Opinion by Judge KLEINFELD; Concurrence by Judge BERZON.

We address three issues: 1) whether a certificate of appealability is needed to appeal a district court’s order denying a writ of habeas corpus arising out of a state’s denial of parole; 2) whether federal constitutional law imposes on the states a requirement for some quantum of evidence to support a state’s denial of parole; and 3) whether, even if there is no general federal quantum of evidence requirement, applicants for parole in California, under the state’s current laws, may obtain federal habeas review of whether there is “some evidence” supporting a negative parole decision.

FACTS

In 1978, Hayward’s girlfriend was out shooting pool on a “girls’ night out,” while he stayed home. While she was out, a man acted abusively toward her. There are varying accounts. It is not clear whether the man slapped Hayward’s girlfriend, or she spat in his face, or he spat in hers, or whether, as Hayward once claimed, the man chased her out into the parking lot, tore off some of her clothes, and tried to rape her. Subsequently, Hayward spent months keeping his eyes open for the man so he could exact revenge.

Several months later, Hayward got a call saying that the man who abused his girlfriend, Tom Strauss (also known as Tom O’Connor), was at the Buccaneer Bar. Hayward enlisted some of his fellow gang members to go with him to the bar because he wanted to “kick his ass.” One of Hayward’s gang knocked Strauss down. Then Hayward stabbed Strauss twelve times in the back, killing him. In 1980, Hayward was sentenced to fifteen years to life for murder.

*550Since completing the first fifteen years of his sentence, Hayward has repeatedly been denied parole. He now petitions for a writ of habeas corpus, claiming that he is constitutionally entitled to be paroled.

Hayward phrases his petition for a writ of habeas corpus as a challenge to then-Governor Gray Davis’s 2003 decision to overturn a grant of parole by the California Board of Prison Terms.2 That decision was not the end of the state proceedings, but is worth summarizing. The Board of Prison Terms had found Hayward was suitable for parole, but wrote, in accord with California law: “[ijnmate not to be released until Governor exercises review authority.”3 The Governor, exercising his discretionary review authority under California law, denied parole and explained his disagreement with the Board of Prison Terms.

After weighing a multitude of discretionary factors, Governor Davis concluded that “Hayward would pose an unreasonable risk to public safety if released at this time.” One factor was Hayward’s “particularly heinous crime.” Hayward stalked his victim for months. When he located the victim he arranged for members of his motorcycle gang to join him. They set out to subdue the victim, who was drunk and recuperating from two broken arms. Hayward stabbed Strauss in the back twelve times, twice to the hilt. After the stabbing, Hayward fled while his victim bled to death. Two witnesses later said that they and their families received death threats intended to keep them from testifying.

Another factor was Governor Davis’s concern about the sincerity of Hayward’s remorse. For his first fifteen years in prison, Hayward denied responsibility and disparaged the victim, saying “[Strauss’s] family is lucky he’s dead.” Even after Hayward finally admitted to his crime in 1993, he told a psychological evaluator that he felt good about killing Strauss.

The Governor was also concerned by Hayward’s substance abuse and his need for further substance abuse therapy. Hayward began using heroin when he was twelve. He subsequently used LSD, POP, methedrine (methamphetamine), cocaine, marijuana, and excessive alcohol. Prison did not stop Hayward’s criminal drug use. He was disciplined for marijuana possession and admitted that he “ran drugs in prison.”

Before this prison stretch for murder, Hayward was active in a gang (the same gang that helped him murder Strauss). Hayward claims he “retired” from his motorcycle gang while in prison. As with the drugs, though, prison did not end Hayward’s involvement with gangs. The Governor noted that Hayward “received a serious disciplinary report for leading a white racist organization, using the organization to intimidate inmates, directing assaults and advocating violence against black inmates.” Hayward continued his white-racist gang involvement until mid-1989.

*551The Governor also considered Hayward’s extensive criminal history in addition to this murder. The murder was not an aberration. As a juvenile, Hayward was arrested approximately twenty times, starting at age eight. He had about sixteen arrests as an adult. During all the time he was out of prison, he never quit committing serious crimes. Hayward admitted involvement with a criminal group “responsible for [75] to [120] very serious crimes including arson, assault, kidnapping, robbery and possession of a large cache of stolen explosives.”

The Governor took Hayward’s moderately favorable mental health evaluation into account, but weighed it against other factors. A psychological evaluation of Hayward’s mental health found that “historical factors” were on the “negative side.” But, “[o]n the positive side is the lack of overt violence during the last twenty years of incarceration, his being disciplinary free for the last thirteen years, an increased level of maturity and insight, his participation in substance abuse recovery, his participation in self help and spiritual activities, and his being older and mature.” The psychologist’s conclusion was that Hayward posed “a low to moderate risk for future violence in the community.” The Governor, however, thought that the risks Hayward posed “remained] too high to risk releasing him into our community” because of Hayward’s “long criminal history, increasing violence, and gang participation.”

The Governor’s decision was not the end of the case in California. In the California system, if the Board of Prison Terms recommends that a prisoner be paroled, the Governor reviews the recommendation and makes his own decision. That decision is subject to judicial review via a prisoner’s state habeas petition.4 Under California law, the state courts review the Governor’s decision and the record for “‘some evidence’ that an inmate poses a current threat to public safety.”5

The Superior Court of California, County of Los Angeles, reviewing the record, determined that Hayward claimed he went to the bar to beat up the victim, rather than to kill him. The Superior Court noted that, prior to his conviction for killing Strauss, Hayward was acquitted of attempted murder, but his prior offenses ranged from armed robbery (as a juvenile) to assault with a deadly weapon on a police officer. The Superior Court also found that Hayward “received four 115 disciplinary violations for ‘serious’ misconduct, the most recent in 1989.” It then considered Hayward’s expressed remorse, vocational training, and plans for parole. The Superior Court held that although some of the Governor’s findings were not adequately supported by the record, others were, and that was sufficient to justify denial of parole.

Hayward petitioned for a writ of habeas corpus in the California Supreme Court,6 which summarily denied his request. He then petitioned for a writ of habeas corpus in federal district court. The district court denied the petition and we now affirm.

*552ANALYSIS

Both we and the California Supreme Court have been engaged in modification of the law to determine what limits there are on denial of parole. We have two decisions, Irons v. Carey7 and the panel decision in this case,8 saying that due process requires “some evidence,” not merely a discretionary judgment, if parole is denied. We address these propositions below. After oral argument in this case,9 two California Supreme Court cases, In re Lawrence10 and In re Shaputis,11 held that under California law, some evidence of future dangerousness is a necessary predicate for denial of parole. Hayward’s appeal addresses what if anything the federal Constitution requires as a condition of denial of parole.12

I. Certificate of appealability.

This case is an appeal from a district court order denying a writ of habeas corpus to a state prisoner who seeks habeas relief after California denied him parole. Our jurisdiction to review the denial arises under 28 U.S.C. § 2253(a).13 However, a petitioner may not appeal a “final order in a habeas corpus proceeding in which the detention complained of arises out of process issued by a State court,”14 unless a “circuit justice or judge issues a certificate of appealability.”15 We will only issue a certificate of appealability “if the applicant has made a substantial showing of the denial of a constitutional right.”16

Hayward did not request or obtain a certificate of appealability from the district court or, initially, from this court. He argues that a certificate is unnecessary on the theory that his detention “arises out of’ an administrative determination, the denial of his parole by the Board of Prison Terms, not a judicial determination “issued by a State court.” 17 He has support for his theory in our decision in White v. Lambert.18 We held in White that a certificate of appealability is not required “when a state prisoner challenges an administrative decision regarding the execution of his sentence,” as opposed to the fact of his conviction.19 Our sister circuits are divided on whether a certificate is needed to appeal denial of habeas relief in such circumstances.20

*553The standard for a certifícate of appealability is lenient.21 Hayward need only “ ‘sho[w] that reasonable jurists could debate’ ” the district court’s resolution or that the issues are “ ‘adequate to deserve encouragement to proceed further.’ ”22 This showing requires “something more than the absence of frivolity,”23 but something less than a merits determination (which we lack jurisdiction to make, absent a certificate of appealability).24 The requirement of a certificate of appealability serves as a threshold requirement, a sine qua non to screen out prisoner petitions that ought not to take up additional judicial resources beyond those already consumed before state courts, federal magistrate judges, and federal district judges.25

Hayward was justified, because of our decisions in White26 and Rosas v. Nielsen,27 in proceeding without seeking a certificate. The statute requires a certificate of appealability if “the detention complained of arises out of process issued by a State court.”28 We reasoned in those cases that a prisoner challenging denial of parole is challenging confinement resulting from an administrative decision, not confinement “arising out of process issued by a State court,” so the statute does not require the certificate. Our sister circuits have generally rejected our reasoning.29

The stronger literal argument supports requiring a certificate. The statutory requirement of a certificate of appealability turns on “the detention complained of.” On the one hand, what keeps the prisoner in prison is denial of parole. On the other hand, what put him there was a state court judgment convicting and sentencing him. Several of our sister circuits reason that the “detention” referred to by the statute is the state court decision that put the would-be parolee in prison, not the administrative decision not to let him out.30 The California Board of Prison Terms does not have the authority to detain prisoners, just to release them from detention (subject to gubernatorial approval), so its decision cannot be the “detention complained of.”31 Even if one were to treat the power to release as the power to detain, the administrative decision by the parole board or by the governor is not the last word in the *554California system, and neither can deny release if the Superior and Supreme Court of California overturn their .decisions on review. Thus even if we were to look at what kept the prisoner confined as opposed to what caused him to be confined in the first place, a court decision is the target of this habeas petition in federal court.

If we look to function, the better construction of the certificate of appealability statute likewise requires a certificate of appealability. We are unable to think of a purpose Congress might have had for a contrary reading. What the requirement of,a certificate of appealability does, and all it does, is screen out of. the federal appellate courts claims that are not even debatable among reasonable judges, which is to say, frivolous claims. We cannot think of a reason why Congress might want to confer on-us jurisdiction to hear frivolous prisoner challenges to administrative decisions but screen out frivolous prisoner challenges to court decisions. Neither could the District of Columbia Circuit: “We do not think the Congress intended to limit federal review of a sentence imposed by a state court while allowing an unfettered appeal from a parole decision declining to decrease the time served thereunder.”32

It is also considerably more practical for the petitioners themselves to have a clear, simple, across-the-board requirement. Habeas jurisdiction, more than most, is a maze of curlicues and cul-de-sacs, with nowhere near enough straight, clearly marked roads. Usable law needs to be clear enough so that people trying to do the right thing can. Prisoners most often petition pro se, so they especially need clear road signs. Complicating jurisdiction rules for habeas appeals merely increases the randomness and decreases the justice of dispositions. The more distinctions and exceptions we build into the already overly complicated subject of habeas jurisdiction, the more the outcomes result from chance rather than the merits. The statute is best read to mean that a state prisoner seeking to appeal denial of a petition for a writ of habeas corpus to a federal court of appeals must get a certificate of appealability.

This simple rule would be troubling if it screened out cases that ought to get review, but the standard and procedure for certificates of appealability protects against that. All a prisoner needs is an issue debatable by reasonable jurists.33 And if the prisoner neglects to request a certificate of appealability before going forward, the court of appeals can grant him one sua sponte.34

We therefore overrule those portions of White and Rosas which relieve a prisoner from obtaining a certificate of appealability from administrative decisions such as denial of parole and prison transfer. A certificate of appealability is necessary to confer jurisdiction on this court in an appeal from a district court’s denial of habeas relief in a § 2254 case, regardless of whether the state decision to deny release from confinement is administrative or judicial. Hayward needs a certificate of appealability if we are to maintain jurisdiction over this case. We may issue such a certificate sua sponte>35 Hayward -had followed our *555prior decisions when he proceeded without a certificate, and he has “made a substantial showing of the denial of a constitutional right”36 in the sense that his claim is debatable among reasonable jurists, so we hereby certify for appeal the issue of whether “some evidence” was a constitutional sine qua non for the state’s denial of parole.

II. Constitutionality of California’s denial of Hayward’s parole.

A. No federal “some evidence” rule in the air.

Our decision today arises from Irons v. Carey,37 and our earlier decision in this case.38 These decisions may be read to mean that a parole-eligible prisoner has a constitutional right to be released unless there is “some evidence” of future dangerousness. If there is any right to release on parole, or to release in the absence of some evidence of future dangerousness, it has to arise from substantive state law creating a right to release. To the extent our prior decisions including Biggs v. Terhune,39 Sass v. California Board of Prison Terms,40 Irons v. Carey41 and our panel decision in this case42 might be read to imply that there is a federal constitutional right regardless of whether state law entitles the prisoner to release, we reject that reading and overrule those decisions to the extent they may be read to mean that.43

B. Good time.

The proposition that the Supreme Court has required “some evidence” of anything derives from a misunderstanding of the differences between “good time” and parole44 We speak now to the commonalities of good time and parole, and address the particularities of California law in the next section of this opinion.

Prisoners usually get released before their sentences are over. In many jurisdictions, there are two paths to early release, good time and parole. They differ.

Good time is a prison discipline device. Prisoners get a certain number of days of good time for each month or year of their terms, and lose days if they misbehave in prison. The misbehavior does not have to be criminal, and the credit does not depend on predictions of good behavior outside prison. For example, federal prisoners serving terms of more than a year get up to fifty-four days of good time per year, credited at the end of the year, for “exemplary compliance with institutional disciplinary regulations.”45 Accordingly, a federal prisoner sentenced to ten years imprisonment, who obeys the rules in pris*556on, will be released after serving nine years and four months.

States likewise typically assign mandatory good time to all prisoners, which they can lose in increments for discipline violations in prison. Good time is designed to give prisoners an incentive to obey prison rules. The right to good time and the losing of it generally depend on how the prisoner behaves in prison, not what he did to be sent there or how authorities think he will behave after he gets out.

Because it is ordinarily a fixed, specific entitlement lost on the basis of misconduct, good time is a right to liberty, that is, release from prison, that can be taken from the prisoner only with due process of law. The “some evidence” standard is the quantum of due process that a prisoner charged with a discipline violation that would cost him good time is constitutionally entitled to.46 If his misconduct cannot be proved by at least some evidence, he is entitled to his good time. This standard gives the prisoner some protection against the risk of mistake, losing his good time for misconduct he did not commit.

Many states use good time more or less like the federal system.47 For example, Alaska state prisoners, with some exceptions, are “entitled to a deduction of one-third of the term of imprisonment rounded off to the nearest day if the prisoner follows the rules of the correctional facility.”48 The California scheme is considerably more complex than the federal or typical state system and has changed frequently, but generally prisoners with determinate sentences, “shall be credited with a one-fourth reduction on their term of imprisonment, unless all or part of such good behavior credit is denied or forfeited as a result of disciplinary action in the amounts listed in section 3323.”49 The California statute has a schedule of how much good time can be lost for different offenses, for example, up to 360 days for murder or rape, 180 days for other prison misconduct that could be prosecuted as a felony, 90 days for misdemeanor misconduct, and 30 days for a “serious disciplinary infraction” as defined by regulation.50

Though the details vary from state to state, good time statutes and regulations have several things in common. First, good time is a right, not a discretionary award.51 This “liberty interest,” as our cases call it, is not a procedural right to be considered for a discretionary benefit. Rather, this is a liberty interest of the most fundamental sort, the prisoner’s right to walk out the prison gate and hear it *557clang behind him.52 Second, good time is lost after discipline proceedings for violations of prison rules committed while behind bars.53 Third, the number of days is arithmetically calculable, so a prisoner knows when he enters prison that if he complies with the rules, he will be released a certain number of days before the end of the term to which he was sentenced.54 Fourth, any loss of good time is historical, not predictive. Good time is taken away because of something the prisoner has already done, as adjudicated in a discipline proceeding, not for something he may do in the future. Prisons use the entitlement to good time, and reductions in good time, to give the felons in their charge an incentive to behave themselves while in prison, to avoid a situation where a prisoner might think he has nothing to lose.55

C. Parole.

Parole, by contrast with good time, typically involves “purely subjective appraisals” that turn on a “discretionary assessment of a multiplicity of imponderables.”56 Parole, unlike good time, has been abolished in the federal system.57 California and many other states still use it.58 In Alaska, for example, an eligible prisoner may be granted “discretionary parole”59 after serving one-third of the term of imprisonment to which he was sentenced.60 Parole is at the discretion of a board that considers numerous factors, such as whether there is a “reasonable probability” that the prisoner, once released, will not violate any laws or parole conditions, “pose a threat of harm to the public” and “release of the prisoner would not diminish the seriousness of the crime.”61 Thus in *558the not a typical Alaska system, a prisoner may get out in one third of the time he is sentenced to, if he conforms to prison rules during that time and appears likely to continue to behave himself after release.62

Parole statutes differ from good time statutes first, in that they are at least in part prospective and predictive. Instead of or in addition to looking back to a fact, whether the prisoner broke a law or rule, they look forward, based in part on a judgment about how the inmate is likely to behave if released. Second, they depend on highly subjective and discretionary judgments, because the future is not yet a fact and can never be proved like a historical fact. Third, an inmate can control the outcome of good time credits by behaving himself in prison, but cannot control the outcome of the parole board hearing by his behavior (though it may influence their discretionary judgment). He can be turned down for parole at the parole board’s discretion, even when he would have an absolute entitlement to good time.

For example, suppose a prisoner is imprisoned for committing a very serious crime. The nature of the crime, his psychological and psychiatric assessments, and his own letters and statements to prison authorities, establish that he is highly likely to commit the same kind of crime or something worse if he gets out. In a state with good time credits, the felon would still get out before the end of his term if he complied fully with all prison rules. But because of the risk to public safety, he would almost certainly not be paroled. He would get out only as early as his good time entitled him to, and not any earlier despite the possibility of parole to a similarly situated prisoner with a better prognosis.

D. The constitutional distinction between good time and parole.

The relevant Supreme Court decisions require “some evidence” for denial of good time, but do not require it for denial of parole, and they carefully distinguish *559good time from parole.63 There is no general federal constitutional “some evidence” requirement for denial of parole, in the absence of state law creating an enforceable right to parole.

Hayward argues that the Supreme Court holding requiring “some evidence” to justify denial of parole is Superintendent v. Hill.64 That decision is not on point, because the holding in Hill addresses good time, not parole. Hill relied on Wolff v. McDonnell,65 the Supreme Court’s seminal “good time” revocation case. Wolff explains that good time “is qualitatively and quantitatively different from the revocation of parole or probation.”66 Wolff holds that revocation of good time requires, inter alia, “that there must be a ‘written statement by the factfinders as to the evidence relied on and reasons for the disciplinary action.’ ”67 Wolff explained that “the provision for a written record helps to insure that administrators, faced with possible scrutiny by ... perhaps even the courts, where fundamental constitutional rights may have been abridged, will act fairly.”68 Relying on Wolff, Hill holds that, to take away a prisoner’s good time, due process requires there to be “some evidence” of a disciplinary infraction.69 It does not hold that “some evidence” is required for denial of parole.70

The Supreme Court decisions addressing parole determinations71 suggest (though they do not explicitly hold) that the “some evidence” requirement does not apply to denial of parole. In fact, the court has expressly distinguished good time from parole.

Greenholtz v. Inmates of Nebraska Penal & Correctional Complex, addressing a claimed right to parole, holds that “[tjhere is no constitutional or inherent right of a convicted person to be conditionally released before the expiration of a valid sentence.”72 The Court distinguishes parole from parole revocation, because revocation is a “wholly retrospective factual question,” 73 but release on parole “depends on an amalgam of elements, some of which are factual but many of which are purely subjective appraisals.”74

*560Greenholtz emphasizes that parole is a discretionary,, predictive decision.75 Parole decisions are “ ‘equity’ type judgments] that cannot always be articulated in traditional findings” because “the choice involves a synthesis of record facts and personal observation filtered through the experience of the decisionmaker and leading to a predictive judgment as to what is best both for the individual inmate and for the community.”76 The Court rejects the due process argument for a “some evidence” standard: “nothing in the due process concepts as they have thus far evolved that requires the Parole Board to specify the particular ‘evidence’ in the inmate’s file or at his interview on which it rests the discretionary determination that an inmate is not ready for conditional release.”77 The state parole statute at issue in Greenholtz arguably created a “liberty interest” to “some evidence” because it said that the prisoner “shall” be paroled “unless” certain negative conditions applied. But the Court rejected the argument. “The Constitution does not require more” than an opportunity to be heard and a statement telling the prisoner why he was not paroled.78

Board, of Pardons v. Allen79 likewise addresses “shall ... unless” language in a parole statute. Though the language does indeed establish a “liberty interest” protected by the Due Process Clause,80 the Court nevertheless holds that the parole board retains discretion to make a subjective decision. Allen reaffirms that “the release decision is ... ’necessarily subjective ... and predictive,’ ” and “the discretion of the Board is ‘very broad.’ ”81 The prisoner’s interest in parole ripens into an entitlement only after the parole board has made the findings that under the statute entitle him to it, which is to say, perhaps tautologically, that a prisoner is entitled to parole only if the parole authority has made the discretionary decision that under the state standard he is entitled to parole.82

By way of contrast, good time is neither subjective nor predictive.83 Loss of good time is instead “a sanction for serious misconduct” in prison, applied after an administrative adjudication.84 Good time is not, as parole is, a “discretionary assessment of a multiplicity of imponderables, entailing primarily what a man is and what he may be rather than simply what he has *561done.”85 Good time entitlement depends on a straight-forward historical determination of what the prisoner has done.

In Wilkinson v. Austin, the Court, addressing prison transfers, distinguishes between parole and good time.86 Wilkinson likens prison transfer to parole and distinguishes the good time “some evidence” decisions, because, like parole and unlike good time, transfer is discretionary.87 Accordingly, a prison transfer decision requires only the “nonadversary procedures set forth in Greenholtz,” not the “more formal adversary — type procedures” set forth in Wolff.88

Thus, in the absence of state law establishing otherwise, there is no federal constitutional requirement that parole be granted in the absence of “some evidence” of future dangerousness or anything else.

III. The California scheme.

Although the due process clause does not, by itself, entitle a prisoner to parole in the absence of some evidence of future dangerousness, state law may supply a predicate for that conclusion. “[D]espite the necessarily subjective and predictive nature of the parole-release decision, state statutes may create liberty interests in parole release that are entitled to protection under the Due Process Clause.”89 This poses two questions for us, whether the California parole scheme creates such an interest, and, preliminarily, whether it is necessary to decide whether it does.

The California parole statute provides that the Board of Prison Terms “shall set a release date unless it determines that the gravity of the current convicted offense or offenses, or the timing and gravity of current or past convicted offense or offenses, is such that consideration of the public safety requires a more lengthy period of incarceration for this individual.”90 The crucial determinant of whether the prisoner gets parole in California is “consideration of the public safety.”91

In California, when a prisoner receives an indeterminate sentence of fifteen years to life, the “indeterminate sentence is in legal effect a sentence for the maximum term, subject only to the ameliorative power of the [parole authority] to set a lesser term.”92 Under the California parole scheme, the prisoner has a right to a parole hearing and various procedural guarantees and rights before, at, and after the hearing;93 a right to subsequent hearings at set intervals if the Board of Prison Terms turns him down for parole;94 and a right to a written explanation if the Governor exercises his authority to overturn the Board of Prison Terms’ recommendation *562for parole.95 Under California law, denial of parole must be supported by “some evidence,” but review of the Governor’s decision is “extremely deferential.”96

Subsequent to Hayward’s denial of parole, and subsequent to our oral argument in this case, the California Supreme Court established in two decisions, In re Lawrence97 and In re Shaputis,98 that as a matter of state law, “some evidence” of future dangerousness is indeed a state sine qua non for denial of parole in California. We delayed our decision in this ease so that we could study those decisions and the supplemental briefs by counsel addressing them. As a matter of California law, “the paramount consideration for both the Board [of Prison Terms] and the Governor under the governing statutes is whether the inmate currently poses a threat to public safety.”99 There must be “some evidence” of such a threat, and an aggravated offense “does not, in every case, provide evidence that the inmate is a current threat to public safety.” 100 The prisoner’s aggravated offense does not establish current dangerousness “unless the record also establishes that something in the prisoner’s pre- or post-incarceration history, or his or her current demeanor and mental state” supports the inference of dangerousness.101 Thus, in California, the offense of conviction may be considered, but the consideration must address the determining factor, “a current threat to public safety.”102

Because the California “some evidence” standard is exactly the same as the one Hayward urges as a federal constitutional standard, the doctrine of constitutional avoidance103 counsels not deciding whether the California parole scheme establishes a predicate for imposing it as a matter of federal constitutional law. State law already does what Hayward would have federal constitutional law do. We therefore do not decide whether a right arises in California under the United States Constitution to parole in the absence of some evidence of future dangerousness. Even if Hayward were correct that he had a federal constitutional right to “some evidence,” it would make no difference, since he has the right to parole in the absence of “some evidence” of future dangerousness under state law.

Since the “some evidence” requirement applies without regard to whether the United States Constitution requires it, we in this case, and courts in this circuit facing the same issue in the *563future, need only decide whether the California judicial decision approving the governor’s decision rejecting parole was an “unreasonable application”104 of the California “some evidence” requirement, or was “based on an unreasonable determination of the facts in light of the evidence.” 105 And that conclusion is one that could not be reached on this record, regardless of what level of “AEDPA deference” we applied. The California Superi- or Court concluded that the governor’s rejection of parole was based on “some evidence” of future dangerousness because of “the nature of the commitment offense” and “the somewhat unfavorable psychological and counselor reports.” The reports that impressed the Superior Court said that Hayward would pose a “low” to “moderate” risk of danger if released, as opposed to “no” or merely “low” risk. The court found that the record supported the governor’s judgment that the murder was premeditated and extremely vicious, and could not be attributed merely to “stress” about Hayward’s girlfriend as the parole board had suggested. These two factors combined, in the state court’s view, to establish some evidence of future dangerousness to the public from this murderer.

CONCLUSION

What we are left with is California law establishing a right that Hayward contends is a federal constitutional right, a right to parole in the absence of some evidence of his own future dangerousness to the public. There was some evidence of future dangerousness, so his parole was denied, and the district court correctly denied the writ. The right in California to parole in the absence of some evidence of one’s future dangerousness to the public arises from California law. We overrule any decisions suggesting that the federal constitution imposes a requirement of “some evidence” of future dangerousness without regard to state law.

We have purposely avoided constraining other states to conform to the California system. Other states may have different parole systems or, like the federal system, no parole at all. A state may make parole a matter of grace, or require consideration not only of the inmate’s future dangerousness if released, but also the need for deterrence of others. Or it may decide that the imprecision and unreliability of predictions of future dangerousness make that an inappropriate criterion for parole boards. Or it may decide that justice, respect for victims, and reaffirmation of societal norms,106 require denial of parole to inmates who have committed terrible crimes regardless of whether they remain dangerous. Or it may decide that it is worth freeing dangerous inmates or inmates who have committed especially heinous crimes to avoid paying medical expenses during their old age. A state may act on the view that age diminishes a prisoner’s inclination to harm others, or that it diminishes only his ability to run fast after he does. It may decide that post-conviction behavior affords a good basis for prediction, or that behavior in prison, where he is guarded and on television monitors 24 hours a day, does not predict post-release conduct as well as how the prisoner behaved the last time he was free. It may rely heavily on prison administrators, psychologists and counselors, or conclude that they cannot see into the prisoner’s soul and are likely to be misled by *564articulate prisoners who have learned what to say.

We do not intimate that any of these policy choices is constitutionally required or prohibited. Procedurally and substantively, the states have considerable room for play in the joints. The states have “flexibility in deciding what procedures are needed in the context of postconviction relief,”107 and state postconviction relief procedures may be upset by a federal court “only if they are fundamentally inadequate to vindicate the substantive rights provided.”108

Affirmed.

join, concurring in part and dissenting in part:

Hayward contends that the Governor of California denied him release on parole without “some evidence” of his future dangerousness and that doing so violated the Due Process Clause. In the substantive part of its opinion that actually pertains to the merits of this case, Part III, the majority declines to decide whether the federal Constitution independently establishes a “some evidence” standard as part of the minimum process due to California prisoners denied parole, because the state’s constitutional law already does so. The majority then reviews the evidence of Hayward’s future dangerousness and concludes that some evidence supported the denial of parole. I concur in Part III of the opinion and offer in Part III of this concurrence some additional bases for the conclusion reached, further explaining, in particular, why federal district courts and this court in the future may reach “some evidence” questions arising in California parole cases on federal habeas, as we do here.

Before it reaches the issues actually raised by Hayward’s petition, however, the majority offers a lengthy disquisition on state prisoner release systems in general, apparently to rebut a constitutional proposition upon which Hayward never relies- — • that all state parole systems must impart a protectable liberty interest because all “good time” statutes do so. See Maj. Op. at 555. But in intimating1 that the opposite is true- — -that parole systems do not uniformly create liberty interests — the majority commits the very error that it purports to correct, relying on broad generalizations about types of prisoner release systems instead of focusing on the particular state scheme at issue.

In fact, the Supreme Court’s precedents make obvious that the majority knocks down the most feeble of straw men when it disapproves a uniform federal rule regarding the. “some evidence” issue on parole, divorced from the particulars of each state’s statutory and decisional law. It is clear beyond cavil that the process due under the United States Constitution depends not on the label given to a particular early release program but on its characteristics under state law. The majority’s disembodied discussion of prisoner release systems in Part II is thus entirely superfluous.

More importantly, deciding constitutional issues in the air is simply not our job as judges. We are here to decide cases, not to issue advisory opinions on questions that may arise in another state at another time. Moreover, the majority’s lengthy *565discussion of the non-issue of a blanket federal “some evidence” requirement for parole decisions is replete with errors and misdirection. For these reasons, I cannot join Part II of the opinion. I also part ways with the majority on the jurisdictional question, to which I turn first.

I.

The majority concludes that 28 U.S.C. § 2253(c)(1)(A) obliges a state prisoner to obtain a Certificate of Appealability (COA) to obtain review of a denial of parole. Although the matter is not free from doubt, for the reasons stated in White v. Lambert, 370 F.3d 1002 (9th Cir. 2004), I would hold that “a COA is not required when a state prisoner challenges an administrative decision regarding the execution of his sentence.” Id. at 1010.

The majority ignores the significance of the statutory language of 28 U.S.C. § 2253(c)(1)(A) in the context of a challenge to prison administrative decisions. As we stated in White, “[h]ad Congress intended that every state prisoner obtain a COA before appealing, irrespective of the nature of the challenge, it easily could have said so.” Id. at 1012. Instead, the statutory language of § 2253(c)(1)(A) provides that a COA is necessary only when “the detention complained of arises out of process issued by a State court,” 28 U.S.C. § 2253(c)(1)(A) (emphasis added), indicating that the inquiry turns on the “target of the prisoner’s complaint,” not merely his status as a state prisoner. White, 370 F.3d at 1011. As we explained in White, had Congress intended the COA requirement to apply to all state prisoners because the relevant “detention” is “the state court decision that put the would-be parolee in prison,” see Maj. Op. at 553, it would have used the language it used in other provisions of the habeas statute — for example, in the exhaustion requirement, which applies to “a person in custody pursuant to the judgment of a State court,” 28 U.S.C. § 2254(b)(1). That language means precisely what the majority would hold the quite different language of § 2253(c)(1)(A) to mean. See White, 370 F.3d at 1011.

In the parole release context, the most plausible reading of § 2253(c)(1)(A) is that the “detention complained of’ is the petitioner’s continued imprisonment as a result of the parole board or Governor’s denial of parole, an administrative decision that does not “arise[ ] out of’ process issued by a state court. Accordingly, I would hold that because Hayward challenges “an administrative decision regarding the execution of his sentence,” White, 370 F.3d at 1010, he need not have obtained a COA to invoke this court’s jurisdiction.

I agree with the majority, of course, that if there is a COA requirement, a COA should issue in this case. In the end, then, the answer to the COA question does not matter to the outcome. Moreover, as the majority indicates, it is hard to conceive of a case in which requiring a COA would preclude review of a meritorious case. This court considers a notice of appeal as a request for a COA, and a COA issues unless the case is, in essence, frivolous.

Still, the COA requirement adds steps to the appeal of a potentially meritorious habeas case for the district court, the petitioner, and this court. We should not expand this requirement beyond what Congress intended. As that is what the majority has done, I dissent from its COA holding.

II.

The majority next engages in its abstract, superfluous discussion of state prisoner release systems, surveying a range of such systems — including, presumably be*566cause it is the home of the author of the opinion, Alaska’s — and distinguishing between the general categories of “good time” and “parole.”

This extended discussion is dicta of the most objectionable type. Hayward has never argued that the Due Process Clause requires some evidence of future dangerousness for every parole denial without regard to state law. Nor has this court ever so held. The best the majority can do in explaining its generic discussion is to indicate that some of our opinions might “imply” — not hold — that the federal Due Process Clause on its own imposes procedural requirements for all state parole decisions, no matter whether the underlying state law makes the decision purely discretionary or not. Maj. Op. at 555. But we don’t usually sit en banc to overrule implications. In any event, there are no implications in our case law of the kind the majority conjures up to overrule.

In fact, as I explain, it is beyond dispute that the scope of due process liberty interests is determined by the attributes of the particular system created by the particular state, not by whether the system is designated as “good time” or “parole.” The upshot is that, if it has any effect at all, Part II of the majority opinion will only sow confusion.

A.

The Supreme Court has established a two-part framework for analyzing claims that the state has deprived an individual of a protected liberty interest without due process, a framework the majority does not acknowledge. First, we are to ask whether an individual has a protected liberty or property interest with which the state has interfered. Second, if the state has interfered with such an interest, we are to ask whether the procedures attending that deprivation of liberty were constitutionally sufficient. See Ky. Dep’t of Corr. v. Thompson, 490 U.S. 454, 460, 109 S.Ct. 1904, 104 L.Ed.2d 506 (1989).

Critically, although an individual claiming a protected liberty interest must have a “legitimate claim of entitlement” to that interest, the interest typically does not arise from the federal Due Process Clause itself but from “ ‘the laws of the States.’ ” Id. at 460-61, 109 S.Ct. 1904 (quoting Hewitt v. Helms, 459 U.S. 460, 466, 103 S.Ct. 864, 74 L.Ed.2d 675 (1983)); see also Vitek v. Jones, 445 U.S. 480, 488, 100 S.Ct. 1254, 63 L.Ed.2d 552 (1980) (“[S]tate statutes may create liberty interests that are entitled to the procedural protections of the Due Process Clause of the Fourteenth Amendment.”). Hayward has always maintained that his liberty interest in parole arises from the specifics of California’s parole scheme, not — as the majority would have it — directly from some ephemeral, uniformly applicable federal law.

To determine whether the California parole scheme gives rise to a constitutionally protected liberty interest, we are guided by the Supreme Court’s elucidation in the parole context of generally applicable due process principles. “In the air,” as the majority states, Maj. Op. at 555, an inmate does not have a “constitutional or inherent right ... to be conditionally released before the expiration of a valid sentence.” Greenholtz v. Inmates of Neb. Penal & Corr. Complex, 442 U.S. 1, 7, 99 S.Ct. 2100, 60 L.Ed.2d 668 (1979); Maj. Op. at 559-60. It is equally clear, however, that when a state’s parole scheme “creates a presumption that parole release will be granted” when or unless statutorily designated findings are made, an inmate does have a constitutionally protected expectancy of release on parole. See Greenholtz, 442 U.S. at 12, 99 S.Ct. 2100; Bd. of *567Pardons v. Allen, 482 U.S. 369, 377-78, 107 S.Ct. 2415, 96 L.Ed.2d 303 (1987).

Contrary to the majority’s fundamentally misdirected analysis, then, the scope of federal constitutional liberty interests is determined by the attributes of the particular system created by the state, not by whether the system is designated as “good time” or “parole” or something else. As the Supreme Court has emphasized, each state’s parole statute “has [a] unique structure and language and thus whether [the] state statute provides a protectible entitlement must be decided on a case-by-case basis.” Greenhottz, 442 U.S. at 12, 99 S.Ct. 2100. It is therefore quite irrelevant how Alaska, for example, has structured its parole system; that the federal system has no “parole” program; or that there are differences between “good time” systems and “parole” systems generically defined. See Maj. Op. at 555-61. What would matter for purposes of Hayward’s claim that some evidence of future dangerousness is a component of the minimum process guaranteed by the Due Process Clause in this context is the nature and scope of the liberty interest provided by California’s parole system, considered in detail rather than superficially. See Greenhottz, 442 U.S. at 12, 99 S.Ct. 2100; Allen, 482 U.S. at 377-78, 107 S.Ct. 2415.

The opinion announces that it overrules a handful of this Circuit’s decisions recognizing protected liberty interests in parole release “[to the extent that they] might be read to imply that there is a federal constitutional right regardless of whether state law entitles the prisoner to release.” Maj. Op. at 555. But each of the cited decisions grounded its inquiry and holding — quite expressly — in the particular characteristics of the state scheme at issue, which in each case was the California scheme. See Irons v. Carey, 505 F.3d 846, 850 (9th Cir. 2007) (“[T]he Supreme Court ha[s] clearly established that a parole board’s decision deprives a prisoner of due process with respect to this interest [created by California Penal Code section 3041] if the board’s decision is not supported by some evidence in the record.” (emphasis added) (internal quotation marks omitted)); Sass v. Cal. Bd. of Prison Terms, 461 F.3d 1123, 1127 (9th Cir. 2006) (“[T]here is no constitutional or inherent right of a convicted person to be conditionally released before the expiration of a valid sentence.... [But] if a state statute uses mandatory language (‘shall’) to create a presumption that parole release will be granted when the designated findings are made, the statute creates a liberty interest in parole.” (internal citations and quotation marks omitted)); Biggs v. Terhune, 334 F.3d 910, 915 (9th Cir. 2003) (“Because the California parole scheme vests in every inmate, a constitutionally protected liberty interest ..., the requirements of due process are satisfied if ‘some evidence’ supports the [parole] decision.” (emphasis added)). Accordingly, there is no implication — much less any holding— for this en banc panel to “overrule.”

B.

The majority’s discussion of parole is an object lesson in why the due process inquiry — or, for that matter, any judicial inquiry — cannot sensibly be conducted in the air. The opinion repeatedly asserts that prisoners may be denied parole at the state’s discretion, whereas good time is nondiscretionary. Maj. Op. at 556-58. But whether either is so depends on what state law provides. States may — as California has done — create an indeterminate sentencing scheme and require that prisoners who have served their minimum term will be released if certain circumstances exist. The Supreme Court has made clear that official discretion in the *568due process context is not all-or-nothing but relative, and that limitations on official discretion such as those California has adopted for its parole release system can give rise to a federal liberty interest. See Olim v. Wakinekona, 461 U.S. 238, 249, 103 S.Ct. 1741, 75 L.Ed.2d 813 (1983) (“[A] State creates a protected liberty interest by placing substantive limitations on official discretion.”). Again, whether there are such limits on discretion turns on the details of the particular state scheme. See Allen, 482 U.S. at 375, 107 S.Ct. 2415 (“[A]n official has discretion when the standards set by a statutory or regulatory scheme cannot be applied mechanically.” (internal quotation marks omitted)). Divorced from context, then, generalizations about “discretion” only mislead.

More fundamentally, because the majority does not offer its general observations on parole and good time in the course of deciding any question that is actually raised by Hayward’s petition, the observations are advisory and so useless to the proper development of the law. See Flast v. Cohen, 392 U.S. 83, 96-97, 88 S.Ct. 1942, 20 L.Ed.2d 947 (1968) (“[T]he rule against advisory opinions [ensures] that clear concreteness provided when a question emerges precisely framed and necessary for decision.... ”). The opinion announces that the good time and parole contexts “differ,” and discusses the differences between the two at some length. But because, as I have explained, neither good time nor parole creates due process rights without reference to the particular state system at issue, one is left wondering what the relevance of this discussion of Platonic ideal types might be. It is clear — even if little else about Part II is — that these generalizations do not require future courts to ignore Supreme Court precedent, our precedent, and the actual characteristics of the state schemes they are called on to evaluate.

III.

As I have indicated, I am nevertheless in general agreement with Part III of the majority opinion, which holds cognizable on habeas review claims that California parole denials were made without “some evidence” of future dangerousness. The majority declines to employ the Supreme Court’s framework for determining whether California parole statutes create protected liberty interests, relying instead on the fact that California’s own constitutional law requires that “some evidence” under-gird parole denials. Although I would approach the question a bit differently, I believe the majority’s ultimate conclusions sound and offer here some additional considerations in support of them.

The California Supreme Court has held that, as a matter of state constitutional law, some evidence must support the Governor’s decision denying parole. See In re Lawrence, 44 Cal.4th 1181, 82 Cal.Rptr.3d 169, 190 P.3d 535, 547-48 (2008); In re Shaputis, 44 Cal.4th 1241, 82 Cal.Rptr.3d 213, 190 P.3d 573, 580 (2008). But the Court has consistently refrained from deciding whether the federal Constitution also so requires. See, e.g., In re Rosenkrantz, 29 Cal.4th 616, 128 Cal.Rptr.2d 104, 59 P.3d 174, 205 n. 12 (2002) (“Because we conclude as a matter of California law that the ‘some evidence’ standard of review is applicable to judicial review of a Board’s decision denying parole, we have no occasion to determine whether the same standard is also mandated under federal constitutional principles.”). By denying relief without addressing the federal issues, the California courts have implicitly determined that the federal Constitution does not impose procedural or evidentiary standards more stringent than those imposed by the California Constitution. Be*569cause Hayward does not contend otherwise, and we have never so held, this case does not present an occasion to disturb the California Supreme Court’s implicit holding on that question.

It is likewise appropriate for the majority to rely on the California Supreme Court’s articulation of state constitutional requirements in determining what the federal Constitution requires in this case. Federal due process rights typically arise from state law entitlements, see Thomson, 490 U.S. at 460, 109 S.Ct. 1904, and there is no reason to think that state constitutional law may not place mandatory limits on official discretion, giving rise to “a presumption that parole release will be granted” and creating a protected liberty interest, Allen, 482 U.S. at 377-78 & n. 10, 107 S.Ct. 2415, just as statutes may do. That is exactly what the California Supreme Court has determined the California Constitution does with regard to parole denial, see Lawrence, 82 Cal.Rptr.3d 169, 190 P.3d at 548, and the majority correctly relies on that decisional law as defining a quantum of evidence requirement, without deciding whether the interest would exist in the absence of that decisional law. Moreover, the arbitrary disregard of process guaranteed by state constitutional law may violate the federal Constitution’s guarantee of due process. See Hicks v. Oklahoma, 447 U.S. 343, 346, 100 S.Ct. 2227, 65 L.Ed.2d 175 (1980). For that reason, the majority is correct to review the state court decision here for compliance with the California Constitution’s requirement of “some evidence” of future dangerousness. The federal Due Process Clause requires at least that much.

Applying the “some evidence” rule in light of the consideration that review of that factual determination is governed by AEDPA’s deferential standards, see 28 U.S.C. § 2254(d), I conclude that the state courts’ decision that Hayward was properly denied parole was not an unreasonable conclusion on the record of the pertinent parole proceeding. In interpreting the statutes and regulations governing California’s parole system, the California Supreme Court recently held that its caselaw “recognize[s] that the paramount consideration for both the Board and the Governor under the governing statutes is whether the inmate currently poses a threat to public safety and thus may not be released on parole.” Lawrence, 82 Cal.Rptr.3d 169, 190 P.3d at 552 (citing In re Dannenberg, 34 Cal.4th 1061, 23 Cal.Rptr.3d 417, 104 P.3d 783, 786 (2005); Rosenkrantz, 128 Cal.Rptr.2d 104, 59 P.3d at 202). The suitability and unsuitability factors may guide the Board or the Governor in their determination that a prisoner remains dangerous to the public, but “[i]t is not the existence or nonexistence of suitability or unsuitability factors that forms the crux of the parole decision; the significant circumstance is how those factors interrelate to support a conclusion of current dangerousness to the public.” Id. at 553. “Accordingly ... the relevant inquiry is whether some evidence supports the decision of the Board or the Governor that the inmate constitutes a current threat to public safety, and not merely whether some evidence confirms the existence of certain factual findings.” Id.

In light of Shaputis, 82 Cal.Rptr.3d 213, 190 P.3d at 580, I cannot conclude that the state trial court unreasonably determined that there is some evidence to support the Governor’s decision. I note, in addition to the factors cited by the majority, that the attack leading to Hayward’s conviction was not an isolated incident of brutality. The murder was, instead, the culmination of a long history of violence and criminal activity stretching back to as early as age eight. See id., 82 Cal.Rptr.3d 213, 190 P.3d at 584 *570(describing the crime of conviction as the “culmination” of several years of violent behavior). Hayward had been arrested at least twenty times prior to the murder, on charges ranging from armed robbery, to assault and battery, to attempted murder. He previously served a prison term for assault with a deadly weapon. As in Shaputis, these circumstances reasonably establish that the attack was not committed during a period of “emotional stress that was unusual or unlikely to recur.” Id. The assault on Hayward’s then-girlfriend occurred eight months before the murder, and Hayward claims that his initial feelings of extreme rage over the incident subsided as time went on. Ultimately, according to the evidence, Hayward inflicted the fatal violence only after the drunk victim lobbed a beer bottle at his companion. See Cal.Code Regs. tit. 15, § 2402(c)(1)(E) (noting that a factor favoring an unsuitability determination exists where “[t]he motive for the crime is inexplicable or very trivial in relation to the offense”).

These static historical factors suggest that, without any doubt, Hayward once was an extreme danger to public safety. But, after more than fifteen years of denying stabbing his victim, Hayward ultimately accepted responsibility for the crime in 1993, more than fifteen years ago. And, although there is some indication that he made statements suggesting he lacked remorse in a 1997 psychological evaluation, more than ten years have elapsed since that time. Moreover, the record does not reveal any violence or serious discipline within the last fifteen years of Hayward’s incarceration. As the California Supreme Court has recently indicated, to focus completely on unchanging factors such as the commitment offense and preincarceration history is at odds with a parole system that assumes, as its basic premise, that some rehabilitation is at least possible. See Lawrence, 82 Cal.Rptr.3d 169, 190 P.3d at 559 (“[I]t is evident that the Legislature considered the passage of time— and the attendant changes in a prisoner’s maturity, understanding, and mental state — to be highly probative to the determination of current dangerousness.”). Over time, the ability of the commitment offense and pre-incarceration history to demonstrate current dangerousness diminishes. See id., 82 Cal.Rptr.3d 169, 190 P.3d at 560.

Still, Hayward’s most recent psychological evaluation reasonably can be read to indicate that Hayward had not yet reached that point, and the Governor so read it. The evaluation concluded that measures of both static and dynamic risk factors associated with violence placed Hayward in “a category that represents a low to moderate risk for future violence in the community.” These factors include the violent criminal history just described, as well as early adjustment problems in the prison community, “episodes of relationship instability,” including current estrangement from two of his daughters, and a history of substance abuse problems, including prior use of cocaine and marijuana. The evaluations are far from damning — they cite Hayward’s “increased level of maturity and insight, his participation in substance abuse recovery” — and they indicate that the “historical risk factors now seem to be adequately contained.” Nevertheless, the evaluation concludes that Hayward presents a “low to moderate risk” of danger to the community. The state court did not unreasonably determine that this evaluation, and the evaluation’s assessment of Hayward’s commitment offense and his violent history, are evidence that Hayward’s release would pose a possibly “moderate” *571risk to public safety.2

I accordingly join Part III of the court’s opinion and its conclusion that the petition for habeas corpus must be denied.