statement. At a minimum, policy statements, along with the Board’s actual practices, provide important instruction as to how the Board interprets its enabling statute and regulations, and therefore whether, as a matter of fact, the amendment to Rule 475-3-05(2) created a significant risk of increased punishment. It is often the case that an agency’s policies and practices will indicate the manner in which it is exercising its discretion. Cf. INS v. Yueh-Shaio Yang, 519 U. S. 26, 32 (1996) (observing that the reasonableness of discretionary agency action can be gauged by reference to the agency’s policies and practices). The Court of Appeals was incorrect to say the Board’s policies were of no relevance in this ease. Absent a demonstration to the contrary, we presume the Board follows its statutory commands and internal policies in fulfilling its obligations. Cf. United States ex rel. Accardi v. Shaugnessy, 347 U. S. 260, 266-268 (1954). In Morales, we relied upon the State’s representation that its parole board had a practice of grant*257ing inmates’ requests for early review. See 514 U. S., at 512-513 (citing Reply Brief for Petitioner, O. T. 1994, No. 93-1462, p. 3, n. 1). The policy statement here, by contrast, is a formal, published statement as to how the Board intends to enforce its Rule. It follows a fortiori from Morales that the Court of Appeals should not have disregarded the policy. Absent any demonstration to the contrary from respondent, we respect the Board’s representation that inmates, upon making a showing of a “change in their circumstance[s]” or upon the Board’s receipt of “new information,” may request expedited consideration. App. 56.
Court of Appeals’ analysis failed to reveal whether the amendment to Rule 475-3~.05(2), in its operation, created a significant risk of increased punishment for respondent. Respondent claims he has not been permitted sufficient discovery to make this showing. The matter of adequate discovery is one for the Court of Appeals or, as need be, for the District Court in the first instance. The judgment of the Court of Appeals is reversed, and the case is remanded for proceedings consistent with this opinion.
It is so ordered.