Office of the Commissioner of Baseball v. LOC

United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT Argued November 18, 2025 Decided September 22, 2026 No. 24-1259 OFFICE OF THE COMMISSIONER OF BASEBALL, ET AL., APPELLANTS v. LIBRARIAN OF CONGRESS, ET AL., APPELLEES AMAZING FACTS, INC., ET AL., INTERVENORS Consolidated with 24-1260 On Appeals of a Final Determination of the Copyright Royalty Board Elisabeth S. Theodore argued the cause for appellants Joint Sports Claimants. With her on the briefs were Philip R. Hochberg, Jeremy W. Dutra, Daniel A. Cantor, Michael Kientzle, and Daniel R. Yablon.  2 Ronald G. Dove, Jr. argued the cause for appellant Public Broadcasting Service. With him on the briefs were Dustin Cho, Daniel G. Randolph, and Pierre Anquetil. Sonia M. Carson, Attorney, U.S. Department of Justice, argued the cause for appellees. With her on the brief were Yaakov M. Roth, Assistant Attorney General, and Daniel Tenny, Attorney. Matthew J. MacLean argued the cause for intervenors Canadian Claimants Group, Commercial Television Claimants, Program Suppliers, and Settling Devotional Claimants. With him on the brief were Jessica T. Nyman, Michael A. Warley, Caroline M. Block, Arnold P. Lutzker, Lawrence K. Satterfield, Victor J. Cosentino, Preetha Chakrabarti, Gregory O. Olaniran, and Lucy H. Plovnick. Philip R. Hochberg, Jeremy W. Dutra, Daniel Cantor, Elisabeth S. Theodore, Michael Kientzle, and Daniel Yablon were on the brief for intervenors Joint Sports Claimants. Ronald G. Dove Jr., Dustin Cho, Daniel G. Randolph, and Pierre Anquetil were on the brief for intervenor Public Broadcasting Service. Before: SRINIVASAN, Chief Judge, CHILDS, Circuit Judge, and ROGERS, Senior Circuit Judge. Opinion for the Court filed by Chief Judge SRINIVASAN. Opinion concurring in part and dissenting in part filed by Senior Circuit Judge ROGERS. SRINIVASAN, Chief Judge: Section 111 of the Copyright Act establishes a compulsory licensing scheme under which  3 cable television systems may distantly retransmit copyrighted broadcast programming in exchange for depositing statutory royalties into a common pool. The Act charges the Copyright Royalty Board with the arduous task of allocating that pool among the competing copyright claimants. This consolidated appeal concerns the allocation of the royalties collected for 2014 through 2017. The appellants in this case—the Office of the Commissioner of Baseball and the Public Broadcasting Service—appear on behalf of two adversely positioned groups of copyright claimants. Although both challenge the Royalty Board’s allocation of the statutory royalties, and both contend that the allocation was arbitrary and capricious and unsupported by the record in various ways, their arguments naturally differ since they are competing claimants to the same pool of collected fees. We reject nearly all of appellants’ challenges to the allocation of the royalty fund. The Board reasonably evaluated the two principal valuation methodologies and explained the adjustments it made to determine the relative marketplace value of the competing categories of broadcast programming. But the Board left unexplained the critical, final step that followed: how exactly it merged the results of those two methodologies to calculate the final allocation percentages among each claimant group. That final step determined the outcome we review, and yet we are unable to discern precisely how the Board arrived at the final allocation. We therefore vacate the final determination and remand for further explanation.  4 I. A. The Copyright Act balances two important principles: “ensuring the protection of intellectual property and encouraging the free flow of information.” Indep. Producers Grp. v. Libr. of Cong. (IPG I), 792 F.3d 132, 135 (D.C. Cir. 2015) (citation omitted). The Act effectuates that balance in part by providing for the compulsory licensing of copyrighted material in certain situations. See generally 17 U.S.C. §§ 111–122. The compulsory licensing scheme at issue here permits cable systems to distantly retransmit television broadcast programming in exchange for a statutorily prescribed fee. Id. § 111. The classic use case is if a cable subscriber wants to watch programming broadcast elsewhere in the country—say the subscriber grew up in one community and moves far away from home but wants to continue watching her home team’s games carried on a local broadcast station there. Section 111 of the Copyright Act provides how the statutorily prescribed fees paid by cable systems to retransmit broadcast programming are calculated and pooled before they are ultimately redistributed among the copyright holders. Generally, a cable system pays a percentage of the “gross receipts paid by subscribers” over a six-month period for the “basic service of providing secondary transmissions of primary broadcast transmitters.” Id. § 111(d)(1)(A), (E)–(F). The percentage increases with the number of distant signals the system retransmits—that is, signals carried “beyond the local service area of [the] primary transmitter.” Id. § 111(d)(1)(B)(i). But even a system that carries no distant signals must pay at least the statutory minimum fee. Id. § 111(d)(1)(E)(ii), (F)(i). Cable systems deposit those fees with the Register of Copyrights. Id. § 111(d).  5 The Copyright Royalty Board is then responsible for determining how to distribute those fees among the appropriate copyright owners. Id. § 801(b)(3). Each July, any copyright owner (or their agents) claiming a share of that year’s royalty fees must file a claim with the Board. Id. § 111(d)(4)(A); 37 C.F.R. § 360.2. If claimants agree on how to distribute the fees, the Board authorizes the Library of Congress to distribute them accordingly. 17 U.S.C. §§ 111(d)(4)(B)–(C), 801(b)(7). Absent such agreement, the Board “conduct[s] a proceeding to determine the distribution of royalty fees.” Id. § 111(d)(4)(B). That proceeding unfolds in two phases. In Phase I, claimants group themselves into categories based on the type of programming they own—such as sports, public television, or devotional (religious) shows—and the Board allocates the overall royalty pot among those categories according to their relative marketplace value. See IPG I, 792 F.3d at 135. In Phase II, the Board distributes each category’s allocation among the individual copyright owners within that category. See id.; see also 37 C.F.R. § 351.1(b)(2)(ii)(C). Both phases are adversarial, trial-like proceedings in which the claimants conduct discovery and submit evidence, including testimony and other evidentiary materials. At the conclusion of each phase, the Board issues a final determination allocating (Phase I) or distributing (Phase II) the contested portions of the royalty fund. See 17 U.S.C. § 803(c). The determination must be in writing, “supported by the written record,” and “set forth the findings of fact relied on” by the Board. Id. § 803(c)(3). The Librarian of Congress then publishes the determination in the Federal Register and distributes the royalty fees. Id. § 803(c)(6). A claimant wishing to challenge the Board’s determination may seek judicial review in this court within 30 days of its publication in the Federal Register. Id. § 803(d)(1).  6 B. On March 20, 2023, the Board began the evidentiary proceedings to determine the allocation of the pool of royalty fees for cable retransmission for 2014 to 2017 among six claimant groups: Joint Sports Claimants (JSC); Public Television (PTV); Settling Devotional Claimants (SDC); Program Suppliers (PS); Canadian Claimants Group (CCG); and Commercial Television Claimants (CTV). Over the course of the one-month proceeding, the Judges on the Board admitted hundreds of exhibits, heard live testimony from dozens of witnesses, and considered the claimant groups’ competing evidence. See Distribution of Cable Royalty Funds, 89 Fed. Reg. 54166, 54168 (June 28, 2024) (J.A. 448). As in prior proceedings, that evidence went to determine the “relative marketplace value” of each claimant’s programming, a term the Judges have long understood to mean valuations “that simulate [relative] market valuations as if no compulsory license existed.” Id. The Judges’ task, in other words, is to construct a “hypothetical market” that approximates the relative values that would emerge in an unregulated marketplace. Id. (citation omitted). Over time, claimants have developed different methods for modeling the hypothetical market. In recent decades, two approaches have predominated: regression analysis and constant-sum surveys. All six claimant groups in this proceeding relied on one of those two approaches in support of their allocation proposals. Modeling the counterfactual free market proved uniquely challenging in this cycle. The advent of streaming services enabled viewers to access content without a cable subscription. Most significantly, the retransmission market experienced a “commercial earthquake” in the midst of the proceedings. Id.  7 at 54220 (J.A. 500). In 2015, WGNA—“by far the most distantly retransmitted channel,” and the source of nearly all the sports programming available under the Section 111 licensing scheme—converted from a broadcast station to a cable network. Id. WGNA carried games of Chicago sports teams. WGNA’s conversion from a broadcast station to a cable network took it out of the statutory licensing market. See id. In light of those fundamentally changed circumstances, the Judges concluded that neither the regression analyses nor the survey evidence alone could adequately model the hypothetical market. They therefore treated both approaches as “useful” and “equally weighted” them in determining the relative marketplace value of each claimant group’s programming. Id. at 54232 (J.A. 512). 1. The Judges began by considering the claimants’ competing regression analyses. Regression analysis is a “method of determining the relationship between two or more variables.” Id. at 54218 (J.A. 498) (citation omitted). In this context, a regression examines the relationship between the programming that cable systems choose to retransmit and the royalties they pay, in an effort to “reveal[] preferences” about the relative value of different programming. Id. at 54266 (J.A. 546). The underlying idea is that a cable system’s willingness to pay more to retransmit programming indicates that it places greater value on that programming. See id. Several experts offered regression analyses, but the Judges ultimately relied on the analyses of PS’s expert, Dr. Cleve Tyler. Dr. Tyler conducted numerous different analyses, two of which are relevant here. One analyzed data from all cable  8 systems. The other, known as a “sensitivity test,” analyzed data only from systems that paid more than the statutory minimum fee. The methodology underlying the two analyses was the same: estimating a cable system’s willingness to pay for different types of programming by treating the statutory royalty rate as a measure of the value the system assigned to additional programming minutes. The WGNA conversion, however, changed the usefulness of the two models. After WGNA’s conversion to a cable network in 2015, many cable systems’ gross receipts fell below the statutory minimum royalty fee. Some of those systems did not retransmit programming to the full extent permitted by the license their minimum fee purchased, while others paid the minimum fee without retransmitting any distant signals at all. That posed a problem for Dr. Tyler’s regression, which sought to infer relative marketplace value from cable systems’ revealed preferences—that is, from their decisions about which programming to retransmit. Those decisions are most informative when they reflect affirmative choices in response to changes in cost. A system that pays only the minimum fee and retransmits no distant signals, however, provides little observable behavior from which to infer its preferences among programming types. That limitation prompted Dr. Tyler to conduct the “sensitivity” regression that excluded cable systems that paid only the statutory minimum fee. Although he expressed some reservations about its economic premise, he considered the sensitivity test “reasonably robust” and “sufficiently reliable” to inform allocation of the 2014–2017 royalty fund. Id. at 54172 (J.A. 452). The Judges concluded that the “dramatic increase in the number of minimum-fee-only” cable systems following WGNA’s 2015 conversion rendered regression analyses that included those systems “less reliable and . . . [of] only very  9 limited economic evidentiary weight.” Id. at 54171, 54177 (J.A. 451, 457). They therefore adopted the model based on data from all cable systems for 2014, when minimum-fee-only systems were less prevalent, and switched to the sensitivity test for 2015 to 2017. They then made several additional adjustments to the regression to account for idiosyncrasies in particular claimants’ circumstances, referred to as Adjustment A, Adjustment B, and Adjustment C. Adjustment A. The first adjustment addressed an anomalous increase in CCG’s allocation under the sensitivity test. Under Dr. Tyler’s model using data from all cable systems, CCG’s allocation rose from 6.5% in 2014 to 15.2% in 2017. Under the sensitivity test, it rose from 7.6% to 34.6% over the same period. The Judges attributed that difference to the distinctive characteristics of Canadian programming: its geographic availability is limited to cable systems within a 150-mile band below the U.S.–Canadian border, and its French-language content carries particular value in that region. Id. at 54229 (J.A. 509). Demand for that programming tended to cause cable systems in the region to pay above the minimum fee, causing CCG to be disproportionately represented in the sensitivity test. The Judges therefore reduced CCG’s allocation by reverting to the model using data from all cable systems and distributing the difference proportionally among the remaining claimants. That adjustment increased JSC’s annual allocation by about 0.25% and PTV’s by between 1.64% and 3.71%, depending on the year. Adjustment B. The second adjustment excluded data from “must-carry” broadcast stations that cable systems are statutorily compelled to carry. Federal law mandates that cable systems “carry the signals of qualified noncommercial educational television stations,” and prohibits systems from receiving “monetary payment . . . in exchange.” 47 U.S.C.  10 § 535(a), (i)(1). The Judges reasoned that, because the regression was premised on a cable system’s willingness to pay, programming that a system was obligated to carry could not reliably reveal associated preferences. See Distribution of Cable Royalty Funds, 89 Fed. Reg. at 54189–90 (J.A. 469–70). The Judges nevertheless gave PTV an opportunity to show how must-carry stations could be incorporated into the regression without distorting the results. The Judges concluded, however, that PTV had failed to identify which stations were subject to must-carry requirements or offer a viable method for incorporating that data into the regression. Without that information, the Judges relied on unrebutted evidence from JSC’s expert regarding the number of must- carry stations, reduced PTV’s allocation by the corresponding percentage for each year, and recalculated the other claimants’ shares accordingly. The adjustment reduced PTV’s allocation by roughly 3% and increased JSC’s allocation by a fraction of a percent. Adjustment C. The third adjustment addressed a limited exception to the exclusion of minimum-fee cable systems from the sensitivity test. Although the test generally treated minimum-fee payments as uninformative of cable systems’ economic preferences, PTV contended that cable systems paying only the minimum fee between 2015 and 2017 revealed a preference when they chose, after WGNA’s conversion, to keep retransmitting PTV programming they had previously received alongside WGNA. The Judges agreed and adjusted PTV’s allocation upward. The adjustment increased PTV’s allocation to approximately 17% in 2015 (a 5.1% increase), 22% in 2016 (a 6.7% increase), and nearly 23% in 2017 (a 7.0% increase). The other claimants’ allocations declined correspondingly, with JSC’s falling by only tenths of a percentage point.  11 The table below shows the claimants’ shares by year under the regressions, after incorporating the various adjustments to the regression results. Table 1: Royalty Allocations Based on Dr. Tyler’s Regression Analyses 2014 2015 2016 2017 CCG 6.55% 12.90% 13.00% 14.35% CTV 11.38% 12.37% 15.94% 12.12% PS 26.80% 44.87% 37.51% 40.39% SDC 4.33% 10.62% 9.95% 9.54% PTV 13.36% 16.96% 22.06% 22.98% JSC 37.48% 2.30% 1.56% 0.61% 2. The Judges next considered the other principal valuation methodology: survey evidence. A constant-sum survey elicits hypothetical valuation judgments by asking respondents to allocate a fixed pool of resources among competing options. The Bortz Survey, one such survey, has been a mainstay of Section 111 royalty-distribution proceedings for decades. JSC commissioned Bortz Media to conduct the survey here, which asked hundreds of cable systems to “value the various types of non-network programming on the distant signals” they carried by “allocat[ing] a percentage of a finite dollar amount to each of the program categories.” Id. at 54241 (J.A. 521) But the survey did not query cable systems that carried no distant signals or signals from only one programming category, on the theory that those systems could not provide comparative value judgments. Recognizing that this methodological choice biased the survey results downward for PTV and Canadian programming, the claimants proposed various adjustments.  12 In prior proceedings, the Judges addressed this issue by applying the “McLaughlin Adjustment,” which assumed that cable systems carrying only PTV or Canadian programming would assign 100% of their value to that programming category. Id. at 54242 (J.A. 522). JSC did not advocate for the McLaughlin Adjustment here, instead proposing alternative adjustments. Relevant here, Adjustment One applied the McLaughlin Adjustment for 2014 but not for 2015–2017, on the theory that cable systems that took steps after WGNA’s conversion to continue retransmitting PTV programming previously carried alongside WGNA would have assigned less than 100% of their value to that programming if surveyed after 2014. Although the Judges recognized limitations in the McLaughlin Adjustment, they nevertheless adopted it. The resulting allocations are shown in the table below. Table 2: McLaughlin-Adjusted Bortz Survey Royalty Allocations 2014 2015 2016 2017 CCG 1.0% 1.8% 1.3% 1.2% CTV 25.2% 19.2% 15.3% 17.2% PS 21% 18.4% 17.8% 14.8% SDC 5.4% 4.4% 5.0% 3.9% PTV 8.4% 43.6% 48.4% 48.2% JSC 39% 12.7% 12.2% 14.8% 3. The Judges turned to allocating shares of the royalty fund with the assistance of the two valuation methodologies. Because Dr. Tyler’s sensitivity test and the Bortz Survey produced a wide range of potentially reasonable allocations, the Judges sought to “reconcil[e] . . . these two useful (albeit imperfect) approaches” by weighting the regression and survey results for each claimant in each year according to the comparative utility of each of those results for the specific  13 claimant category. Id. at 54256 (J.A. 536). So, for instance, the regression results might be given relatively greater weight for one claimant category but the survey results might be given relatively greater weight for another category. The Judges then adjusted the resulting shares proportionally to ensure that they totaled 100%. After the evidentiary proceedings, the Judges sought additional evidence concerning the PBS–National Cable & Telecommunications Association (NCTA) agreement governing the carriage of certain PTV stations by cable systems. PTV submitted the agreements but moved for reconsideration on multiple grounds. On September 5, 2023, the Judges denied the motion as moot, concluding that the documents need not be entered into the record. The following day, the Judges issued an initial determination setting forth each claimant group’s percentage allocation of the 2014–2017 cable royalties. JSC and PTV both petitioned for rehearing. The Judges denied the petitions but corrected several undisputed arithmetic errors. In June 2024, the Judges issued a final determination allocating the funds among the six claimant groups:  14 Table 3: Final Royalty Allocations 2014 2015 2016 2017 6.19% 14.59% 14.6% 15.77% CCG $13,976,255 $30,291,018 $30,311,780 $32,740,874 20.55% 19.78% 17.36% 17.5% CTV $46,399,360 $41,066,233 $36,041,952 $36,332,613 21.21% 28.29% 25.53% 23.29% PS $47,889,559 $58,734,264 $53,004,092 $48,353,517 4.85% 6.74% 7.01% 5.83% SDC $10,950,700 $13,993,246 $14,553,806 $12,103,950 11.07% 19.18% 24.78% 25.25% PTV $24,994,692 $39,820,544 $51,446,980 $52,422,770 36.13% 11.42% 10.72% 12.36% JSC $81,577,075 $23,709,625 $22,256,320 $25,661,205 Both JSC and PTV promptly appealed the Board’s Phase I allocation, and we consolidated the two cases. The four remaining claimant groups—SDC, PS, the CCG, and the CTV—intervened, as did JSC and PTV in the portions of the appeal initiated by the other. II. We review decisions of the Copyright Royalty Judges under the familiar standards of the Administrative Procedure Act (APA) to assess if their decision is arbitrary or capricious, contrary to law, or not based on substantial evidence. See 17 U.S.C. § 803(d)(3) (incorporating by reference 5 U.S.C. § 706). Our review is “highly deferential,” Intercollegiate Broad. Sys., Inc. v. Copyright Royalty Bd., 571 F.3d 69, 79 (D.C. Cir. 2009), and, in royalty-allocation proceedings, we ask only whether the Judges’ allocations fall “within a zone of reasonableness,” Christian Broad. Network, Inc. v. Copyright Royalty Tribunal, 720 F.2d 1295, 1304 (D.C. Cir. 1983) (citation omitted). The Judges must of course “make reasoned decisions supported by the written record before them,” Settling Devotional Claimants v. Copyright Royalty Bd., 797 F.3d 1106, 1121 (D.C. Cir. 2015), but perfection is not required. We have long upheld a  15 “rough-justice approach” under which the agency may rely on “relevant and creditable methodological evidence, even if it [is] ‘far from perfect.’” Id. (quoting Nat’l Cable Television Ass’n, Inc. v. Copyright Royalty Tribunal, 724 F.2d 176, 184 (D.C. Cir. 1983)). Both JSC and PTV challenge how the royalty pie was divided, contending that the Judges’ application of rough justice was far more rough than just. JSC and PTV, however, unsurprisingly perceive different flaws in the Judges’ approach. We first consider JSC’s challenges to the regression and survey analyses and then consider PTV’s, neither of which we find persuasive. We last consider JSC’s challenge to the final allocation based on a combination of those two inputs, which does persuade. A. JSC argues that the Judges’ reliance on Dr. Tyler’s regression analyses and the Bortz Survey was arbitrary and capricious in three ways. 1. JSC begins by challenging the Judges’ reliance on Dr. Tyler’s sensitivity test on three grounds: (i) that it failed to accurately measure relative marketplace value of programming, (ii) that it drew from an unrepresentative dataset limited to the small minority of cable systems paying above the minimum fee, and (iii) that it produced implausible and statistically insignificant results that the Judges’ adjustments only exacerbated. None succeeds.  16 a. JSC contends that Dr. Tyler’s regressions rest on an economically incoherent foundation, such that they are incapable of determining the relative marketplace value of each claimants’ programming. Both regressions treated the statutory royalty rate per subscriber as the dependent variable and incremental programming minutes as the independent variable, with the regression purporting to “represent the incremental impact on the [statutory royalty rate] for each . . . minute.” Distribution of Cable Royalty Funds, 89 Fed. Reg. at 54198, 54225 (J.A. 478, 505) (citation omitted). In JSC’s view, those variables cannot bear the weight the regressions place on them because the Judges’ own findings establish that cable systems neither consider the statutory royalty rate nor the relative quantity of programming when selecting programming for retransmission. We are unpersuaded. The statutory royalty rate was a reasonable dependent variable. The statutory formula does not vary a station’s royalty rate based on the programming it carries: a station’s rate is the same whether it carries mostly sports, religious programming, or something else. JSC therefore reasons that the rate reflects only how the statute treats the station, not the value cable systems place on its programming. But the regressions did not use the statutory rate as a measure of programming value. They instead measured whether cable systems were willing to incur higher royalty costs to carry more of particular types of programming. If cable systems valued all programming equally, one would expect the relationship between royalty costs and programming minutes to be uniform across all systems. Yet, the data showed otherwise. Based on nearly 20,000 observations and more than 2,000 unique pricing relationships, the regressions found substantial variation in the relationship between royalty costs and  17 programming minutes. Had the regressions “merely mimick[ed] the statutory formula,” OCB Br. 22, as JSC suggests, the variation would be difficult to explain. The Judges understood that the regression was imperfect. Indeed, the Judges credited testimony that the “amount of money at issue regarding section 111 royalties is essentially de minimis” to cable systems. Distribution of Cable Royalty Funds, 89 Fed. Reg. at 54177 (J.A. 457). But the fact that royalty costs did not drive cable systems’ carriage decisions did not make those decisions irrelevant. The Judges reasonably concluded that cable systems’ choices could still reveal their relative preferences. After all, the systems still had to decide which signals to retransmit, and those choices were the behavior from which the regression sought to infer relative value. Nor was there anything arbitrary about the regressions’ independent variable. If anything, minutes of programming by category are a natural, pertinent measure. Consider a cable system that consistently chooses a signal with twice as many minutes of sports programming over another with half as many. Why? The evident answer is that the system places greater value on sports. There is nothing arbitrary or capricious about drawing that inference. But even if the variable was less than perfect, the Judges did not overlook its limitations. They recognized that some categories may have value “not well- correlated with overall program minutes,” id. at 54234 (J.A. 514), and they supplemented the model with the Bortz Survey to address that limitation. That is reasoned decisionmaking. b. JSC next objects that the Judges’ reliance on the sensitivity test deprived it of fair notice. Because cable systems paying  18 only the minimum fee constituted the lion’s share of the data, JSC argues, excluding those systems left the regression resting on a small and unrepresentative subset of cable systems. Worse, JSC says, no expert proposed the test. Neither contention holds up. Taking JSC’s second contention first, its suggestion that it lacked fair notice of the sensitivity test because no expert proposed it is difficult to square with the record. One of the central issues in the proceeding was the “use of above- Minimum Fee evidence as a building block for the ascertainment of relative value.” Id. at 54269 (J.A. 549). JSC itself had an expert undertake an additional regression to “separate[] out minimum fee systems from” Dr. Tyler’s analysis, precisely to test his regression using all cable systems, putting the very distinction on which the sensitivity test turned squarely before the Judges. Id. at 54201 (J.A. 481). JSC cannot now claim surprise at a distinction its own expert drew. Nor would it matter if the Judges had gone beyond the parties’ precise proposals. The Judges were not “strictly limited to choosing from among the proposals set forth by the parties” and could “modify proposals set forth by the parties, or [] suggest models of their own.” Johnson v. Copyright Royalty Bd., 969 F.3d 363, 381–82 (D.C. Cir. 2020). What matters is whether the claimants had notice of the methodology—not whether the Judges adopted the precise model a party preferred. Here, notice plainly existed. The Judges reasonably determined that the sensitivity test focused on the cable systems whose behavior was most probative of willingness to pay—the very thing the regression sought to measure. As Dr. Tyler explained, the test “restrict[ed] the data to those [cable systems] in which we have the highest degree of confidence that the [system] is considering tradeoffs between different stations,” thereby  19 isolating genuine marketplace preferences rather than the inertia associated with paying only the statutory minimum. J.A. 216. JSC objects that those systems were not “representative of the larger universe of cable systems.” OCB Br. 32. But the question was not whether these systems looked like the universe of cable systems; it was whether their behavior provided reliable evidence of a willingness to pay for additional programming. A cable system paying above the minimum fee may have differed from other systems precisely because it was the one making tradeoffs that revealed relative programming preferences. The Judges were entitled to credit Dr. Tyler’s expert judgment that those data provided the most useful evidence of programming preferences. c. Finally, JSC contends that it makes no difference which regression model the Judges used because both the full-dataset regression and the sensitivity test yielded results that were “absurd” in substance and “often statistically insignificant.” OCB Br. 33. The two models did not produce the sort of results JSC describes. The results JSC labels “absurd” are not as implausible as it supposes. Consider sports programming. Dr. Tyler’s sensitivity test found sports to be the least valuable programming category during 2015–2017 and suggested that its incremental value could be zero. See Distribution of Cable Royalty Funds, 89 Fed. Reg. at 54228 (J.A. 508). JSC responds that the result cannot be right because sports programming is plainly valuable, pointing to the Super Bowl, the World Series, and other marquee events. But no one disputes that. The question is not what sports are worth in the abstract; it is what the sports programming at issue here— which does not include the Super Bowl and World Series and  20 other such events—was worth in the § 111 distant- retransmission market. As to that category of sports programming within the purview of the § 111 retransmission market, the marketplace evidence gives reason to think it was worth less than JSC suggests. After WGNA’s conversion, cable systems retransmitted significantly less sports programming, even when adding another channel would have cost them nothing. The marketplace thus did not behave as though the sports programming at issue here was uniquely valuable. The supposed absurdities in the Canadian and devotional programming results were no mystery to the Judges, either. The sensitivity test produced an anomalously high allocation for Canadian programming and exceptionally high per-minute values for devotional programming. See id. at 54257 (J.A. 537). But the Judges recognized why. They understood that Canadian programming’s niche appeal, coupled with its relative scarcity, could cause it to tilt the sensitivity test. They likewise recognized that devotional programming is a category whose value “might not show up well in regressions.” Id. They therefore did not treat the regression results as dispositive for either category. Infomercials involve the same story. True, the sensitivity test attributed value to infomercials— programming all parties agree has no real-world value. See id. at 54216 (J.A. 496). But infomercials were only one type of programming within the broader PS category, and the regression estimated relative values across broad programming categories based on observed carriage decisions. It did not, as JSC urges, purport to assign an intrinsic value to individual programs standing alone. JSC’s statistical objection is not without force, but it has less force than JSC submits. JSC contends that the sensitivity test produced “wide confidence intervals” such that many of  21 the estimates were “statistically indistinguishable from each other.” OCB Br. 37. It is true that the test could not establish that sports programming had positive value at any conventional confidence level. The Judges acknowledged as much, noting that the sports estimates for 2016 and 2017 turned positive only at the 55% level. See Distribution of Cable Royalty Funds, 89 Fed. Reg. at 54231 (J.A. 511). But that limitation did not make the regression useless, much less render the Judges’ reliance on it arbitrary. The Judges’ task was not to establish the relative marketplace value of each claimant with scientific certainty, but to make reasonable estimates from the evidence before them. The Judges reasonably found regression analysis useful for those purposes: it produced an estimate for each category, with the confidence interval reflecting the uncertainty surrounding each estimate. A wide interval made the estimate less precise, not useless. And in any event, we do “not attempt to decide the merits of the methodologies,” sitting instead “as a panel of generalist judges,” not “as a panel of statisticians.” AEP Texas N. Co. v. Surface Transp. Bd., 609 F.3d 432, 443 (D.C. Cir. 2010). 2. JSC also challenges the Judges’ use of the Bortz Survey. JSC itself chose the Bortz Survey, rather than a regression, as its preferred measure of the relative marketplace value of claimants’ programming. The survey, as discussed, has a known limitation: it does not query cable systems that carry only a single programming category, because, in the survey’s view, those executives would have no basis to offer “comparative value judgments.” Distribution of Cable Royalty Funds, 89 Fed. Reg. at 54240 (J.A. 520). The Judges have long addressed that limitation through the McLaughlin Adjustment, which assumes that cable systems retransmitting such programming “would assign a relative value to [such  22 programming] of 100%.” Id. at 54242 (J.A. 522). JSC concedes that an adjustment was necessary. It objects only to the one the Judges chose, contending that the McLaughlin Adjustment was arbitrary and that other alternatives were available. We disagree. The Judges reasonably used a longstanding adjustment to address a known limitation in JSC’s own preferred methodology. a. JSC first takes issue with the Adjustment’s assignment of a 100% valuation to PTV programming, some of which cable systems were required by law to carry. JSC sees an inconsistency with the Judges’ recognition elsewhere that must-carry signals have no marketplace value, pointing in particular to the Judges’ exclusion of must-carry signals in Adjustment B to the regression results. The two adjustments, however, answer different questions. Adjustment B asks how much value a cable system would place on a signal if it were free to choose whether to carry it, and a carriage decision compelled by law tells us nothing about that. The McLaughlin Adjustment asks something else: who should get credit for the royalties paid by a system that carried only one claimant’s programming? The Bortz Survey cannot answer that question for those systems because it does not sample cable systems that carried distant signals from only one category. Without an adjustment, their royalties would instead be apportioned based on the revealed preferences of systems that carried multiple categories—potentially giving a share to categories that the system itself did not carry. The McLaughlin Adjustment avoids that mismatch by assigning the royalties to the only category the system carried. It makes no claim about what  23 that programming was worth. It claims only that the programming the system chose to carry was not worthless. That most PTV-only cable systems paid only the minimum fee does not change the analysis. In a regression, the prevalence of minimum-fee systems matters because royalty payments serve as a measure of actual willingness to pay. The Bortz Survey works differently. It does not infer value from what systems actually paid: it involves “hypothetical” rather than “actual decisionmaking,” id. at 54267 (J.A. 547) (emphases omitted), asking respondents to allocate a hypothetical budget to isolate relative value from the fee structure generating the payments. The minimum fee a system actually paid reveals little about how it would value programming in the hypothetical exercise. JSC also contends that the WGNA conversion render