Sabermetrics 101: WAR and WPA

This is the first installment of a series on sabermetrics by Josh Strohman, an MLB contributor for The Walk On Blog.

For the past 20 years, baseball has shifted away from traditional statistics like batting average and RBI’s for the higher mathematical sabermetrics. Sabermetrics aims to better evaluate a player’s past performance and predict future performance.

Early on, sabermetrics were almost exclusively used by forward thinking baseball clubs that wanted to get an edge on their competition. Today, sabermetrics have become commonplace in teams and in the media. They are regularly used in television broadcasts, written media and player profiles.

Despite their penetration into the mainstream, many people are still unfamiliar with these statistics and what they are telling them. In these articles, I will attempt to explain how some of the commonly used sabermetric statistics are derived, how they are used, and how to read their values.

For today’s discussion, we will dive into two of the most common sabermetric statistics, Wins Above Replacement (WAR) and Win Probability Added (WPA)

Wins Above Replacement (WAR)

I would guess that the vast majority of people, at the very least, are familiar with WAR, but most don’t truly understand it. I have had many conversations with baseball purists about how they dislike the stat because they view it as imperfect, and honestly, they’re right.

However, it is vastly superior to any other traditional statistics when we attempt to determine a player’s value to a baseball team. So, let’s take a quick look on at what WAR actually tries to accomplish.

How WAR is determined

Let me start by saying that WAR is calculated slightly different depending on where you look. Because of this, the actual WAR value of a player may differ depending on the source.

For example, in 2015 Mike Trout had a FanGraphs WAR (fWAR) of 9.3, but a Baseball References WAR (brWAR) of 9.6. Don’t get caught up on the differences, they’re insignificant. My recommendation would be to pick one source of WAR and stick with it.

There is one more thing that we need to address before we jump into the WAR calculation and that is replacement level players. Think of a replacement level player as a AAA player that gets pulled up to replace an injured MLB player. They wouldn’t be great, but they would fill the position.

WAR tells us how many additional games the team would win over the course of a season with the MLB player in the lineup opposed to the replacement level player.

Alrighty, now let’s take a look at how WAR is determined. Full discloser, WAR can be complicated, so I am going to attempt to simplify it.

For position players, WAR is calculated using the batting runs, base running runs, fielding runs prevented (or allowed), and the players position. It then adds in how many runs a replacement level player is worth and converts the total into games. Lastly, they add park adjustment which removes the advantage that players that play in hitter happy parks like Colorado’s Coors Field and vice versa.

Pitching WAR is more complicated than position player WAR; however, all you need to know is that Pitching WAR is determined using runs allowed or FIP (Fielding Independent Pitching) and fielding runs, scaled to innings pitched.

WAR: Things to keep in mind

As I mentioned, WAR is not a perfect tool to use when determining player value. One player may have a slightly greater WAR, but that does not mean the player with the greater WAR is necessarily more valuable.

For example, last year Nolan Arenado (COL 3B) had a brWAR of 6.7 and Anthony Rendon (LAA 3B) had a brWAR of 6.4. This difference is small enough to be considered negligible and we could say these players had comparable seasons; neither was any more or less valuable to their team.

Because WAR is league, position and park adjusted, it allows us to compare players over different variables.

Lastly, WAR is accurate. I know that I made a point that WAR is imperfect (and it is); however, it is very accurate. One fun exercise to determine the accuracy of WAR is to add up all the WAR’s of the players that played for a specific team and then add 43.5 (the number of wins a team with all replacement level players would have). You should get a number of wins close to team’s Pythagorean record (the number of wins a team should have based on the runs they scored and allowed).

For example, the 2019 Tampa Bay Rays had a Pythagorean record of 93-69. If we add up all the WARs as I specified above, you get a record of 89-73. No, it’s not perfect, but it’s the best metric we currently have.

Tiers of WAR values

Differences in WAR calculations make it somewhat difficult to put WAR values into tiers; however, I will try to put them in tiers that make sense regardless of where you choose to look for WAR values.

Win Probability Added (WPA)

WPA is a very straightforward statistic that allows us to examine how important a player’s performance was, rather than simply how they performed. If that seems a little confusing at the moment, it’s okay because it will begin to make more sense as we dive into the nuts and bolts of WPA.

How WPA is determined

WPA is a running total of a player’s effect on a team’s win probability from plate appearance to plate appearance. You can easily figure this statistic out at home on your couch with a win probability chart. The best way to learn this statistic is to dive right to into an example.

Let’s say that Nelson Cruz comes to the plate while the Twins have a 40% chance of winning. Cruz hits a home run, and the Twins win probability jumps to 60%. Because the Cruz increased the Twins win probability by 20%, Cruz’s WPA would increase by .20.

The opposite would occur for the pitcher (let’s say it’s the Yankee’s Luis Severino) that gave up the home run. Because the Yankee’s win probability decreased by 20%, Severino’s WPA would decrease by .20.

As you can see, it’s a pretty uncomplicated statistic that tells a story of how a player’s plate appearances over the course of a game, season or career have affected the probability of his team’s victory.

WPA: Things to keep in mind

The most important thing to realize about WPA is that the context of the performance is everything. A home run in the eighth inning will affect win probability far more than a home run in the second inning. Because of this, WPA takes into account how significant the performance was, rather than simply how the player performed.

Despite this, we cannot misconstrue WPA for a statistic that tells us how clutch (or unclutch) a player is. This is because WPA is an additive stat. This means that a player that plays more will have more opportunities to increase (or decrease) their WPA.

Because WPA deals with win probability, a zero WPA is considered to be average. This differs from WAR, where zero indicates a replacement level player.

Lastly, WPA is not predicative of future success. Because different situations yield different changes in win probability, WPA does not necessarily tell us the value of a player. Rather WPA tells a story about how a player helped his team over the course of a game, season or career.

Tiers of WPA values

As mentioned, WPA is not predictive; however, we can still put WPA scores into tiers in order to tell us how a player has affected win probability. Note that these tiers are for MLB regulars for an entire season.

Stay tuned for next week for when we continue our sabermetrics discussion with BABIP and OPS+.