Troy Percival's windup drove me crazy. He had a fast, high leg kick and then seemed to tumble down the mound like Chris Farley on a coffee table. The windup seemed so quick, so erratic, so nearly unrepeatable, that I was surprised pitches ever made it to their destination (and near the end of his career, they typically didn't).
Sadly, in a phenomenal career, marked by a 3.17 ERA and a 3.87 FIP, Percival's time with the Rays featured a ~6.00 FIP and scores of fans in the hospital with cardiac and pulmonary issues. His bombastically bad tenure in Tampa -- and the team's outstanding results in spite of it -- has made me suspicious of bullpens and their usefulness.
Also, as a part-time fan and interested follow of the Chicago Cubs, I must yearly endure off-season signings most painful: Average relievers given above-market deals. Meanwhile, Andrew Friedman continues to cobble together useful bullpens like a Junkyard Wars episode. So, in effect, it seems: Bullpens don't matter.
If Friedman makes his bullpen -- year after year -- on the cheap, and Jim Hendry makes his bullpen -- year after year -- on a Hollywood budget (complete with Michael Bay explosions), then why bother investing in bullpens!? The Rays made it to the World Series with a 6.00-FIP closer!
So, let us endeavor, together, to discover the true value of bullpens.
In this first act, let's look at the last four years of bullpennery. The first question we need to ask is simply: Is there any correlation between a bullpen's results and a team's record?
To explore this question, I have created an interactive chart below. Go ahead, scroll down the page and take a look at it. Go on.
(The numbers next to the data points are win totals for that team.)
Observe the sliders in the middle -- these are filters we can use to change the data set. By moving the 'Pen ERA slider, we can look at the win-loss record of just team with, say, ERAs ranging from 4.09 (just above league average for this period of 'pens) to 6.16.
If we play around with the W-L% slider, then we can constrain our field to just winning teams (.501 to .636 W-L%).
Toy around with these functions, and then continue reading for some analysis.
NOTE: For our purposes, ERA does a sufficient job. We are not necessarily looking for the most talented bullpens, but rather how a bullpen's results affect a team's W-L record. Runs allowed (RA/9) is probably an even better measurement than earned runs allowed, but for now, just shut up.
ALSO NOTE: I've included the FIP to ERA chart on the bottom to reassert we are not examining bullpen talent, per se, but bullpen results.
Observations:
- I expected the red regression line would be as flat as Hak-Ju Lee's chest, meaning bullpens have zero affect on a team's success. I was, clearly, wrong. No matter how many regressions I ran, the result were similar: Bullpens are not useless.
- If we constrain our field to just winning teams, the regression line becomes much flatter. Smaller ERAs have increasly less affect on the team's success. This could mean a variety of things...
- Such as: There are diminishing returns on elite bullpens. This coincides with the results of my polynomial regressions, where the net results of improved ERAs nearly flat-lined after a 3.50 ERA.
- Math talk, science words, blah blah blah, regressionary co-linear, but multiple causality? Yes, but only in quantum reversal kittens.
- If we constrain ourselves to look at teams with only below-average ERAs (i.e. >4.08), then we see: (a) lots of losing team, but (b) a few very good teams (including two 97-win teams and two 94-win teams).
