WAR Primer Volume 1 – Position Player Components
This post is part of a series that includes material originally written for The New York Yankees All-Time All-Stars but had to hit the cutting room floor prior to publication. For other posts in the series you can search for the label "Author's Cut" on this site.
Ranking each Yankee player by position throughout the team's long history was no small task. To make the decisions easier, WAR (Wins Above Replacement) and its close relative WAA (Wins Above Average) were the two main performance metrics used. At a high level, the concept of these metrics are simple. They attempt to quantify how many of his team's wins a player was personally responsible for.
Before we can start to think about wins, though, we need to start with runs. There are three main run components that are used in the wins calculations for position players and they are readily available online thanks to the wonderful folks at baseball-reference.com - an invaluable resource for anyone interested in the data side of baseball.
In addition to publishing refreshed data everyday, the Baseball Reference team has also gone to great lengths to provide detailed explanations on each calculation made for the run components. Below is my attempt to distill those explanations down to something a bit more manageable for the average fan while still being as thorough as possible.
If reading this leaves you wanting more, Baseball Reference's WAR Explainer is highly recommended.
Once the run values are calculated, some additional runs are added or subtracted depending on park factors (a player that calls homer-happy Coors Field home gets a penalty while a San Diego Padre would get a bonus for playing at offensively challenged Petco Park) and how the player’s run rate compares to league average. To be fair, the runs added or subtracted are scaled based on the number of plate appearances the player made in a given year. There is a lot more complexity within the calculations which are not worth getting into here as they only minimally affect the results. Rest assured, the people at Baseball Reference give careful consideration to the Rbat calculation to arrive at the most accurate number possible.
While far from being a perfect system, TZR does the best it can to estimate runs saved by fielders using the limited fielding data available prior to 2003. For seasons where play-by-play data is available (generally back to 1953), that level of detail is used to infer things like range factors, outfield arm ratings, double play ability, etc. When data of that detail is not available, more crude measures are taken to arrive at best estimates for Rfield.
On the flip side of the coin, DRS provides a level of detail that was previously unimaginable. The main difference between DRS and all fielding metrics that came before it is that it’s observational. That is, trained analysts watch enhanced video tape of every fielding play that occurs and score them for things such as fielding range, difficulty of play made, and proper positioning. Further enhancing the metrics are pitched and batted ball data which allow for velocity and location components to be accounted for. It also gives the unique opportunity to quantify a catcher’s ability to frame pitches and handle a pitching staff. Given the thoughtful consideration and science serving as the backbone of DRS, it’s safe to say that Rfield is more reliable and accurate for seasons since 2003 than it is for seasons prior.
The second baserunning event component involves what a player does on base when a teammate puts the ball in play. There are over 30 different scenarios that can occur depending on what base the runner is on and what happens when the batter puts the ball in play. All of them result in the base runner either advancing a base, making an out, or staying put. For seasons where play-by-play data is available, the number of times a player advances, fails to advance, or runs into an out in these scenarios can be totaled and adjusted to the league average. Those extra bases and extra outs are then converted to runs. Just as we saw with stolen bases, the negative impact of running into an out far outweighs the corresponding benefit of advancing an extra base.
Since the ability to avoid a double play is mainly tied to a player’s ability to run, it’s also captured as a part of this metric. As with the other components, a player’s rate at which they ground into double plays is adjusted to the league average, scaled to the player’s total number of opportunities and converted to runs via a multiplier. The runs coefficient can vary from year to year, but in general the value of avoiding a double play versus grounding into one is just shy of half a run.
Ranking each Yankee player by position throughout the team's long history was no small task. To make the decisions easier, WAR (Wins Above Replacement) and its close relative WAA (Wins Above Average) were the two main performance metrics used. At a high level, the concept of these metrics are simple. They attempt to quantify how many of his team's wins a player was personally responsible for.
Before we can start to think about wins, though, we need to start with runs. There are three main run components that are used in the wins calculations for position players and they are readily available online thanks to the wonderful folks at baseball-reference.com - an invaluable resource for anyone interested in the data side of baseball.
In addition to publishing refreshed data everyday, the Baseball Reference team has also gone to great lengths to provide detailed explanations on each calculation made for the run components. Below is my attempt to distill those explanations down to something a bit more manageable for the average fan while still being as thorough as possible.
If reading this leaves you wanting more, Baseball Reference's WAR Explainer is highly recommended.
Rbat - Runs from Batting
Number of runs better or worse than average a player was as a hitter. Rbat is adjusted for league averages and home ballpark and is calculated by assigning a run value to each of the ways that a hitter can contribute, positively or negatively, when making a plate appearance. The run value assignments are based on actual results found throughout baseball history and are shifted from year to year accordingly. For instance, a home run has been worth, on average, about 2 runs throughout baseball history. However, the actual value used in 2000 - when home runs were fairly common - is 1.95 whereas the value used for 1968 was 2.34 due to home runs being a relatively scarce and valuable commodity at the time.Once the run values are calculated, some additional runs are added or subtracted depending on park factors (a player that calls homer-happy Coors Field home gets a penalty while a San Diego Padre would get a bonus for playing at offensively challenged Petco Park) and how the player’s run rate compares to league average. To be fair, the runs added or subtracted are scaled based on the number of plate appearances the player made in a given year. There is a lot more complexity within the calculations which are not worth getting into here as they only minimally affect the results. Rest assured, the people at Baseball Reference give careful consideration to the Rbat calculation to arrive at the most accurate number possible.
Unsurprisingly, Babe Ruth leads all Yankees with 1,215 career Rbat.
The Sporting News Archive/Wikimedia Commons/Public Domain
Rfield - Runs from Fielding
Number of runs better or worse than average a player was for all fielding. Fielding of balls in play, turning double plays, outfield arms, and catcher defense are all included. One key distinction here is that two different systems are used to arrive at the Rfield number. Total Zone Rating (TZR) is used for all seasons prior to 2003 and the far more sophisticated Defensive Runs Saved (DRS) for all seasons since 2003.While far from being a perfect system, TZR does the best it can to estimate runs saved by fielders using the limited fielding data available prior to 2003. For seasons where play-by-play data is available (generally back to 1953), that level of detail is used to infer things like range factors, outfield arm ratings, double play ability, etc. When data of that detail is not available, more crude measures are taken to arrive at best estimates for Rfield.
On the flip side of the coin, DRS provides a level of detail that was previously unimaginable. The main difference between DRS and all fielding metrics that came before it is that it’s observational. That is, trained analysts watch enhanced video tape of every fielding play that occurs and score them for things such as fielding range, difficulty of play made, and proper positioning. Further enhancing the metrics are pitched and batted ball data which allow for velocity and location components to be accounted for. It also gives the unique opportunity to quantify a catcher’s ability to frame pitches and handle a pitching staff. Given the thoughtful consideration and science serving as the backbone of DRS, it’s safe to say that Rfield is more reliable and accurate for seasons since 2003 than it is for seasons prior.
Rbrdp - Runs from Baserunning and Grounded into Double Plays
Number of runs better or worse than average a player was for all baserunning events and avoiding grounding into double plays. There are two components for the runs associated with baserunning events. The straightforward component is base stealing. Just as hits are converted to a run value when calculating Rbat, stolen bases can be assigned a run value. Conversely, getting caught stealing has a negative run value and is much more costly than the corresponding benefit of a stolen base. In general, a player needs to be successful at least 3 out of 4 times in order for their stolen bases to be worthwhile. A player’s total run value based on stolen bases and times caught stealing is then adjusted for the league average.The second baserunning event component involves what a player does on base when a teammate puts the ball in play. There are over 30 different scenarios that can occur depending on what base the runner is on and what happens when the batter puts the ball in play. All of them result in the base runner either advancing a base, making an out, or staying put. For seasons where play-by-play data is available, the number of times a player advances, fails to advance, or runs into an out in these scenarios can be totaled and adjusted to the league average. Those extra bases and extra outs are then converted to runs. Just as we saw with stolen bases, the negative impact of running into an out far outweighs the corresponding benefit of advancing an extra base.
Since the ability to avoid a double play is mainly tied to a player’s ability to run, it’s also captured as a part of this metric. As with the other components, a player’s rate at which they ground into double plays is adjusted to the league average, scaled to the player’s total number of opportunities and converted to runs via a multiplier. The runs coefficient can vary from year to year, but in general the value of avoiding a double play versus grounding into one is just shy of half a run.
With career totals of 135 Rfield and 65 Rbrdp, Brett Gardner ranks first all-time in both among Yankees.
Keith Allison/Wikimedia Commons
I only come here for the war puns :-(
ReplyDeleteThey're good for absolutely nothing.
Delete