CUDAHY, Circuit Judge.
The Division of Enforcement (Division) of the Commodity Futures Trading Commission (CFTC or Commission) need present only circumstantial evidence — and only a preponderance of it — to establish liability for trade practice violations. See, e.g., Reddy v. CFTC, 191 F.3d 109, 118 (2d Cir.1999); In re Buckwalter, [1990-1992 Transfer Binder] Comm. Fut. L. Rep. (CCH) ¶ 24,995 at 37,684 (CFTC Jan. 25, 1991). Like most trade practice enforcement actions, this case involves a set of undisputed facts and the competing inferences that can be drawn from those facts. See, e.g., In re Reddy, [1996-1998 Transfer Binder] Comm. Fut. L. Rep. (CCH) ¶ 27,-271 at 46,210 (CFTC Feb. 4, 1998). The Commission says the circumstances here support an inference of non-competitive and pre-arranged, and therefore illegal, trading. Wayne Elliott, Francis Maritote, J. Brian Schaer and Jonathan A. Sion (the petitioners) disagree, as did an ALJ who heard the case in the first instance. The CFTC reversed the ALJ, finding against the petitioners, who now seek our review. Because the Commission is presumptively better able to weigh the relative strengths of the competing inferences and because its conclusion is reasonable, we affirm its order granting the enforcement action.
I.
Bringing a common combination of charges,1 the Division alleged, in essence, that the petitioners had executed 32 noncompetitive, non-bona fide trades. Pursuant to 17 C.F.R. § 10.66(d), the Division’s investigator, Hugh Rooney, presented his direct testimony in the form of a written report filed before the hearing. Rooney’s testimony consisted of two parts. First, he described what happened. The petitioners did not contest this description, and we summarize this factual testimony immediately below. Second, Rooney offered expert opinion testimony about the inferences to be drawn from these facts. The petitioners bitterly opposed Rooney’s analysis and conclusion. We discuss the issues raised by this testimony later. See Part III, infra.
Rooney provided an undisputed factual introduction to the petitioners, the futures [928]*928market and the charged trades. He told the following story. In 1991, the petitioners were members of, and registered floor traders at, the Chicago Board of Trade (CBOT). Each was an experienced, high-volume trader in wheat futures (among other commodities) for his own account (a “local” trader). Elliott had traded wheat futures for almost 20 years. Maritote had 15 years -experience. Schaer and Sion were newer to the pits, but still active, high-volume traders.
Rooney also explained that wheat contracts are delivered in March, May, July and September. Trading during these delivery months is considerably more risky than during non-delivery months. Contract delivery on the CBOT is on a “first-in-first-out” basis: those traders with the oldest contracts are at the front of the delivery line; those with newer contracts are at the back. Traders can reposition themselves in the delivery line (a practice called “freshening”) by liquidating old open positions in the nearby delivery month and buying equivalent volume contracts for future delivery (called the deferred month).2 As a trader’s delivery date nears, the price differential between the contracts on the deferred month and those on the nearby month narrows. The longer a trader waits to freshen his position, the greater the potential profit. If he fails to successfully freshen, however, he risks arrival at the front of the line where he will be forced to accept delivery and pay the carrying charges for storing and insuring the grain. Because of these risks, only some of the 75 to 100 wheat futures traders remain active during the wheat delivery months. Each of the four petitioners was a member of this relatively elite crew.
At the end of February 1991, each of the petitioners was close to the front of the delivery line for March wheat contracts. Each also held positions in May and July wheat contracts. Over eight trading days immediately prior to and during the March delivery cycle, the petitioners, trading among themselves, sold and repurchased the same spreads in identical quantities at the same price differential in a series of 32 trades.3 None of the petitioners suffered a loss or realized a profit from these trades; their net positions did not change. Summary descriptions of the trades follow:
1. February 25. Elliott sold 1000 March-July spreads at 21 cents to Mari-tote. Elliott bought 1000 of the same spread at the same price — 450 from Maritote and 550 from Sion. Sion got his 550 from Maritote, also at 21 cents. Schaer was not involved.
2. February 26. Elliott sold 735 March-May spreads at 10.75 cents to Maritote. Elliott bought 735 of the same spread at the same price from Sion. Maritote sold 735 of the same spread at the same price to Sion. Schaer was not involved.
3. February 27. Trading only March-May spreads at 11.5 cents, Elliott bought 1500 from Maritote, sold 2500 to Sion and then bought another 1000 from Maritote. Maritote bought 2500 of the [929]*929same spread at the same price from Sion. Schaer was not involved.
4. March 4. Elliott bought 1000 March-May spreads at 9.5 cents from Sion, then, turned around and sold them to Maritote. In two separate trades, each of 500, Maritote sold 1000 of the same spread at the same price to Schaer. Schaer sold them to Sion.
5. March 6. Elliott bought 2500 March-May spreads at 10.75 cents from Sion and sold them to Schaer. Schaer sold the same quantity of the same spread at the same price to Sion. Mari-tote was not involved.
6. March 7. Elliott sold 2500 March-May spreads at 10 cents to Schaer and bought the same quantity of the same spread at the same price from Sion. Sion bought 2500 from Schaer. Mari-tote was not involved.
7. March 8. Maritote bought 500 March-May spreads at 9.5 cents from Sion and sold them to Schaer at the same price. Elliott bought 2000 March-May spreads at 9.5 cents from Sion and sold them to Schaer. Schaer sold all 2500, in two trades, to Sion.
8. March 18. Elliott bought 1800 March-May spreads at 8.75 cents from Schaer and sold them to Sion. Sion sold 2600 of the same spread at the same price to Schaer and then bought 800 back. Maritote was not involved.
This was the Division’s substantive case against the petitioners — the undisputed facts surrounding the trades. The Division presented no direct evidence that the petitioners had pre-arranged or non-competitively executed any of the trades. The evidence was entirely circumstantial.
Rooney concluded from this series of facts — the mere circumstances — that the petitioners “engaged in a trading scheme to freshen their March 1991 wheat futures positions to avoid, decrease and/or delay delivery of cash wheat. Furthermore, this was accomplished through noncompetitive trading and wash sales.” R. 53, at 310. Rooney relied on seven “significant charac-support this conclusion. Id. at 272.
1. Size. Rooney opined that all but one of the charged trades were large trades (over 500,000 bushels) and that the petitioners accounted for an abnormal percentage of the total trade volume. See id. at 310-11.
2. No gains or losses. Rooney found it “very unusual” and “remarkable” that none of the petitioners made a profit or suffered a loss in any of the 32 trades over the eight days: “the lack of any profit or loss on any of the trades suggest that the trades were noncompetitive and wash sales.” Id. at 311.
3. No other traders involved. Rooney thought that it was “unusual” that no other traders participated in any of the 32 trades. Id. at 312-14. On the February and March 4 trading days, there were between 25 and 41 traders trading March 1991 spreads. On March 6, only eight traders swapped March 1991 spreads. Thirteen got into the action on March 7, but only seven participated on March 8 and 13. See id. at 313. Despite the limited trading population, Rooney concluded that the “absence of any participation by other traders or public customers suggests that no other traders had any genuine opportunity to participate in any other trades. In other words, the [petitioners] deliberately excluded other market participants and therefore their trades were noncompetitive.” Id. at 314.
4. Trade configurations. Rooney concluded that the circular trading pattern — between three or four of the petitioners — was indicative of non-competitive trading and constituted an effort by the petitioners to avoid any market risk. See id. “In general, competitive trading will rarely, if ever, result in a trade configuration with such overall precision and symmetry. It is virtually impossible for these configurations to occur in a competitive market. Competitive trading is char[930]*930acterized by a certain randomness.... [S]ince price should be the only reason to prefer one trader over another trader, competitive trading is usually dispersed among numerous traders and brokers. The consistent repetition of the same trading partners indicates a lack of genuine price competition and prearrangement.” Id. at 315.
5 and 6. Audit trail irregularities. The fifth and sixth significant factors both relate to the reporting and recording of the trades. Rooney concluded that trading irregularities and audit trail deficiencies such as out-of-sequence trades and late trading cards, although not unusual in volatile markets, suggested non-competitive trading in this case because none of the trades involved third parties (brokers or customers or dual traders) and none involved price fluctuations. See id. at 317-19. The petitioners also recorded the trades in unusual fashions, using single cards or consecutive lines. See id. at 319-20.
7. Trader motivation. Rooney noted a similarity of motivation: all of the trades appear to have been executed to allow the traders to freshen their positions and avoid delivery. See id. at 320-21. He did not think that this factor was particularly significant: “I do not believe it is essential to isolate a motivation, as I have here, in order to conclude that the trades were noncompetitive and wash sales.” Id. at 321.
The petitioners did not contest Rooney’s factual account; they admitted making the charged trades in order to freshen their positions. But they also asserted — each petitioner testified at the hearing — that all of the trades were completed competitively by open outcry in the pit and denied that any of the trades were pre-arranged. Character witnesses testified on behalf of each petitioner, vouching for his reputation for honesty. The petitioners did, however, challenge Rooney’s conclusion that the circumstances gave rise to an inference of non-competitiveness or pre-arrangement strong enough to establish liability.
Before the hearing — recall that Rooney’s direct testimony was presented in written form and filed in advance of the hearing — the petitioners argued that Rooney’s opinion lacked foundation and was unreliable and therefore should be excluded under Fed.R.Evid. 703. See R. 75. The ALJ, repeating the Division’s response to this motion, see R. 77, rejected this argument on the ground that the petitioners could cross-examine Rooney about the foundation for his conclusion at the hearing and thereby test its reliability there. Thus, it denied the motion. See R. 79. '
So, at the hearing, the attorneys for each of the petitioners took turns probing for weaknesses in the bases of Rooney’s ultimate conclusion. They poked many holes in his opinion and extracted some significant concessions. Under cross-examination, Rooney acknowledged that his report was based on limited information. He admitted, for example, that he did not interview any other wheat traders to determine whether the trades had been conducted by open outcry, see R. 124, at 186-87, and that he had not conducted any historical analysis to determine whether lack of profit was an unusual occurrence during delivery months (despite, in his second significant characteristic, concluding that the lack of profit was unusual enough to warrant an inference of liability), see id. at 175. Rooney also admitted that it was impossible to know how many traders were in the wheat futures pit attempting to trade, as opposed to actually trading, during the relevant time periods (a concession which might undermine his reliance on the first, third and fourth significant characteristics). See id. at 97-98. More, he acknowledged that he had used the eight charged trading days as his entire sample for purposes of his analysis of delivery period trading (another admission which might negate reliance on the first four characteristics), see id. at 152-54, and conceded that, in his opinion, it would be impossible to quantify the likelihood of the observed audit trail irregularities (casting [931]*931doubt on his reliance on characteristics numbers five and six), see id. at 189. These and other Rooney concessions amounted to an admission that his conclusion was not based on a rigorous statistical analysis. In fact, it was impossible for the petitioners to cross-examine Rooney about the factual and statistical bases for many of his conclusions because the conclusions were based on computerized CBOT records which Rooney had studied online but had not reduced to hard copies and had not produced for the petitioners. See id. at 72-75.4 The Division apparently managed to produce portions of some of these documents during a break in Rooney’s cross-examination, see id. at 80-82, but the petitioners were never given all of the relevant background information.5
The ALJ dismissed all three charges. Relying on the petitioners’ testimony as well as the testimony of two other traders, the ALJ determined that all of the trades had been conducted by open outcry and that market conditions, rather than prearrangement, had limited participation during the risky delivery cycle. He was unpersuaded by Rooney’s seven significant characteristics, concluding that the trading patterns amounted to only insubstantial circumstantial evidence.
The Division appealed to the full Commission, which reversed the ALJ’s dismissal of the charges. The Commission discounted the traders’ testimony that the trades had been executed by open outcry and focused instead on the “precision and symmetry” of the trading configurations. It concluded that this pattern evidence belied the randomness normally associated with competitive trading and therefore granted enforcement.6
II.
The petitioners frame the question on appeal as one of sufficiency of the evidence^ — whether the circumstantial evidence was enough to establish liability. See Petitioners’ Br. at 16. As we noted at the start, circumstantial evidence can suffice to meet the Division’s burden. See, e.g., In re Buckwalter, [1990-1992 Transfer Binder] Comm. Fut. L. Rep. (CCH) at 37,684. “[H]owever, the Division must do more than present suspicious circumstances suggesting the possibility of knowing wrongdoing. It must establish that the existence of these factual elements is more probable than their nonexistence.” In re Rousso, [1996-1998 Transfer Binder] Comm. Fut. L. Rep.(CCH) ¶ 27,133 at 45,-308 (CFTC Aug. 20, 1997) (internal quotations and citations omitted); see also In re Abrams, [1994-1996 Transfer Binder] Comm. Fut. L. Rep. (CCH) ¶ 26,479 at 43,136 (CFTC July 31, 1995) (“If both innocent and culpable inferences are equally supported by the record, the Division fails in its burden of proof.”). Evidence of unusual trading patterns, like that presented [932]*932here, commonly gives rise to an inference of culpability. See, e.g., In re Reddy, [1996-1998 Transfer Binder] Comm. Fut. L. Rep. (CCH) at 45,308; In re Rousso, ¶ 27,133 at 45,308 (“a pattern marked by characteristics unlikely to occur in an open and competitive market was indicative of noncompetitive trading”) (citations omitted); In re Bear Steams & Co., [1990— 1992 Transfer Binder] Comm. Fut. L. Rep.(CCH) ¶ 24,994 at 37,663-64 (CFTC Jan. 25,1991).
Two standards of review can apply to appeals from the Commission. If the question presented is “of the sort that courts commonly encounter, de novo review is proper.” Ryan v. CFTC, 145 F.3d 910, 916 (7th Cir.1998) (citations and internal quotations omitted). On the other hand, if the Commission’s decision was peculiarly within its area of expertise, we apply a deferential standard and will affirm so long as the decision is reasonable. See Cox v. CFTC, 138 F.3d 268, 271-72 (7th Cir.1998); LaCrosse v. CFTC, 137 F.3d 925, 929 (7th Cir.1998). Determining which standard to apply is no easy matter, however, because “if the question involves an interpretation within the specialized knowledge of the agency, a court should not automatically abandon heightened review.” Maloley v. R.J. O’Brien & Assoc., Inc., 819 F.2d 1435, 1441 (8th Cir.1987). “When the agency diet is food for the courts on a regular basis, there is little reason for judges to subordinate their own competence to administrative ‘expertness.’ ” Hi-Craft Clothing Co. v. NLRB, 660 F.2d 910, 915 (3rd Cir.1981). Thus, courts have exercised plenary review over common law issues, see, e.g., Morris v. CFTC, 980 F.2d 1289, 1294 (9th Cir.1992) (reviewing de novo a waiver of “pivotal issue” in a fraudulent inducement claim), constitutional issues, see, e.g., LaCrosse, 137 F.3d at 929-32 (reviewing de novo a double jeopardy issue), and issues within an agency’s special expertise but which are also a regular part of the court’s jurisprudential diet, see, e.g., Maloley, 819 F.2d at 1440-42 (reviewing de novo the Commission’s determination that the respondent had failed to exercise “reasonable diligence” in discovering the fraud because the court had often considered what evidence supported a finding of “reasonable diligence”). On the other hand, courts have applied the deferential standard to Commission determinations of the evidence necessary to prove violations of various sections of the Commodity Exchange Act (the Act), see, e.g., Monieson v. CFTC, 996 F.2d 852, 859-61 (7th Cir.1993) (§ 13(b)); Morris, 980 F.2d at 1295-96 (§ 4(b)); Silverman v. CFTC, 549 F.2d 28, 29-33 (7th Cir.1977) (§ 4(b)), as well as Commission rules, see, e.g., Monieson, 996 F.2d at 861-62 (CFTC Rule 166.3).
We believe the deferential standard applies here. Deciding whether a particular set of circumstances supports an inference of non-competitive trading on the futures markets is an issue peculiarly within the Commission’s area of expertise. See Ryan, 145 F.3d at 923 (“[N]either a criminal jury nor the Seventh Circuit, however authoritative their declarations, can claim expertise in the conduct of trading at the CBOT .... [T]he Commission [is] certainly an expert in matters relating to trading ....”) (Cudahy, J., concurring). While this court often faces sufficiency issues, the particular and peculiar nuances of the issue in this situation—involving, as it does, the enforcement of CFTC trade practice rules, see, e.g., Monieson, 996 F.2d at 859-62—are not “food for the court on a regular basis.” Hi-Craft, 660 F.2d at 915. We have found no published opinions in which a federal court has considered whether a particular set of futures market trading facts gives rise to an inference of pre-arrangement or non-competitive trading. The Commission, on the other hand, regularly considers the question and has already drawn a line, even if difficult to define, between evidence that establishes culpability and that which does not. See discussion supra at 931-32. We are not inclined to second-guess the Commission on this issue, nor do we relish redrawing the line ourselves. In short, the Commission is uniquely well positioned [933]*933to make these sorts of fact-specific, inference-laden determinations.7
In this case, the Commission’s decision passes muster. After noting that, “as the Commission has repeatedly found, direct evidence of noncompetitive trading is rarely available, and [that] the circumstantial analysis of trading patterns is an appropriate starting point for proving noncompetitive trading,” R. 175, at 13, the Commission analyzed the common characteristics of the charged trades. The series of trades, it found, reflected “a precision and symmetry not generally found in competitively executed trades.” Id. at 14. The interrelated trading pattern was also critical, and the Commission highlighted the fact that some of the trades were against the economic interest of the petitioners: “If a single participant in the ring-like transaction failed to play his part, none of the participants would succeed. In these circumstances, it is unlikely that a trader would repeatedly act against his own economic interest without advance knowledge that the other participants would continue trading the spread, enabling him to offset his newly-acquired long position.” Id. at 15. In the Commission’s view, an open and competitive market simply could not have “so precisely accommodate[d] the financial and timing goals of the four [participants] acting independently of each other.” Id. at 18. Therefore, “[b]ased on the totality of circumstantial evidence in the record, it is unlikely that petitioners would have known one another’s precise needs without some meeting of the minds or could have achieved this kind of trading symmetry in a volatile market by competitive execution.” Id. at 21.
Thus, the Commission was satisfied that the trading pattern—spreads in exactly identical quantities at the same price—was more likely than not the result of prearrangement and non-competitive trading. It carefully considered all of the evidence, articulating its reasons for discounting some and accepting some. In the end, the Commission determined that the Division had presented sufficient facts to give rise to an inference of culpability, and we defer to this considered, supported and reasonable decision.
III.
The issue of Rooney’s opinion testimony remains. The fight over the admissibility of this testimony did not reach this court. Despite the pre-hearing motion, the aggressive cross-examination and the competing expert testimony, the petitioners made no objection to the admission of Rooney’s report at the hearing, did not argue admissibility in their post-hearing briefs and did not raise the admissibility issue with the Commission during the Division’s appeal. Thus, the admissibility óf Rooney’s testimony does not appear to be inescapably before this court. Indeed, we doubt that the petitioners have properly preserved the issue for appeal.
Had the petitioners sustained their attack and properly challenged the ALJ’s decision to admit Rooney’s opinion testimony, we might have been inclined to agree with them. Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 589, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), held that reliability is the touchstone for expert testimony based on “scientific, technical, or other specialized knowledge” (quoting Fed. R.Evid. 702). In order to “ensure that any and all scientific testimony or evidence admitted is not only relevant, but reliable,” the factfinder must act as gatekeeper. Id. “Proposed [scientific] testimony must [therefore] be supported by appropriate validation—i.e., ‘good grounds,’ based on what is known.” Id. at 590, 113 S.Ct. 2786. Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999), extended this gatekeeping function to all expert testimony. The petitioners’ [934]*934effective cross-examination of Rooney exposed his opinion — and therefore his ultimate conclusion — as unreliable. The Division, through Rooney, presented what was essentially a “numbers” case: the circumstances surrounding the charged trades, it alleged, were statistically improbable. But it provided no reliable statistics to support this charge (and, indeed, could not even produce for the petitioners some of the raw numbers upon which it allegedly relied). It conducted no analysis of the liquidity of the market or of the usual number of traders swapping spreads during wheat delivery cycles; no analysis of the likelihood that delivery cycle freshening trades would result in no profit for the traders; and no analysis of the likelihood of making mistakes on trading cards or other audit trail irregularities — that is, it failed to justify Rooney’s assertion that the “significant characteristics” were at all significant. It was a numbers case without meaningful numbers, and, had the respondents objected to the admission of Rooney’s opinion testimony, we think that Kumho might well have compelled the ALJ to sustain the objection.8
Note, however, that we say only that we might have been inclined to agree with a properly preserved challenge to the admission of Rooney’s expert testimony. Our analysis is qualified because it is not plain that Rooney’s opinion testimony played any role in either the ALJ’s or the Commission’s disposition of this matter. Recall that the ALJ was unpersuaded by the seven “significant characteristics” upon which Rooney relied, see R. 153, at 7-11:
Taken as a whole, these characteristics do not prove that the challenged transactions were the result of an implied or express agreement to prearrange trades. The trade characteristics cited by the Division reflect a natural outcome of trades that can occur in a constricted market among perceptive and intelligent large volume traders who want to secure a later delivery date.
Id. at 11. On the other hand, the ALJ did state that, “[o]n balance, [Rooney] provided an accurate factual description of the ... transactions and of the [CBOT] delivery process.” Id. at 7 n. 9 (emphasis added). The Commission also appears to have drawn a distinction between Rooney’s factual testimony and his opinion testimony. It mentioned Rooney’s opinion testimony only once, see R. 175, at 14, and, in that singular instance, not as support for its conclusion. Instead, the Commission simply drew its own inferences from the undisputed accounts of the trades, focusing on the trading symmetries which it thought were indicative of non-competitive and pre-arranged trading. See id. at 12-23. The patterns, not Rooney’s opinion, swayed the Commission.
It appears then that both fact-finders9 discounted Rooney’s unreliable opinion testimony. This is an appropriate approach. Daubert and Kumho were decided in the context of admissibility, but the principle for which they stand — that all expert testimony must be reliable— should apply with equal force to the weight a factfinder accords expert testimony. See Libas v. United States, 193 F.3d 1361 (Fed.Cir.1999). Thus, a fact-finder should employ the reliability benchmark in situations, as here, in which unreliable expert testimony somehow makes it in front of the factfinder, and assign the unreliable testimony little if any weight. Both the ALJ and the Commission appear to have complied with this principle, ignored Rooney’s unreliable opinion and drawn their own inferences from the undisputed facts. We therefore decline to rule on the admission of the opinion testi[935]*935mony in the first instance; if error, it appears to have been harmless.
IV.
We recognize the seeming illogic of rejecting (to the extent we do) Rooney’s inference on the one hand, see Part III, supra, while endorsing (also to the extent we do) the Commission’s on the other, see Part II, supra. Both of their conclusions were, after all, based on the same set of undisputed facts. From our perspective, however, there is a cognizable difference. The condemnation of the admission of Rooney’s opinion testimony would amount to an application of an evidentiary rule and recent interpretations of it. We would simply be telling a factfinder what evidence it may properly consider, something we regularly do. The Commission’s decision, on the other hand, reaches us in a different posture: the question is not whether one piece of evidence is sufficiently rehable to be considered in the first place, but rather whether a certain quantum of specialized evidence was sufficient to raise a strong enough inference to establish liability for a violation of various provisions and rules, the enforcement of which was entrusted by Congress to a specialized agency. For precisely this reason, as we have already said, we choose not to second-guess the Commission’s inferential conclusion. So, while we might prefer to see a more statistically rigorous investigation of this issue, the Commission’s analysis of the circumstantial evidence was adequate.
Y.
If our dissenting colleague were a member of the CFTC, his opinion, offered as the opinion of the agency, might well survive judicial review as reasonable and as supported by substantial evidence. The issue before us, however, is not whether the CFTC here might have applied a different analysis and reached a different result; the issue is whether the analysis the agency did apply and the result it did reach have a rational basis and are supported by substantial evidence. 5 U.S.C. § 706. Wisdom strongly counsels and the Administrative Procedure Act demands appropriate deference to the Commission, an independent regulatory agency charged with the difficult task, among others, of detecting pre-arranged trading on its regulated exchanges. Direct evidence of prearrangement is rarely available and regulatory authorities are forced either to rely on circumstantial evidence or to abandon their statutory responsibilities. See, e.g., Reddy v. CFTC, 191 F.3d at 118. The effect of requiring the elaborate sort of critique proposed by the dissent is, as a practical matter, to abandon any effort to detect pre-arrangement.
We defer to the Commission in part because of its “expertise” or perhaps more precisely, as the dissent would have it, its specialized knowledge. The expertise or specialized knowledge of the agency is institutional, not personal. The dissent is off-base in exploring the occupational backgrounds of members of the Commission. As the dissent illustrates, it is a tempting but ultimately unrewarding exercise to make this a contest between the expertise of the commissioners and the expertise of the judges. The agency employs experts in the various subjects with which it must grapple and some of its staff advise the commissioners. Courts, on the other hand, employ no experts to advise the judges although it has been suggested from time to time that judges too should have expert advisors. But in the interest of fairness, courts generally eschew the idea of kept experts whispering in the judges’ ears. And to try to run agencies like courts is what Alfred Kahn deplores as the “over judieialization” of the regulatory process. 2 AlfRed E: Kahn, The ECONOMICS of Regulation 87 (1971) (quoting Harry M. Trebing, A Critique of the Planning Function in Regulation, Pub. Util. FoRtnightly (March 16,1967)).
This distinction between agencies and courts is one reason we reject our colleague’s suggestion that we follow antitrust precedent in this case. The dissent [936]*936argues that the petitioners’ trading patterns suggest “oligopolistic interdependence,” a dynamic often found in antitrust cases in which “repeat interactions lead small numbers of players to act as if they had agreed.... ” Dissent at 944. This court has refused to punish oligopolistic interdependence without some evidence of agreement, and the dissent urges us to follow suit here. See JTC Petroleum Co. v. Piasa Motor Fuels, Inc., 190 F.3d 775 (7th Cir.1999).
But the courts are responsible, in treble damage actions and otherwise, to resolve private and public antitrust disputes in the larger economy; the CFTC has a different task. It must detect prearrangement in a much smaller and more intimate sphere. These are, in some respects, similar responsibilities, but that does not mean that the doctrines that bind the courts in antitrust matters can be transposed unmodified to guide the regulation of commodity exchanges. At least on the surface, relations among traders on the exchange floor are quite different than relations among large business corporations selling in the same market. Traders rub elbows constantly and may have lunch together often. Theirs is the intimacy of the poker game as opposed to the remoteness of rival armies in the field. It does not strike us that, on the question whether parallel action shows agreement, the doctrine of the courts in antitrust cases should bind the surveillance activities of the Commission. We think that the Commission is free to draw its own conclusions, as long as they are reasonable under all the circumstances. And it is free, the dissent’s policy criticisms notwithstanding, to punish even pre-arrangement that hurts no customers. See Dissent at 942-43. That is a policy choice, and we don’t think it is any of our business.
In addition to setting policy within specialized mandates, agencies must decide what resources they can afford to commit to one or another of the many determinations .they have to make and must take some things as a given. This is one reason we are not moved, beyond the concerns noted earlier, by our colleague’s complaints about evidentiary weaknesses in this case. For instance, the dissent complains that the record contains no evidence of “usual” trading patterns during the delivery months, without which the Commission could not conclude that the trades in question here were unusual. Our colleague apparently requires as a remedy “statistical analysis” or “empirical inquiry.” Dissent at 942. But this is putting a burden on the Commission that exceeds what is routinely accepted in other areas of decision making. Witnesses are often (perhaps even usually) allowed to testify in comparative terms without establishing a baseline. As an oversimplified example, a police officer can testify that he was suspicious of a driver because he thought it unusual that a car was driving slowly and not using turn signals. The officer would be allowed to draw inferences from these facts without presenting evidence that cars usually drive faster on that particular street (much less evidence of the normal speed at which they drive). The factfinder could rely on its own experience to conclude that this sort of behavior was out of the ordinary. In the case before us, of course, the facts are more specialized, and so is the factfinder. We only ask whether the decision was non-arbitrary. Based on other like decisions, we conclude that it was.
In any event, the record is not wholly silent on the trading baseline for these delivery months. The CFTC did attempt a comparative analysis. The Commission reviewed petitioner Schaer’s freshening activities on a day when he was not charged with unlawful trading in order to glean characteristics of “normal” trading. The agency noted that Schaer had freshened about a million bushels of wheat through ten “round-turn trades” opposite ten other traders over several of the 15-minute time brackets into which the CBOT divides the trading day (ALJ Dec. at 9 n.13), all at slightly different prices. R. 175, at 7. Against this backdrop, the agency then [937]*937reviewed petitioner Elliott’s charged trades on the same day, during which he freshened 2.5 million bushels. Elliott sold the entire 2.5 million bushels to a co-petitioner, who sold the entire amount to a second co-petitioner, who then re-sold the amount to Elliott in two trades. Id. at 7. These transactions took place over three consecutive trade brackets—45 minutes— with no profit or loss for any participating trader. Id. The Commission concluded that the first incident of “innocent” freshening “differed markedly from the characteristics of the challenged trading.” Id. at 7, 14. So the Commission has invoked a baseline quite understandable even to inexpert judges. Our point is not that this baseline is an exemplar, but that the CFTC did not totally neglect a contextual analysis of the market as the dissent suggests.
Similarly, the Commission properly relied on the irregular audit trail to support a finding of pre-arrangement. The dissent says irregular trading cards prove nothing because traders frequently alter them to correct mistakes made during “fast-breaking events.” Dissent at 941. But the Commission did not draw the inference of pre-arrangement from the correction of mistakes made in the hurly-burly of trading, as the dissent suggests. Instead, the Commission was swayed because the traders recorded many of the purportedly random transactions on consecutive lines of the same card, which is consistent with the inference that the trades were pre-arranged. R. 175, at 6.
In the same vein, the dissent derides the Commission’s reliance on steady price action during the delivery months as evidence of pre-arrangement. Our dissenting colleague is, of course, loyal to the efficient market hypothesis, and insists that the non-volatility of prices can mean only an absence of new information about supply and demand. But if prices are set by prearrangement, would the efficient market hypothesis apply? And, in any event, is that hypothesis holy writ binding on the CFTC in these circumstances? Cf. Tennessee Gas Pipeline Co. v. FERC, 926 F.2d 1206 (D.C.Cir.1991) (Williams, J.) “If the [Federal Energy Regulatory] Commission proposes to reject ... the Efficient Market Hypothesis ..., we ... assume that it is free to do so.” Id. at 1211.
One additional point on the sufficiency of the evidence. The Commission was particularly impressed that the opening trades on most of the charged days were against the economic self-interest of the trader— he bought the spread when, to freshen, he would have wanted to sell. And the other traders passed up the opportunity to turn a quick profit at the opening trader’s expense. The dissent says this seemingly irrational behavior can easily be explained by modern game theory. See Dissent at 943-44. The traders were “repeat players,” who predictably forfeited a one-time profit in order to remain viable players in the pit. The logic of this argument is fine, but it fails for a number of reasons. First, it was not argued on appeal and would therefore be an improper basis on which to reverse the Commission. Second, the Commodity Exchange Act did not enact Dean Douglas Baird’s Game Theory and the Law. See Lochner v. New York, 198 U.S. 45, 75, 25 S.Ct. 539, 49 L.Ed. 937 (1905). (Holmes, J., dissenting). Third, if the traders’ strategic considerations were so obvious, we might be safe in believing that the Commission considered and rejected this explanation. Fourth, none of the traders (or other witnesses) testified that they passed up a quick profit in order to stay in good graces in the wheat pit. This last point is quite salient since, if offered, direct evidence of trading strategy would seem more probative than probabilities ascribed to game theory.
Our dissenting colleague also wonders how the CFTC could overturn the ALJ’s credibility determination without a good reason. See Dissent at 940-41. But there was a good reason. The ALJ identified two crucial factors purportedly supporting his finding of no pre-arrangement: first, the traders said they acted on publicly [938]*938available information, and, second, they said that they had no choice but to trade among themselves because no one else was interested in participating. The CFTC found that the evidence contradicted both of these factors. The publications on which the traders supposedly relied, the “CBOT Deliveries Last Trade Date Assigned Reports” and the “Issue and Stop Listing,” were “unavailable on at least some of the trading days at issue and on other relevant days contained only partial data.” R. 175, at 10. And on some of the charged days early in the delivery cycle, a substantial number of traders did participate. The CFTC explained its decision to disregard the ALJ as follows: “[w]here as here, the ALJ’s credibility findings are strongly contradicted by other compelling evidence, and the issue to be decided involves derivative inferences rather than testimonial inferences, the Commission will review de novo the inferences to be drawn from the record. See N.L.R.B. v. Stor-Rite Metal Products, Inc., 856 F.2d 957, 964 (7th Cir.1988).” R. 175, at 21.
The dissent takes the CFTC to task for ignoring Stoller, and chastises us for creating a conflict with it. This is puzzling, particularly because that case is not even cited in the petitioners’ main brief. Be that as it may, in Stoller, the CFTC, without any evidentiary hearing sanctioned a trader for making “wash sales” with intent to avoid a bona fide market transaction. 834 F.2d at 264. The Second Circuit held that, since the facts had never been found, the trader was entitled to a hearing to establish that he did not collude with others to wholly avoid market risk. See id. at 265. The court further held that without notice to the public through rulemak-ing, the Commission could not sanction as “wash sales” freshening operations that sought to minimize market risk but were not pre-arranged to wholly negate it. See id. at 267. The Second Circuit explained that the historical definition of “wash sales” concerned “transactions that were virtually risk-free, often prearranged, and intentionally designed to mislead.” Id. at 266. In the present case, the Commission is sanctioning these traders after finding as a fact after hearing that their conduct fits this historical definition; there therefore is no need for notice to the public or a new rule proclaiming a broader concept of “wash sales.” The issue here, unlike Stol-ler, is whether substantial evidence supports the Commission’s findings. There is no conflict with Stoller.
Finally, our colleague accuses us of a Chenery violation in “filling gaps” not addressed by the Commission. See Dissent at 944, citing SEC v. Chenery Corp., 318 U.S. 80, 88, 63 S.Ct. 454, 87 L.Ed. 626 (1943). Admittedly, the CFTC did not digress into such matters as game theory, antitrust dynamics or Stoller. Nor did we address or analyze these factors in earlier sections of this opinion. The present section merely responds to the arguments of the dissent. Since the dissent does not necessarily adhere to the main lines of the Commission opinion, this responsive section by way of rebuttal must dwell on some matters not found in the Commission opinion. Therefore, the decision of the Commission is
Affirmed.