Stealing bases when it counts

One of my earliest motivations to set up my own play-by-play database was to see if I could determine which players stole bases when his team really didn’t need then. I’m not sure why this particular brand of stat-padding bugs me more than any other, but it was enough to inspire one of my first Retrosheet-fueled witch hunts.

I’m going to bet you know exactly what I’m talking about. Your team is up five or six runs, it’s the bottom of the eighth. Johnny Quickpants is on second and he really has no business thinking about stealing third. But he steals the base anyway. Why? I suppose if the base is available with little doubt of being thrown out then you should take it, sure. But in your gut you suspect Johnny just wants to see some figure between 40-50 under the “SB” column on the back of his baseball card, regardless of whether it helps his team or not.

I was reminded of this old crusade recently when Tom Tango rekindled those old flames of fury by asking to see “how many ‘empty steals‘ runners make” in response to my last article.

My original attempts at unearthing this brand of fleet-footed stat-padders was limited to analyzing changes in run expectancy. But now, with access to the Fangraphs win probability database, we can view a player’s stolen bases through the prism of run expectancy, Leverage Index and Win Probability Added as well.

For those not familiar with these metrics, our Hardball Times glossary defines Win Probability Added (or WPA) as “the impact each specific play has on the team’s probability of winning” and Leverage Index (LI) as “a measure of how critical a specific batting situation is.”

Let’s get to it then by looking at the average Leverage Index for all players that made at least 100 successful stolen base attempts since 1972. And for the time being let’s leave out those steals in which the player was the trailing runner in a double steal. (If you’d like, we can address this in the comments—how much credit should the trailing runner be given in a double steal? Half? Less than half? None?)

Here are the players who stole bases most often in non-critical situations:

Lowest average LI on steals (min 100)

Rnk NAME SB* SB_LI SB_raa SB_WPA
1 Damion Easley 107 1 0.16 0.014
2 Kirby Puckett 125 1 0.14 0.014
3 Bernie Williams 141 1.01 0.145 0.013
4 Howard Johnson 227 1.02 0.158 0.014
5 Jason Kendall 178 1.04 0.163 0.015
6 Ivan Rodriguez 125 1.05 0.161 0.014
7 Ryan Theriot 116 1.05 0.167 0.017
8 Jim Gantner 130 1.06 0.158 0.017
9 Dick Schofield 116 1.06 0.153 0.016
10 Grady Sizemore 132 1.06 0.185 0.017

Hall of Famer Kirby Puckett stands in a virtual tie with journeyman infielder Damion Easley with the lowest average LI for their stolen bases. Their shared average LI of just “1.0″ suggests they committed some of the most needless and boring bag-thefts one can imagine in the game of baseball.

Naturally the pair also demonstrate a weaker average WPA on their stolen bases at just .014, significantly lower than the league average mark of .019. Yet long time New York Yankees center fielder Bernie Williams returns the lowest figure of the group with an average of just .013 WPA for his 141 non-trailing stolen bases.

Of course, one could certainly argue that it may have been difficult for Williams to contribute much WPA as a member of a franchise that seemed to always have a big lead. He was, after all, a staple in the lineup of arguably the most successful dynasty of all time. But Bernie did not add much run expectancy with his stolen bases either, with just .014 runs above average in his career, tenth lowest amongst qualifying base stealers.

This suggests that Bernie’s steals weren’t simply less important because of the score, but also when considering base and out state. Not all steals are created equal in the eyes of run expectancy, after all. A steal of third base with two outs will get you very little love from RE24 (just .05 runs or so), but a steal of second with no outs increases your team’s run expectancy enormously (about .3 runs).

Coincidentally, another New York superstar, David Wright currently owns the lowest average RE24 on his stolen bases at just .0139 runs.

I suspect that some of the players on this list simply made a habit of stealing a base when the defense wasn’t expecting it. When the game is in a low leverage state, the middle infielders might be caught daydreaming, or the first baseman may not feel the need to hold a slow-footed base runner. In these cases a Jason Kendall or Ivan Rodriguez might see an opportunity and ask himself, why not steal a base here?

Should they be penalized for this? Absolutely not. They still increased their team’s chances of winning and scoring runs, just to a lesser degree than average. If, instead, we wanted to look only at players who were true stat-padding base-stealers, we might opt to raise the minimum requirement a bit. In that case we find that players like Edgar Renteria, Reggie Sanders, and Lloyd Moseby to be the most offending, as they all ammassed at least 250 steals with an average LI below 1.16 during their careers.

Conversely, here are the speedsters with the highest average LI on stolen bases minimum 100 since 1972:

Highest average LI on steals (min 100)

Rnk NAME SB SB_LI SB_raa SB_WPA
1 Matt Alexander 100 2.2 0.185 0.039
2 Bobby Brown 109 1.68 0.168 0.028
3 Willie Bloomquist 124 1.65 0.192 0.029
4 James Mouton 109 1.6 0.162 0.023
5 Eric Davis 339 1.6 0.164 0.022
6 Miguel Dilone 267 1.58 0.18 0.025
7 Larry Lintz 116 1.54 0.177 0.024
8 Rodney Scott 201 1.54 0.186 0.024
9 Damian Jackson 131 1.51 0.183 0.023
10 Tim Raines 793 1.5 0.177 0.022

Here we find a handful of players with just enough to creep over the 100 steals requirement, and that might include one of the world’s most famous replacement level icons in Willie Bloomquist. Regardless of their low totals, I still find figures like Matt Alexander’s record high .039 career average stolen base WPA absolutely astounding. As you might have already guessed, the reason for this grossly impressive figure is an egregious amount of pinch running.

As it turns out, no player since 1950 has seen more pinch running appearances than Matt Alexander. In fact, using Retrosheet play-by-play files, I count a whopping 85 of his 103 (total) stolen bases that were achieved as a pinch runner. Similarly, other players from the list like Miguel Dilone (176) and Larry Lintz (140) also amassed some of the highest pinch running appearances in the Retrosheet era as well.

So that a player like Eric Davis was able to achieve an average LI of 1.6 over the course of an incredible 339 steals is hugely impressive. But it almost seems like small potatoes compared to the timeliness exhibited by legendary speedster Tim Raines. ‘Rock’ held onto an average LI of 1.5 over the course of his 23 year career and nearly 800 non-trailing stolen bases!

But to fully appreciate the timeliness with which Raines used his speed, we should compare him to the other great base stealers of this era. Among all 38 players with at least 350 non-trailing stolen bases, Raines beats them all in average Leverage Index and WPA. Other base stealing greats like Ricky Henderson, Vince Coleman, and Willie Wilson all had average LI’s of just 1.4.

Of course the true mark of superior base stealing skills reveals itself when we include caught stealing LI and WPA as well. Raines’ average WPA on all his stolen base attempts ranks fifth best since 1972. He trails a handful of active base stealers that have yet to lose their speed, and don’t have nearly the amount of raw career stolen base totals to their credit– those four are Carlos Beltran, Jayson Werth, Chase Utley, and Brett Gardner.

More to come

Extracting stolen bases from a player’s LI, WPA, and RE24 totals is something I intend to pursue quite a bit in the future here at THT. But if you’d like a sneak peek at some of the data, here’s a google doc with most of the data I used for today’s article along with some other interesting information. If there’s any thing you’d suggest this line of inquiry ought to include or do differently, please drop a line in the comments.

References & Resources
Thanks to Fangraphs and Retrosheet.

0000
« Previous: The Baseball Reliquary and the Eternals 2013
Next: THT forum: What can we do better? »

1. Marc Schneider said...

I wonder if you could do something similar with RBIs.  For example, a guy comes up in the 8th inning with his team ahead (or behind) by six runs, runner on third, less than two outs, he hits a ground ball that gets the runner in but has no discernible affect on the game-or, in some cases, actually reduces the team’s probability of winning by making an out and effectively killing a rally.  Of course, no one is suggesting the guy shouldn’t be trying to drive the run in, but that always seemed to me a reason why RBI is such a misleading stat.

2. James Gentile said...

@Cyril, Thanks for the links. I’d like to tweak the method a bit and make it *wholly* representative of what a player contributed—this means splitting the credit on double steals and strike-em-out-throw-em-outs and so on. Slowly but surely we’ll get there.

@Marc, that’s a terrific idea and I may follow up on that! It could be useful to help explain the flaws of RBI to those who still swear by it, but I’m skeptical that such a fan is going to welcome the concept of Leverage Index with open arms.

@John, that’s a great point. And I certainly don’t blame anyone for trying to make a little extra cash. Though don’t discount that we may see some saber-leaning arbitrators in the (distant) future.

3. Cyril Morong said...

You’re welcome

4. John Fox said...

I don’t think “Johnny Quickpants” (from 1st paragraph) is looking to pad his statistics for the back of his baseball card so much as he is looking to pad them for the next time his contract is up, especialy if it is in arbitration.  An arbitrator is likely to be swayed by big numbers of steals, not likely to delve into when they were relatively meaningless steals.

5. cooldrive said...

Perhaps a player would steal in a high leverage, non-crucial situation to make the defense(s) more apprehensive and likely to err in future crucial situations.
Didn’t Cobb run into outs at 2nd base after singling in lopsided games in order to get the outfielders thinking “Oh, it’s that crazy Cobb again,” putting extra pressure them?

6. Gunter Angermayr said...

While I don’t really have any numbers available to delve into this, there surely must be some impact on base stealing by a manager’s inclination to call for or refrain from calling for stolen base attempts. This is probably more significant in high-leverage situations. My gut feeling tells me that in high-LI situations, a baserunners’s decisions will be codetermined by a manager’s strategic decisions more so than in low-LI situations.