kWAR: An Introduction to a New Pitching Metric
By: Max Greenfield and Jason Fixelle
The advent of defense independent pitching statistics introduced us to the notion that there are numerous factors to pitching success that are outside of the pitcher’s control. A pitcher like Max Scherzer can throw his slider down and away to a right-handed hitter outside of the strike zone and the hitter could still find a way to dunk it over the second baseman’s head into right field. As baseball thinking has evolved over time, more measures have been created to better identify the areas within a pitcher’s control (i.e., FIP, xFIP, SIERA, and pCRA). Each of these metrics try to narrow down a pitcher to those outcomes for which pitchers should be evaluated on. While the previously listed metrics consider in some form the impact of certain batted ball events, kwERA eschews all batted ball events in favor of a focus on strikeouts and walks. While subsequent iterations of kwERA have included intentional walks and ground balls, for now, we decided to stick with the original kwERA to create a WAR statistic around it called kwERA WAR or kWAR.
What is kwERA?
Conceptualized by GuyM and popularized by Tom Tango, kwERA is an ERA estimator based solely on a pitcher’s strikeouts and walks. Completely ignoring balls in play is both a feature and a flaw of the metric. Avoiding home runs or even hard contact may be events that fall within some modicum of a pitcher’s control, but they are also subject to randomness and variability. However, strikeouts and walks are the only outcomes completely within a pitcher’s control. The more a pitcher controls in a game, the more likely he is going to be successful.
How is kWAR calculated?
In short, exactly how fWAR (FanGraphs WAR) is, but with kwERA replacing FIP as the central component. The kwERA for each player is adjusted into an RA9 scale the same way FIP is in fWAR. Once the kwERA has been park-adjusted, the process is the same.
When calculating fWAR, FanGraphs utilizes a specifically designed park factor composed soley for each of the FIP components (i.e. HRs, walks, strikeouts, and infield fly balls, which for purposes of FIP are included as strikeouts). When applying FanGraph’s FIP park factor in order to calculate fWAR, Fangraphs assumes that pitchers play half of their games in their home ballpark and half in collectively neutral parks, creating some issues for players who appear more often in ballparks with extreme park factors. (i.e., Coors Field)
With kWAR, we have taken a similar approach to FanGraphs by creating a park factor that utilizes the park factors of statistical components utilized by our driving ERA estimator. FanGraphs provides park factors for each of strikeouts and walks which were utilized in our calculation. It should be noted that we utilized the 2018 strikeout and walk park factors for purposes of our calculations, as these park factors for 2019 have yet to be made available. However, given that the park factors for strikeouts and walks were the same in both 2017 and 2018 and only slightly different in 2016, we feel comfortable that these should generally reflect the impact of the ballpark on our analysis. We will provide an updated leaderboard when the 2019 park factor components become available.
Unlike fWAR, the park factors used for calculating kWAR do not assume that each pitcher has played half of their games in their home ballpark and the remaining games in a neutral ballpark. Rather, kWAR applies the park factors on a weighted average basis for each component to create a composite park factor unique to each player. As noted above, after the park adjustments are applied, kWAR follows the same process as fWAR with respect to adjusting for league, the use of a dynamic RPW, leverage, etc. For a more in-depth
discussion of the remaining steps to calculate fWAR (and as a result, kWAR), please visit FanGraphs.
What does kWAR do?
By focusing on outcomes solely within a pitcher’s control, kWAR attempts to better estimate a player’s true talent level relative to fWAR. Balls in play and home runs have some randomization and variability in them. What our goal is with this metric, is to give a closer evaluation of a player’s true talent level. Strikeouts are good, walks are bad, and the more strikeouts and less walks, the better the pitcher is most likely going to be, both presently and on a go-forward basis.
This metric can be used with other WAR values to give you a good sense of a pitcher’s overall ability. As with other WAR metrics, it is not the end all be all for a pitcher. kWAR and kwERA should be used with other metrics to figure out how valuable a player is. Given that kwERA has previously proven to be more predictive of future ERA than FIP, it is possible that kWAR may be more predictive of future performance than fWAR. Once prior season data is collected and calculated, we plan to test kWAR’s ability to predict future performance and its year to year reliability.
The Results
In our calculations for 2019, Gerrit Cole had a 10.1 kWAR and a 1.32 kwERA in 2019, both of which were tops in baseball amongst qualified starting pitchers in 2019. Justin Verlander also graded out well with a 1.74 kwERA and 9.0 kWAR. The two spearheaded a dominant 2019 Astros staff that made its way to the World Series. Both values are significantly higher than their respective fWAR, as FIP’s inclusion of home run prevention (something that both Cole and Verlander struggled with, ranking in the bottom 30 among qualified starting pitchers by HR/9) weighted down their elite strikeout and walk numbers.
Similarly, Chris Sale ranked 6th in all of baseball in kWAR despite his higher ERA in 2019 because of a particularly good strikeout rate. If Sale can recover successfully from Tommy John surgery, he should still be considered a force to be reckoned with and may have the potential to bounce back to his dominant form. On the other hand, Zack Wheeler accumulated a 3.5 kWAR, significantly less than his 4.7 fWAR season. As expected, Wheeler’s FIP (and as a result, his fWAR) was buoyed by Wheeler’s strong home run prevention rate with the 13th best HR/9 during 2019. Without the inclusion of home run prevention, Wheeler’s value (and kWAR) goes down. Since Wheeler has not historically limited home runs at an above-average rate relative to the league average, it’s possible the Phillies new front line starter is in for some regression to the mean and a decline in relative to his strong 2019 season.
When looking at relievers, Josh Hader and his astounding 0.50 kwERA led the way in 2019. That value was 0.55 better than the next closest pitcher, with a minimum 30 innings pitched. Hader’s low kwERA value and above-average workload for a reliever resulted in a 5.1 kWAR, nearly 2 wins higher than the next reliever.
Zack Britton 2019 resulted in a 4.37 kwERA and a -0.4 kWAR, well below the 0.9 fWAR he accumulated. Britton’s 8.6 K-BB% ranked in the bottom 30 in the league among qualified relievers but his 77.2% GB% was the highest among qualified relievers this past season, resulting in strong home run prevention numbers consistent with his career performance. While Britton is no longer in the same elite status for relievers as he was during his dominant run with the Orioles, kWAR casts doubt as to whether turn back the clock as his command drags down his K-BB% and. However, Britton’s outlier status, owing to his unique ability to induce groundballs at an absurd rate, may be a reason that kWAR would undervalue his performance.
The full leaderboard can be found at the bottom of this post. The results are for only 2019.
Notes
- All data, including the component park factors used to create the composite park factors applied to calculate kWAR, were pulled from FanGraphs. FanGraphs has not yet released the 2019 park factors, but as noted above the 2016–2018 K and BB park factor value by park were all within .01 of one another. The leaderboard will be updated to the extent the 2019 component park factor values differ from their 2018 counterparts.
- Any player who did not record a strikeout or walk was given a park factor of one. This only affects position players who pitched and certain pitchers who saw less than 3–6 innings of work.
- The leaderboard does include position players because it is for any player who threw a pitch in 2019 so enjoy some sweet Mike Ford stats!
Moving Forward
From here, the next step is to compile and calculate more years of data so we can then test the reliability of the statistic and its ability to predict future performance against other WAR metrics. It is unclear how many years we plan to do for testing, but the more years the better. We believe that this stat gives a good indication of true talent level and we want to see how our intuition measures up. We will also consider utilizing additional iterations of kwERA such as GBkwERA, to test for the most reliable and predictive WAR statistic. To do this, we will need help. We are seeking any additional help that can be offered during this time. My email and Twitter are at the bottom of the post, please contact me if you wish to help us with this project.
Thank you for reading and I hope to give you more information on the statistic as time goes on. A big thank you to Jason Fixelle for working with me on this project. His knowledge of FanGraphs and excel gave this project a lot of life. If you want to work with us to help expand this metric, even more, contact me on Twitter through @GreenfieldMax18 or email me at greenfield.max1404@gmail.com. Hope you are staying safe during this pandemic and wish you and your family good health. Thank you for reading!
Full Leaderboard here
