Who knew that simple math could be so tricky? The parties to this dispute, the State of California and two California Indian tribes, signed Gaming Compacts intend ed “to initiate a new era of tribal-state cooperation” with respect to gaming in the state. Central to the Compacts is a formula to calculate the number of gaming devices California tribes are permitted to license. How to interpret this opaquely drafted and convoluted formula has preoccupied the parties for some time, as the result has significant economic implications. Indeed, math and money have led to a breakdown in the cooperative spirit envisioned by the Compacts.

The Compacts stem from the Indian Gaming Regulatory Act (IGRA), passed by Congress in 1988 and designed “to provide a statutory basis for the operation of gaming by Indian tribes as a means of promoting tribal economic development, self-sufficiency, and strong tribal governments.” 25 U.S.C. § 2702 (1). IGRA sets out three classes of lawful gaming; at issue here are slot machines and other gaming machines (“gaming devices”) that are included under Class III. 25 U.S.C. § 2703 (7)(B)(ii), (8). A tribal-state gaming agreement, known as a “compact,” is required to conduct Class III gaming under IGRA. 25 U.S.C. § 2710 (d)(1)(C).

In 1999, California and approximately 60 California Indian tribes signed substantively identical bilateral Gaming Compacts that authorized Class III gaming. The total number of slot machines allowed was restricted by contract language that authorized the continued operation of existing machines, permitted tribes who were not yet operating machines to operate up to 350 machines, and provided a formula for a limited license pool for the remaining machines. The primary issue in this appeal is the interpretation of the formula for the license pool, a mere two paragraphs in a 215-paragraph agreement. Unfortunately, these provisions are not a model of clarity. As a consequence, California and certain tribes have been mired in disputes for much of the period since the bilateral Compacts were signed.

This appeal springs from a disagreement between California and plaintiff Cachil Dehe Band of Wintun Indians of the Colusa Indian Community and plaintiffintervenor Picayune Ranchería of the Chukchansi Indians (“Picayune”) (collectively, “Colusa”). The parties agree that the formula sets a ceiling on the number of licenses in the pool. But the tribes claim the formula permits more licenses, while California maintains that it sets a lower limit. Acknowledging that the formula language is ambiguous, California and Colusa each offered different interpretations. The district court adopted yet a different formulation, introduced by Colusa as an alternative way to calculate the license pool. The parties also submitted extrinsic evidence purporting to explain their calculations.

Such a posture would normally suggest that summary judgment is inappropriate, even though contract interpretation is a matter of law. Nonetheless, both parties agreed that the matter should be decided on cross motions for summary judgment. An additional twist is that the parties’ extrinsic evidence does not support their interpretations of the formula. As a result, we interpret the Compact de novo based on the plain meaning that adheres closest to the contract language.

We affirm in part the grant of summary judgment to Colusa because we agree that the limit on licenses exceeds that recognized by California. However, our interpretation of the governing provisions differs slightly from the district court’s formulation. We also affirm the denial of California’s motion for summary judgment. Finally, we uphold the remedy ordered by the district court of a license draw open to all eligible tribes, administered according to the process delineated in the Compacts. Before we wade into the somewhat mind-numbing discussion of numbers, it is useful to provide a background context for the formula.

Background

I. Compacts under IGRA

Following a successful ballot initiative permitting California Indian tribes to run “Nevada and New Jersey”-type casinos, and in response to the likely imminent invalidation of that initiative, then-Governor Gray Davis invited California tribes to negotiate Class III gaming compacts. By that time — April 1999 — a number of California tribes were already operating gaming devices, although without authorization under IGRA. These tribes operated around 19,000 devices state-wide. In late August 1999, the California Supreme Court invalidated the ballot initiative permitting casino operation by Indian tribes. See Hotel Employees & Rest. Employees Int’l Union v. Davis, 21 Cal.4th 585 , 88 Cal.Rptr.2d 56 , 981 P.2d 990 (1999). California and the tribes, including Colusa, continued negotiating, however, intending to condition execution of the Compacts on the ratification of a constitutional amendment that would exempt Indian tribes from the prohibition on Class III gaming.

The final Compact negotiation sessions were held on September 8 and 9, 1999, and continued into the early hours of September 10. Late on September 9, the lead negotiator for California presented the entire draft Compact to the tribal representatives for approval. The representatives were given until September 10 to sign letters of intent to enter into bilateral Compacts with California. The Compacts required legislative ratification, and the end of the legislative session was fast approaching. Colusa’s Chairman signed the tribe’s letter of intent in the early hours of September 10. In total, about 60 tribes (the “Compact Tribes”), including Colusa and Picayune, entered into bilateral Class III gaming Compacts with California. These Compacts are substantially identical. See Artichoke Joe’s Cal. Grand Casino v. Norton, 353 F.3d 712 , 717-18 (9th Cir.2003).

The California legislature ratified the agreements in September 1999. On the same day, the Governor’s Press Office released an “information sheet” regarding the number of licenses available, stating that the Compacts authorized 44,448 gaming devices total statewide, including those already in operation. California voters ratified the constitutional amendment in March 2000, enabling the Compacts to be executed. Colusa and Picayune’s Compacts went into effect on May 16, 2000.

While the Colusa Compact includes a variety of provisions relating to the operation and licensing of Class III gaming devices, the only provisions at issue in this appeal relate to the aggregate number of gaming devices authorized statewide in addition to those already in operation as of September 1, 1999, i.e., the size of the “license pool.” The Compact provides a formula for determining that number, at § 4.3.2.2(a)(1):

The Compact defines Non-Compact Tribes as federally-recognized tribes that are operating fewer than 350 gaming devices, whether or not the tribe executed a Compact with the State. § 4.3.2(a)(1). In other words, some tribes are both Compact and Non-Compact Tribes under the agreement. Section 4.3.1, which is referenced by § 4.3.2.2(a)(1), states:

For convenience, we refer to §§ 4.3.1 and 4.3.2.2(a)(1) as the “License Pool Provisions.”

A Compact Tribe may operate a certain number of gaming devices without securing licenses from the pool. This initial “free pass” covers either 350 devices or the number of devices the tribe was already operating as of September 1, 1999, whichever is larger. § 4.3.1. Colusa operated 523 gaming devices as of September 1, 1999, so it was permitted to continue operating all of those devices without licenses. A Compact Tribe must secure a license from the statewide pool for each additional device above the Tribe’s “free pass” number, up to the maximum of 2000 devices per tribe. § 4.3.2.2(a).

The licenses are allocated from the license pool to Compact Tribes that request them according to a detailed draw process. § 4.3.2.2(a)(3). The draw process, which includes tiers of priority for different tribes, was designed to skew the distribution of the available licenses towards those Compact Tribes that did not yet conduct large gaming operations. § 4.3.2.2(a)(3).

II. Administration of the License Process

For the first two years, the license draw process was administered by an accounting agency engaged by the Compact Tribes, Sides Accountancy Corporation (“Sides”). In 2001, following complaints regarding the accounting and administration of the draws and Sides’s unwillingness to provide certain information to California, Governor Davis issued an executive order creating the California Gambling Control Commission (“the Commission”), which took over the licensing process.

During its administration, Sides issued 29,398 licenses, exceeding by about 25% the 23,450 number that would have been available in the pool according to the Governor’s summary “information sheet,” which limited the total devices to 44,448 statewide, including those already in operation. After taking control, in 2002 the Commission evaluated the various interpretations of the License Pool Provisions that had been suggested and adopted an interpretation that authorized a license pool of 32,151 licenses. This number surpassed the number of licenses that Sides previously issued, allowing all of those licenses to remain valid. According to the Commission — and California, which has adopted the Commission’s formulation— 2753 licenses remained in the license pool for distribution after the Commission took charge.

The Commission notified the Compact Tribes that it would conduct a draw in September 2002. Colusa was placed in the third priority tier for the first draw, in which it requested and received 250 licenses. For the second draw in December 2003, Colusa was placed in the fourth priority tier. Although Colusa requested 377 licenses, it received none. In October 2004, the Commission conducted a third draw, and Colusa requested 341 licenses. Colusa was again placed in the fourth priority tier, and received only 73 licenses. From the three draws, Colusa secured 323 licenses in total; when added to the 523 devices in operation, the licenses drawn from the pool allowed Colusa to operate 846 devices, well under the individual limit of 2000 devices per Compact Tribe.

III. The Colusa Lawsuits

Immediately following the December 2003 draw, Colusa requested that California meet and confer regarding, among other issues, the size of the license pool and Colusa’s placement within the lower priority tier for the 2003 draw. Following an unsuccessful meeting, Colusa initiated suit in October 2004.

Colusa’s initial complaint alleged five claims for breach of the Compact. Only one claim — California’s unilateral determination of the aggregate number of licenses authorized by the Compact under the License Pool Provisions — is at issue here; in its complaint, Colusa sought a declaration that the license pool consisted of “more than 62,000 Gaming Device licenses,” and requested that the court order California to immediately issue 377 licenses to Colusa.

The district court dismissed four of the claims on the ground that Colusa was required to join other Compact Tribes as necessary parties under Federal Rule of Civil Procedure 19; the fifth claim was dismissed for failure to exhaust remedies. On appeal, we reversed and held that the non-party Compact Tribes did not have a protectable legal interest in the size of the license pool and were not required parties that must be joined under Rule 19. Colusa I, 547 F.3d at 972. We affirmed the dismissal of the unexhausted claim. Id. at 968 n. 3.

While Colusa I was pending, California negotiated Compact amendments with at least five tribes. Neither Colusa nor Picayune entered an Amended Compact, although Colusa negotiated regarding a potential amendment. The Amended Compacts provided up to 22,500 additional gaming devices outside of the aggregate limits established by the original Compacts.

In June 2007, also during the time Colusa I was pending, Colusa filed a second suit, alleging California breached the Compact by refusing to conduct another round of draws, miscounting multi-station games as equal to the number of terminals, and failing to negotiate in good faith. The district court consolidated Colusa’s two actions. In January 2009, Picayune intervened, alleging the Commission breached the Compact by miscalculating the size of the license pool.

The parties cross-moved for summary judgment on the issue of the size of the license pool. On April 22, 2009, the district court granted summary judgment to Colusa on the aggregate number of gaming devices authorized by the Compact and on Colusa’s placement within the priority tiers. The court entered final judgment on all claims on August 19, 2009, and ordered California to conduct a draw of the remaining licenses in the pool that would be open to all eligible Compact Tribes. California’s request to stay the order for thirty days was denied. California timely appealed and then moved for a stay of the district court’s remedy order. Following denial of the stay motion, in October 2009, California conducted a license draw open to all eligible Compact Tribes. In that draw, 1878 licenses were issued to ten Compact Tribes, including Colusa and Picayune.

Analysis

General principles of federal contract law govern the Compacts, which were entered pursuant to IGRA. Kennewick Irrigation Dist. v. United States, 880 F.2d 1018 , 1032 (9th Cir.1989). In practical terms, we rely on California contract law and Ninth Circuit decisions interpreting California law because we “discern, and the parties note, no difference between [California] and federal contract law.” Idaho v. Shoshone-Bannock Tribes, 465 F.3d 1095, 1098 (9th Cir.2006) (employing Idaho contract law to interpret a tribal-state compact that was to be “construed in accordance with the laws of the United States”).

The California Court of Appeal recently reviewed the court’s role in interpreting contracts, according to California law:

Wolf v. Walt Disney Pictures and Television, 162 Cal.App.4th 1107, 1125-26 , 76 Cal.Rptr.3d 585 (Cal.Ct.App.2008).

I. Interpretations of License Pool Provisions Offered by Parties

The License Pool Provisions are repeated here, to facilitate their explanation. The calculations fall naturally into two steps:

§ 4.3.2.2(a)(1).

Section 4.3.1 provides:

§ 4.3.1.

A. The Parties’ Original Interpretations

From the beginning of this litigation, Colusa and California have agreed on the operation of Step 1 of the § 4.3.2.2(a)(1) sum. The Compact defines “Non-Compact Tribes” as federally-recognized tribes that were operating fewer than 350 gaming devices as of September 1, 1999, whether or not they ultimately entered Compacts. § 4.3.2(a)(i). There are 84 such tribes. Step 1 is therefore:

The parties vigorously dispute the operation of Step 2. That step is defined by the following formula:

In the original summary judgment submissions, California and Colusa champí oned two competing interpretations of Step 2. The parties agreed, however, on these important predicates:

The value of X — the number of tribes that should be multiplied by 350 in order to aggregate § 4.3.1(b) — -is where California and Colusa parted ways. California claimed X should be 16, because 16 tribes operated fewer than 350 gaming devices, but more than zero, as of September 1, 1999. It argued that a tribe cannot be “authorized” to operate zero gaming machines, and so the tribes operating zero machines should not be counted. California maintained the license pool therefore contained:

Colusa disagreed, and contended in its original summary judgment filing that X equaled 84, because 84 reflected those California tribes that operated fewer than 350 devices on September 1, 1999, including those operating zero. Colusa claimed that the plain meaning of “operate” under § 4.3.1 would cover those tribes operating no devices, as well as those operating between zero and 350 devices. According to Colusa, the license pool contained:

B. The District Court’s Adoption of Colusa’s “Alternative” Formulation

In a footnote in its original summary judgment filing, and expanded on in its reply brief, at argument, and through supplemental briefing, Colusa introduced another interpretation of the License Pool Provisions, which the district court termed the “alternative formulation.” The alternative formulation employs the same interpretation of Step 1. But Step 2 identifies whether each Compact Tribe, treated as an individual Compact Tribe, would have the limit on permissible unlicensed devices set by § 4.3.1(a) or set by § 4.3.1(b). Aggregate numbers are separately calculated for §§ 4.3.1(a) and 4.3.1(b) based on those individual limits. The lesser of these two aggregate numbers is then employed in Step 2 as the “lesser number authorized under Section 4.3.1.”

Applied, the alternative formulation under Step 2 proceeds as follows.

To begin, the district court acknowledged that “[t]he parties do not dispute that the meaning of [§ ] 4.3.2.2(a) is unclear and susceptible to varying interpretations.” Before interpreting the License Pool Provisions as a matter of law, the court concluded that the parties’ extrinsic evidence shed no light on their mutual intention at the time of contracting. The court was also influenced by its observation that the original formulations of both parties forced a “strained reading of the Compact language.” In interpreting an ambiguous term as a matter of law, the court stated its obligation as “providing] an interpretation that will make an agreement lawful, operative, definite, reasonable, and capable of being carried into effect.” Badie v. Bank of Am., 67 Cal.App.4th 779, 800 , 79 Cal.Rptr.2d 273 (Cal.Ct.App.1998). The court then concluded the alternative formulation reflected these necessary qualities, and that of the three interpretations offered “the alternative formulation most accurately follows the language of § 4.3.2.2(a)(1), giving the words their ordinary meaning.” The conclusion that 42,700 licenses are available in the pool provided the court’s basis for ordering a new license draw, which took place in October 2009.

II. De Novo Review of the License Pool Provisions

On de novo review, we agree with the district court that the License Pool Provisions are ambiguous and reasonably susceptible to more than one interpretation. U.S. Cellular Inv. Co. v. GTE Mobilnet, Inc., 281 F.3d 929 , 933-934 (9th Cir.2002) (“We also review de novo the determinations of whether contract language is ambiguous, Tyler v. Cuomo, 236 F.3d 1124, 1134 (9th Cir.2000), and “[w]hether the written contract is reasonably susceptible of a proffered meaning.” Brindersom-Newberg Joint Venture v. Pac. Erectors, 971 F.2d 272 , 277 (9th Cir.1992)).” That the parties and the district court each, in good faith, divine multiple results from the same formula underscores this ambiguity. For example, the opaque language of the provisions permits more than one interpretation of how the word “authorized” should be understood when interpreting § 4.3.1 in the context of the § 4.3.2.2(a)(1) sum. The fact that the term “authorized” is not used in the text of § 4.3.1 only exacerbates the ambiguity.

Given this ambiguity, we are permitted to consider extrinsic evidence when interpreting the Compact as a matter of law if the language of the provisions is reasonably susceptible to the interpretation of the party proffering the evidence. See In re Bennett, 298 F.3d 1059 , 1064 (9th Cir.2002). We use a two-step process to determine whether extrinsic evidence passes the reasonable susceptibility barrier. The district court summarized the process in its summary judgment order:

Athough we ultimately decline to consider the extrinsic evidence, in light of the parties’ extensive briefing on the issue, we first address our rationale on this point before turning to interpretation of the Compact.

A. The Parties’ Extrinsic Evidence

California offers extrinsic evidence to support its view that the parties’ mutual intention was the Compact would authorize no more than approximately 23,450 devices in the license pool, because the parties agreed on a statewide cap of approximately 44,500 devices total. The difficulty with this evidence, as we explain below, is that the foundation for this 23,450 number is at odds with the plain language of the contract and with an interpretation of part of the formula that is now agreed upon by both parties.

Colusa offers extrinsic evidence to counter California’s position and demonstrate that the Compact Tribes never intended that the license pool would be limited to 23,450 based on a statewide cap of around 44,500. Colusa claims that the Tribes originally believed the license pool would, in fact, be around 56,000 devices, not including the devices already in use.

B. Use of the Parties’ Extrinsic Evidence

The district court admitted the parties’ extrinsic evidence, but then concluded that the parties’ submissions were ultimately to no avail, because “the circumstances under which the Compact was entered into do not aid[ ] the court in discerning the parties’ intentions” as “the submissions of the parties reveal that there was no clear consensus between the parties regarding the maximum number of Gaming Devices allowed under the Compact at the time the agreements were executed.”

California contends on appeal that the district court erred in admitting the extrinsic evidence because the credibility of the conflicting evidence should have been tested by the jury. According to California, the district court erred by skipping that step and interpreting the contract as a matter of law.

California is correct that when there is a material conflict in extrinsic evidence supporting competing interpretations of ambiguous contract language the court may not use the evidence to interpret the contract as a matter of law, but must instead render the evidence to the factfinder for evaluation of its credibility. See City of Hope Nat’l Med. Ctr. v. Genentech, Inc., 43 Cal.4th 375 , 75 Cal.Rptr.3d 333 , 181 P.3d 142, 156-57 (2008) (“Interpretation of a written instrument becomes solely a judicial function only when it is based on the words of the instrument alone, when there is no conflict in the extrinsic evidence, or a determination was made based on incompetent evidence.”).

That general principle comes with a key preliminary caveat: The extrinsic evidence must qualify for admission. In this case, we never reach the question of what happens when the court is faced with conflicting admissible extrinsic evidence, because neither California’s nor Colusa’s evidence qualifies for admission in the first place. If credited, California’s evidence, drawn heavily from the “information sheet,” would show that the parties agreed upon a statewide cap of around 44,500 devices, which would mean the license pool should not exceed around 23,450 licenses. Not surprisingly, California’s briefing glosses over the actual operation of the formula to reach 23,450 in the license pool and emphasizes instead the ultimate conclusion of the total cap on devices. However, a close read of the extrinsic evidence shows that California’s rationale underlying its original interpretation does not give effect to the definitions in the Compact or to all of the language of the formula. The language of the formula simply is not susceptible to the interpretation suggested by California. The evidence therefore fails at the provisional step of the admission procedure. “The test of whether parol evidence is admissible to construe an ambiguity is not whether the language appears to the court to be unambiguous, but whether the evidence presented is relevant to prove a meaning to which the language is ‘reasonably susceptible.’ ” Winet v. Price, 4 Cal.App.4th at 1165 , 6 Cal.Rptr.2d 554 (internal citation omitted). California’s extrinsic evidence is not relevant because the language of the License Pool Provisions is not reasonably susceptible to the interpretation California proposes.

In briefing before the district court, California proposed a formulation, different from its original interpretation, one that would permit 32,151 licenses in the pool and around 53,000 devices statewide. California’s current formulation is the same one adopted by the Commission, which selected the formulation as the most reasonable of the interpretations suggested to that point. California acknowledges that there is a gap between the 23,500 limit on the license pool, allegedly reflecting the parties’ original intention, and its current conclusion of 32,151 licenses, although it suggests the numbers are close in theory. California states that the reason for the mismatch is that, after taking control, the Commission made a good faith effort to interpret the Compact in a way to accommodate the excess licenses issued by Sides. Unlike the district court, we acknowledge that California’s later effort through the Commission to accommodate the licenses issued by Sides is not the benchmark for the state’s original intention under the Compact. However, that acknowledgment does not provide an alter native entry for California’s extrinsic evidence. California continues to urge an application of the Compact that results in a license pool of 32,151 devices, but is not able to demonstrate a connection between its extrinsic evidence regarding original intention and that interpretation. Further, the language of the License Pool Provisions is not reasonably susceptible to an interpretation that would produce a license pool of 32,151 devices.

Colusa’s extrinsic evidence suffers from the same infirmity as that of California. The tribe’s evidence tends to show that the Compact Tribes believed the license pool included around 56,000 licenses. Colusa offers no connection between this extrinsic evidence regarding its initial intentions and the alternative formulation it now supports, which concludes that only 42,700 licenses are available in the pool. Like California’s proffer, Colusa’s evidence may not be admitted to construe the License Pool Provisions.

“It is [ ] solely a judicial function to interpret a written instrument unless the interpretation turns upon the credibility of the extrinsic evidence.” Parsons v. Bristol Dev. Co., 62 Cal.2d 861 , 44 Cal.Rptr. 767 , 402 P.2d 839, 842 (1965). California and Colusa, however, have not provided credible evidentiary support for their interpretations of the disputed provisions. “Even if we indulge in every reasonable inference that can be drawn from this evidence, the ... ] issue [of the interpretation of the License Pool Provisions] cannot be reasonably construed as turning on the credibility of such insubstantial evidence.” New Haven Unified Sch. Dist. v. Taco Bell Corp., 24 Cal.App.4th 1473, 1483 , 30 Cal.Rptr.2d 469 (Cal.Ct.App.1994) (concluding that, although defendant introduced extrinsic evidence, it was “of marginal relevance,” and therefore the appellate court should retain interpretation as a judicial function).

In Nat'l Union Fire Ins. Co. of Pittsburgh, Pa. v. Argonaut Ins. Co., 701 F.2d 95, 97 (9th Cir.1983), we concluded that summary judgment on an ambiguous insurance contract claim was permissible under California law when there was a lack of evidentiary support for competing interpretations of the contract language. That reasoning applies equally here, when considering California’s argument:

Id. at 97 .

Because the extrinsic evidence submitted by both parties is inadmissible, no genuine issue of material fact precluded the district court from interpreting the ambiguous License Pool Provisions as a matter of law or precludes us from doing the same. We turn to that challenge now.

C. Interpretation of the License Pool Provisions

We have the task of interpreting the ambiguous License Pool Provisions as a matter of law without the help of extrinsic evidence to support the proposed formulations. But we are not left to do so on a blank slate. The language of the License Pool Provisions — coupled with the backdrop of the Compact, common sense, and an eye toward mathematical consistency— permit us to interpret the Provisions in a manner that gives meaning to all of the terms. The district court concluded that neither of the parties’ offerings provided an acceptable interpretation of the License Pool Provisions. We agree that those interpretations are not reasonable. For example, the original interpretations appear to ignore the language of Step 2 of the § 4.3.2.2(a)(1) sum that requires finding the “difference between 350 and [the lesser number authorized under § 4.3.1].”

Nor does the alternative formulation, adopted by the district court, provide a viable choice. The alternative approach treats “the difference between 350 and” language of Step 2 as an afterthought in the aggregation process, to be subtracted from whichever aggregate number is deemed “the lesser number authorized under Section 4.3.1.” Yet it is not a natural reading of § 4.3.2.2(a)(1) that only half of Step 2 was designed to aggregate, and the other half was intended merely to make a minor adjustment. A more reasonable reading of Step 2, one that would give meaning to the “difference between 350 and” language, includes both elements in the aggregation process.

Concluding the alternative formulation is unsatisfactory, we turn a fresh eye to the meaning of the § 4.3.2.2(a)(1) sum. To reiterate, the § 4.3.2.2(a)(1) sum sets the “number of machines that all Compact Tribes in the aggregate may license pursuant to this Section.” In other words, the sum calculates the size of the license pool.

Because Step 1 is undisputed, we turn to Step 2, which requires calculating “the difference between 350 and the lesser number authorized under Section 4.3.1.” A reasonable reading of Step 2 within § 4.3.2.2(a)(1) is that it contemplates application of Step 2 to all Compact Tribes in order to calculate how many machines “all Compact Tribes in the aggregate may licence.” If all of Step 2 is not applied, then part of the Step 2 language is left hanging as surplusage or an afterthought. If Step 2 is not applied to each Compact Tribe, then the aggregation is incomplete. Therefore, to give Step 2 meaning and consistent application, it makes sense to apply all of Step 2 to each Compact Tribe in order to create an aggregate number.

Standing on its own, § 4.3.1 is written so as to apply to a single Compact Tribe, and sets the number of devices the Tribe may operate without securing licences. It states a tribe “may operate no more Gaming Devices than the larger of the following [two options].” Accordingly, § 4.3.1 calculates two numbers for each Compact Tribe. The Tribe may operate either the smaller or the larger number of devices calculated in §§ 4.3.1(a) and 4.3.1(b) without securing licenses. Stated another way, the Tribe is “authorized under Section 4.3.1” to operate either number of devices. Of course, within the context of setting individual limits, the larger of these two numbers would be the pertinent number for the individual Tribe, as that sets the limit of “free pass” devices.

Step 2 of the § 4.3.2.2(a)(1) sum references the same numbers calculated under § 4.3.1 on its own. However, under Step 2 the key number is the smaller number calculated for each Tribe, “the lesser number authorized under Section 4.3.1.” § 4.3.2.2(a)(1) (emphasis added). Step 2 (“the difference between 350 and the lesser number authorized under Section 4.3.1”) is accomplished by individually identifying the two numbers for each Compact Tribe, selecting the smaller, and finding the difference from 350. Colusa serves as an example:

Simple math allows us to save time by grouping similar Compact Tribes according to how many devices they operated as of September 1, 1999, and using the same basic steps. We can then add the group totals to reach the same overall total for Step 2 that would result from aggregating individual calculations. The aggregate calculation would go as follows:

Group 1: Compact Tribes that operated more than 350 devices as of September 1, 1999 = 23 Tribes.

Group 2: Compact Tribes that operated between 1 and 350 devices as of September 1,1999 = 16 Tribes.

Group 3: Compact Tribes that operated zero devices as of September 1, 1999 = 23 Tribes. A reasonable reading of the License Pool Provisions contemplates including these Tribes in the aggregation process. This interpretation accounts for “all Compact Tribes,” in conformance with § 4.3.2.2(a)(l)’s purpose of establishing the aggregate limit for licenses available for “all Compact Tribes.” It also better reflects how an individual Compact Tribe that operated no devices is covered by § 4.3.1 standing on its own, and reasonably retains that coverage when applying § 4.3.1 within the context of § 4.3.2.2(a)(1).

To complete Step 2, we add the results of each of the component calculations:

To solve the entire § 4.3.2.2(a)(1) sum, we then add 29,400 (Step 1) and 10,801 (Step 2) to produce 40,201, “the number of machines that all Compact Tribes in the aggregate may license pursuant to” § 4.3.2.2(a)(1).

We conclude that this interpretation, specifically that 40,201 licenses were authorized for distribution statewide through the license draw process, serves to make the License Pool Provisions “lawful, operative, definite, reasonable, and capable of being carried into effect.” Badie, 67 Cal.App.4th at 800 , 79 Cal.Rptr.2d 273 . It deserves observation that, although we select a slightly different route than the district court, our ultimate conclusion regarding the size of the license pool differs from that reached by the district court by only about 2500 licenses.

III. Remedy of a License Draw Open to All Eligible Compact Tribes

On summary judgment, the district court concluded that the size of the license pool was 42,700 licenses. Based on this conclusion, the court determined that 10,-549 licenses remained in the license pool beyond the 32,151 licenses California claimed were authorized. Colusa requested the district court to order a draw for all Compact Tribes to distribute the additional licenses and the district court ordered that within 45 days California conduct “a draw of all available gaming device licenses, in accordance with the court’s April 22[,2009,] order, and in which all eligible Compact Tribes may participate.” California conducted a draw in October 2009, following an unsuccessful motion in this court to stay the order. In that draw, 1878 licenses were requested by ten interested Compact Tribes, including Colusa and Picayune, and all requested licenses were issued.

California contends that the district court did not have the discretion to order a draw open to all eligible Compact Tribes, claiming the draw would impermissibly extend relief to non-parties. We review the district court’s choice of remedy for abuse of discretion. United States v. Alisal Water Corp., 431 F.3d 643 , 655 (9th Cir.2005).

It is worthwhile to recap the license draw scheme created by the Compact to provide context for reviewing the district court’s remedy. The Compact between California and Colusa creates the right for Colusa to draw from a pool of gaming device licenses that are available to all Compact Tribes. § 4.3.2.2. (“The Tribe, along with all other Compact Tribes, may acquire licenses to use Gaming Devices in excess of the number they are authorized to use under Sec. 4.3.1, but in no event may the Tribe operate more than 2,000 Gaming Devices .... ” § 4.3.2.2(a)). The Compacts executed by the Tribes that are not parties to this suit all include an identical right to draw from that same license pool. See, e.g., Picayune Compact, § 4.3.2.2(a). The Colusa Compact and the other Compacts set out identical instructions for how the series of license draws must be administered. § 4.3.2.2(a)(3); see, e.g., Picayune Compact § 4.3.2.2(a)(3). Once a round of draws is announced, any interested Compact Tribe may request licenses from the pool. The requesting Tribe will then be placed into a priority tier for distribution, depending on how many devices the Tribe currently operates. § 4.3.2.2(a)(3). That priority tier governs the maximum licenses the Compact Tribe may request for that round. § 4.3.2.2(a)(3)(i)-(v). The licenses are allocated according to the priority tiers. § 4.3.2.2(a)(3). Rounds are to continue until Tribes stop making draws, and to resume again once a Tribe requests additional licenses. § 4.3.2.2(a)(3)(vi).

The Colusa Compact does not provide for a draw process for select tribes rather than for all interested Compact Tribes that are eligible to apply for additional licenses. See § 4.3.2.2(a)(3). Neither do any of the other 1999 Compacts. See, e.g., Picayune Compact § 4.3.2.2(a)(3). This circumstance is no surprise, as common sense dictates a closed draw may not be conducted under the scheme created by the Compacts. The license pool operates to distribute the available licenses among all of the California tribes that took the step of entering Compacts and were not already operating 2000 devices, the maximum number permitted. One purpose of the elaborate priority tier system was to skew distribution of new machines to those Tribes that did not already conduct extensive gaming operations. Allowing a limited draw would permanently undermine the intended distribution process, as it would use up a portion of the license pool only for the benefit of a subset of the Compact Tribes. Because a cap is set on the license pool, licenses siphoned off for a limited draw could not be recouped later.

Colusa claims injury based on the denial of licenses in earlier draws. To be effective, any remedy must allow Colusa the opportunity to apply for some of the 10,549 licenses remaining in the pool. The district court ordered the only effective remedy derived from the Colusa Compact — an open draw of available licenses— and did not abuse its discretion by turning to the process for management of the license pool agreed upon by both California and Colusa. Indeed, the district court demonstrated prudence by mirroring its relief on a system agreed upon by the participants in the 1999 Compacts, instead of crafting relief that may have had adverse effects on non-party Compact Tribes.

Although California argues it was improper to afford relief that benefits non-party Compact Tribes, an exception to the requirement of limiting relief to the parties applies when effective relief is otherwise unavailable. See Bresgal v. Brock, 843 F.2d 1163, 1170-71 (9th Cir.1988) (“[A]n injunction is not necessarily made over-broad by extending benefit or protection to persons other than the prevailing parties in the lawsuit — even if it is not a class action — if such breadth is necessary to give prevailing parties the relief to which they are entitled.”). See also, e.g., Easyriders Freedom F.I.G.H.T. v. Hannigan, 92 F.3d 1486, 1501-02 (9th Cir.1996) (upholding injunction from enforcing helmet law without reasonable suspicion against all motorcyclists, instead of against just fourteen individual plaintiffs, since California Highway Patrol policy is set on statewide level, and so in practical terms it is unlikely to expect officers would inquire whether an individual was one of the prevailing plaintiffs before issuing a citation).

Our decision in Colusa I contemplates the open draw remedy ordered by the district court. There we concluded that Colusa need not join other Compact Tribes as parties because those tribes did not have a protectable legal interest in the size of the license pool. 547 F.3d at 971-72. The district court observed that Colusa I supported its remedy:

The district court correctly read Colusa I to assume that a decision within this litigation resulting in a larger license pool than California had yet recognized would be followed by a remedy ordering the eventual distribution of the remaining licenses to all Compact Tribes that applied and were eligible to receive them. Although we concluded that non-party Compact Tribes did not have a protectable legal interest under Rule 19, we discussed how their economic interests would nonetheless be affected by an adjudication that the license pool was smaller or larger than California previously maintained:

547 F.3d at 971.

We anticipated what was surely the response by some non-party Compact Tribes to the district court’s remedy, that “those [non-party Compact Tribes] who intend to expand their gaming operations and compete with the dominant gaming tribes will gladly accept an increase in the size of the license pool.” Id. at 971.

Finally, we preempted California’s contention in this appeal that the remedy deprived the state of its right to “litigate the size of the license pool under different facts in other pending and future cases.” Colusa I anticipated that California would be liable for a single number of licenses in the statewide pool, not separate numbers for separate litigants based on their respective situations. Notably, we declared that any “inconsistent conclusions with respect to the size of the license pool created under the 1999 compacts” that were reached in separate district courts “could be resolved by an appeal to this court.” 547 F.3d at 972 n. 12. Through this decision, we have indeed removed any danger that California will face inconsistent interpretations regarding the size of the license pool, at least as it obtains under the 1999 Compacts.

In sum, to provide effective relief to Colusa, the district court appropriately ordered a license draw according to the process the parties agreed to in the Compacts. Through ordering the open draw, the district court did not abuse its discretion by misapplying the law or “rul[ing] in an irrational manner.” See Alisal, 431 F.3d at 655 (internal quotations omitted). The open draw falls within the exception to the restriction against extending relief to non-parties as the benefits to non-party Compact Tribes were incidental to providing effective relief to Colusa. Our opinion in Colusa I bolsters this conclusion.

Conclusion

The License Pool Provisions that California and Colusa included in their Compact as a foundation for establishing Class III gaming in California are murky at best. The multiple interpretations offered in this litigation underscore this reality. The language of the License Pool Provisions is not reasonably susceptible to the interpretations advanced by the parties, which do not give effect to the structure and explicit terms of the Compact. Because the Provisions are not reasonably susceptible to the parties’ interpretations, the extrinsic evidence submitted by the parties is inadmissible. Taking a fresh look at the Compact, we come to an interpretation that is lawful, operative, definite, reasonable, and capable of being carried into effect. We conclude that, under §§ 4.3.1 and 4.3.2.2(a)(1), the Compacts authorize 40,201 licenses for distribution through the license draw process. As this number exceeds the limit employed by California and proffered in its cross-motion for summary judgment, we affirm in part the grant of summary judgment to Colusa and the denial of California’s cross-motion for summary judgment. We also affirm, as being within the district court’s discretion, the order of a license draw open to all eligible Compact Tribes.

AFFIRMED IN PART AND REVERSED IN PART. Each party shall bear its own costs on appeal.

. The number of tribes that entered Compacts in 1999 varies slightly according to source: 63 according to the record in a predecessor case, Cachil Dehe Band v. State of California, 547 F.3d 962, 966 (9th Cir.2008) ("Colusa I”); 57 according to the district court and California; and 61 according to Colusa. . Picayune joined in Colusa's relevant summary judgment filings. The Rincon Band of Luiseno Indians, the San Pasqual Band of Mission Indians, the California Association of Tribal Governments, and the California Nations Indian Gaming Association filed amicus curiae briefs. Rincon Band and San Pasqual have pending litigation in the district court raising substantially similar issues to Colusa's litigation. . California’s formulation is the same interpretation that the Commission adopted in 2002 after assuming control of the draw process. . Colusa counted all Non-Compact Tribes, i.e., tribes who entered Compacts and those that did not. . In summary, California’s evidence consists of declarations from California negotiators stating that they repeatedly advised tribal participants of California's intention to limit the number of devices statewide at about 44,500; that the idea of a licensing pool was discussed with key tribal representatives before drafting § 4.3.2.2; that a draft of the subsection was presented to a group of tribal attorneys for comment, and modified upon the request of tribal representatives; that the tribal representatives asked no questions about the meaning of § 4.3.2.2(a)(1) after receiving the text of the subsection; that the ''information sheet” from the Governor's Press Office, stating the total number of gaming devices authorized by the Compacts was 44,448, provoked no questions or complaints from tribes; and that the Commission adopted 32,151 as the size of the license pool after assuming control from the Sides administration, which had already issued 29,398 licenses. . Through declarations of tribal representatives and counsel, Colusa maintains that California drafted § 4.3.2.2(a)(1) on its own, without the input of the tribes; that when the assembly of tribal representatives was presented with the text of § 4.3.2.2(a)(1) on September 9, 1999, various tribal leaders asked the State negotiators to explain the provision, and the State negotiators refused to do so; that Colusa's Chairman believed, after discussing § 4.3.2.2(a)(1) with other tribal leaders and without the help of explanation by the State negotiators, that the provision would authorize about 56,000 gaming devices in addition to those already in operation, and he signed the letter of intent on the basis on that understanding. Although Colusa initially offered a contract interpretation that produces this number, it has now abandoned that original formulation in favor of the alternative formulation, again highlighting the problematic nature of the contract language and the extrinsic evidence. .The Court continued: Defendants present evidence that the State’s intention was to limit the aggregate number of devices at approximately 45,000, including those already in operation at the time the compacts were signed. As such, only approximately 23,000 devices would be authorized under the Compact. In contrast, Colusa presents evidence that its Chairman understood the Compact to provide for approximately 55,000 additional licenses at the time he signed the Compact. ... Furthermore, the evidence demonstrates that there was no consistent course of conduct between the parties and that there continued to be debate about the number of devices authorized under the Compact.... Accordingly, the court finds that the extrinsic evidence does not reveal a plain intent or meaning that was either understood by the parties at the time the Compact was executed or followed by the parties in their subsequent relationships. . When the approximately 21,000 devices already in operation statewide as of September 1, 1999, are subtracted from 44,500, about 23,500 devices remain for the license pool. . California relies on the "information sheet” issued by the Governor’s Press Office immediately following ratification of the Compacts as the primary expression of the parties’ original intentions regarding the license pool. The information sheet declares there are 23,-450 licenses authorized by the Compacts for the pool. However, the rationale expressed in the information sheet reaching that conclusion suffers from major flaws when compared to the language of the License Pool Provisions, such that the License Pool Provisions cannot be considered reasonably susceptible to the interpretation expressed in the sheet. In brief, the information sheet appears to ignore the Compact's definition of Non-Compact Tribes, which includes any federally recognized California Indian Tribe that was operating fewer than 350 devices as of September 1, 1999, substituting instead the subset of tribes that operated zero devices. Additionally, the sheet does not appear to include Step 2, "the difference between 350 and the lesser number authorized under Section 4.3.1,” in its calculation. .The district court also noted this gap in California's argument: Significantly, no party proffers an interpretation of the Compact that substantiates this number. [That approximately 23,500 licenses were authorized under the Compact for the license pool, reflecting the Defendant's contention that "the State's intention was to limit the aggregate number of devices at approximately 45,000.”]. Rather, the Commission rejected [California’s lead negotiator's] interpretation of the Compact, which assumed "that uncompacted tribes have permanently waived their right under Compact 4.3.1 to deploy up to 350 gaming devices following entry into a Compact with the State.” The Commission noted that such an interpretation contradicts the express language of 4.3.1. The court is not persuaded that Commission's formulation is most reflective of the parties' intent based upon its piecemeal reliance on [the State’s lead negotiator’s] interpretations. . When the approximately 21,000 devices either already in operation as of September 1, 1999 or given a "free pass” under § 4.3.1 are added to the 32,151 licenses California maintains are in the pool, about 53,000 devices would be allowed statewide. . Indeed, in its report selecting California's current formulation, the Commission distanced itself from some of the state’s earlier explanations of its intentions, particularly the lead negotiator's statement that the license pool was based on an assumption that Non-Compact Tribes had impliedly foregone their "free pass” devices. . This challenge is equivalent to that faced by the district court, as the district court appears to have admitted the parties' extrinsic evidence, but explicitly did not rely on it when interpreting the Compact as a matter of law. . To recap in mathematical terms, our interpretation and resolution of the § 4.3.2.2(a)(1) sum reads as follows: (350 x 84) (Step 1)+ [23(350-350) + [16(350)-2849] + 23(350-0) ](Step 2)= 29,400+ (0 + 2751 + 8050) = 29,400+10,801 = 40,201 total . For example, the amici curiae raise the concern that a limited draw might interfere with their bargained-for rights under the individual bilateral Compacts to access licenses in the pool.