Do Bad Teams Earn Good Saves Totals?
The Cubs are in trouble. Their lineup is pretty bad, their rotation only has a couple decent pieces, and their bullpen. Oh, their bullpen. By ERA, it’s only the eighth-worst pen in the league, but FIP (2nd worst) and xFIP (worst) tell a better story.
After Carlos Marmol blew up, they tried Rafael Dolis. His wildness relieved him of his duties around the same time the team decided Michael Bowden needed more time in the minors. Lefty James Russell and right Shawn Camp, both better cast as supporting, lower-leverage pieces in a better pen, are now sharing the role, with the also-underwhelming Casey Coleman looking in, ready for his chance.
Each update on the situation does goad reaction from the saves-hungry hoards, but there’s a more important question lurking behind. Should we care at all about messy situations like those in the Cubs pen right now?
This is a multiple-part question that needs more investigation, but step one is to look at the link between saves and winning percentage. Obviously, your team needs to win games to save any, but do more wins necessarily mean more saves?
Anecdotally, it seems like saves come out of bad pens. After all, the 2004 Diamondbacks lost 111 games, and Greg Aquino still managed decent ratios and 16 saves for them. Then again, not one pitcher logged as much as six saves for the record-holding 119-loss 2003 Detroit Tigers. So let’s run the numbers and find out how much of the variance in saves is explained by win percentage. Here is the relationship between saves and winning percentage for all teams in baseball since the free agency era begin in 1974:
The line is nice and updward, and the p-value is less than .0001, but we knew that there would have to be a positive relationship between wins and saves, and we knew that they would be tied to each other. Again, you need a win to even have a chance at a save. The news here might be that the slope of the line is modest. The r-squared value here is .183, meaning that about 18.3% of the variance in saves can be explained by winning percentage.
If team winning percentage is only about one-fifth of the secret sauce for saves, how much should we think about bad teams? Plenty.
Look at the bottom of the chart. There are about as many losing teams with fewer than 20 saves than there are winning teams. Blowouts in either direction don’t produce saves. And the intercept is at 20, so even the worst teams can reasonably be expected to cobble together at least 20 saves chances. If all those saves go to one player, that’s a decent pickup. It looks like drafting a closer on a winning team versus one on a losing team is like getting a closer with 30-save potential versus one with 40-save potential, to reduce this to a slogan.
So, yes, bad teams earn enough saves to continue to track their bullpens. Next, we’ll have to find a way to examine bad pens on bad teams in particular, and maybe even look at what happens to a pen in which a lefty might be their best pitcher. Then maybe we’ll really know if we can look away from this particular bad bullpen and focus on more rewarding pursuits.
Not meaning to nitpick, but it was the 2003 Tigers that lost 119 games.
“Next, we’ll have to find a way to examine bad pens on bad teams in particular…”
It seems like it would be helpful to look at the actual ability of closers in question. Not just bad pens on bad teams, but good pens (or good closers) on bad teams. Take a guy like Joakim Soria. He’s been on a bad team, but he’s a great closer.
So I might look at “save chances” as my response variable. If that has a slope even closer to 0, then as a nerdy sabermetrician, I would find the best FIP guys (or best K% guys, or whatever) on the worst teams and buy low.
Yeah, that’s what I was thinking, spend your money on Soria or League or Hanrahan instead of getting into a bidding war for Mariano.
you should really look at save OPPORTUNITIES and not just saves on the y-axis.
Here are the problems with using save opps — They occur in the seventh and eighth innings (see blown saves by setup men) and therefore don’t accurately reflect closer situations as much as bullpen situations.
Shouldn’t we be looking at save opportunities vs. wins? If a good closer on a bad team saves 30 games out of 30 and a bad closer on a good team saves 30 games out of 40 then both teams will be on the 30 save horizontal line – implying there is no difference between winning % and saves – which is incorrect in this example.
http://www.fangraphs.com/fantasy/index.php/more-about-bad-teams-and-saves/
As much as good teams win more games, some of the best teams actually offer less opportunities than medicare teams due to potent offenses and blow outs.
My totally unscientific rule of thumb has been to take closers from teams that have good overall pitching. The idea is that win or lose, you’re looking for the 3-2 game rather than the 8-4 game.
Looking at the past three years, the top ten teams in W/L have averaged 45.2 saves, while the bottom ten are at 36.1. Going off team ERA rather than W/L, it’s 43.0 to 36.3. Obviously there’s overlap as the ERA correlates to W/L, but each on their own seem to be relatively equal in terms of predicting saves.
I wonder…within the top/middle/bottom 10 teams in W/L, is there another correlation to to number of save opps within the smaller group?
How would the chart look if you use run differential instead of wins?
I think a problem you have with bad teams that have even good closers is that you can get vulture losses from time to time in what might have otherwise been blown save situations.