It’s HoF season again, and I’ve never been satisfied with the dicsussion around relievers. I wanted something that attempted to quantify excellence at the position while still being a counting stat, and what better way to quantify excellence at RP than by comparing to the average closer? (Please treat that as a rhetorical question)
I used the highly scientific method of defining the average closer as the aggregate performance of the players who were top-25 in saves in a given year, and I used several measures of wins. I wanted something that used all events (so not fWAR) and already handled run environment for me, and comparing runs as a counting stat across different run environments is more than a bit janky, so that meant a wins-based metric. I went with REW (RE24-based wins), WPA, and WPA/LI.
IP as the denominator instead of PA/TBF because I wanted any (1-inning, X runs) and any inherited runner situation (X outs gotten to end the inning, Y runs allowed – entering RE) to grade out the same regardless of batters faced. Not that using PA as the denominator would make much difference.
The first trick was deciding on a baseline Wins/IP to compare against because the “average closer” is significantly better now than 1974, to the tune of around 0.5 normalized RA/9 better.
I used the regression as the baseline Wins/IP for each season/metric because I was more interested in excellence compared to peers than compared to players who were pitching significantly different innings/appearance. WPA/LI/IP basically overlaps REW/IP and makes it all harder to see, so I left it off.
For each season, a player’s WAAC is (Wins/IP – baseline wins/IP) * IP, computed separately for each win metric.
Without further ado, the top 20 in WAAC (REW-based) and the remaining HoFers. Peak is defined as the optimal start and stop years for REW. Fangraphs doesn’t have Win Probability stats before 1974, which cuts out all of Hoyt Wilhelm, but by a quick glance, he’s going to be top-5, solidly among the best non-Rivera RPs. I also miss the beginning of Fingers’s career, but it doesn’t matter.
|Career WAAC based on||REW||WPA||WPA/LI||Peak REW||Peak Years|
|Dennis Eckersley (RP)||3.4||-0.3||5.2||7.1||1987-1992|
|John Smoltz (RP)||3.2||9.0||3.5||3.2||2001-2004|
|Kenley Jansen (#20)||3.1||2.6||3.5||4.3||2010-2017|
|Lee Smith (#25)||2.5||0.9||1.8||4.5||1981-1991|
|Rollie Fingers (#64)||0.7||-6.0||0.8||1.9||1975-1984|
|Bruce Sutter (#344)||0.0||2.4||3.6||4.3||1976-1981|
Mariano looks otherworldly here, but it’s hard to screw that up. We get a few looks at really aberrant WPAs, good and bad, which is no shock because it’s known to be noisy as hell. Peak Gossage was completely insane. His career rate stats got dragged down by pitching forever, but for those 10 years (20.8 peak WPA too), he was basically Mo. That’s the longest imitation so far.
Wagner was truly excellent. And he’s 3rd in RA9-WAR behind Mo and Goose, so it’s not like his lack of IP stopped him from accumulating regular value. Please vote him in if you have a vote.
It’s also notable how hard it is to stand out or sustain that level. Only one other player is above 3 career WAAC (Koji). There are flashes of brilliance (often mixed with flashes of positive variance), but almost nobody sustained “average closer” performance for over 10 years. The longest peaks are (a year skipped to injury/not throwing 10 IP in relief doesn’t break it, it just doesn’t count towards the peak length):
Rivera 17 (16 with positive WAAC)
Wagner 15 (12 positive)
Hoffman 15 (9 positive)
Wilhelm 13 (by eyeball)
Smith 11 (10 positive)
Henke 11 (10 positive)
Fingers 11 (8 positive, giving him 1973)
O’Day 11 (7 positive)
and that’s it in the history of baseball. It’s pretty tough to pitch that well for that many years.
This isn’t going to revolutionize baseball analysis or anything, but I thought it was an interesting look that went beyond career WAR/career WPA to give a kind of counting stat for excellence.