Running it Twice (or Thrice): Double-Header, Triple-Header, and Reverse Baseball Arbitration

Michael J. Hasday*

This Essay illustrates how the “Running It Twice” concept that makes poker games less of a gamble can also be used in another forum where large amounts of money can be at stake: arbitrations. I introduce three new forms of arbitration based on this concept: Double-Header Baseball Arbitration, Triple-Header Baseball Arbitration, and Reverse Baseball Arbitration. In this Essay, I show that that these new forms of arbitration are superior to current methods because they result in what the average or median qualified arbitrator would award—thereby making arbitration more accurate, predictable, and fair.


Imagine you are on the game show, Heads or Tails. The rules of this game show are quite simple. The game show host flips a coin. Heads you get a million dollars. Tails you get nothing. Being a risk averse sort, you would probably want to sell your “spot” on this game show for something approaching $500,000, the expected value of your payout.[1] But, unfortunately, that is not permitted.

However, you are allowed an option that is almost as good: you can have the game show host flip the coin up to 100 times, with each “head” proportionately less rewarded. For example, you could have the host flip the coin twice, with the payout for heads set at $500,000; ten times, with the payout for heads set at $100,000; or 100 times, with the payout for heads set at $10,000 (the payout for tails for every version remains at $0.) Under each of these options, the expected value of the game remains the same for you: $500,000. Pursuant to the probability theorem known as the Law of Large Numbers,[2] however, the probability of your receiving a payout that approximates the $500,000 expected value of the game increases as the number of coin flips increases.[3] For example, you have a 0% probability of receiving a payout between $400,000 and $600,000 if you flip the coin just once, 50% if you flip the coin twice, 66% for 10 flips, and 96% for 100 flips.[4]

Although the game show Heads or Tails is fictional, the option of allowing multiple flips is analogous to a mechanism called “Running It Twice” that is sometimes used in high stakes cash poker games when tens, if not hundreds, of thousands of dollars are on the line based simply on the next one or two cards in the deck.[5] When there are two players left in the hand, the final card or cards have not yet been dealt, and at least one of the players is “all-in,” meaning there are no strategic moves left in the hand, some casinos allow the players to agree to run the remaining cards twice and award half the pot to the winner of each run. This can result in a split pot if each player wins one of the two runs.[6] The objective of the “Running It Twice” procedure is a payout that better reflects the true odds, or expected value, of the situation.[7]

This Essay illustrates how the “Running It Twice” concept that makes poker games less of a gamble can also be used in another forum where large amounts of money can be at stake: arbitrations. I introduce three new forms of arbitration based on this concept: Double-Header Baseball Arbitration, Triple-Header Baseball Arbitration, and Reverse Baseball Arbitration. In this Essay, I show that that these new forms of arbitration are superior to current methods because they result in what the average or median qualified arbitrator would award—thereby making arbitration more accurate, predictable, and fair.


Most arbitrations are either conventional or final offer[8]—the latter of which is also called “baseball arbitration” because Major League Baseball uses it.[9] Both types of arbitrations generally use either one arbitrator, or a panel of three arbitrators, to decide the case.[10] In conventional arbitration (CA), the arbitrator(s) can award whatever relief they believe is most accurate—regardless of what the parties have proposed.[11] In final offer arbitration (FOA), the arbitrator(s) have to select the proposal of one of the parties as the award.[12] The premise behind FOA is that this method results in the parties making more reasonable proposals, which then increases the chance of settlement.[13] FOA also offers the advantages of “speed, low cost and simplicity.”[14] FOA is often used when the ultimate award is a number, such as a salary or an amount of contractual damages.[15]

Both forms of arbitration have been criticized on accuracy grounds.[16] The common criticism of CA is that the arbitrators will attempt to “split the difference,” or come up with an award that is halfway between the parties’ proposals, regardless of the merits of the case.[17] Accordingly, where one party submits an extreme proposal and the other party offers a reasonable (or less extreme) proposal, the resulting “split the difference” award will be inaccurate in the direction of the party that submitted the extreme award.[18] Critics of FOA, on the other hand, say that it suffers from the opposite problem.[19] FOA creates an incentive for both parties to offer reasonable proposals, which means the accurate award is likely to be close to halfway between the parties’ proposals.[20] Nevertheless, FOA rules prohibit arbitrators from making a “split the difference” award, and therefore requires “the arbitrator [to] choose the lessor of two evils rather than what [is] actually fair.”[21]


This Essay’s first proposal is a variation of FOA,[22] which I call Double-Header Baseball Arbitration (DHBA). My proposal plays out like regular FOA, except that two different arbitrators—independent of each other—decide which of the parties’ proposals to award. If both agree on a proposal, then that is the award. If they disagree, then the award is the midway-point between the two parties’ proposals.

To see how my proposal would improve the accuracy of arbitration awards in FOA, let’s take a real-world example of the recent arbitration between the New York Yankees and their star reliever, Dellin Betances, in which Mr. Betances’ salary for the 2017 season was at stake. Mr. Betances asked for $5 million, while the Yankees countered with $3 million. Forced to choose one of these numbers, the panel of three arbitrators, deciding the matter collectively, sided with the Yankees, so Mr. Betances’ salary was set at $3 million.[23]

If the Yankees had used DHBA, they would have employed only two arbitrators and each of them would have reached her decision independently of the other. If the two arbitrators had both sided with the Yankees, the award would still have been $3 million, if both had sided with Mr. Betances, the award would have been $5 million, and if the arbitrators had split, the award would have been $4 million.

Although the implementation of DHBA is simple, its accuracy-improving effect is substantial. Let’s imagine that in the Betances-Yankees arbitration, for example, there are five approved arbitrators:[24] three of whom would decide in favor of the Yankees—let’s call them A, B, and C, and two of whom would decide in favor of Mr. Betances—let’s call them D and E. Thus, the average award of these five arbitrators is $3.8 million, i.e., the average of three awards of $3 million and two awards of $5 million. Accordingly, if one arbitrator decided this case, the average “error” under this process—defined as the absolute value of the difference between the single arbitrator award and the average award of $3.8 million—is a whopping $960,000.[25]

You can improve the accuracy only a bit by using a panel of three arbitrators instead of one arbitrator. There are now ten possible panels: ABC, ABD, ABE, ACD, ACE, ADE, BCD, BCE, BDE, and CDE. We can assume that all but three of these panels (ADE, BDE, and CDE) would decide in favor of the Yankees.[26] Thus, the average error for a three-arbitrator panel, deciding the matter collectively, is $920,000.[27]

Now compare what happens under DHBA. Here, there are again 10 possible panels: AB, AC, AD, AE, BC, BD, BE, CD, CE, and DE. We can assume that AB, AC, and BC would decide in favor of the Yankees, DE would decide in favor of Mr. Betances, and the remaining six panels would split. Thus, the average error under this process is $480,000,[28] which is exactly half of the average error when there is a single arbitrator deciding the case and a little more than half of the average error when there is a three-arbitrator panel deciding it collectively.

The accuracy-improving mechanism of “Running It Twice” can be incorporated into any FOA in which the dispute is limited to a number. Furthermore, in the case of Major League Baseball or any other arbitration using a panel of three arbitrators, DHBA will substantially reduce administrative costs because the parties will hire one fewer arbitrator.

Finally, DHBA lessens a major problem with FOA—that it favors the party with the deeper pockets and/or the party who is a repeat player, which each can submit an optimal proposal or number because it has the resources to accept greater risk or can spread the risk over numerous arbitrations.[29] Under FOA, the less-resourced or more risk averse party may submit a suboptimal proposal or number that increases its odds of winning the particular arbitration, but is a “loser” in the long run because when the less-resourced or more risk averse party wins, its “win” may be limited.[30]

To illustrate this concept, let us assume that a Major League Baseball player’s salary will be determined by one arbitrator out of a pool of three possible arbitrators, but the parties do not know which of the arbitrators will be assigned to the arbitration. Let us further assume that the player can accurately predict that the team will ask for $1 million, that two of the three arbitrators will determine that the player’s “true” value is $1.3 million, and that the third arbitrator (who has a reputation for favoring the players) will determine that the player’s “true” value is $2 million. In this case, assuming that numbers have to be submitted in increments of tenths of a million and that each arbitrator will pick the number that is closest to its determination of the player’s “true” value, the “optimal” number for the player to submit will be $2.9 million, even though the player will “lose” two of the three arbitrations. The three results of the arbitrations will be $1 million, $1 million, and $2.9 million, giving the player an average result of $1.63 million. Here, the “win” is essentially so profitable that it more than makes up for the two “losses.” But because the player has less resources than the team and is engaged in only one arbitration compared to the multiple arbitrations the team is engaged in, the player may prefer instead to “play it safe” and submit a number of $1.5 million. This offer will ensure victory in each arbitration and guarantee an award of $1.5 million, even though that is $130,000 less than the average result the optimal number will generate.

DHBA allows the less-resourced or more risk averse party to play the FOA game more optimally because it essentially spreads the risk over two arbitrations. Spreading the risk allows the less-resourced or more risk averse party to compete on a more level playing field with the party that has the deeper pockets and/or the party that is a repeat player. In the example above, for instance, let us assume all the same facts except that two of the arbitrators are now assigned to the case because the parites are using DHBA. If the player now submits the optimal number of $2.9 million, the resulting awards will be $1 million, $1.95 million, and $1.95 million, giving the player the same $1.63 million “average” award, but with less of a disparity among the awards. Because it is now less risky for the player to submit the optimal number, the player is more likely to do so, taking away a “profit center” that the team has under FOA. Indeed, it is for this very reason that some of the most well-resourced poker professionals turn down requests from their less-resourced peers to “Run It Twice”; they sense, correctly, that their ability to take on risk gives them an advantage that the “Running It Twice” procedure reduces.[31]

A critic of DHBA might argue that by enabling the less-resourced party to submit an optimal number, the parties will be less likely to settle than under regular FOA because the difference between the two parties’ numbers will now be larger. A critic might also argue that the parties will be less likely to settle because DHBA removes some of the uncertainty of the result that arbitrator inconsistency causes. The parties’ incentive to settle in DHBA should be the same as in FOA,[32] however, just at a number that is fairer to both sides. Even if DHBA decreases the settlement rate to some degree in practice, parties may believe that the increased benefits in accuracy and fairness are worth the trade-off.


As you may be thinking, you do not have to stop with running it just twice.[33] In Triple-Header Baseball Arbitration (THBA), three arbitrators each decides the case independently of the others. If all three agree on a number, then that is the award; but, if the arbitrators split 2-1, the award is set at the applicable two-thirds point between the parties’ proposals. Using the earlier Betances-Yankees example once more, we can assume that one of the panels (ABC) would rule unanimously in favor of the Yankees, six of the panels would rule 2-1 in favor of the Yankees (ABD, ABE, ACD, ACE, BCD, and BCE), and three panels would rule 2-1 in favor of Mr. Betances (ADE, BDE, and CDE). Thus, the average error is now reduced to $320,000,[34] which is exactly one-third of the average error when there is a single arbitrator deciding the case and a little more than one-third of the average error when there is a three-arbitrator panel deciding it collectively. Of course, the parties will have to decide for themselves whether the accuracy boost in “Running It Thrice” versus “Running It Twice” is worth the administrative cost of hiring the additional arbitrator.


The concept of “Running It Twice” can also be used in CA. If the possible award is limited to a number, the process can play out as it normally does except that two arbitrators will be deciding the case independently and the final award will be the mid-point of the two arbitrators’ tentative awards.

Furthermore, the concept can be used in CA even if the possible award is more complicated than just a number. The two arbitrators could, upon deciding the matter independently, conference to generate a compromise final award. But if that seems too arbitrary, there is another alternative available—what I call Reverse Baseball Arbitration (RBA). In this variation, each of the three arbitrators decide the case independently. After the three tentative awards are revealed, the parties simultaneously strike one of the tentative awards without knowing which tentative award the other party decided to strike, and the remaining tentative award becomes the final award.[35] Thus, instead of the arbitrators being required to choose one of the parties’ proposals (baseball arbitration), the parties will be required to select one of the arbitrators’ proposals (which is the reverse of baseball arbitration).

The essential benefit of RBA over CA with a three-arbitrator panel (3CA) is that it guarantees that the median award (or better) prevails.[36] In contrast, the unpredictable panel dynamics of 3CA can result in more extreme awards. To appreciate why, it is important to understand the two main ways that the three arbitrators are selected in 3CA: the party-appointed method and the strike-and-rank method. In the party-appointed method, both parties select one of the arbitrators who then mutually agree on the third arbitrator.[37] Under this selection method, 3CA has been known to lead to a de facto “final offer” situation, where the two party-appointed arbitrators remain inflexible with their polar opposite positions in favor of the party that appointed them, and the third arbitrator is forced to select one of these “positions” as the final award.[38]

In the strike-and-rank method, the arbitration tribunal supplies the parties with a list of possible arbitrators; the parties then separately strike the unacceptable arbitrators, rank the others, and the panel becomes the highest mutually ranked arbitrators on both parties’ lists.[39] For example, the well-known arbitration tribunal JAMS provides 10 names to each side, and the parties are each given three strikes, thus guaranteeing that there will be enough ranked/unstruck arbitrators left to form a three-arbitrator panel.[40] Because the strike-and-rank method selects all three arbitrators in a more neutral fashion, the concern here is group polarization rather than a de facto “final offer” situation.[41] Group polarization refers to the well-established phenomenon that “members of a deliberating group predictably move toward a more extreme point in the direction indicated by the members’ predeliberation tendencies.”[42] What this means in the context of 3CA is that if all three arbitrators even slightly favor Party A over Party B before deliberation, post-deliberation they may issue an award to Party A that is more extreme than any of the arbitrators would award to Party A individually. Phrased another way, “group polarization may result in group decisions that are more extreme than the median of the individual members’ views.”[43]

By eliminating the potential problems which group polarization poses, the strike-and-rank method is superior to the party-appointed method in terms of its stronger ability to bring about a median arbitration award. The following example makes this clear. Imagine that for a given case there is a universe of 10 qualified arbitrators—let us call them Arbitrator 1, Arbitrator 2 and so on until you get to Arbitrator 10. Arbitrator 1 is the most Party A-favorable, Arbitrator 2 is the second-most Party A-favorable and so on until you get to Arbitrator 10 who is least favorable to Party A. Assume that Arbitrators 1 to 7 will side with Party A and Arbitrators 8 to 10 will side with Party B. Under the party-appointed method, assuming the parties can accurately predict the arbitrator leanings, Party A will appoint Arbitrator 1 and Party B will appoint Arbitrator 10. There is nothing in the rules, however, that requires the mutually selected—and all-important—third arbitrator to be Arbitrator 5 or 6. Arbitrator 10 may simply “dig in” and refuse to agree to any arbitrator other than Arbitrator 8 or 9. Arbitrator 1 may then relent and accept Arbitrator 8 or 9 as the third arbitrator, or the party-appointed arbitrators may stalemate, which then allows the overseeing arbitration tribunal[44] or court[45] to select the third arbitrator, which could potentially be Arbitrator 8 or 9. Accordingly, the resulting three-arbitrator panel could very well consist of Arbitrator 1, Arbitrator 8, and Arbitrator 10, and the eventual winner could be Party B, although the median arbitrator would side with Party A.

Moreover, even if there is no bias in the selection of the third arbitrator—and Arbitrator 5 or 6 is selected—the panel dynamics could result in an award far from the median arbitrator award, but this time in the opposite direction (i.e., towards Party A). If a de facto “final offer” situation occurs, the final award could favor Party A, but in a more extreme fashion than what Arbitrator 5 or 6 would individually issue.

In contrast, under the JAMS strike-and-rank method, assuming the parties can accurately predict the arbitrators’ leanings, Party A will strike Arbitrators 8-10, Party B will strike Arbitrators 1-3, and the panel will be selected from Arbitrators 4 to 7—all of whom would side with Party A. Here, the panel’s award would be a median arbitration award if it reflected the average of the individual views of the panelists. But, taking into account group polarization, the three centrist arbitrators (who all agree that Party A should win) may convince each other that the award should be more extreme in favor of Party A than what any of them would individually issue.

Putting it all together, parties can obtain the best of both worlds by combining RBA with the strike-and-rank selection method. The panel would again be selected from Arbitrators 4 to 7. After the parties make their tentative strikes, however, the panel would make a final award in line with what Arbitrator 5 or 6 would issue, i.e., the median arbitrator award. It is possible that the parties could obtain the same result by selecting one arbitrator using the strike-and-rank method. But, the success of that scenario in producing a median arbitrator award is entirely dependent on the two parties being able to accurately strike-and-rank the arbitrators. Under this RBA and strike-and-rank selection method combination, the parties have more leeway to make mistakes in their strikes-and-ranks without jeopardizing the mutually desired end result of a median arbitrator award.

Finally, even apart from issues of panel dynamics, RBA is superior to 3CA because RBA’s party-strike mechanism optimizes the award by allowing the parties, instead of the arbitrators, to determine the median award. For example, imagine an arbitration in which the claimant is seeking both a monetary award and injunctive relief. In this arbitration, both parties may prefer Award A, in which the claimant receives the full monetary award requested but is denied injunctive relief, over Award B, in which the claimant is denied the monetary award but is granted the injunctive relief. However, while Award B will never prevail in RBA, 3CA may lead to Award B based simply upon the arbitrators’ mistaken collective belief that the parties prefer it to Award A.

To be sure, arbitrators (and their associations) might resist RBA, arguing that it will undermine the legitimacy of the arbitration system because it will show that the same case can lead to wildly different outcomes. But, considering that arbitration already allows for dissenting opinions,[46] it is unlikely that any arbitrator resistance on these grounds would overcome party demand for this procedure.


Current arbitration methods are sub-optimal because they produce awards that do not reflect what the average or median qualified arbitrator would issue. The methods proposed in this Essay can make arbitration less a roll of the dice. They also enhance the forum’s ability to produce accurate, predictable, and fair results.

Moreover, unlike proposals for reform of the court system, these procedures do not require parties to wait for any “official blessing” before employing them in their arbitrations. Because arbitration is famously a “creature of contract,”[47] parties can write these procedures into their agreements on an individual basis.[48] Indeed, the official rules of JAMS and the American Arbitration Association explicitly allow for party-agreed procedures in lieu of the tribunal’s official rules.[49] And if demand for these procedures is established, it is likely that some of the professional arbitration organizations (as competitors in the marketplace) will incorporate them as standard options that the parties may select.[50]

In short, parties can, and should, improve their arbitrations by “Running It Twice (or Thrice)”.

© 2018, Michael J. Hasday.

* Associate, Anderson & Ochs, LLP. Juris Doctor, Northwestern Pritzker School of Law, 2001. Thanks to Carol M. Hasday, Joni Hasday, Robert J. Hasday, and Hannah Rubashkin for their helpful comments. All views expressed, and any mistakes made, are my own. Correspondence may be sent to:

[1]. To calculate the expected value of a game, each possible outcome of the game is multiplied by its probability. The sum of all of these products is the expected value of the game.

[2]. See Scott Ryan, Wisdom of Crowds: Tests of the Theory of Collective Accuracy 12 (2014) (unpublished Ph.D. dissertation, Univ. of Conn.) (on file at Univ. of Conn. Library), (“The Law of Large Numbers states that as a sample size increases, the sample mean converges to the population mean.”).

[3]. A series of coin tosses is an example of a binomial experiment, which is defined as an experiment with repeated trials, each of which has only two possible outcomes, with a constant probability that a particular outcome will occur in any given trial, and with the outcome of each trial independent of the outcome in any other trial. See Statistics and Probability Dictionary, Stat Trek, (last visited Feb. 17, 2018).

[4]. The probabilities associated with the possible outcomes in a binomial experiment constitute a binomial distribution and can be determined using a binomial calculator. See Binomial Calculator: Online Statistical Table, Stat Trek, (last visited Feb. 17, 2018). For example, if the probability of success on a single trial is 50% and there are 100 trials, the calculator indicates that the probability of fewer than 40 successes is 0.0176. Since the distribution is symmetrical, the probability of more than 60 successes is also 0.0176, so the probability of between 40 and 60 successes is approximately 96% (1 – (0.0176 x 2)).

[5]. The “Running It Twice” procedure was used on the televised game show High Stakes Poker, which aired on the cable network, GSN, from 2006 to 2011. See Wikipedia, High Stakes Poker, Clips from the program using this procedure can be found on See, e.g., Negreanu Harman High Stakes Poker Season 4 Run It Twice, YouTube (Oct. 5, 2007),

[6]. See Poker Wiki, Running It Twice,; 2017 World Series of Poker Official Live-Action Rules 15-16,

[7]. See Poker Wiki, supra note 6 (“Rather than let a large pot depend on a single turn of cards, both players may decide to reduce their risk by running it twice, which makes the odds of winning get closer to the theoretically ‘correct’ odds.”); Dan Harrington & Bill Robertie, Harrington on Cash Games: Volume II: How to Play No-Limit Hold ’em Cash Games 333 (1st ed., 2010) (“Running it twice is a fair way of reducing the variance in the game. The expectations of the two players are unaffected by running the hands multiple times, but the swings are lessened because instead of a guaranteed result that one player will win the whole gigantic pot, there’s now a significant chance that the pot will be split.”).

[8]. See Daniel Mihai Nedelescu, Experimental Studies of Arbitration Mechanisms and Two-Sided Markets 11 (Fall 2013) (unpublished Ph.D. dissertation, Purdue Univ.),, (“When solving disputes, Conventional Arbitration (CA) and Final-Offer Arbitration (FOA) are the most commonly used methods of arbitration.”).

[9]. Josh Chetwynd, Play Ball? An Analysis of Final-Offer Arbitration, Its Use in Major League Baseball and Its Potential Applicability to European Football Wage and Transfer Disputes, 20 Marq. Sports L. Rev. 109, 110 (2009).

[10]. See American Arbitration Association, Commercial Arbitration Rules and Media-  tion Procedures, Rule 16(A) (Oct. 1, 2013), [hereinafter AAA Rules]; Chetwynd, supra note 9, at 115.

[11]. Benjamin A. Tulis, Final-Offer “Baseball” Arbitration: Contexts, Mechanics & Applications, 20 Seton Hall J. Sports & Ent. L. 85, 86 (2010).

[12]. Id.

[13]. Id. at 89, 90. Whether FOA, in fact, leads to more settlements is a matter of dispute. See David L. Dickinson, A Comparison of Conventional, Final-Offer, and “Combined” Arbitration for Dispute Resolution, 57 Indus. & Lab. Rel. Rev. 288, 292 (2004) (noting “the continued debate as to whether FOA will produce higher or lower dispute rates than CA”).

[14]. Joost Pauwekyn, Baseball Arbitration to Resolve International Law Disputes: Hit or Miss? 22 Fla. Tax. Rev. (forthcoming 2018) (manuscript at 14),

[15]. International Centre for Dispute Resolution, Final Offer Arbitration Supplement-ary Rules 7 (Jan. 1, 2015), [hereinafter Supplementary Rules] (“Each party’s final offer shall be a single monetary amount that includes all breaches, controversies and claims arising out of or relating to the contract . . .”); Pauwekyn, supra note 14, at 2–3. (“FOA works best in bilateral disputes over a number (e.g. salary figure or intra-company transfer price.”).

[16]. See Max H. Bazerman & Daniel Kahneman, How to Make the Other Side Play Fair, Harv. Bus. Rev., Sept. 2016, at 79–80.

[17]. See id.; Tulis, supra note 11, at 88; Chetwynd, supra note 9, at 111.

[18]. See Bazerman & Kahneman, supra note 16, at 79 (noting that under CA, “[t]he more unreasonable your offer is . . . the better you are likely to fare”).

[19]. See id. (noting the criticism that under FOA, the arbitrator “may [be] prevent[ed] from choosing a number he or she believes is truly fair”); Amy Lok, Final-Offer Arbitration, 10 ADR Bulletin 61, 62 (Mar. 1, 2008), (“The main weakness of final-offer arbitration is that the decision may not be the most equitable . . .”).

[20]. See Pauwekyn, supra note 14, at 6–7.

[21]. Baxerman & Kahneman, supra note 16, at 80.

[22]. A handful of variations to FOA have been previously proposed—such as Combined Arbitration, Double Offer Arbitration, Amended Final Offer Arbitration, and Alpha-Final Offer Arbitration—but none based on the “Running It Twice” concept. See Nedelescu, supra note 8, at 5–10 (providing a literature of review of previous proposals). The most prominent variation—which has actually been used in the real world as opposed to just in laboratory experiments—is Night Baseball Arbitration, which plays out like regular FOA, except that the arbitrator(s) issue their tentative award without knowing what the parties have proposed, with the final award being the party proposal that is closest to the tentative award. See Gerald Lebovits & Lucero Ramirez Hidalgo, Alternative Dispute Resolution in Real Estate Matters: The New York Experience, 11 Cardozo J. Conflict Resol. 437, 442 (2010).

[23]. See Mike Axisa, Yankees President Rips Dellin Betances After Winning Arbitration Hearing, CBS Sports (Feb. 18, 2017),

[24]. In reality, there is a pool of about 15 arbitrators that are mutually agreed upon by the league and the players’ union from which the panel of three arbitrators is chosen. See Chetwynd, supra note 9, at 126–27.

[25]. The average error is computed by determining the error of each possible arbitrator (or arbitrators) that could be assigned to hear the case, and then dividing the sum of these errors by the number of possible arbitrator configurations.  In this example, since Arbitrators A, B, and Cs error would be $800,000 ($3.8 million – $3 million), and Arbitrators D and E’s error would be $1.2 million ($5 million – $3.8 million), the average error is $960,000 ([(3($800,000) + 2($1.2 million)] / 5).

[26]. The assumption is that if two of the three arbitrators would decide one way, the panel collectively would decide that way.

[27]. Since the error of each of the seven Yankees panels is $800,000 ($3.8 million – $3 million) and the error of each of the three Betances panels is $1.2 million ($5 million – $3.8 million), the average error is $920,000 ([(7($800,000) + 3($1.2 million)] / 10).

[28]. Since the error of each of the three Yankees panels is $800,000 ($3.8 million – $3 million), the error of the one Betances panel is $1.2 million ($5 million – $3.8 million), and the error of each of the six split panels is $200,000 ($4 million – $3.8 million), the average error is $480,000 ([3($800,000) + 1($1.2 million) + 6($200,000)] / 10).

[29]. See Pauwekyn, supra note 14, at 17.

[30]. Id.

[31]. See Dave Woods, High Stakes Heroes #1: Barry Greenstein – Running It Once and The Truth Behind that Pot with Tom Dwan, Poker Player Magazine (Mar. 10, 2014),

[32]. See Cary A. Deck, Amy Farmer & Dao-Zhi Zeng, Amended Final-Offer Arbitration is Improved Arbitration: Evidence from the Laboratory 3 (May 20, 2005), (noting that “based on pure theory parties should settle when information is symmetric regardless of the convergence or divergence of offers”).

[33]. Hands are sometimes run three times in high stakes poker games. See, e.g., gtdtours Poker Moments – Patrick Antonius vs. Jamie Gold – Huge Pot on High Stakes Poker, YOUTUBE (Dec. 4, 2009),

[34]. Since the error of the one unanimous Yankee panel is $800,000 ($3.8 million – $3 million), the error of each of the six majority Yankees panels is $133,333 ($3.8 million – $3,666,667), and the error of each of the three majority Betances panels is $533,333 ($4,333,333 – $3.8 million), the average error is $320,000 ([1($800,000) + 6($133,333) + 3($533,333)] / 10).

[35]. In the (presumably) rare case that both parties select the same tentative award to strike, that tentative award would be stricken and there would be a second round of strikes. If both parties select a different tentative award to strike in the second round, those strikes would be nullified, and the final award would be selected from the two remaining tentative awards by the arbitrator whose tentative award was stricken in the first round.

[36]. The average award is the appropriate benchmark for DHBA and THBA because under FOA, an individual arbitrator can only issue one of two possible awards. For RBA, the appropriate benchmark is the median award because arbitrators in CA have complete discretion in the awards they issue, so an extreme award can skew the “average.” It should be noted that the theoretical justification for the accuracy of the median award is the Condorcet Jury Theorem rather than the Law of Large Numbers. See Cass R. Sunstein, Group Judgments: Deliberation, Statistical Means, and Information Markets, 80 N.Y.U. L. Rev. 962, 974 (2005).

[37]. See Alan Scott Rau, Integrity in Private Judging, 38 S. Tex. L. Rev. 485, 497–98 (1997).

[38]. See id. at 502 n.65.

[39]. See Angela Zambrano & Robert Velevis, When ‘I Pick, You Pick, They Pick’ Goes Wrong, Sidley Austin LLP 3-4 (June 12, 2017, 12:25 PM EDT),

[40]. Comprehensive Arbitration Rules and Procedures, Rule 15, JAMS (July 1, 2014), [hereinafter JAMS RULES] (last visited July 29, 2018).

[41]. See Catherine A. Rogers, A Window into the Soul of International Arbitration: Arbitrator Selection, Transparency, and Stakeholder Interests, 46 Victoria U. Wellington L. Rev. 1179 (2015) (arguing that the party-appointed method should be maintained because other methods are susceptible to group polarization).

[42]. Cass R. Sunstein, Deliberative Trouble? Why Groups Go to Extremes, 110 Yale L.J. 71, 74 (2000).

[43]. Christopher R. Drahoz, A Behavorial Analysis of Private Judging, 67 L. & Contemp. Probs. 105, 125 (2004).

[44]. See, e.g., AAA Rules, supra note 10, at Rule 14(a)&(b).

[45]. See Zambrano & Velevis, supra note 39 (describing a case where the plaintiff prevailed in an arbitration, two to one, after its party-appointed arbitrator rejected all of the arbitrators that the other party-appointed arbitrator proposed, which included five former judges, and the third arbitrator was selected by the court).

[46]. See Pedro J. Martinez-Fraga & Harout Jack Samra, A Defense of Dissents in Investment Arbitration, 43 U. Miami Inter-Am. L. Rev. 445, 458 (2012) (noting “the widespread acceptance of dissents in arbitration, both in the rules and practice of numerous arbitration institutions”).

[47]. See Edward Brunet, The Minimal Role of Federalism and State Law in Arbitration, 8 Nev. L.J. 326, 326, n.2 (2007) (noting the hundreds of “state and federal cases using the phrase ‘arbitration is a creature of contract’”).

[48]. See Edward Brunet, Replacing Folklore Arbitration with a Contract Model of Arbitration, 74 Tul. L. Rev. 39, 47–50 (1999) (finding that under U.S. Supreme Court decisions, “the content of arbitration is dictated by the parties through their contract and not by any generic definition or arbitration procedure”).

[49]. See JAMS RULES, supra note 40, at Rule 2(A) (“The Parties may agree on any procedures not specified herein or in lieu of these Rules that are consistent with the applicable law and JAMS policies . . .”); AAA Rules, supra note 10, at Rule 1(a) (“The parties, by written agreement, may vary the procedures set forth in these rules.”).

[50]. See Brunet, supra note 48, at 41 (“Different arbitration providers compete vigorously for the right to supply arbitration services.”); Supplementary Rules, supra note 15, at 3 (“In response to feedback from the international and domestic business community, the International Centre for Dispute Resolution (ICDR) and the American Arbitration Association (AAA) have created a set of supplementary rules called Final Offer Arbitration Supplementary Rules.”).