About

Methodology

Data Collection

The data set used to build the visualizations and the predictive model was composed of player statistics and contract details. In constructing the data set, I decided to take ten players from each free agent class since 1977, the first year of free agency. This data set consists only of free agent contracts. No contract extensions are included for the purpose of maintaining consistency throughout. The contract figures are based on guaranteed salary only, so no bonuses, incentives, or option years were included. To ensure that most of the players selected received substantial contracts, I sorted the players by the number of wins above replacement (WAR) they had accumulated over the past three seasons and chose the top ten. By doing this, I was able to diversify the data set. All sorts of players were included, and it was not limited to a certain position or age group. If I had selected the players based on the size of the contract they received, as one might first think to do, I likely would have excluded older players who were nearing the end of their careers. This would have made the data less representative of the true nature of the free agent market.

Why WAR? What is WAR?

WAR is a statistic that has become wildly popular among baseball fans in the twenty-first century. To reiterate, WAR stands for wins above replacement, and it essentially represents how many wins a player is worth when compared to a replacement-level player. For instance, Aaron Judge accrued 10.6 WAR with the New York Yankees in 2023. That means that if Judge was replaced by a replacement-level player, the Yankees would have won about 10.6 fewer games. One of the best parts about WAR is that it can be retroactively applied. Although not created until around 2010, WAR values can be given to past players just by looking at their stats.

The most confusing aspect of WAR is the formula behind it, or perhaps the lack of such a concrete one. Most sources that calculate WAR do so differently. The three main calculations come from Baseball Reference, FanGraphs, and Baseball Prospectus. Like nearly all of the data collected for this project, the WAR numbers used here came from Baseball Reference for the purpose of consistency. Baseball Reference’s calculation of WAR is too complicated to include here, but the links to the full calculations are included for batters and pitchers. According to Baseball Reference, a WAR above 8 signifies an MVP-level player, a WAR above 5 corresponds to an All-Star-caliber player, a WAR higher than 2 signifies a starting-caliber player, and a WAR from 0-2 denotes a substitute. A player with a subzero WAR would be considered below replacement-level.

Data Limitations

Advanced Statistics

As I mentioned before, the great thing about WAR is that it is an advanced stat that can be retroactively applied. Unfortunately, the same cannot be said for other metrics that no doubt play a role in the contracts players receive. Expected statistics are becoming more and more popular as teams try to predict a player’s future performance rather than focus on past performance. Expected batting average (xBA), expected earned run average (xERA) and expected weighted on-base average (xwOBA) are just a few of these predictive stats. However, they haven’t been available until recently. These metrics were only able to be calculated as technology has developed, and they cannot be applied retroactively. Again, because I wanted continuity in my data set, I did not include any expected values. I only included predictors that pertain to every observation in the data set.

This philosophy also rules out any Statcast data. According to Major League Baseball, “Statcast is a state-of-the-art tracking technology that allows for the collection and analysis of a massive amount of baseball data, in ways that were never possible in the past.” Fully implemented in all 30 ballparks in 2015, Statcast cameras track nearly everything that happens on the field, from the spin rate and movement of pitches to the efficiency of the route an outfielder takes to a fly ball. For hitters, two metrics that have become mainstream are launch angle and exit velocity of the ball off the bat. Teams often look at these measurements when trying to project how many home runs or extra-base hits a player will rack up. As useful as the data are, they cannot be applied to seasons before 2015. Thus, I did not include it.

The main goal of this project is to track the evolution of free agency through time, and so even though expected statistics and Statcast metrics certainly play a role in how much money players earn presently, that isn’t something that can be tracked through time.

Defense

Another limitation of this data set is its omission of defensive statistics. The only stat included that accounts for defense is WAR, so the data leave a lot to be desired in this regard. There are two reasons for my decision to leave out defensive metrics. First, many defensive statistics have the same problem with consistency. In today’s game, the most popular ways to measure defense are through advanced stats like Defensive Runs Saved, Ultimate Zone Rating, and Statcast’s Outs Above Average (OAA). These are considered the best stats available, but they’ve been created relatively recently and can’t be applied to past players. Secondly, the metrics that have been around since the 1970’s (or can be applied retroactively) are, quite frankly, poor ways to measure defensive performance. For the longest time, the error was the preferred stat, and people have been keeping track of errors since the 1800’s. In short, a player is charged with an error if he fails to make a play that one would expect him to make. But what does that really mean? How do we know if a player should be expected to make a play? The decision rests on the opinion of the scorekeeper, and so an error is too subjective to be reliable. Additionally, errors often penalize players with superior defensive range. Reaching a batted ball and misplaying it would count as an error, while not even being able to reach the ball would not. A player with greater range will be exposed to more opportunities to make errors. Stats such as fielding percentage and total zone rating are also recognized as significantly flawed fielding metrics.

Another factor that throws a wrench into analysis of defensive statistics is the unique function of the catcher. The catcher’s role is unlike that of any other position player. When a ball is put in play, the catcher functions much like any other infielder and is subject to the same metrics. The difference lies in the instances where the ball is not put in play. While the other fielders do nothing, the catcher’s job has just begun. He is first responsible for framing the pitch if it is close. When a catcher frames a pitch outside the strike zone, he subtly moves his glove back into the zone after catching the ball, giving the impression that it was a strike all along. The best framers can consistently trick umpires into calling more strikes. Below, watch catcher Willson Contreras take this outside pitch and successfully present it as a strike (via The Athletic).