Reintroducing Expected Contract Value: ECV 1.1 & 2.0

Approximately one year ago, we introduced Expected Contract Value (Part 1, Part 2, Part 3, Part 4 & Part 5), a contract value metric that assigns probabilities as to contract termination outcomes – and therefore player earnings – on the basis of contract characteristics. We have made revisions to the metric as introduced a year ago, and we have also created a second version to be utilized for certain purposes.

Expected Contract Value:

To quickly summarize the basic premise, NFL contracts are difficult to value due to their generally non-guaranteed nature and the variety of ways in which they can be structured. The stated value of a contract is not determinative of the amount of money the player will actually receive pursuant to the contract, and the differing salary cap treatment under the NFL CBA of otherwise identical contract amounts dictates that seemingly similar contracts may produce different outcomes with respect to actual earnings.

NFL stakeholders recognize the problematic nature of contract valuation, but they respond by utilizing conceptually flawed (i.e. over-inclusive or under-inclusive) valuation metrics – such as guaranteed money and three-year payout – that highlight relevant valuation considerations but do not holistically capture a contract’s value. Subjective analysis is difficult to perform on a large scale and is subject to human bias, as it is extremely difficult to manually synthesize all of the components of every contract while placing the appropriate amount of emphasis on each. As a result, industry analysis focuses on non-comprehensive value indicators, and contracts are potentially designed inefficiently.

Expected Contract Value produces an output representing the probability a player will remain under contract, for each season of a contract, at the moment of signing. Expected Contract Value represents a probabilistic approach, forecasting expected outcomes as opposed to predicting actual earnings. Because the version of Expected Contract Value we introduced a year ago (“ECV 1.0”) assigned probabilities for future contract seasons, its inputs consisted solely of variables certain at the time of signing. As a result, seemingly important variables such as performance and health were omitted, and the metric instead focused on characteristics of the contract itself and characteristics inherent to the player. The idea was to forecast the outcome of a scenario in which a team must make a decision with respect to a contract without knowing which player the contract belongs to. In effect, ECV 1.0 incorporated an average of all possible outcomes with respect to performance, health, conduct and other matters, as its inputs encompassed a wide variety of such outcomes.

ECV 1.0 utilized six input variables, further explained in Part 2 of the introduction series:

  • Save:Avg: The ratio of the amount of cap savings that the team would realize upon releasing the player to the average annual value of the player’s contract.
  • Save:Cap: The ratio of the amount of cap savings that the team would realize upon releasing a player to the player’s cap number if the team does not release him.
  • Cap:Avg: The ratio of the player’s cap number if the team does not release him to the average annual value of the player’s contract.
  • Contract:Complete: The ratio of completed seasons to total seasons in a contract.
  • Age:Peak: The ratio of the player’s age to a fixed denominator representing a theoretical peak age.
  • Delay:Dead: The ratio of the amount of potential dead money incurred upon a release that could be deferred to the following year to the total potential dead money incurred upon a release.

ECV 1.1:

After further considering the theory behind each of the inputs, as well as testing the metric on historical data samples, we have decided to make several changes to ECV 1.0. We will refer to the revised metric as “ECV 1.1,” and we will apply ECV 1.1 to newly signed contracts beginning this offseason (such as Zach Ertz, Lane Johnson, Vinny Curry and Travis Kelce).

First, we have removed Save:Cap, Cap:Avg and Delay:Dead as inputs. With respect to Delay:Dead, we determined that it does not have much of an effect at all on the outputs, and at times produces odd results. With respect to Save:Cap and Cap:Avg, we arrived at the somewhat counter-intuitive conclusion that a player’s salary cap number is not relevant for the purposes of contract termination decisions. This is because a player’s salary cap number typically incorporates fixed amounts (such as prorated signing bonus amounts) that cannot be manipulated or avoided via team action.

For example, if Player X has a $5 million salary cap number, and $1 million of that amount consists of prorated signing bonus, then only the $4 million attributable to base salary, roster bonus, etc. is relevant for decision-making purposes. The $1 million amount is “stuck” on the team’s salary cap; it cannot be traded, deferred or avoided by any method. We will ignore the $1 million sunk cost salary cap space and instead focus on the amounts that can be manipulated through team action. As a result, Save:Cap and Cap:Avg have been removed, and Save:Avg remains. We view the simplicity of utilizing only three input variables as an attractive characteristic of ECV 1.1.

Second, we have separated players into positional groupings to account for the possibility that teams utilize different criteria to make contract termination decisions based on perceived differences in value as between positions, as well as different aging patterns as between positions. The positions groupings are Quarterbacks, Running Backs, Receivers (WR/TE), Offensive Lineman, Front Seven, Defensive Backs and Specialists. We are of the view that it would not be productive to further distinguish positional groupings (i.e. offensive tackles v offensive guards, cornerbacks v safeties), as the impact on the outputs would likely be minimal and the sample sizes would be greatly reduced.

ECV 2.0:

We have now developed a modified version of Expected Contract Value for the purpose of more accurately forecasting current-season outcomes. We refer to this version as “ECV 2.0,” and we will apply ECV 2.0 to 2016 contract seasons in order to determine the likelihood of each player being released (as we have done in the team offseason preview series).

To create ECV 2.0, we added an additional input variable for each input contract season consisting of the player’s “Approximate Value” (AV) as described on Pro Football Reference in the season immediately preceding the given input contract season. At the time that ECV 2.0 is applied, each players “Prior AV” is known with certainty, which allows us to “update” the Expected Outcome without deviating from the principal of only utilizing known inputs. As a result, each player’s 2015 performance is incorporated into the 2016 ECV 2.0 expected outcomes, and health is incorporated to the extent that a player is unable to accumulate AV during games that he misses due to injury.

The primary effect of incorporating prior season performance is to avoid odd outcomes with respect to very good players, very bad players, and players who have clearly aged better than would typically be expected. We are not arguing that AV is any sort of perfect performance metric, but our goal is not to split hairs as to rankings of which players are better than other players. AV presents a simple metric that is uniform across positions, available historically, and seems to produce generally intuitive results. This is sufficient for the purposes of ECV 2.0, although we are open to using a different performance input variable if we become aware of a better one meeting such criteria.

ECV Results:

Historically, Expected Contract Value (ECV 1.1) has correlated more strongly with actual earnings than stated contract value, guaranteed money, or three-year payout. We calculated each of these metrics for 100 contracts that would have expired no later than the 2015 season by their original terms. The applicable players collectively earned approximately $2.1758 billion, representing 68.1% of the aggregate stated contract value. The aggregate Expected Contract Value of these historical contracts was approximately $2.1119 billion.

As between Expected Contract Value and actual earnings, the coefficient of correlation, r, was .932, as compared to .868 (stated contract value), .822 (three-year payout) and .762 (guaranteed money). We note that stated value is more strongly correlated with actual earnings than three-year payout or guaranteed money, both of which are used as a result of the inadequacy of stated contract value. Only 24 of the 100 historical contracts were terminated after exactly three contract seasons, further casting doubt on the merits of three-year payout.

We calculated ECV 2.0 for 480 players signed for the 2015 season under preexisting contracts and subsequently tracked whether the 2015 contract seasons remained in effect on original terms (keep in mind that the Expected Outcomes posted last year on this website were calculated using ECV 1.0). If we separate the 2015 contract seasons into quartiles based on ECV 2.0 expected outcomes, the results are as follows:

  • Expected Outcome 100-75%: 95% remained under contract
  • Expected Outcome 74.9-50%: 71% remained under contract
  • Expected Outcome 49.9%-25%: 61% remained under contract
  • Expected Outcome 24.9-0%: 44% remained under contract

ECV 2.0 tends to overestimate to some degree the likelihood that each contract will be terminated, but the Expected Outcome of the 2015 contract seasons and the actual outcomes are positively correlated. Anecdotally speaking, ECV 2.0 appears to generate more success identifying which underperforming players will not be released due to contract considerations than identifying which adequately performing players will be released due to contract considerations. The 2015 contract seasons are now included as inputs for the purpose of recalculating the models for 2016 (now bringing the total sample size to approximately 2,900 contract seasons), which we expect will  improve the accuracy of the Expected Outcome forecasts.

We are in the process of publishing the team-by-team 2016 ECV 2.0 Expected Outcomes in the offseason preview series, but we will publish the complete list in its own post tomorrow. We will also publish ECV 1.1 calculations for all of the sizeable contracts signed this offseason and continue to provide a variety of other types of analysis based on Expected Contract Value.

Expected Contract Value was created by Bryce Johnston and Nick Barton.

Bryce Johnston earned his Juris Doctor, magna cum laude, from Georgetown University Law Center in May 2014, and currently works as a corporate M&A associate in the New York City office of an AmLaw 50 law firm.  Before becoming a contributor to, Bryce operated for 10 NFL offseasons, appearing multiple times on 610 WIP Sports Radio in Philadelphia as an NFL salary cap expert. Bryce can be contacted via e-mail at or via Twitter @NFLCapAnalytics.

Nick Barton is  a junior at the McDonough School Business at  Georgetown University.  He is majoring in Finance and Operations and Information Management. Nick currently interns with an NFL team . His prior work experience includes interning with CollegeSplits and Dynamic Sports Solutions, and working as a research assistant for the Center of Applied Research of the Apostolate.