This appeal has its genesis in the material modification of parole laws by the Pennsylvania legislature in 1996 and corresponding changes in the parole decisionmaking policies of the Pennsylvania Board of Probation and Parole ("Board").1 As a consequence, the post-1996 parole regime placed primary consideration on the risk to public safety by the parole petitioner as the dominant factor in evaluating parole applications. The United States District Court for the Eastern District of Pennsylvania held that the Board retroactively applied this policy change adversely to the parole applications of Louis Mickens-Thomas ("Thomas"), in violation of the Ex Post Facto clause. The Commonwealth timely appealed; Thomas cross-appealed on his claim that the Board violated his due process rights when it denied his parole applications. We affirm.2

A. Pre-1996 Parole Considerations in Pennsylvania

Thomas is currently serving a life sentence for the 1964 rape and murder of a 12-year-old girl in Philadelphia, Pennsylvania. The parties agreed to vacate the original guilty verdict because of the unreliability of the expert whose testimony connected fibers and microscopic particles found on the victim to Thomas. In 1967, the state trial court granted Thomas a new trial; in 1969, he was again convicted.3 His second conviction was upheld by the Pennsylvania Supreme Court in 1972. However, Thomas still professes innocence. Thomas is presently 74-years-old and has been in prison for nearly 40 years. His current efforts to seek release on parole have garnered the strong support of prisoner advocates, incurred the equally vehement opposition of the Philadelphia District Attorney, and have attracted considerable media scrutiny.

Life sentences in Pennsylvania presumptively exclude any possibility of parole. The only exception occurs when the governor-appointed Pardons Board recommends commutation of the inmate's sentence by majority vote, and the Governor subsequently approves the commutation. Thomas was one of only 27 sentences commuted by former Governor Casey out of nearly 3000 life terms being served during his tenure as governor. By the terms of his commutation. Thomas became eligible for parole on July 21, 1996. In recommending commutation, the Pardons Board noted Thomas's attainment of a college degree, his participation in Alcoholics Anonymous, his participation in sex-offender therapy, the support of the Corrections Department, the long length of time served, the numerous recommendations from scholars, religious, and community leaders, and Thomas's overall maturity and stability.

Following a commutation, a prisoner seeking to be released must still submit to the same parole procedures applicable to all other prisoners. Furthermore, the parole must first be approved by the Board, which virtually has unreviewable power to grant or deny the parole application. Around the time of Thomas's eligibility for parole, new appointments of then-Governor Ridge were placed on the Board in 1995;4 a parolee from the Pennsylvania prison system was arrested for murder in New Jersey in 1995; and in early 1996 a Pennsylvania Senate committee, in view of the New Jersey arrest, strongly recommended that the Board place added emphasis on community safety. In December 1996, Pennsylvania enacted a change in its law concerning the Board's mission, which arguably placed greater emphasis on public safety as a criterion for parole release.

In December 1996 the Pennsylvania legislature modified the law governing parole in Pennsylvania. The new language, inserted into the aspirational introductory provision of the Pennsylvania parole statutes, provides that the public safety must be considered "first and foremost" in the Board's execution of its mission. The relevant statute, in its post-1996 form, provides as follows:5

§ 331.1. Public policy as to parole

The parole system provides several benefits to the criminal justice system, including the provision of adequate supervision of the offender while protecting the public, the opportunity for the offender to become a useful member of society and the diversion of appropriate offenders from prison. In providing these benefits to the criminal justice system, the board shall first and foremost seek to protect the safety of the public. In addition to this goal, the board shall address input by crime victims and assist in the fair administration of justice by ensuring the custody, control and treatment of paroled offenders.

The 1941-1996 statute, in effect at the time of Thomas's conviction, made no specific mention of public safety. It provided:

The value of parole as a disciplinary and corrective influence and process is hereby recognized, and it is declared to be the public policy of this Commonwealth that persons subject or sentenced to imprisonment for crime shall, on release therefrom, be subjected to a period of parole during which their rehabilitation, adjustment, and restoration to social and economic life and activities shall be aided and facilitated by guidance and supervision under a competent and efficient parole administration, and to that end it is the intent of this act to create a uniform and exclusive system for the administration of parole in this Commonwealth.

To assess this modification of the statute, one must regard this change in the context of recent policy statements issued by the Board and other government officials. Other events coincident with the 1996 revision must also be considered to determine whether, in practice, the parole policies of the Commonwealth have undergone any substantive changes.6 The Board's 1989 Manual of Operations and Procedures recognized that "[p]robation and parole services must consider that offenders can change their behavior patterns when desirous, capable, and given the opportunity, help, dignity, and respect they deserve as human beings." The Manual goes on to state that, in considering an inmate for parole, the Board must "weigh[] numerous factors relative to the welfare of the client and the safety of the community," including seriousness of the offense; length of the sentence; institutional adjustment (behavior and program adjustment); and assessment of the effect of rehabilitation services while incarcerated. Whether the individual can be safely supervised in the community, personality characteristics, any history of family violence, strength of the parole plan (home and employment), testimony from victims, and opinions of the sentencing judge and prosecuting attorney must also be considered.

In reaching its parole decision, "[t]he Board ... feel[s] that an individual should be given every consideration for parole at the expiration of the minimum sentence." The 1990 Board-authored "Parole Decision Making Guidelines: A Statement on Policy, Procedure and Philosophy," stated that "[a]n eligibility for parole expresses a philosophy of presumed release unless information reviewed demonstrates by its preponderance that the public safety interests of the community outweigh the liberty interests of the inmate."

A decisional protocol called the "Parole Decision Making Guidelines" played a critical role in the Board's decisionmaking process pre-1996. The Guidelines provided an objective prediction of the likelihood of a successful parole by assigning numerical values to various criteria, based on historical patterns of parolee recidivism rates. According to the Board-authored 1991 "Guidelines: A Special Report Prepared for the House Judiciary Committee," the Guidelines considered factors which, based on a study of past instances of recidivism, were predictors of future recidivism in potential parolees. For example, historically, those convicted of theft offenses have high recidivism rates. Thus, they are assigned a higher score; similarly, substance abusers are high recidivists, and correspondingly, they have higher scores assigned in the Guidelines. If enough of these negative predictors are present, the aggregate score will fall above a threshold value, and the Guidelines will recommend against parole.

In addition to risk of recidivism, "risk to the community" is also a relevant factor under the Guidelines in determining whether parole is warranted. Thus the Guidelines, in addition to calculating the risk of recidivism, add additional points for offenders whose post-parole conduct might include violent behavior, categorizing them as having "high assaultive behavior potential." Therefore, a person who initially committed a violent crime will be evaluated as a parole candidate under the Guidelines, based upon his or her risk to the public as a function of both the likelihood of recidivism and the severity of the crime he or she might commit as a recidivist. The Board's 1990 Statement on Policy, Procedure and Philosophy notes that

[i]nherent in the concept of risk management is the notion that some crimes, although less likely to be repeated, have more serious consequences than others if repeated. In other words, although some offenders may be low risk from the viewpoint of recidivism and parole supervision failure, the stakes are high in terms of protecting the public if their new crime is violent or serious in nature.

According to the House Judiciary Report, the Guidelines are "related to an empirically sound, criterion-referenced policy assessment which evaluates each inmate in terms of criminal justice policy and normative past practice in parole decision making." Although the Guidelines aspire to objectivity in parole decision making, the Board periodically has countermanded their recommendations. The 1990 Statement on Policy, Procedure and Philosophy declared that the Guidelines protocol typically determined "approximately eighty percent of the decisions rendered." The Statement notes, however "that the structuring of discretion does not eliminate it; the burden of the decision remains with the decision maker to evaluate the merit of each case."

Thus, before 1996, about 20% of such decisions were decided contrary to the Guidelines recommendation. Nonetheless, a discretionary Guidelines departure must give full weight to all the factors, both for and against release. The Report stated that decisions which countermand the Guidelines are "rarities," and "require written explanation to justify the policy exception." In reaching a decision outside the Guidelines, the Board cannot merely recite factors already incorporated in the Guidelines analysis; instead it must consider "consequential characteristics" of the applicant that are "objectively unique" to the factors already incorporated into the Guidelines analysis.7 A Guidelines worksheet is provided along with the formal numerical protocol, to permit elucidation of non-Guidelines factors reached in a parole decision, as well as a checklist (Part IV.B. of the Guidelines) for designation of common non-Guidelines factors which may have been significant in reaching a decision contrary to the Guidelines recommendation.8

In early 1995, Robert "Mudman" Simon, who had been released on parole from his Pennsylvania prison sentence, was arrested and charged with murder in New Jersey.9 In mid-1995 Simon's release led to the publication of a "System-Wide Assessment" of the Board, by the Pennsylvania Inspector General, in which it was determined that more emphasis on public safety and on the nature of the underlying crime was needed in making parole decisions. The System-Wide Assessment noted that the Board management generally "has emphasized [inmate] interests over community protection," and that one faction of Board officials perceives "reintegration into society [as] the primary mission," while another faction "focuses ... on the protection of society."

The Chairman of the Pennsylvania Senate Judiciary Committee, reporting on the investigation of the parole of Simon, published in February 1996, noted that following a 1989 prison riot in Pennsylvania, "reduction in prison overcrowding through the parole process became an objective of [the Board] coequal with assuring public safety." The Report also noted that the Board recently "has set out to establish public safety as a priority including a more careful review of parole eligible cases." Accordingly, the Report recommended that "[e]xisting statutes and policies should be looked at, amended, restated and/or purged as necessary to effectuate the new corrections philosophy of Pennsylvania [emphasizing] public safety...." (emphasis added.)

The Board's self-assessment report, entitled "Fiscal Years 1995-1997 Biennial Report," stated that "[i]n recent years, the Governor and General Assembly have mandated through statute that the foremost concern for the Board must be the protection of the safety of the public...." The Report went on to note recent "heightened awareness and concern for public safety," which prompted it to institute more careful review procedures for cases involving "violent offenders." Thus, both the Judiciary Committee Report of February 1996 and the contemporaneous Biennial Report gave public notice that henceforth the "foremost concern" of the Board would be the safety of the public.

We now turn to the effect of the Board's new policy on Thomas's parole application. When Governor Casey commuted Thomas's sentence, he authorized the Department of Corrections to "prerelease" Thomas, presumably into a transitional facility, prior to Thomas's parole-eligibility date. This the Corrections Department declined to do. Then, in the Board's first review of Thomas's case in September 1996, it also declined to parole him on the ground that, under the new law passed in 1995, "prerelease" was made a necessary precondition of parole. The Commonwealth Court heard Thomas's mandamus action in which he sought to direct the Board to hear his parole petition. In that case, the Board conceded that its application of the "prerelease" law violated the Ex Post Facto clause. The court remanded the case to the Board, and ordered it to hear the merits of the parole application and issue its decision within 10 days. See Mickens-Thomas v. Pennsylvania Board of Probation and Parole, 699 A.2d 792 (Pa.Cmwlth.Ct.1997).

Soon thereafter, in August 1997, the Board issued its decision, relying in part on its Decision Making Guidelines to reach a conclusion as to parole eligibility.10 Thomas received a Guidelines-based recommendation for release, along with the recommendations of all voting Department of Corrections institutional staff, including the prison counselor and housing officer. Thomas demonstrated his participation in pre-release counseling, including Alcoholics Anonymous and sex offender therapy, as well as participation in college courses and job training. Thomas also had post-release support networks in place. Nonetheless, the Board denied Thomas parole in 1997. The Board stated its reasons for the denial in a formal letter to Thomas called the "Board Decision." The reasons given were as follows: "Assaultive instant offense. Very high assaultive behavior potential. Victim injury. Unfavorable recommendation from District Attorney. Conviction of prior assault offense." Many of these factors were automatic designations; for example, Thomas's past crime was a sex offense, which caused him to be automatically classified on the Board Decision as having "very high assaultive behavior potential."

The Board's 1997 Decision urged Thomas to secure the following before his next application review: investigation of a home plan; the availability of out-patient sex-offender treatment; participation in a program plan prescribed by Department of Corrections officials; maintenance of a good conduct record; a continuing institutional recommendation for parole; and an evaluation by mental health professionals, with experience with sex offenders. The Board made these recommendations in spite of Thomas's apparent compliance with all of the Board's suggestions prior to the hearing. For example, he had a good conduct record and the endorsement of prison staff.

The Board urged a psychiatric examination, despite the existence of a pre-commutation 1993 psychiatric report supporting Thomas's release.11 There were other psychological evaluations in his file that did not contraindicate release. Although Thomas had engaged in sex offender therapy, there is some indication, based on handwritten notes on the Board's decision making worksheet, that the Board may have been troubled by Thomas's presence in a "deniers" group — those who deny responsibility for the underlying offense — rather than an "admitters" group. The Board Decision makes no specific mention of the admitter-denier distinction, and its internal notes make only passing mention of the issue.12 Later, in 1997, the Pennsylvania Supreme Court heard Thomas's habeas petition but summarily denied it.

Thereafter, Thomas apparently complied with all of the Board's prerequisites stated in its 1997 Decision. He maintained the positive recommendation of corrections authorities, who once more unanimously recommended his release and noted that he was in compliance with treatment programs. The prison counselor, corrections officer and psychologist all endorsed his release. He continued to participate in a sex-offender therapy program — although it was a "deniers" program — along with an Alcoholics Anonymous program. Post-release support networks were in place. And the Guidelines assigned Thomas a risk-assessment score which militated in favor of release.

Despite his compliance with essentially all of the Board's conditions, it again denied parole in March 1998, stating: "Assaultive instant offense. Very high assaultive behavior potential. Victim injury. Your need for counseling and treatment." In this latest Board Decision, it again advised Thomas to seek counseling and treatment, to participate in prescribed programming, to maintain a clean record and obtain institutional recommendation for the purposes of his next application. Unlike the 1997 decision, the 1998 decision recommended no specific sex offender treatment, nor mentioned in its internal decision making worksheet that Thomas was in a "denier" group. Moreover, despite the comment that Thomas needed "counseling and treatment," psychiatric and psychological evaluations did not contraindicate his release. Presumably in response to the 1997 Board Decision's admonition that Thomas needed to be evaluated by a mental health professional, the Board noted, in its worksheet, that a 1998 psychological evaluation showed Thomas to be an "average risk candidate." The Department of Corrections psychologist, in the 1998 Vote Sheet, noted "No Psychological Contraindications" for release.

A 1996 psychological evaluation did show an "antisocial personality," and "possible sexual preoccupation and psychosexual immaturity." However, the decisionmaking worksheet does not reflect that the Board was deeply concerned with those findings, and instead only made mention, in a handwritten notation, of the 1998 psychological report's conclusion that Thomas was an "average risk candidate." No notation was made of two earlier psychological evaluations, which had more clearly favored release. One 1993 report called Thomas "a good candidate for commutation from the psychological perspective." A second 1993 psychiatric report added there was no "psychiatric contraindication [to commutation]" and that Thomas "has developed significantly during his years of imprisonment." In December 1999, Thomas sought a writ of habeas corpus in the United States District Court.

Again, the Board denied his parole in 2000, during the pendency of these habeas corpus proceedings. The Board gave as its reason the cryptic statement that it "has determined that the mandates to protect the safety of the public and to assist in the fair administration of justice cannot be achieved through your release on parole." Again, all voting members of the Department of Corrections institutional staff, including his counselor and work supervisor, unanimously recommended his parole. Again, he demonstrated a continued record of good conduct in prison and participation in sex offender therapy and all other programming prescribed by the Department of Corrections. Nonetheless, the Board denied parole. Again, the Board advised Thomas to maintain his Department of Corrections recommendation as a precondition for consideration at his 2002 parole hearing. But, in 2000 the Board revived its 1997 recommendation that Thomas should undergo sex offender therapy as a suggested pre-condition for release. Handwritten notes again show that the Board may have been concerned that he was in a "denier" group — although, once more, no mention of this concern is made in the formal Board Decision. The lack of admitter therapy is simply stated in a neutral, non-critical way in the Guidelines worksheet.

Finally, although the decision-making guidelines had assigned Thomas a favorability score that counseled in favor of parole on both the 1997 and 1998 applications, the 2000 decision, despite no evidence of changes in his situation, reached a different outcome. The Board interviewer classified Thomas as a habitual substance abuser on the Guidelines form, which increased Thomas's risk score by 3 and placed him in an unfavorable category for release. The Board did not indicate why it made this material alteration to what appears to be a boilerplate risk-assessment protocol. Thomas apparently did have a record of alcohol abuse prior to his incarceration in 1964 (for which he attended Alcoholics Anonymous while in prison) but it is unclear why, if past alcohol abuse over forty years ago was a relevant factor, it had not been considered on his two prior Guidelines evaluations.

Moreover, the Guidelines were modified since his last application, with a score of 2 now added to Thomas's overall score as a result of "Victim Injury" (the past two evaluation forms assigned only a score of "1" for Victim Injury). As a direct result of these changes Thomas's score ascended to nine, placing him in a category exceeding seven. Therefore, the Guidelines contraindicated parole. In summary, the Board denied Thomas parole a total of three times, in 1997, 1998 and 2000, although he complied each time with all of the Board's recommendations, except for his continued enrollment in the deniers group. The Board denied parole to Thomas alone of all 266 prisoners whose life sentences had been commuted.

In Thomas's current habeas petition, he alleges that the Board denied his parole in violation of the Ex Post Facto clause, by applying retroactively the revised December 1996 parole statute. According to Thomas, he had a constitutional expectation that his parole petition would be evaluated under the laws in effect when he was convicted. The District Court agreed that the Board violated the Ex Post Facto clause by applying the 1996 statutory mandate. However, the Court declined to rule outright that Thomas would have been paroled under the prior rule; instead, it remanded the case to the Parole Board to rehear the matter under the pre-1996 laws. Thomas also asked that the District Court order his release on the grounds that his due process rights were violated. The District Court held that, although Thomas had complied with all the seeming prerequisites for relief as prescribed by the Board, the presence of any evidence sufficient to show that the Board based its decision on a rational and good faith exercise of discretion, vindicated its action. Thus, the District Court concluded, there was no due process violation. See Mickens-Thomas v. Vaughn, 217 F. Supp. 2d 570 (E.D.Pa.2002).

The District Court remanded the case to the Board to apply its pre-1996 parole policies to the Thomas petition. The Board appealed, and Thomas cross-appealed, on the denial of his due process claim and on the court's failure to grant him release outright as a result of the Ex Post Facto violation.

The Ex Post Facto clause of the United States Constitution applies to a statutory or policy change that "alters the definition of criminal conduct or increases the penalty by which a crime is punishable." California Dep't of Corrections v. Morales, 514 U.S. 499, 506 n. 3, 115 S. Ct. 1597, 131 L. Ed. 2d 588 (1995). A new law or policy violates the Ex Post Facto clause (1) when it is retrospective, i.e., when it "appl[ies] to events occurring before its enactment," and (2) when it "disadvantage[s] the offender affected by it." Weaver v. Graham, 450 U.S. 24, 29, 101 S. Ct. 960, 67 L. Ed. 2d 17 (1981); see Coady v. Vaughn, 251 F.3d 480, 488 (3d Cir.2001). As to the first criterion for an Ex Post Facto violation, the Board strenuously argues in its brief that "the 1996 amendments... did not change the Board's standards for determining parole." First, it asserts that Pennsylvania's statement of public policy for parole, 61 P.S. § 331.1, refers to the supervision of parolees, rather than to conditions of release.

The foregoing argument has little merit. The statute unequivocally has been interpreted by Pennsylvania courts to express broad and general aspirations of Pennsylvania's parole policy. See Stewart v. Pennsylvania Bd. of Probation and Parole, 714 A.2d 502, 508 (Pa.Cmwlth.Ct. 1998) ("Section 1 of the Parole Act, 61 P.S. §§ 331.1, ... enunciates the state's public policy concerning parole...."). The essential matter before us is not whether the statute on its face pertains to parole decisionmaking, but whether, in practice, the new language has altered the fundament for reviewing parole applications. See Garner v. Jones, 529 U.S. 244, 256, 120 S. Ct. 1362, 146 L. Ed. 2d 236 (2000). We look beyond the language of the statute and examine the Board's pronouncements of policy and its public statements that shed light on the interpretation of its statutory mandate. These suggest that after 1996 the Board gave foremost importance to the public safety factor. This is confirmed by the report in the September 20, 2000 Harrisburg Patriot-News, when then-Board chairman William Ward observed that legislative changes around 1995 recast the Board's mission to put public safety first.

The Board also asserts that its policy historically has placed equal emphasis on the interests of the inmate and the interests of public safety, and it points in its brief to us to statutory language in effect in the 1940s to prove this point: "whenever in its opinion [1] the best interests of the convict justify or require his being paroled and[2] it does not appear that the interest of the Commonwealth will be injured thereby," a prisoner will be granted parole. The Board correctly notes that the potential risk to public safety in granting parole has always been a consideration in the decisional process. It claims that other provisions of the parole statute have, under both the earlier and the current versions, required that the Board "consider the nature and circumstances of the offense committed, [and] the general character and background of the prisoner." 61 P.S. § 331.19.13 However, to state that public safety was always a consideration does not mean that the Board gave it the same weight after 1996 in the decisional equation.

The record is convincing that after 1996, the Board applied to the public safety interest far greater weight. The evidence here demonstrates that since 1996, the Board has given special weight to the risk to public safety. Pre-1996, a prisoner could be denied parole because of public safety concerns only if those concerns together with other relevant factors outweighed, by a preponderance, the liberty interests of the inmate. The 1996 policy change placed first and foremost the public safety to the disadvantage of the remaining liberty interest of the prisoner.

The Pennsylvania courts have suggested that the 1996 public safety directive has caused the Board to review the petitions of violent offenders with redoubled scrutiny: "As a result [of statutory and policy changes in 1996], violent offenders are subjected to a more stringent standard of review for parole eligibility than nonviolent offenders. The purpose behind the classification and the disparate treatment between the violent and nonviolent offenders is the protection of public safety." Myers v. Ridge, 712 A.2d 791, 799 (Pa.Cmwlth.Ct. 1998). Furthermore, the policy change around 1996 took place in the ambience of numerous policy statements that shed light on the Board's interpretation of its statutory mission: it clearly viewed its statutory mandate to require special emphasis on public safety.

Our attention is directed to the 50th Anniversary Report of the Board (1991), which states in its concluding paragraph that "protection of society" is the Board's "primary goal." Read in context, however, this passage applies to the board's supervision of parolees. This same passage provided that "conditional release" permits the Board to meet its goal of protecting society. An earlier section of the document states that "[t]he immediate goal of parole supervision is the protection of society," by closely supervising the parolee and setting "conditions" for continued release, pertaining to work, health, education or other needs, that ensure smooth reintegration and, hence, the public safety. (emphasis added). Upon analyzing this language, it is obvious that the Board meant in this report that "conditional release," with fixed conditions for continued parole, is designed to safeguard the public after a parole has been granted. Thus, this 1991 anniversary report sheds no light on the post-1996 Board treatment of "public safety" as a factor before parole is granted.

The statistical evidence is quite staggering here, and strongly confirms the change in policy in 1996: of the 266 historical instances of commuted sentences on which the Board has kept records, all were granted parole on the first or second application. Many, if not most, of these original sentences were for violent crimes. Doubtless, these earlier Parole Boards spanned a wide spectrum of political and penological philosophies. Yet, the gubernatorial grant of commutation of sentence had such significance that the Board agreed to parole every commutee on his or her first or second application. The Thomas application is distinguished from these 266 cases only by the intervening policy directive of 1996, emphasizing public safety.

In addition to these statistics, substantive declarations of Board policy strongly support the proposition that, after 1996, the Board applied a new standard. A 1996 report by the Legislative Judiciary Committee strongly exhorted the Board to reform its parole policies by placing greater stress on public safety. A 1997 self-assessment by the Board specifically noted that during the 1995-1997 period, public safety became the Board's new "foremost concern." The new Guidelines, implemented between 1998 and 2000, placed more weight on "Victim Injury." The 2000 Board Decision denying Thomas's parole noted that its action was consistent with the Board's "mandate" to protect the public. This language did not appear on earlier Board decisions and reflects its new parole policy.

These declarations stand in bold contrast to the pre-1996 policies, which commanded that the Board give weight to various factors in the parole process, such as Department of Corrections staff recommendations, educational accomplishments, job training, and therapy programs. This factor-based approach strongly suggests that dispositive weight should not be given to any one factor.14 Pre-1996, release upon eligibility for parole was presumed, and any decision to deny parole based on public safety considerations had to be supported by specific reasons, which outweighed those factors favoring release. Prior to 1996, a Board recommendation contrary to the Guidelines required that the Board have "appropriate reasons for [its parole denial] decision." Because the pre-1996 Guidelines already factored in the risk to public safety vís-a-vís relevant recidivism indicators, the Board after 1996 could not give added, and certainly not exclusive, weight to public safety in overruling the Guidelines.

We conclude, then, that prior to 1996, the Board's concern for potential risks to public safety could not be the sole or dominant basis for parole denial under the existing Guidelines. Considerations of public safety were already incorporated into its Guidelines analysis; the Board had to point to "unique" factors as a basis for its rejection of the Guidelines. Moreover, the Board had to weigh all factors, militating for and against parole, and make its decision on the totality of the factors pertinent to parole, and give appropriate weight to the interests of the inmate. Heavy foot application on one factor could not have been the basis of granting or rejecting parole. Policy declarations in and after 1996 demonstrate that Board stance shifted and that, indeed, post-1996 considerations of public safety became the dominant concern of the Board.

The possession of a discretionary component in a parole policy does not per se exempt it from constitutional scrutiny. "The presence of discretion does not displace the protections of the Ex Post Facto clause." Garner, 529 U.S. at 253, 120 S. Ct. 1362; cf. Winsett v. McGinnes, 617 F.2d 996, 1007 (3d Cir.1980) (en banc) (holding that prison officials' "discretion must be exercised consistently with the purpose and policy" governing early release program to satisfy due process). A Parole Board policy, although partly discretionary, is still subject to ex post facto analysis when there are sufficiently discernible criteria to suggest to a reviewing body that the new retroactive policies are being applied against the offender's interest.

In this case, as in our Winsett decision, a prison release authority is not permitted to circumvent its constitutional obligations merely because it has some discretion: "[I]t is by no means clear that the [relevant authorities] may, under the rules, invoke any criterion [they] choose[]." Winsett, 617 F.2d at 1006. Rather, the Board has, by both its past decisions to grant parole for commuted sentences, and by its formal declarations of policy, expounded discernible parameters that govern its discretion. See id. Here, the changes in parole policy can be shown to have been applied to Thomas's parole application, even though the Board possessed some discretion both before and after the 1996 policy change.

Although we are unable to express precisely what moved the Board to deny Thomas's petition, there is significant evidence that it acted upon policies that were established after Thomas's crime and conviction. Although discretion inheres within the Board's parole authority, and new Board members may carry new ideas regarding the exercise of that discretion, and old Board members may change their mind in the light of new considerations, Thomas is nevertheless entitled to have the Board give genuine consideration and due regard to the factors prescribed by the Board's pre-1996 policies. We agree that the Board is entitled to learn from past experiences and mistakes. This is so, just as a legislature might determine sentences for some crimes are too light and order judges to weigh certain factors more heavily in rendering a sentence. That a Board or legislature may learn from experience does not mean that those who were sentenced at an earlier juncture may now be more severely re-sentenced in the light of newly-found wisdom. This is precisely what the Ex Post Facto clause prohibits. Under the Board's reasoning, a determination, founded on newly discovered experience, could, by virtue of the Board's exalted discretion, forever deny a prisoner's preexisting right to parole consideration.

Although some discretion might still exist within the pre-1996 parameters, a parole decision that fails to address any of the criteria mandated by Board policy, such as institutional recommendations, willingness to undergo counseling and educational achievement, and instead utterly ignores all factors counseling in favor of release, falls outside of the realm of the legitimate exercise of discretion under the pre-1996 policies. Inference instructs us that the Board inappropriately relied on policies implemented in 1996, rather than the parole policies in place at the time of Thomas's crime and conviction.

In its briefs to this court, the Board purports to have reasons apart from public safety for its Thomas decisions. Those reasons appear to be asserted primarily as a post hoc defense to the allegations made in these proceedings. The Board argues that Thomas's "instant assault offense" and "very high assaultive potential," both functions of the nature of the past crime, were not the primary bases on which the decision to deny parole was made. It claims that. upon reading Thomas's file, "it is easy to see why the Board found the [arguments against releasing Thomas] so persuasive." It asserts that numerous reasons were considered in support of its decision, including an unfavorable recommendation from the District Attorney, lack of sex offender therapy, and questionable psychological evaluations. However, these were not bona fide considerations in the decisions to deny parole. Our analysis shows that the primary basis for the parole denials was the risk of potential harm to public safety.

Under the Guidelines, Thomas was entitled to parole at his hearings in both 1997 and 1998. He is the only prisoner out of 266 commuted sentences who was not granted parole in his first or second application. The voting members of the Department of Corrections staff unanimously recommended Thomas for release at each application. The pre-1996 policies place significant weight on factors relating to an inmate's potential to adapt to life on the outside, and on the recommendations of the institutional staff. The pre-1996 policies suggest that no single factor should be controlling in a decision to deny parole to an applicant. Moreover, the pre-1996 Decision Making Guidelines were given significant, although not dispositive weight. A departure from the Guidelines required a recitation of unique factors, outweighing those in the Guidelines analysis. The Board Decisions on each of Thomas's parole hearings rely heavily on "high assaultive behavior potential," which relates primarily to the nature of the original offense, despite many other significant factors favoring parole.

Thus, reviewing the pre-1996 documents pertaining to parole, it becomes evident that, although the risk of potential danger to the public has always been a factor, it became the controlling feature of the Board's decision after 1996. The Board defaulted in its duty to consider factors other than the underlying offense and risk to public safety; it has failed to address any of the factors favoring release.

In 1997, soon after the Department of Corrections denied Thomas pre-release, the Board denied Thomas parole on the very ground that he had not undergone a pre-release phase. The statutory pre-release requirement was adopted after Thomas's sentence was commuted. The Board later conceded in a state court suit initiated by Thomas that the law concerning the pre-release requirement should not have been applied to him. The Board's actions in this respect conveniently disregarded the Ex Post Facto clause to support its decision to deny parole.

We have carefully analyzed the Board's reports of disposition of Thomas's parole applications. The Board's 1997 and 1998 Decisions denied Thomas parole on the basis of several summary factors, including the severity of his underlying offense, his potential for future assaults, a prior assault offense, adverse recommendation from the District Attorney,15 and Thomas's need for counseling and treatment. The Board also set forth suggestions, presumably to improve Thomas's next effort for parole, including participation in prescriptive programming, good prison conduct, sex offender therapy, and positive psychological evaluations. All of these appear to have been met prior to the 1997 and 1998 decisions: All voting officials from the Department of Corrections recommended parole in both 1997 and 1998; he had complied with all prescriptive programming; a 1993 psychological and a psychiatric evaluation, made in anticipation of his commutation hearing, recommended release. The 1998 Department of Corrections Vote Sheet showed "No Psychological Contraindications" to release; he had participated in sex offender therapy; he had job training; and he had a post-release support network in place. Rather than explain in what manner its recommendations had not been met, or what additional steps needed to be taken, or whether some insurmountable barrier existed to Thomas's parole, the Board essentially reiterated the same recommendations for improving Thomas's parole candidacy in each subsequent Decision.

Given its indifference to Thomas's efforts to improve his parole candidacy, and its repeated reliance on Thomas's "instant offense" and his potential for future "assaultive behavior," despite the Guidelines' finding that Thomas was not a recidivism risk, the Board appeared to rely exclusively on the nature of the underlying offense and the potential danger to the public if Thomas were released. However, the Board, in its briefs to this court, suggested it had other reasons than public safety, and submitted that Thomas had only participated in "denier" sex offender therapy, rather than "admitter" therapy; that is, he was engaged in a form of therapy for offenders who refused to admit their crimes. This concern did not appear in the formal Board Decisions and, therefore, must be disregarded.

The Board's own internal notes shed light on its deliberations in this regard. We recognize that the Board is not required to share its specific reasons for denying parole. The Board's internal files in 1997 and 2000, however, merely noted, in a neutral way, that Thomas participated only in denier therapy and denied guilt for his crime without further comment or discussion of how this factor may have outweighed others favoring release. Significantly, the Board in 1997 and 1998 failed to mention lack of responsibility (or any other factor) in the section of the Guidelines worksheet where specific space is allotted to provide unique reasons for departing from a Guidelines recommendation. Instead, we have only the terse Board Decision and the Board's handwritten notes from which to glean its rationale for the parole denial.

In contrast to the scrawled notation of Thomas's lack of admitter therapy in its 1997 and 2000 worksheets, the Board underscores this point now in its briefs to us. Similarly, it discusses in its briefs how the benefits of the inmate sex offender therapy program are not fully realized, unless the inmate admits guilt for his or her crimes. However, the original, official deliberations showed that the Board failed to consider these matters at the times it reviewed Thomas's applications.

Moreover, the recommendation that Thomas receive sex offender therapy, which appeared on the 1997 report, did not appear on the 1998 Decision or worksheet. Then, inexplicably, the recommendation for sex offender therapy reappeared on Thomas's 2000 parole-refusal report. This casts still more doubt on the genuineness of the concern. It is also not clear that the Board's renewed concern over Thomas's "denier" therapy was ever properly communicated to Thomas, given that the reasons for denial in the Board Decision are vague and boilerplate. They nowhere mention the admitter-denier issue.

According to its briefs, the Board, in its 1998 and 2000 Decisions, may have relied on a 1996 psychological report that showed "evidence of possible sexual preoccupation and psychosexual immaturity," as well as an "antisocial personality." However, there is no evidence that the report recommended against release, and the Board never weighed explicitly the report against the balance of all the other favorable recommendations for release by counselors and corrections staff. No reference was made to two 1993 pre-commutation reports by a psychologist and a psychiatrist, respectively, both strongly recommending commutation.

Furthermore, a 1999 psychological report, although acknowledging the negative factors cited in the 1996 report, never expressly recommended against release. On the contrary, the report made suggestions as to how to structure Thomas's parole once granted. Meanwhile, a prison psychologist, on the 1998 Department of Corrections Vote Sheet, recommended release and noted that there were "no psychological contraindications" against release. The Board's own worksheet in 1998 merely noted that psychological evaluations showed Thomas to be an "average risk candidate."

In addition, many of the factors listed in the 1997 and 1998 Decisions were automatic designations. For example, "very high assaultive behavior potential" is assigned to a parole applicant whenever an applicant is convicted of a sexual offense. There is no indication whatever that the Board seriously contemplated the gravity of the public safety threat; nor is there any evidence that the Board followed its own procedures by pointing to factors independent of the Guidelines that counseled against granting parole.

In 2000, the Board again denied Thomas's parole, this time because "the mandates to protect the safety of the public and to assist in the fair administration of justice cannot be achieved through your release on parole." Although, again, it suggested sex offender therapy, favorable recommendations from Department of Corrections officials, prescriptive programming and continued good conduct, the decisive element of the Board's decision protecting "the safety of the public."

In 1997 and 1998 the Guidelines protocol resulted in a conclusion that Thomas should be released, but he was not. In 2000, the protocol recommended against parole. The data entered into the chart for past substance abuse changed, and thus he was then classified as a habitual offender with history of past abuse. The record shows some alcohol abuse by Thomas, but no drug use. Moreover, this reclassification on the Guidelines worksheet increased his overall objective score and placed him in the range of cases where the Guidelines recommended against parole. There is no evidence that alcohol abuse should, suddenly, as of the 2000 report, be given such significance: The Guidelines recommendation in 2000 is not worthy of consideration because it appears to have been deliberately designed to achieve a non-parole decision.16

Most forcefully, the 2000 decision report highlighted the Board's new rationale for denying parole, a rationale which implicitly pervaded all of the Board Decisions on Thomas's application: it bluntly stated that Thomas's release interfered with its mission "to protect the safety of the public." Although public safety had been a part of the Board's pre-1996 criteria, it had never been an exclusive, or even the most important, criterion. However, the Board does not attempt to offer any other explanation for its 2000 decision, while, in 1997 and 1998 it summarily recapitulated: "assaultive offense," potential for "assaultive behavior," and "victim injury" as reasons for its parole determinations. The reliance on these factors, and its failure to credibly consider any other factors, leads us to the ineluctable conclusion that the Board relied almost exclusively in 1997, 1998 and 2000 on the nature of the past offense and the potential danger to public safety.

The Board protests that the "assaultive potential" designation does not "require an automatic parole refusal." On the record before us, however, we do not agree. The Guidelines did not show him to be a recidivism risk. The Board did not consider seriously psychological contraindications or any other non-Guidelines factor that might have militated against parole. The Board's denial of Thomas's parole, despite its claims that the decision was the result of the discretion vested in it by the pre-1996 policies, exceeded any reasonable interpretation of the applicable policies. It appears that the Board was applying the new policy. Now, belatedly, in its briefs, the Board seeks to eviscerate the grounds for its decisions with a gloss of compliance with the pre-1996 policies. This will not do.

Since oral argument in this matter, the Board has called to our attention the recent Pennsylvania Supreme Court decision, Winklespecht v. Pennsylvania Board of Probation and Parole, 813 A.2d 688 (Pa.2002). The Board cites this case in support of the proposition that § 331.1's concern with "protect[ing] the safety of the public," added "nothing new to the parole process and[has] always been [an] underlying concern[]." Id. The Pennsylvania Supreme Court held that § 331.1 does not change Pennsylvania policy as to the criteria for parole "[n]or did the addition of this [new] language create a new offense or increase the penalty for an existing offense." Focusing on the added language to § 331.1 concerning "protect[ing] the safety of the public" and "assist[ing] in the fair administration of justice," the court concluded that these concepts have always been underlying concerns.

This decision, made after the Board's actions on Thomas's parole, came too late to alter the Board's view of the statutory amendment on the outcome of this case. Not having the benefit of the Supreme Court decision, the evidence before us shows that the Board interpreted § 331.1 to mandate foremost the consideration of public safety. The Board mistakenly construed the 1996 statutory change to signify a substantive change in its parole function. See Gall v. Parker, 231 F.3d 265, 304 (6th Cir.2000). As we noted previously, a public statement of the Board chairman and Board policy declaration confirm this substantive change in Board policy. The Pennsylvania Commonwealth Court too understood the 1996 amendment to enact a substantive change in Board policy. See Stewart, 714 A.2d at 508; Myers, 712 A.2d at 799. The Board's actions and policy pronouncements demonstrate a marked added weight on public safety concerns, uninfluenced by the subsequent Court interpretation of the statute.

E. Adverse Impact of Retrospective Policy on Thomas

As to the second Ex Post Facto criterion, that the change must adversely affect the offender, the Board argues that Thomas, having been sentenced to life, "had no legitimate expectation of ever being paroled." It notes that, during the 1970s, only 10% of life sentences were commuted and paroled. That figure diminished to less than half a percent in the 1990s. The Governor's power to grant commutation was in his absolute discretion, and thus, according to the Board, Thomas's eligibility for parole was entirely speculative. The Board does not dispute that the possibility of parole at sentencing based on some explicit criteria gave rise to a liberty interest. Hence, the procedures for reviewing parole applications must be constitutionally sound.

Garner held that the Ex Post Facto clause prohibited the application of post-conviction laws to prisoners that would result in a significant increase in the chances of prolonged incarceration. 529 U.S. at 251, 120 S. Ct. 1362. Prisoners are entitled to know the range of punishments available at the time of sentencing, and during the adjudication of their case, so that they can plea bargain and strategize effectively: The Ex Post Facto clause "(1)... prevents legislatures from interfering with the executive and judicial roles of prosecution and punishment; and (2) it assures that legislative acts give fair warning of what actions will be punished and the degree to which they will be punished." Coady, 251 F.3d at 487-88. Therefore, an offender, prior to his conviction and sentencing, is entitled to know not only his maximum possible punishment, but also his or her chances of receiving early release, since this too is a relevant factor in the plea bargaining calculus. An adverse change in one's prospects for release disadvantages a prisoner just as surely as an upward change in the minimum duration of sentence.

The possibility of commutation existed at the time of Thomas's conviction and sentence. The relevant criterion for determining the applicability of ex post facto analysis is the effect of new policies on "eligibility for reduced imprisonment," rather than any fixed guarantee of release. Lynce v. Mathis, 519 U.S. 433, 445, 117 S. Ct. 891, 137 L. Ed. 2d 63 (1997) (emphasis added). Therefore, a sentence that contained the right to parole consideration would give rise to a constitutional expectation that the parole guidelines extant at the time of the crime would be applied. See Garner, 529 U.S. at 250, 120 S. Ct. 1362. Eligibility for a commutation of a life sentence entails the possibility of parole, albeit a more distant possibility than for sentences that carry the possibility of parole ab initio. It also gives rise to the expectation that the parole criteria in effect at the time of the crime will be applied.

The Board contends that there was never a "significant" possibility, given the unlikelihood of commutation, that Thomas would ever be paroled. Garner, 529 U.S. at 256, 120 S. Ct. 1362. Indeed, as the Board contends, in most cases of life sentences in Pennsylvania, parole will never be an option as commutations are quite rare. However, as unlikely as these initial prospects for parole might have been, the application of the new parole policies in Thomas's case rendered them even more remote. The new policy "substantially increased the period of incarceration;" it reduced the possibilities of ever obtaining release.

The Board's reliance on California Dept. of Corrections v. Morales, 514 U.S. 499, 508-09, 115 S. Ct. 1597, 131 L. Ed. 2d 588 (1995) is misplaced. Although the parole policy change in Morales wrought a small change in the average duration of a prison sentence, the change was held to be too minuscule to rise to a constitutional violation. Morales considered the effect of a procedural change in parole law, which provided for a greater wait-period between first and second hearings. Id. at 507, 115 S. Ct. 1597. Here the substantive criteria for parole release have changed.

Lynce v. Mathis set forth a key distinction between the Morales case and this petition. In Lynce, a law in effect at the time of conviction provided that if a prison population reached approximately 98% of its capacity, good conduct credits needed for early release could be acquired at an accelerated pace. The law was changed during the prisoner's sentence, and his accelerated credits, earned during a time of over-98% prison capacity, were cancelled and parole denied. 519 U.S. at 438-39, 117 S. Ct. 891.

The Lynce prison officials argued that, at the time of conviction, it was entirely speculative whether the prison would become overcrowded during the petitioner's incarceration, and thus he was excluded from ex post facto protection under the holding in Morales. However, Lynce distinguished Morales, because there was no evidence in Morales that the change affected the petitioner's own sentence detrimentally. Lynce, 519 U.S. at 447, 117 S. Ct. 891. In Lynce, the population did exceed 98% during petitioner's incarceration, and by his own conduct the prisoner achieved enough credits for good behavior, so that he became eligible for release under the old rules. The change in policy had the effect of increasing the punishment in his individual case and thus violated ex post facto. Id.

Thus, under Lynce's reasoning, the parole change substantially impacted Thomas in violation of the Ex Post Facto clause. Moreover, Thomas is entitled to the benefits of his good behavior in prison; the opportunity to reduce his sentence through commutation, no matter how speculative, existed at the time of Thomas's crime. Thomas successfully attained a commutation of his sentence; he was entitled to corresponding reduction in sentence. We, therefore, hold that to retroactively apply changes in the parole laws made after conviction for a life sentence in Pennsylvania that adversely affect the release of prisoners whose sentences have been commuted, violates the Ex Post Facto clause.

Thomas argues that the Board's handling of his application without any real consideration of its merits, in violation of the Board's own procedures, offends due process, and that this court should itself order him released. Although the Board has not given due consideration to the relevant factors, this can be explained by its misguided reliance on the post-1996 criteria. Moreover, we are exceedingly reluctant to usurp the Board's functions and, except in our review capacity, substitute our own judgment for that of the parole Board. Although there were flaws and oversights in the Board's consideration of Thomas's applications, we are not entirely convinced that the Board is unable to give Thomas a fair hearing in light of the important considerations we have set forth in this opinion.

Ordinarily, the Board's decision to parole or deny parole to a prisoner is based on the consideration of many factors, with no one factor being dispositive. We expect that, on remand, the Board will not be defensive, but instead will fairly consider Thomas's application in the light of our observations and Ex Post Facto prohibitions. If the Guidelines recommend release, the Board should fairly consider the weight of this recommendation. A decision contrary to a Guidelines recommendation must be buttressed by unique factors which outweigh the Guidelines endorsement. Moreover, release on parole is a Board policy presumption, and parole should be granted unless countervailing negative factors affirmatively outweigh reasons supporting release.

In conclusion, the Order of the District Court is hereby affirmed, with directions to remand the matter to the Board for further proceedings consistent with this opinion, including a new hearing for Thomas and the Board's written decision thereon within 45 days after the mandate of this court.