Vince Coleman was the last baseball player to steal 100 bases in a season in 1987. The highest total over the past 10 years was Kenny Lofton's 75 swipes in 1996. Where has all of the speed gone?
Stolen bases have gone out of fashion faster than Rickey Henderson's lime green batting gloves. The real reason that totals have decreased is because managers are sending out the runners less, and increasingly relying on the long ball. But is this an effective strategy?
Any sabermetrician will tell you that "the value of the stolen base is generally too small to warrant the risk of being thrown out." So why does anyone steal bases? Why, therefore, was Dave Roberts stealing second with no outs in the ninth inning, jeopardizing the greatest comeback in history? Stealing bases late in games, in highly leveraged situations, can drastically increase the probability of winning and, therefore, should be used in the proper late game situations.
The above graph shows that the mean frequencies of stolen base attempts per team per season have been falling, while homerun frequencies have been rising. The period of the graph, from 1995-2004, was also known as the "juiced era." The value of the stolen base is smaller in the "juiced era" for many reasons, including the fact that it was never very valuable, but also because a home run drives in the runner no matter if he is on first or second.
By running the 1995-2004 data through a regression function to calculate the "linear weights" of each offensive possibility (single, double, triple, etc.), a function was derived in which the coefficients show how many runs on average each offensive event was worth.
The equation is: Runs = 46.27 + (.565)1b + (.711)2b + (.986)3b + (1.44)Hr + (.32)Bb + (.215)Sb - (.33)Cs - (.125)Out.
We can see that stealing a base adds .215 runs and getting caught stealing subtracts .33 runs. These numbers are consistent with historical values as computed by "Total Baseball." Thus, the value of stealing a base has remained constant, while its frequency has drastically decreased. A comparison of the value of stolen bases and home runs yields the fact that seven stolen bases without being caught are equal to the value of one home run.
The breakeven stolen base rate is the percentage a player has to steal at to neither add runs nor take runs away from his team. From the regression, this is seen as 65.15 percent based on the coefficients of SB and CS.
During the "juiced era" teams only stole at 69.15 percent frequency, which also shows that on the macro level, per steal, the expected number of runs added is only a mere .047. This clearly illustrates how insignificant stealing bases has been over the past ten years: the average number of runs added per stolen base attempt was 47/1000 of a run.
Two statistics are also helpful in evaluating base stealers: Stolen-Base Runs Added (SBRA) and Stolen-Base Efficiency (SBE). These stats were computed as follows:
SBRA= ?SB x (SB) + ?CS x (CS)
SBE= SBRA/(SB +CS)
where ?SB = .125 = the number of runs added per stolen base (seen as the coefficient of stolen bases from the multiple linear regression model).
?CS = -.33 = the number of runs added (subtracted) per time caught stealing (seen as the coefficient of caught stealing from the MLRM).
Using Stolen Base Runs Added and Stolen Base Efficiency, I determined the top base stealers of the past 10 years.
Top 5 SBRA1. Tony Womack, 1999: 11.19 2. Scott Podsednik, 2004: 10.763. Tony Womack, 1997: 10.594. Kenny Lofton, 1996: 10.525 Chuck Knoblauch, 1997: 10.03
Top 5 SBE1. Carlos Beltran, 2001: .1982. Doug Glanville, 1999: .185 2. Eric Byrnes, 2004: .1854. Darin Erstad, 2004: .1835. Luis Rivas, 2004: .181
This list shows the best single season performances over the past 10 years. The insignificance of the stolen base is seen once again: the best SBRA's only resulted in adding a single win to their team's tally (using Pythagorean win percentages). In this era, the top 50 base stealers managed to add only 3 to 4 runs per season, on average. Some players, such as Coco Crisp, Luis Castillo and Alex Sanchez, added no runs, or actually took runs away from their respective teams via the steal.
The second list is something more useful, because it shows the most efficient stealers. These players are useful because when a steal is needed late in a game or in the playoffs, they will probably come through. Still, stealing bases is not the main aspect of these players' game.
We can think of the lists as two groups: SBRA are players who just try to steal as many bases as possible. In other words, they steal above the break even point. The SBE players are the stealers that sabermetricians could cope with: they do not steal as frequently, steal above an 85 percent success rate, and can be counted on for those micro steals for which the significance can not be disputed.
A second graph - one too complicated to be reprinted in black and white - serves as an accurate depiction of how valuable many of the so-called "premier" base stealers are to adding runs to their team. The scatter plot includes the top 50 base stealers from each of the past 10 seasons with stolen base runs added on the y-axis and stolen base attempts on the x-axis. For example, Juan Pierre in 2004 stole 45 bases but was caught 24 times. According to the regression coefficients, his steals added .215 x 45 = 9.675 runs, but his number of times caught stealing cost his team 24 x -.33= -7.92 for a total runs added of 1.755.
Every player below zero in stolen-base-runs-added hurt his team by getting caught stealing and losing runs for his team. For example, Luis Castillo in 2003 who stole 21 bases and was caught 19 times.
So, the next time you see managers letting Podsednik, Pierre, and Castillo run wild on the base paths, think about the value of a stolen base. The green light on the base paths should be reserved for those who steal with extremely high efficiency like Carlos Beltran and for those circumstances in which steals can change the outcome of a game.
