OPINION OF THE COURT
BECKER, Circuit Judge.This case, before us for the second time, presents the question whether the Adult Guidelines for Parole Decisionmaking of the United States Parole Commission constitute “laws” within the meaning of the ex post facto clause of the United States Con*1158stitution.1 In early 1980, appellee, Lawrence Forman, was given a “presumptive release date” based on Commission guidelines promulgated in 1979, see 28 C.F.R. § 2.20 (1979), as applied to Forman’s convictions for offenses committed between 1967 and 1974. In United States ex rel. Forman v. McCall, 709 F.2d 852 (3d Cir.1983) (“Forman /”), a panel of this court held that application of the 1979 guidelines to Forman was retrospective and to For-man’s detriment. The panel further held that the guidelines would constitute laws for purposes of the ex post facto clause if they are applied without “substantial flexibility.” The panel remanded the case for development of a full factual record concerning the manner in which the Commission applied the guidelines in practice.
On remand, the district court accepted written submissions by the parties and held a hearing during which it considered statistical evidence and expert testimony. The court found that the evidence revealed the absence of substantial flexibility in the application of the guidelines, and accordingly held that the guidelines constitute “laws” for ex post facto purposes. We have carefully reviewed the record developed on remand and conclude that it unequivocally demonstrates that the guidelines are applied with substantial flexibility. We therefore reverse.
I.
The procedural history of this case was recounted at length in Forman I; a brief description of Forman’s situation and our prior opinion will therefore suffice.
Forman was convicted of evading more than $2,000,000 in taxes between 1967 and 1974. When he received his initial parole hearing, in early 1980, the Parole Commission applied the then-current parole guidelines, promulgated in 1979. See 28 C.F.R. § 2.20 (1979). The guidelines comprise a grid on which “offense characteristics” (designed to measure the severity of an offense) are plotted against “offender characteristics” (designed to measure the likelihood of recidivism) to yield a “customary” range of time that the offender is to serve before being released from prison on parole.2 See Forman I, 709 F.2d at 857. See generally Warren v. United States Parole Commission, 659 F.2d 183, 189-93 (D.C.Cir.1981), cert. denied, 455 U.S. 950, 102 S.Ct. 1454, 71 L.Ed.2d 665 (1982) (detailed history of the guidelines). The Commission determined that under the 1979 guidelines, the “customary range” appropriate for Forman was forty to fifty-two months in prison.
Forman petitioned for a writ of habeas corpus in the district court for the Middle District of Pennsylvania pursuant to 28 U.S.C. § 2241 (1976), claiming that the application of the 1979 guidelines violated the constitutional proscription against ex post facto laws. The district court agreed and ordered the Commission to afford Forman a new parole hearing to be conducted in accordance with the guidelines in effect at the time of the 1976 sentencing. The Commission complied, and Forman was ultimately released on parole after signing a Certificate reserving to the Commission the right to reincarcerate him should the district court’s order be reversed or vacated on appeal, which promptly followed.
The Forman I panel first held that the law in effect at the time of the offense is the relevant law for purposes of ex post facto analysis. Forman I, 709 F.2d at *1159856-57. See Weaver v. Graham, 450 U.S. 24, 28-31, 101 S.Ct. 960, 964-65, 67 L.Ed.2d 17 (1981). In order to determine whether the retrospective application of the 1979 guidelines was detrimental to Forman, the panel looked to the 1974 guidelines, those in effect at the time of commission of For-man’s offense, stating:
It thus appears that the Commission’s application of the 1979 instead of the 1974 guidelines resulted in the establishment of a different “customary” range of incarceration in Forman’s case: under the 1979 guidelines, Forman’s worst-case prognosis was forty to fifty-two months, as compared with a thirty-six month worst-case presumptive minimum under the 1974 guidelines. Moreover, For-man’s best-case prognosis was twenty-four to thirty-six months under the 1979 guidelines but twelve to sixteen months under the 1974 guidelines.
Forman I, 709 F.2d at 859. The panel thus concluded that the retrospective application of the 1979 guidelines was detrimental to Forman. The panel therefore went on to discuss the question whether the guidelines were “laws”.
On that issue, following the precepts of Geraghty v. United States Parole Commission, 579 F.2d 238 (3d Cir.1978) (“Geraghty I ”), vacated and remanded on other grounds, 455 U.S. 388, 100 S.Ct. 1202, 63 L.Ed.2d 479 (1980), the panel held that, if administered without sufficient flexibility, the guidelines could be considered laws for ex post facto purposes. This position has since been rejected by every other circuit that has addressed the issue.3 However, it remains the law of this circuit until overruled by the in banc court. See Third Circuit Internal Operating Procedures, Chapter VIII C.
In reaching its decision, the panel rejected the proposition that the guidelines could not be laws for ex post facto purposes simply because they are administrative regulations, and not statutes, Forman I, 709 F.2d at 859 (citing Geraghty I, 579 F.2d at 266).4
The panel noted, but rejected, the argument that the guidelines fall within the class of regulations, “possibly excluded” from the ambit of the ex post facto clause, consisting of “no more than general statements of policy, interpretive rules, or rules relating to agency practice or procedure.” Forman I, 709 F.2d at 859 n. 17 (citing Pickus v. United States Bd. of Parole, 507 F.2d 1107, 1112-14 (D.C.Cir.1974) (rejecting contention that guidelines are merely statements of general policy and holding their promulgation to be subject to rulemaking provisions of Administrative Procedure Act)). On the contrary, the panel recognized that the guidelines play an important role in the parole process:
“Unbounded discretion probably does not exist in the Commission’s decisionmaking; the guidelines provide perimeters that may be overstepped only upon a *1160showing of good cause, see 18 U.S.C. § 4206(c) (1976), and changes in the guidelines appear to shape the exercise of that discretion.”
Forman I, 709 F.2d at 861 (footnote omitted).
However, the panel declined to adopt an approach, then advocated by Forman, which looks only to the statutory language and legislative history for enlightenment as to the ex post facto effect of the guidelines. The panel thus rejected the position that simply because all parole decisions must fall within the guidelines or depart from them only upon a showing of “good cause,” see 18 U.S.C. § 4206(c), the guidelines affect all parole decisions, the Commission may never ignore them, and therefore they constitute a law for ex post facto purposes. The panel relied instead upon the approach taken in our earlier opinion in Geraghty I, and reaffirmed the reasoning of that opinion, which treated the ex post facto question as essentially one of fact:
Geraghty recognized both that the Parole Commission’s “discretion” is severely constricted because the Commission must either follow its guidelines except for good cause or should revise the guidelines when parole decisions outside the regulations become too frequent and that the “ ‘channel for discretion’ ” under the guidelines therefore appeared to be “in actuality an unyielding conduit.”____ Geraghty held that the manner in which the Commission actually applied its guidelines still constituted a question of fact.
Forman I, 709 F.2d at 862 (footnote omitted).
As refined by Forman I, the inquiry thus became, as we have noted, whether the guidelines in fact are applied with “substantial flexibility.” Id. The panel ruled that a Commission practice to accord each inmate individualized treatment would not necessarily be dispositive of the issue, but that the “range and contours of that allegedly individualized treatment,” id. at 861, would also be relevant to the determination of whether the guidelines were merely “a channel for discretion,” or “an unyielding conduit” which constitutes a “law” for ex post facto purposes. The district court was therefore invited to rely on a wide variety of evidence on remand, including statistical evidence concerning the frequency with which parole decisions are made within the guidelines. We now turn to the district court record and the court’s findings.5
*1161II.
After receiving extensive expert reports, briefs, and expert testimony, the district court made the following specific findings of fact:
1. Statistically, the Parole Commission decisions fall within the guideline parameters in 85-90% of the cases, thus [there is] a deviation of only 10-15%.
2. The range and contour of the “individualized decisions” is relatively smooth and, for the greatest part, within the guidelines.
3. Nominally the Parole Guidelines are discretionary but in actuality, discretion is so absent that the guidelines are revamped whenever it appears that a percentage of the Parole Commission decisions fall outside of the guideline parameters.
Opinion of the District Court at 13.
The court explicitly disclaimed reliance on statistical evidence alone, however, and considered additional testimony and other submissions relating to the practice of the Commission in acting upon applications for parole. The court stated that when the Commission guidelines are used, “there is little or no room for discretion.” Id. Applying these findings of fact to the standard set out in Forman I, the court concluded that the “guidelines are woodenly applied, thus constituting an ‘unyielding conduit’ [Forman I, 709 F.2d] at 863, void of substantial flexibility. [7d] at 862.” Opinion of the District Court at 14 (reprinted in Appendix at 179). The Commission challenges on appeal both the district court’s findings of fact and its conclusion of law. Our review of the district court’s conclusions of law is plenary, see Tustin v. Heckler, 749 F.2d 1055, 1060 (3d Cir.1984); we review findings of fact under a clearly erroneous standard of review. Cf. Fed.R. Civ.P. 52(a).
III.
Much of the record developed on remand consists of statistical evidence concerning the frequency with which the Commission renders parole decisions within the guideline ranges computed for inmates. Forman contends that during the period covered by those statistics (October 1, 1977, to March 31, 1983), 79.9% to 86.5% of the parole decisions have fallen within the guideline ranges. Those same statistics show some variation in this frequency among the five regional offices of the Commission; for the same period, the highest and lowest rates of compliance by a single regional office are 90.7% and 76.7%. The district court, as we have seen, found that within-guideline decisions accounted for “85% to 90%” of all parole decisions.6 The Commission asserts that both Forman and the district court are mistaken, and submits that, properly interpreted, the data reveal that decisions within the guidelines were rendered, for all regions combined, in 75.4% of the cases, with a regional variance of between 73.5% and 78.3%. We therefore turn to an evaluation of the government’s contentions regarding the data.
The discrepancy between the Commission’s figures and Forman’s is rooted in the different definitions of “within” and “outside” the guidelines employed by the parties. Forman’s presentation of the data adopts the convention used in the Annual *1162Reports of the Commission, which counts as “outside” the guidelines only those decisions that must be accompanied by statements of “good cause.” See 18 U.S.C. § 4206(c) (Commission must furnish inmate “good cause” justifying a decision outside the appropriate guideline range). All other decisions are included in the “within” category in the annual reports. The Commission, on the other hand, urges that for purposes of determining whether the guidelines are applied with substantial flexibility, a more sophisticated approach to the data is necessary. This approach isolates certain kinds of decisions that are counted as “within” decisions in the annual reports. The Commission submits that for present purposes these cases should be treated as decisions “outside” the guidelines, which, technically, they are, or at least excluded from the data base entirely.
The substance of the Commission’s contention arises from the interaction between the function of the Commission and that of the courts in imposing sentences. The two categories of decisions that account for the bulk of the difference between the Commission’s figures and the district court’s findings are those cases in which the Committee was precluded from exercising its usual latitude by reason of a maximum sentence shorter than an inmate’s computed guideline range or a minimum sentence longer than the range. These categories of decisions were described in detail in an affidavit submitted by Dr. Peter Hoffman, Research Director for the Parole Commission and its predecessor, the United States Parole Board, and a major participant in the development and refinement of the guidelines. Portions of Dr. Hoffman’s affidavit are rescribed in the margin.7
*1163Forman argues that these cases are properly counted as within the guidelines because in deciding them, the Commission “applies the guidelines to the fullest extent possible.” However, the Commission argues persuasively that in these situations a decision that Forman counts as within the guidelines does not in any way indicate the absence of flexibility. Rather, these decisions, which on average accounted for 27% of the Commission’s decisions, see note 7, supra (sum of averages for BELOW CTE and ABOVE MJS classifications), are essentially phenomena resulting from the imposition of relatively long or short judicial sentences. As such, these decisions have almost no significance for the inquiry before the court. For example, where a long minimum sentence precludes the Commission from setting a release date within or below the guidelines, we have no way of knowing the decision which the Commission would have rendered based on its evaluation of the relevant factors. In such a case, the Commission is robbed of any discretion to choose among the “within,” “above,” and “below” categories. All we can know from the data is that the Commission decided not to depart even further from the guideline range than required by the minimum sentence. Similarly, when the maximum sentence falls below the computed range, we can discern from a Commission decision only its judgment whether it was appropriate to go even further below the guideline range.
When these decisions are excluded from the data base altogether, the average annual percentage of within-guideline decisions drops to 75.4%. We agree with the Commission’s argument and we believe that this statistic reflects a representation of the data better suited to present purposes than the data taken by Forman unmodified from the Parole Commission annual reports. We therefore hold that as a matter of law, these categories of anomalous decisions should have been excluded from consideration on the issue of substantial flexibility. Inasmuch as it appears that the district court accepted Forman’s characterization of these decisions as “within” the guidelines, the court’s finding of fact on this issue was clearly erroneous. We further conclude that a 25% deviation from the prescribed guideline ranges — that is to say, a finding of “good cause” in one out of every four cases in which such a finding can have any effect on the inmate’s release date — is strong evidence of “substantial flexibility” in the application of the parole guidelines. We thus hold that, under the data of record, the guidelines are being administered with sufficient flexibility and that they do not constitute “laws” for purposes of the ex post facto clause. See Forman I.
It would appear from the unrebutted testimony of Dr. Hoffman at the hearing that this result is no accident. Dr. Hoffman testified that the Commission takes into account all relevant aggravating and mitigating circumstances in setting a release date. Tr. at 70-130. See also 18 U.S.C. § 4206(a); 28 C.F.R. § 2.19. It is true that Dr. Hoffman conceded the routine nature of assigning offense severities in about half the cases decided by the Commission (Tr. at 69); however, his testimony also revealed that assignment of the salient factor scores, designed to assess the likelihood of recidivism by an inmate, involved a much less mechanical inquiry. Tr. at 73-74. Moreover, Dr. Hoffman testified that *1164in many cases, the Commissioners disagree among themselves as to the appropriate decision. Tr. at 89-90. Such evidence is counter-indicative of an “unyielding conduit.”
Forman argues that other evidence presented at the hearing also tends to establish the absence of “substantial flexibility.” Specifically, Forman pointed to the so-called Principle of Parsimony, a rule contained in the Manual which is distributed to all Commissioners and hearing examiners. Parole Commission Rules and Procedures Manual, 2.23-02 (May 13, 1983). Under the principle, which is presumably what the district court had in mind in finding that the “range and contours of the decisions was relatively smooth,” Opinion of the District Court at 13 (reprinted in Appendix at 178), whenever a decision within the guidelines is recommended, it is expected that the release date will be in the lower half of the guideline range unless one or more factors is present. Forman also submitted evidence showing that in Fiscal Year 1983 approximately 45% of decisions within the guideline range fell within the lower half of the range. We cannot agree that the Principle of Parsimony demonstrates the absence of substantial flexibility. The manual itself stresses that the principle “is intended to provide a methodology to promote analysis, not a mechanical rule.” 8
IV.
In sum, we hold that the Adult Guidelines of the Parole Commission are applied with “substantial flexibility” as required by Forman I. Accordingly, the judgment of the district court will be reversed.9
. “No Bill of Attainder or ex post facto law shall be passed.” U.S. Const, art. I, § 9, cl. 3; cf. id., art. I, § 10, cl. 1 ("No State shall pass any Bill of Attainder [or] ex post facto law”).
. Application of the guidelines assumes satisfactory institutional adjustment by the inmate. Since 1979, the Commission has afforded initial hearings to most prisoners, regardless of their parole eligibility date, within 120 days of confinement. See 28 C.F.R. § 2.12 (1985); 18 U.S.C. § 4208(a) (1982). The decisions rendered at these hearings assume good institutional adjustment in the future. At subsequent hearings, the Commission may accelerate or delay a presumptive release date if an inmate’s institutional adjustment warrants. See 28 C.F.R. § 2.14 (1985); Geraghty v. United States Parole Commission, 719 F.2d 1199, 1207 (3d Cir.1983).
. See Northeast Regional Parole Commission v. DiNapoli, 764 F.2d 143 (2d Cir.1985); Dufresne v. Baer, 744 F.2d 1543 (11th Cir.1984); Paschal v. Wainwright, 738 F.2d 1173 (11th Cir.1983); Richardson v. United States Parole Commission, 729 F.2d 1154 (8th Cir.1984); Roth v. United States Parole Commission, 724 F.2d 836 (9th Cir.1984). See also Portley v. Grossman, 444 U.S. 1311, 100 S.Ct. 714, 62 L.Ed.2d 723 (Rehnquist, Circuit Justice, 1980); Stroud v. United States Parole Commission, 668 F.2d 843 (5th Cir.1982); Warren v. United States Parole Commission, 659 F.2d 183 (D.C.Cir.1981), cert. denied, 445 U.S. 950, 102 S.Ct. 1454, 71 L.Ed.2d 665 (1982); Hayward v. United States Parole Commission, 659 F.2d 857 (8th Cir.1981), cert. denied, 456 U.S. 935, 102 S.Ct. 1991, 72 L.Ed.2d 454 (1982); Priore v. Nelson, 626 F.2d 211 (2d Cir.1980); Zeidman v. United States Parole Commission, 593 F.2d 806 (7th Cir.1979); Rifai v. United States Parole Commission, 586 F.2d 695 (9th Cir.1978); Shepard v. Taylor, 556 F.2d 648 (2d Cir.1977); Ruip v. United States, 555 F.2d 1331 (6th Cir.1977); Joost v. United States Parole Commission, 535 F.Supp. 71 (D.Kan.1982). But see Allen v. Hadden, 536 F.Supp. 586, 595-596 (D.Colo.1982); Note, Application of the Federal Parole Guidelines to Certain Prisoners: An Ex Post Facto Violation, 62 B.U.L.Rev. 515, 518-23.
. But see supra n. 3 and cases cited therein. Cf. Geraghty v. United States Parole Commission, 719 F.2d 1199 (3d Cir.1983) (“Geraghty II ”) (suggesting that guidelines not enacted pursuant to a delegation of legislative or judicial authority, but holding that if they are, the delegation is constitutional).
. On this appeal, the government contends that the guidelines are not promulgated pursuant to a delegation of legislative authority and, therefore, fail to satisfy the essential criterion for a regulation to have the effect of a law. In support of this contention, the government cites Chrysler Corp. v. Brown, 441 U.S. 281, 99 S.Ct. 1705, 60 L.Ed.2d 208 (1979), and Geraghty II. In Brown, the Court held that for an agency regulation to be a "law” it must both be a "substantive" rule and be issued pursuant to a delegation of legislative power. 441 U.S. at 301-303, 99 S.Ct. at 1717-1718. The government asserts that the guidelines are not substantive but rather belong to that class of rules called "general statements of policy.” This issue was decided otherwise in Forman I, 709 F.2d at 859 n. 17, and we are bound by that prior decision. See Third Circuit Internal Operating Procedures, Chapter VIII C.
With respect to the second portion of the Brown test, the government relies on this court’s decision in Geraghty II as authority for the proposition that the guidelines are not promulgated pursuant to a delegation of legislative power. We believe the government misapprehends the significance of Geraghty II for this case. It is true that certain passages in the opinion might be read to suggest a holding along the lines advanced by the government here. See 719 F.2d at 1212 ("because the guidelines establish the time to be served before release in the typical case, they do not conflict with or assume the legislative function of setting maximum penalties for the most heinous ones”) (emphasis in original). As the opinion makes quite explicit, however, the precise issue before the court in Geraghty II was whether the guidelines constitute a standardless, hence unconstitutional delegation of the legislative function. Id. The court answered that question in the negative, specifically refraining from any consideration of the ex post facto effect of the guidelines. On the contrary, the court noted the continuing vitality of the discussion in Geraghty I, relied upon by the Forman I panel, concerning the appropriate inquiry for a court faced with a challenge to the guidelines upon ex post facto grounds. We concede that there may be some tension between the aspects of the opinions in *1161the two Geraghty cases. See Geraghty II, 719 F.2d at 1213 (statement by Circuit Judge Adams sur denial of rehearing in banc) (suggesting that inconsistencies between Geraghty I and II be resolved by full court). At the same time, however, the Geraghty II panel had no more power than we have to overrule the previous decision in Forman I. See O. Hommel Co. v. Ferro Corp., 659 F.2d 340, 354 (3d Cir.1981).
The government argues in the alternative that we are not precluded from reexamining the premises of the holding in Forman I and taking into account what the government perceives to be the impact of the Supreme Court’s holding in Brown. While we doubt that such an examination would be fruitful in any event, in view of our holding today, favorable to the government, we decline the invitation to retrace old ground.
. There is some discrepancy between Forman's submissions and the findings by the district court. Specifically, the finding of "85% to 90%’’ compliance appears to be a rough estimate rather than a precise mathematical average derived from the raw data. We will assume, however, that the district court’s findings are essentially in agreement with Forman’s presentation of the data and their implications.
. The following description applies to categories of cases in which decisions formally outside the computed guideline range were rendered, due to the nature of the judicial sentence.
BELOW CTE ["continue to expiration"] Reflects the number of cases where the prisoner was continued to expiration below the guidelines. This would occur in cases in which a short sentence was imposed resulting in a mandatory release date below the computed guideline range. In these cases, the Commission is limited by the sentence from rendering a decision above or within the parole guidelines. The Commission would have the discretion to render a decision further below the guidelines than by operation of sentence (mandatory release date). If the Commission so chose, these cases would be [classified as] "BELOW DISC,” [decisions below the guidelines accompanied by statements of good cause].
ABOVE MJS ["minimum judicial sentence"] Reflects those cases where the effective parole date or presumptive parole date was established above the guidelines solely because the minimum term (period of parole ineligibility) established by the Sentencing Court precluded the Commission from rendering an earlier release date. For example, this would encompass those situations in which the Commission ordered an offender released after service of seventy-two months on an eighteen year regular adult sentence [minimum parole eligibility is normally one-third] wherein the guideline range was 40-52 months. The Commission would not, in that case, have the discretion to render a decision within or below the computed parole guideline range. The Commission, of course, would have the discretion to render a decision above the parole eligibility date, and further above the parole guideline range____ Those cases would be [classified as] "ABOVE DISC," [decisions above the guidelines accompanied by statements of good cause].
Affidavit of Dr. Peter Hoffman 9-11, reprinted in Appendix at 20-22. These two categories of decisions accounted for a significant number of decisions during the period covered by the data in the record. For all regions of Commission combined, the relevant percentages were:
(FY) 1978 1979 1980 1981 1982 1983 TOTAL
BELOW CTE 26.4 19.4 16.4 22.0 22.4 22.4 21.5
ABOVE MJS 0.1 5.6 10.9 6.5 6.4 6.1 5.7
Two other categories, which together account for the remainder of the difference between the statistics submitted by the respective parties, are much less significant. Dr. Hoffman's affidavit described them as follows:
BELOW 4YR This category is designed to filter out those cases which were continued for a four year reconsideration hearing under the old policy (in effect between September 6, 1977 through March 5, 1979) wherein the date of the four year reconsideration hearing would fall below the computed guideline range. It could not be determined from that information whether a decision below the guidelines, within the guidelines, or above the guidelines at such reconsideration hearing would be rendered.
ABOVE PRL This row reflects those cases in which an effective parole date was rendered *1163above the guidelines because the time in custody at the initial parole hearing precluded the Commission from rendering a decision within or below the guidelines. For example, a prisoner could choose to waive his parole hearing until he had served a period of time on his sentence exceeding his guideline range. This would not be a discretionary decision by the Commission to go above the guidelines. If the Commission exercised discretion to go further above the computed parole guidelines, the case would be [classified] as "ABOVE DISC.”
Affidavit of Dr. Peter Hoffman at 9-10, reprinted in Appendix at 20-21. For all regions, the statistics for these two categories of decisions are as follows (in percentages):
(FY) 1978 1979 1980 1981 1982 1983 TOTAL
BELOW 4YR 1.6 0.6 0 0 0 0 0.4
ABOVE PRL 3.7 2.9 1.8 1.6 1.0 1.4 2.2
Appendix at 29 (Table of Guideline Usage).
. We also discount the significance of the examples in the Manual of factors warranting decisions outside the guidelines, inasmuch as the examples are illustrative, not exhaustive. Parole Commission Rules and Procedures Manual, § 2.20-06, (October 1, 1983). Moreover, Dr. Hoffman testified that in fact decisions are made to go outside the guidelines for reasons not listed. Tr. at 130.
. The Commission contends that the district court applied the "substantial flexibility" test of Forman I in a way that guarantees that any system of guidelines that is not routinely ignored will be found to constitute a "law” for ex post facto purposes. Indicative of this error, in the Commission's submission, is the court’s treatment of certain of the evidence developed on remand. In support of its position, the Commission proffered evidence of a variety of ways in which it exercises what it calls "discretion” in administering the guidelines. As noted by the district court, the Commission argues that it exercises discretion:
1. In the adoption of the guidelines as such.
2. In determining the Offense Severity Classification.
3. In determining the range of months to be served.
4. In the modifications that have been made to the guidelines over the years since their initial adoption.
5. In the computation of the components of the guidelines in determining the seriousness for the offense.
6. In the computation of the salient factor score.
7. In determining where within the guidelines the release decision should fall.
8. In deciding to go outside the guidelines. Opinion of the District Court at 9. As far as the district court’s opinion reveals, however, in evaluating the record the court considered as evidence of "substantial flexibility” only the final item — the frequency with which the Commission chose, for "good cause," to depart from the recommended guideline range in particular cases. The court reasoned that the other examples of "discretion” are "aimed more at ... the discretion used in finding which portion of the guidelines applies to a specific case, rather than to the discretion used in each individual instance of determining when parole will be granted.” Id. at 10. Because we decide in favor of the Commission for the reasons stated in the text, we need not discuss the significance of any of these other ways in which the Commission asserts that it exercises discretion.