NFL contracts are extremely difficult to accurately value, compare, and budget. This difficulty arises primarily from two factors: (1) the generally non-guaranteed nature of the contracts and (2) the variety of types of components which comprise the contracts (signing bonus, roster bonus, base salary, etc.)
Because NFL contracts are generally not guaranteed, the face value of a contract is largely irrelevant, as it is not determinative of either (i) the amount of money that the player will earn under the contract or (ii) the amount of cap room that the team will allocate to the player over the life of the contract. Because NFL contracts may contain a variety of types of components, each having different salary cap implications, NFL contracts lack uniformity of style, as NFL teams are able to structure contracts in many different ways.
In an attempt to overcome these difficulties, NFL stakeholders (media, fans, agents, players, teams) have traditionally resorted to inadequate metrics and methods to value, compare, and budget contracts. In making such attempts, however, the stakeholders demonstrate, at least implicitly, that they realize valuation, comparison, and budgeting of NFL contracts are difficult tasks. They also demonstrate an understanding that the irrelevance of contract face values and the variation in structures creates a need for some type of metric.
Common examples of inadequate metrics used by NFL stakeholders include “guaranteed money,” “three-year payout,” and “average per year (APY).” Each of these metrics does in fact enable valuation, comparison, and budgeting of contracts, but they are inadequate because they are either under-inclusive or over-inclusive of contract value. Because these metrics do not adequately capture value, the results of any comparison utilizing the metrics will be correspondingly tainted, and salary cap budgeting will lack accuracy.
“Guaranteed money” is an under-inclusive metric because it assigns a $0 value to all contract money that is not fully guaranteed, implying that there is a 0% chance that the player will earn the money. This is not accurate, as the likelihood that the player will earn the money is certainly greater than 0%, although less than 100%.
“Three-year payout” is also an under-inclusive metric because it assigns a $0 value to all contract money in years four and beyond, implying that there is a 0% chance that the player will receive that money. This is not accurate, as the likelihood that the player will earn the money is certainly greater than 0%, although less than 100%. “Three-year payout” is also an over-inclusive metric because it implies that there is a 100% chance that the player will earn all of the money in the first three years of the contract. Unless all of the money in the first three years is fully guaranteed, which is rarely the case, this implication is also not accurate; the likelihood that the player will earn the money is certainly less than 100%.
“APY” may be either over-inclusive or under-inclusive, but in most cases it is over-inclusive. By depicting value in the form of a rate metric, rather than a counting metric, APY implies that the timing of payments within a contract are not relevant; all contract money is equally likely to be paid. Those who follow the NFL know from experience that money scheduled to be paid in the later years of a contract is less likely to be earned than money scheduled to be paid in the earlier years of a contract. Therefore, APY is over-inclusive if the money scheduled to be paid in the later years of a contract equals, or is greater to (as is usually the case), the money scheduled to be paid in the earlier years of a contract, as an accurate rate metric would not include all contract money equally in its output. APY will be deceivingly high, as the unlikely money scheduled to be paid toward the end of the contract will artificially raise the average. In theory, APY could be under-inclusive if the money scheduled to be paid in the later years of a contract is dramatically lower than the money scheduled to be paid in the earlier years of a contract, but we are not aware of any contract possessing such a characteristic.
In addition to inadequate metrics, NFL stakeholders also frequently employ other less-than-satisfying methods to value, compare, and budget contracts. Each time a noteworthy contract is signed, the same process takes place. Adam Schefter tweets the general parameters. Pro Football Talk warns everyone to reserve judgment until the true details are revealed. The agent leaks the true details. Over The Cap, Pro Football Talk, Joel Corry, and maybe a few others scrutinize the contract to provide subjective analysis, commenting on the likely outcome of the agreement and the discernible strategies utilized by the parties. One might read that, “Player X’s contract is really for Y amount” or that, “after Z number of years, the contract will be renegotiated.”
Such articles are typically very informative and insightful for the reader. However, this type of analysis is difficult to replicate on a mass scale, and it can be described more accurately as an art than as a science. When comparing two contracts, it is difficult to manually determine which combination of contracts components is superior to the other. Even when such a determination is possible, it is not feasible to place a numerical value on the superiority of one contract in comparison to the other. When comparing dozens or hundreds of contracts, it becomes impossible to manually synthesize all of the components of every contract. The degree to which NFL teams employ advanced proprietary techniques is unknown, but we have no specific reason to think that they do so to any material degree.
Thus there currently exists a gap in NFL contract analysis. Despite the proliferation of advanced metrics across the major North American sports, no such tools exists to evaluate NFL contracts. “Contract Analytics” does not yet exist as a subset of sports analytics generally. This gap can be filled, and the aforementioned subjective analysis greatly supplemented, by an objective metric that can be employed on a mass scale and that is neither under-inclusive nor over-inclusive of value.
To that end, we introduce “Expected Contract Value.” All money included in the face value of a contract has a greater-than-zero percent, but [usually] less than-one-hundred percent, chance of being earned by the player. Expected Contract Value assigns a likelihood within such range for each dollar of the contract, producing an expected value of the contract when the likelihoods are multiplied against each dollar and subsequently summed. In other words, Expected Contract Value determines the probability that a player will remain under contract, for each season of the contract, despite the non-guaranteed nature of the contract, and it then calculates how that probability will affect earnings and/or salary cap numbers.
At first glance, assigning such an expectation to future years of a contract would seem nearly impossible, given that seemingly crucial factors such as the performance and health of the player, as well as the needs and salary cap status of the team, are extremely difficult to project years in advance. But as it turns out, the probability that a player will remain under contract can be determined with a relatively high degree of accuracy for each season of the contract, at the moment that the contract is signed, by taking into account nothing other than the characteristics of the contract itself.
The same variations in contract component types that contribute to the difficulty in valuing, comparing, and budgeting for contracts by traditional methods lend themselves quite well to regression analysis designed to produce probabilities and expected values. By creating a set of variables derived from relationships among contract components in completed or terminated contracts, we have used regression analysis to determine which variables have historically correlated most strongly with whether or not a player remained under contract in any given season of a contract, based on the distribution and timing of the face value dollars of the contract across the various types of contract components. To take an obvious example of a contract characteristic that is taken into account in Expected Contract Value, the amount of dead money that a team would incur upon releasing a player plays a part in the team’s decision as to whether or not to release the player.
The idea behind Expected Contract Value is to forecast what the decision makers of a team would do with a given contract in a given off-season if all they could do is look at the contract, but had no idea which player the contract belonged to. As it turns out, by simulating this seemingly absurd scenario, we can account for approximately 80% of real-life decision-making. We are basically saying, “Look, we may not be able to legitimately measure or value the performance of football players, let alone project performance into the future. But that’s fine, we can completely ignore performance and health and team situation, and still forecast the decisions a team will make about a contract based solely on the characteristics of the contract. So, ‘football people’, tell us what you think the player is worth in terms of dollars and cap with respect to performance relative to other players, however you go about determining that, and we will show you how your future decisions will be affected by the method and timing by which you allocate those dollars and cap numbers over a given contract.”
In doing so, we believe that we have created an objective metric that provides a common ground by which to value all contracts, which in turn allows for easy and accurate comparison, as well as useful salary cap budgeting. In Part 2, we will walk through the inputs and outputs of Expected Contract Value and further expound on the theory. In Part 3, we will demonstrate how Expected Contract Value can be used to compare contracts.
In Part 4, we will demonstrate how Expected Contract Value can be used by teams to budget their salary cap structures years in advance. We encourage readers to leave questions and critiques in the comments section. To the extent that we are unable to respond individually, in Part 5 we will follow-up with a responsive FAQ-style post in which we address the most interesting questions and salient criticisms. Next week, we will post the 2015 expected outcomes for all players with cap numbers greater than $2 million (we don’t yet have the calculations built automatically into the site).
In the coming weeks, months, and years we will use Expected Contract Value to analyze recently signed contracts, anticipate team decision-making, and comment on optimally efficient contract structuring. As our sample size grows and our theory is further developed, the Expected Contract Value formula will likely evolve over time. Our hope is that the metric will become widely used in the industry, and that it will influence the way in which NFL contracts are negotiated, structured, and thought about generally.
Expected Contract Value was created by Bryce Johnston and Nicholas Barton.
Image Courtesy of nfl.com
Georgetown Law Class of 2014
Email Bryce: email@example.com
Follow Bryce on Twitter: @eaglessalarycap
Georgetown University Class of 2017
Email Nick: firstname.lastname@example.org