Sabermetric Mining – Leverage Index

Posted on 07 September 2012 by Blake Murphy

With the baseball regular season winding down, fantasy owners have very little time left to make appreciable gains in the standings. With rosters expanded, out-of-contention teams experimenting, and injuries shutting players down early, September baseball does not always resemble what we see from April through August. Thus, it can be difficult, albeit valuable, to mine for advantages this late in the fantasy season, especially with most trade deadlines having long since passed.

With that in mind, today’s Sabermetric Mining piece will look toward next season a bit more than usual, although there are still practical rest-of-season implications for save chasers. Today, we will examine pLI, or Leverage Index, a statistic that can be used to identify how relief pitchers are deployed and hopefully give us insight into their future saves potential.

The Stat
pLI – Leverage Index is an index of how important a game situation is. Based on the game situation (inning, score, outs, etc), it assigns a grade to how important the situation is when a pitcher pitches, across the totality of their appearances. The higher the Leverage Index, the more the game is “on the line,” when that pitcher is on the mound. A pLI of 1 is a neutral situation, while roughly 10% of situations have an LI of 2 and 60% have an LI of less than 1, per Fangraphs.

inLI – This is Leverage Index broken down to more specific situations, in this case only the LI when a pitcher starts an inning. This is generally a good indicator for closer usage, since few managers will still deploy their closer at any time except the start of a new inning.

gmLI – This is Leverage Index broken down to just when a pitcher enters a game, and thus more often includes runners on base. This is generally a good indicator for identifying relievers that managers trust a great deal as their “firefighters” so to speak, brought in to handle tough situations.

How To Use
Unfortunately with Leverage Index, we are doing an analysis that involves intuition, logic, and attempting to predict the actions of sometimes irrational managers. I won’t get on a tangent about managing to the save rule, but you will soon notice that if a closer is the best reliever on a team, they are sometimes deployed sub-optimally based on game situations.

With that said, we can use LI to attempt to predict future closers. The logic here is that if a manager trusts a reliever in high-leverage situations, they should, in theory, be in line for the closing gig should it open up. It can also help us to identify closers that appear to be closers in name only, those whom managers do not trust a great deal. These are closers that are likely to be replaced with a string of poor performances or with a manager change.

Again, we cannot take pLI as a clear ordering of the bullpen roles. Managers are a funny breed when it comes to bullpen usage, plus we can introduce unintended bias when using a catch-all like pLI due to the deployment of handedness specialists, ground ball specialists, and more, whereby a pitcher can be deployed for a particular skill rather than his overall effectiveness.

Given that we are taking pLI as only a rough indicator of bullpen hierarchy, it suffices as a general means of trying to predict future save opportunities. Our assumption will be that, for the most part, a high pLI or gmLI is indicative of a manager’s trust and thus, a manager’s likelihood of promoting that player to the closer’s role if the opportunity opens up.

Vinnie Pestano – Pestano has long been thought to be a potential closer-in-waiting thanks to strong success, a high strikeout rate, and great peripheral numbers. For the second year in a row, Pestano has a large pLI (1.80) indicating he pitches in situations that are 80% more influential than a neutral situation, on average. He has a gmLI of 1.82, further indicating that he usually enters the game in these high-leverage situations, rather than creating them due to poor performance. His exLI, which I did not explain but is the average LI when a pitcher exits the game, is 1.52, highlighting Pestano’s success in lowering the leverage of situations, on average from 1.82 to 1.52. Pestano is the ultimate firefighter and would make a great closer should Chris Perez stumble in the role.

Addison Reed – Reed is currently the closer, so this is not necessarily in the spirit of the analysis, but it is worth pointing out that Reed leads all relievers with a 1.99 gmLI and is second to Jonathan Broxton with a 2.09 pLI. Basically, Reed is being deployed more optimally than any other closer in baseball. His 4.28 ERA is inflated by some early-season struggles, and it’s clear from these numbers that he has the complete trust of manager Robin Ventura.

Josh Roenicke – There are multiple pitchers that fit this same narrative, but I chose Roenicke because as a prospect he was identified as a potential future closer. This year, Roenicke has a 2.67 ERA over 81 innings, which would lead some to tap him as the potential heir apparent behind Rafael Betancourt. Alas, beyond his 4.43 FIP we also see that Roenicke simply does not have the trust of his manager, checking in with a miniscule 0.67 pLI and a 0.64 gmLI. Even worse, his exLI is 0.89, meaning that he has increased the leverage of situations while pitching. While this could be skewed by things such as the Rockies catching up in games where they’re behind, it could also be indicative that he is getting into trouble, a narrative backed up by him being among the league leaders in Pulls, or times removed in the middle of an inning. Add it all up and the 2.67 ERA is a mirage, not backed up by peripheral pitching stats or his usage pattern.

Carlos Marmol – Marmol has been in and out of the closer role this season, but our leverage stats allow us to examine how he has been deployed overall. Basically, Marmol is our best example of a “closer in name only,” someone who is deployed based on the save rule but not used in important situations. Despite the 17 saves, Marmol is around the league median for reliever pLI with a mark of 1.28, while his gmLI is just 1.04, by far the lowest mark of anyone with at least 10 saves.

Candidates
Potential Buy Low – These pitchers have less than 10 saves on the season but have strong pLI, inLI, and gmLI marks. These are relievers that are trusted a great deal by their managers and may see closing opportunities down the stretch, or, for those of you in keeper leagues, next season.

Potential Sell High – You probably can’t sell off closers in most leagues at this point, but for those of you in keeper formats, these are pitchers with 10 or more saves but poor pLI, inLI, and gmLI marks, indicating they have yet to earn the full trust of their managers.

The saves chase is a difficult but necessary evil in most leagues. The axiom “don’t pay for saves” is a good one, but only if you can effectively identify those players who will be losing or acquiring closer gigs. Leverage Index stats are a good means of evaluating a manager’s trust in a pitcher, as well as a pitcher’s success relative to the game situation. Identifying pitchers deployed in high leverage situations can be a key asset for identifying future closers and thus, future sources of saves.

Follow me on Twitter @BlakeMurphyODC.
All stats courtesy of Fangraphs, for games through September 5.

2 Comments For This Post

  1. Gary Perilloux Says:

    Would appreciate a clearer explanation of how “it assigns a grade.” Obviously, the stat can’t be self-assigning, but this reads like a tautology and — without a more rigorous definition of the kind and degree of “game-on-the-line” situations — one can’t adequately judge the value of this stat. Yes, there’s something simmering here. You may have convinced me to hang on to Rodriguez for the remainder of the year (against my better judgment and against Axford’s steadily improving performance), but I find the “leverage” concept a bit muddled, as in there does not appear to be a clear demarcation between a pitcher who is merely a key setup reliever (and who may never advance to closer) in a bullpen versus a pitcher who is being used in critical positions and who may actually advance to closer. Therein lies a problem that is is not necessarily the “irrational manager” problem, but a problem of precise definition that foils the effectiveness of the statistic. For example, in a closer-by-committee situation, I would place little to no confidence in the Leverage Index and considerably more confidence in the stated intentions of the manager. But this is an interesting primer, and thanks for bringing it to us. (Suggestion: Could you include a team abbreviation in the future for those of us, unlike the guy on “Criminal Minds,” who do not have eidetic memories?)

  2. Blake Murphy Says:

    Hey Gary,
    Thanks for the well thought out response. Allow me to respond in kind of a scattered manner here…

    *In terms of how LI assigns a grade to a situation, I can direct you to this great resource from Tom Tango from Inside the Book – http://www.insidethebook.com/li.shtml . It’s a bit much to skim through, but based on historical Win Expectancies in different game situations, run expectancies in different at-bat situations, etc, it assigns an amount of leverage to any situation, i.e., how large an impact that situation would be expected to have on the final outcome of the game. Obviously, the later and the closer the game, the more “leveraged” the situation is, and the same goes for a bases loaded vs an empty bases situation. For more information on the statistical basis for the grades, Tom Tango had a great three-part piece for The Hardball Times in June of 2006, which I couldn’t possibly do justice in the 1000-2000 word range or this reply. You can find links to all three parts here – http://www.fangraphs.com/library/index.php/misc/li/ .

    *In terms of value, it may be of more value to look at pitcher usage relative to their counterparts in the bullpen (e.g. Axford and K-Rod, specifically when the closer role there was more fluid), although that is obviously a bit larger a time investment.

    *I understand the difficulty in using it as a catch-all stat for potential future closers as it does come with some assumptions. Your point about the “career set up man” is fair, and especially true for lefties (see: Scott Downs’ career). As with all of the Sabermetric Mining pieces, it was meant as an extra tool to add to your evaluations, and it seems like you took it as such anyway, so that’s great.

    *I will add Team abbreviations in future charts (and size them better).

    Hope this clears some of the concerns up. I definitely recommend at least skimming the Tango pieces I linked to. Feel free to email me if you would like to discuss further (or, of course, comment here). Cheers,
    Blake

3 Trackbacks For This Post

  1. Full Spectrum Article – Explaining Leverage Index « The On Deck Circle Says:

    [...] Check it out! If you enjoyed this post, please consider leaving a comment or subscribing to the RSS feed to have future articles delivered to your feed reader. Click here to cancel reply. [...]

  2. Sabermetric Mining – Leverage Index « Blake Murphy Sports Writing Says:

    [...] Sabermetric Mining – Leverage Index Date: September 7, 2012 Original Source: Full Spectrum Baseball Synopsis: The latest Sabermetric Mining piece dives into Leverage Index, a measure of how relief [...]

  3. Sabermetric Mining: Relief Pitchers Part 2 (LI & WPA) | Full Spectrum Baseball Says:

    [...] last week’s Leverage Index piece, I discussed potentially using Leverage Index and its derivative stats (gmLI, inLI, pLI, and exLI) [...]

Leave a Reply

Advertise Here
Advertise Here
BBA