Getting Geeky Part 2: Leverage Index and Your Blood Pressure
Image courtesy of © David Berding-USA TODAY SportsWednesday we talked about the movement away from the closer dominant hierarchy that has ruled MLB bullpens since the 80s. The Twins are leading that movement, and were surprised to find reliever Tyler Duffey casually drop the sabrmetric stat Leverage Index (LI) in an interview. It’s useful for understanding the changes the Twins bullpen is making, so let’s learn a little bit about it.
If you were hooked up to a blood pressure monitor while watching a baseball game, the readings would look a lot like Leverage Index. Things get tense, and Leverage Index goes up. Things relax, it goes down. In fact, Leverage Index is easier than blood pressure – it’s not two numbers over each other. It’s a single number that you compare to 1, which is an average MLB at-bat. Lower than one, your blood pressure is probably OK. If it’s over one, you’re sitting forward in your seat. If it’s over two … maybe just keep the defib handy.
That’s because Leverage Index estimates how important an at-bat is in a game – before the actual at-bat happens – and does so objectively. For instance, you and your potential myocardial infarction know that an at-bat when your team is up by four with one out in the bottom of the fifth inning and the bases empty (LI=0.2) is less important than a one-run lead in the top of the ninth with one out and runners on the corners (LI=5.7). You don’t need LI to tell you that. But how about this?
On August 22nd last year, the Twins were facing the Tigers. Leading 2-1, Trevor May walked Niko Goodrum in the sixth inning to load the bases with two outs, and was replaced by Tyler Duffey. Later, when the game went to extra innings, Taylor Rogers started the top of the 10th, which meant he started it with a runner already on second.
So which reliever was brought into the game at a more critical time?
According to Leverage Index, Duffey was. It was his highest leverage appearance of the season. It had an LI of 4.51, which means it was 4.51 times more important of an at-bat than an average at-bat. Rogers’ was important, too, but it was just 2.5 times more important than an average at-bat.
(By the way, neither reliever got very lucky in that game. Duffey had the tying run score on an error and Rogers watched a ground ball squirt between short and third to lose the lead. But the Twins won the game on a walkoff hit by Max Kepler in the bottom of the tenth.)
The bad news is that it takes a lot of computational power to devise LI values. But the good news is that someone else already did that for you. So if you’re really interested in how LI is devised, dive into this next italics area. Otherwise we’ll meet again at the bottom.
How To Compute Leverage Index
- Chart several decades of play-by-play for MLB games and determine the probability a team has for winning a game in all specific situations. For instance, if you found that road teams that were down by one with runners on the corners and one out in the seventh won 48 out of 100 games, then their chance of winning would be 48%. Do this for all run differentials, innings, outs and men on base situations.
- Now for all those, figure out what could happen next. In the example above, they could hit a single, scoring the runner from third, moving the other runner over, and likely increasing their chance to win. Or he can hit into a double play, ending the inning, scoring nobody, and decreasing their chance to win. Both are pretty big outcomes. So is striking out, or hitting a home run, or all the other things that can happen.
- Now establish odds of all those scenarios happening, and multiply them by the absolute value of the differences from Step 2.
- Do the same computation for all the at-bats in a game. Use that to determine what the average difference is over a whole game.
- For each scenario, divide #3 by #4. If it’s over 1, it’s higher (and more tense) than average. If it’s lower, it’s lower than average.
If you want to dive further into LI, you can also find a primer (that is far more detailed than I provided) here on Fan Graphs. While you’re there, you can find average LIs for various pitchers on their player pages. For instance, the Twins relievers, sorted by the average LI they faced when they entered a game last year is here.
That seems like a good place to start tomorrow, as we take another look at whether Baldelli is truly treating his bullpen differently than most other managers, by looking at LI.
Next: Using a Bullpen With Leverage
- Dman and ToddlerHarmon like this