The 2016 MLB Playoffs may be remembered for the Chicago Cubs finally breaking the curse, but for many baseball fans including myself, it will be known for changing the way we view a bullpen. In recent seasons, managers have been using their bullpens much differently. Players like Dellin Betances and Wade Davis have been used in “fireman” roles, being utilized in the highest-leverage situations of the games.

The storylines around the usage of relievers were bountiful, beginning with Buck Showalter leaving Zach Britton — arguably the most dominant inning-for-inning pitcher in the league — to watch as Ubaldo Jimenez gave up a walkoff, three-run home run to Edwin Encarnacion. The Indians used Andrew Miller for multiple innings at a time, and the lanky left-hander was downright brilliant, pitching to a 1.40 ERA in 19.1 innings and striking out just under 14 batters per nine.

On the other side of the coin, Cubs manager Joe Maddon tried to ride closer Aroldis Chapman to a World Series title, and the flame-throwing lefty blew the lead and almost the title for the Cubs, leading Chapman to criticize his usage by the team. It is clear to many within the game and to many of us who just watch and try to explain the game that bullpen usage is changing. The Cardinals have reportedly discussed stretching out former closer Trevor Rosenthal to use him in multi-inning relief ace/fireman role this upcoming season.

In Brian Kenny’s dream world, we will reach a point where teams employ their bullpens at will, removing the idea of the traditional starting pitcher. This theory he calls “bullpenning” will take advantage of the successes of pitchers the first time through the order and the uptick in ability pitchers see when they throw one inning at a time. While we are not heading to that extreme just yet, it is clear that teams are looking to add as many dominant relievers as they can.

In 2015, I developed a way of determining the value of relief pitchers using a metric I called RWAR. The methodology and the logic behind the statistic can be found here. I have often tried use this metric to determine what it takes to dominate in high leverage situations.


To this end, I recently took every individual relief pitcher season dating back to 2012 and calculated the RWAR. Then I looked at each season that I considered be “high leverage.” For this, I took all pitchers who threw more than 40 innings and finished with a leverage index above 1.5. This gave me a sample of 214 pitcher seasons, ranking from Britton in 2016 at 5.4 wins at the high end to 2012 Heath Bell and 2015 Neftali Feliz tied at the low end, costing their teams 2.2 wins each.

**The next three paragraphs will discuss some slightly higher level statistics concepts***

Using this sample of pitchers, I ran correlations to determine the most significant statistics for relievers. My results showed the strongest correlations were with BB/9, HR/9, GB/FB ratio, and Swinging Strike Percentage. The exact correlations are not significant, but it supports many of the ideas that back up current thinking on what makes a great reliever. Teams typically like hard throwers who induce a number of swings and misses while keeping the ball in the ballpark. It is often hard to string together hits against any reliever, let alone elite closers, so the guys who typically fail are those with extreme homer or walk issues. Walking multiple guys in one inning can lead to the offense only needing one hit to score the tying run, as opposed to the typical two or three hits needed with no one on base.

Using these four metrics, I ran a PCA or principal components analysis. This method is explained here. Since HR/9 and GB/FB ratio have a strong correlation between them, I needed to get a component combining the two. The analysis provided me with two components, and since the first one explains 70% of the data I used. Then with the other two metrics, I ran a linear regression on ADREIP. ADREIP is the efficiency of a late inning reliever and is used in the RWAR calculation.

Using this regression, I then predicted each player’s ADREIP based on their underlying statistics. The regression had an R-squared just below .4, and all components were significant on a 0.05 confidence level. Using this new xADREIP, I found there was a correlation of 0.6 between this value and RWAR. This suggests that this model is a solid indicator of potential success as a late-inning reliever.


Based on this analysis, it is easy to see why Britton is one of the most dominant relievers in the game. He limits walks with above average K numbers and his ground-ball tendencies allow him to limit home runs. Three of Britton’s seasons appear in the top 10 in xADREIP among late-inning hurlers. Chapman and Mark Melancon are also repeats in the top 10 by this analysis, and both signed massive contracts this season to lead the New York Yankees and San Francisco Giants bullpens in 2017. Brad Ziegler, the submariner who has excelled as a closer, finds himself at 11th in the xADREIP stat. Ziegler does not have high K numbers, but his exceptional control and extreme ground ball tendencies have allowed him to succeed in the late innings. This suggests that this analysis can be helpful in finding guys who may succeed despite not having the typical hard-throwing profile.

While it is easier in hindsight to say why one reliever was more successful than other, it is much harder to predict future success. Armed with a knowledge of what makes relievers successful, I will determine if these same underlying peripherals have predictive power. Then, I will take this knowledge to determine players who have the makeup of elite relievers going forward.

About The Author

Paul Mammino

A left handed pitcher at Wagner College, Paul is studying mathematics and is an avid fan of baseball statistics. A contributor for Baseball Essential he formerly contributed for and You can follow him on twitter @paulmammino

Related Posts

Leave a Reply