Since a stolen base when the game is close and/or late is more important than a SB in other situations, it raises the issue of whether or not some players can "pick their spots" and do more of their stealing when it really matters. Just like with clutch hitting, which has been extensively researhed, we could also ask: How much difference would the clutch stealing make or how many extra wins would the most clutch stealers add? I take a look at these questions.
The data comes from Ed Oswalt's "The Baseball Player Value Analysis System." There, he lists data for how many wins a player added with his basestealing. Here is an example of what he means: "In the example of the seventh inning situation above, if the runner steals third, the home team's chance of winning improves from .45253 to .49368. This improvement of .04115 is a contribution of the baserunner." He did this for each player's SBs from 1972-2002. Then he added up the change in value for each SB and caught stealing (CS, which has a negative value) to get base stealing wins (or BS Wins). I sorted Oswalt's data and here are the top 25:
So those are the top clutch base stealers. Of course, they are all, for the most part, known for being good stealers.
To see if players were getting more SB wins than expected, I first calculated the value of all of their SBs and CSs using the run values of .22 and -.38, respectively. These are the values that Pete Palmer uses in the "Baseball Encyclopdia." When those values are applied to a given player's SB and CS totals, the result is called "base stealing runs" or BSR. The correlation between Oswalt's SB wins and Palmer's BSR is .911 (a perfect correlation is 1.00-I used layers with 5,000 or more plate appearances (PAs) from 1972-2002). If you square that, you get the "r-squared," which is .83, meaning that 83% of the variation across players in SB wins is explained by BSR.
Then I divided that BSR value by 10, which is the number of additional runs over the course of a season that it normally takes to add 1 more win (givne the way Oswalt calculated BS wins, this not the most accurate or precise way to do this-see notes at the end). This gives an expected number of wins a player adds from his SBs. That got subtracted from Oswalt's SB wins to see how many more or fewer games a player added from his "clutch stealing" or ability to steal (and not get caught) when it mattered most. The table below shows the top 25 in added SB wins.
Whitaker had 143 SBs and 75 CSs. Using the values of .22 for a SB and -.38 for a CS, he should have 2.96 BSR or "base stealing runs." That divided by 10 (runs added for a win) leaves .296 stolen base wins using the Palmer values. Oswalt gives Whitaker 4.956 SB wins. That exceeds his predicted value by 4.66. But Whitaker had nearly 10,000 career plate appearances (PAs). With 700 or so PAs being a full season, that means he added about .33 wins a year by being able to steal when it mattered most.
The table below shows the top 25 in terms of added SB wins per 700 PA.
So even the most clutch stealers barely add one-third of a win per season. But I wonder if Kent Hrbek and Wade Boggs had reputations for clutch stealing?
I also looked at SB wins and BSR per time reaching first base (walks + HBP + singles). Over this time period, the average player reached 1B 189 per 700 PAs (of course that changed over time as the average on-base percentage changed). The table below shows the number of SB wins above what would be expected based on BSR per 189 times reaching 1B.
It looks like there is not alot of clutch stealing going on. The value of stealing for each player, if you take the game situation into account (inning, closeness of score), is not much different from what we already knew about using Palmer's SB and CS values.
Notes: The higher the league average in runs scored per game, the more additional runs it will take to add 1 win. In some years it is more than 1. Oswalt used each season's data to calculate how much the probability of winning would change with any given event. So a SB in a low scoring environment would probably help more than in high scoring environment. Guys who played more in low scoring years should probably have their BSR divided by a lower "runs per win" value than guys from the higher scoring years. So what I have here is an approximation (using the 10 runs). But if I divided each guy's BSR by a different "runs per win" value based on the league average, I probably would have gotten an even higher correlation between SB wins and BSR and we would see even less "clutch stealing."
Source: The Lee Sinins Complete Baseball Encyclopedia