Designated HitterMay 24, 2007
The Value of the Stolen Base: A Comparison of MLB and NCAA Division I Baseball
By Mike Current and Chad McEvoy

Over the years there has been a great deal of debate amongst baseball insiders and fans over the value of the stolen base. Some, such as longtime Baltimore Orioles manager Earl Weaver, have argued that the stolen base is rarely worth the risk. Others, however, view the stolen base as a valuable means of applying pressure to the opposing team's defense. The question is: Which side is right?

Most past research on the stolen base seems to side with Weaver. Using data from Major League Baseball, researchers have found that stealing at less than a 75% success rate is detrimental to success. Joe Sheehan explains in Baseball Prospectus Basics: Stolen Bases and How to Use Them that when considering stolen bases, one must consider both the cost and the benefit. Therefore, the break-even point for successful base-stealing is so high because outs are more valuable than bases in nearly every instance. For example, the Run Expectancy Matrix created by Baseball Prospectus reveals that a runner on first base with no one out is worth approximately 0.864 runs. A successful steal of second base would raise that figure to 1.173. However, a failed stolen base attempt drops that number to 0.270. In this example, the loss is nearly two times the gain.

In the same article, Sheehan also suggests that the secondary effects of base-stealing, such as putting pressure on the opposing pitcher and defense, do not exist. In fact, he goes as far as to suggest that a runner at first base is more disruptive to the defense than a runner at second base, simply because the first baseman must hold the runner on and the middle infielders are forced to cheat toward second base to have a chance at a double play.

While these findings have been consistently replicated and are generally accepted by Sabermetricians and others when talking about professional baseball, there has been little or no research conducted examining the stolen base at other levels of play. As a Division I college baseball coach, this leads me to wonder: Is the stolen base a more valuable offensive weapon in college baseball than it is at the professional level?

The numbers seem to indicate that the stolen base is more a part of the college game than it is the professional game, even to the casual fan who has taken a few minutes to compare player and team statistics from both levels. For example, in 2006, the Los Angeles Angels led all of Major League Baseball in stolen bases with .91 stolen bases per game. That same season, the average Division I college baseball team stole 1.2 bases per game, with the national leader averaging slightly more than three stolen bases per game.

A deeper analysis of both college and professional statistics is even more revealing. A series of multiple linear regression models were created using data from both NCAA Division I and Major League Baseball. The models used both stolen bases per game and caught stealing per game to predict runs scored, while controlling for base-stealing opportunities. The results were interesting. The first set of regression models, examining the relationship between stolen bases per game and runs scored, revealed that in college baseball, runs per game increased by .295 with each stolen base per game. However, in Major League Baseball, runs per game actually decreased by .208 with each stolen base per game. While it seems strange that a successful stolen base attempt would result in fewer runs scored, it is likely explained by the fact that teams stealing more bases generally do so to compensate for a lack of offensive firepower (i.e. power hitting). Therefore, it is not the stolen base itself that is costing the team runs but the team's overall style of play. The second set of regression models, analyzing the relationship between caught stealing per game and runs scored, indicated that in college baseball, runs per game decreased by .304 with each unsuccessful stolen base attempt per game. In Major League Baseball, the cost of a failed stolen base attempt was even more severe at .845 runs per game.

So what do these findings actually tell us? In the most simplistic sense, they indicate that the stolen base is indeed a more valuable offensive weapon in college baseball than it is in Major League Baseball for two reasons: 1) The reward for a successful stolen base attempt is greater; 2) The cost of an unsuccessful stolen base attempt is less significant. Therefore, because they have more to gain and less to lose, it makes sense for college teams to utilize the stolen base more liberally. However, the fact that college baseball teams attempt considerably more stolen bases per game than do big league teams seems to suggest that many college coaches are already aware of this more favorable "risk/reward" ratio.

That being said, it is also important to acknowledge and understand the limitations of these findings. The biggest weakness of this study is the inability to examine specific situations. Therefore, while the above findings provide information about the big picture, they offer little or no guidance relative to specific in-game strategy decisions. In other words, there are a multitude of factors (i.e. the ability of the base runner, the opponent, the game situation, etc.) that were not considered in this study but are extremely influential in the outcome of any base-stealing attempt. As a result, coaches must remember that the actual "risk/reward" ratio changes with the situation. Below is a more detailed look at factors that must be considered before attempting a stolen base.

The Base Runner
The speed and base-running ability of the runner are extremely important when deciding whether or not to steal a base. It makes the most sense to run when the base runner is fast and has good instincts.

The Hitter
The ability of the hitter at the plate is extremely important. It makes the most sense to attempt a stolen base when the hitter at the plate is a double play threat and/or when the hitter has little chance of driving a runner in from first base.

The Pitcher/Catcher
The ability of the pitcher and catcher to stop the running game is also important. A pitcher that is slow to the plate is much easier to run on than one who is quick. Similarly, a poor throwing catcher is easier to run on than one who throws well.

The Game Situation
Research has repeatedly shown that in the majority of Major League Baseball games, the winning team scores more runs in one inning than the losing team does in the entire game. This revelation backs up Earl Weaver's advice to play for the big inning, especially early in games. Therefore, one-run strategies, such as the stolen base, make the most sense in situations where one run is of great importance (i.e. late in games or in low-scoring games).

Michael Current is an assistant baseball coach at Illinois State University. He graduated from Blackburn College with a degree in Communication and recently completed his master's degree in Sport Management at Illinois State University. Last summer, Current served as an assistant coach with the Yarmouth-Dennis Red Sox in the prestegious Cape Cod League, where his team won the league championship.

Dr. Chad D. McEvoy is an Assistant Professor of Sport Management in the School of Kinesiology and Recreation at Illinois State University, where he is the coordinator of the sport management program. Dr. McEvoy has published articles in journals including Sport Management Review, Sport Marketing Quarterly, International Journal of Sport Management and Marketing, and Sport Management and Related Topics.

Comments

The break-even % in MLB is not close to 75%, but more like 68 or 69%. A little quibble, but probably important for the reasons: throwing errors.

I imagine in college ball, there are more throwing errors per SB attempt than in MLB. That's a huge benefit.

***

Not knowing how the regression was run (exactly what are the parameters used), it's hard to say the impact of the causation of the parameters ("correlation is not causation").

I will point out that the *difference* between the run value of the SB and CS according to your findings is around .60 to .65 runs, be it in MLB or college ball. This is consistent with the actual run values of SB and CS that we reported in The Book.

Furthermore, the breakeven point for any baseball game is based on the run environment. This chart for example shows you the run values of events for various environments:
http://www.tangotiger.net/customlwts.html

It's very possible that the reason college ball has so many SB is simply that the talent level of the runners v the pitcher/catcher is simply that much higher than in MLB. I imagine college ball is filled with Juan Pierre-types, and are not abundant with IRod-types.

That distribution balance is likely what causes the number of SB in college ball.

it is likely explained by the fact that teams stealing more bases generally do so to compensate for a lack of offensive firepower (i.e. power hitting). Therefore, it is not the stolen base itself that is costing the team runs but the team's overall style of play.

Or it is an indication that teams tend to attempt steals in situations where runs are precious and do not attempt steals as often in situations where runs are being scored in bunches.

Which is what makes the analysis by Joe Sheehan banal idiocy:

"For example, the "Run Expectancy Matrix" created by Baseball Prospectus reveals that a runner on first base with no one out is worth approximately 0.864 runs. A successful steal of second base would raise that figure to 1.173. However, a failed stolen base attempt drops that number to 0.270. In this example, the loss is nearly two times the gain."

Absolutely no team randomly attempts to steals bases and there is little reason to think that the situations where a steal is attempted reflect a random sample of situations where runners are on first base. Its pretty obvious that if Frank Thomas is on first he is not likely to steal. He is also probably less likely to score than a speedster. So what you really have when you lump those situations together is meaningless data.

Sheehan also suggests that the secondary effects of base-stealing, such as putting pressure on the opposing pitcher and defense, do not exist




I have one question about the above statement. Does Sheehan's study show anything about the effect on the defense regarding the threat of the stolen base. For instance, if Juan Pierre is on first, but doesn't steal, does that force the defense to act in a certain way to guard against the steal that would mitigate the possibility of the double-play, or free balls in the form of pitchouts?



Thanks.

Actually, the break-even point is probably even a little lower than 68-69%, if we could measure the positive value of sending the runner on hit-and-runs. (Or alternatively, if we could identify the CSs that result from broken H&Rs and pull them out, we would find that the actual success rate on true SBAs is higher than the SB/CS data indicate.) The raw SB/CS data captures much of the cost of broken H&Rs, but not the benefit of successful ones.

I happened to notice that the success rate is up quite a bit so far this year. The SB% is 73.6% so far, compared to 71.4% last year. Maybe it's just a small sample thing, but I thought that was interesting. Maybe managers are reading The Book!

Good point about the busted H&R. I suppose we could look at it in game state situations where H&R would be more prevalent than not (similar to how Andy looked at the Protection issue, by looking at game situations where it was a likely IBB situation).

As for managers reading The Book, only Ned Yost has been reported by the media to have read it. It would be interesting to see if his tendencies have changed.

Some comments about the above posts.

Run Matrix: My purpose for including information about the run matrix in my article was not to suggest that it is the end-all-be-all. As I discussed at length at the end of the article, anyone who knows anything about the game of baseball understands that every situation is different and must be treated as such. A coach or manager who makes decisions based soley on that matrix is likely going to make many more poor decisions than he will make good ones.

Joe Sheehan: There have been several comments about Joe Sheehan's article. My purpose for including that information in my article was simply to present arguments from people on both sides of the debate. It was just background information. I actually tend to agree with Ambiguity's post. As a college baseball coach, I have far too often witnessed first-hand the negative effect that the threat of the stolen base has on the pitcher and defense. For example, when the runner on first base is a threat to steal a base, pitchers are oftentimes forced to be quicker to the plate than they are comfortable with, costing them velocity and command. As for whether there has acutually been any research done on this concept I do not know. However, I would guess that if Sheehan was willing to make a statement like that in his article, that he has research to back it up.


Busted H&R: The point about the busted H&R and its impact on the break-even point is very interesting. I have never heard that point made before, but it seems to be valid.

Thanks for putting hard data to something I have been thinking about for some time. I coach at much lower levels than D1 (travel ball and HS through 16U) and I had more or less came to the same conclusion (without the data) that the risk/reward was higher the higher the level of baseball played. We have employed a system where we "run until were caught" philosophy and I have found that we will win more often than not when we are able to get bases at little cost. I just now know the statistics behind what I observed.

As others have discussed we weigh the situation with the risk/reward. We do however time the opposing teams pitcher to catcher to second times (since they vary so much at this level) and with the known stealing times of our players we weigh the value of the base vs. the cost of loosing the runner. As you go down from MLB and D1 baseball the level of play drops significantly, so although the hard data is difficult to come by, it makes sense that the risk/reward would drop as you move from MLB ball all the way down to the local league ball. So at the MLB level steals are relatively rare, moderate at college, more at HS, and kids running wild at the local little league on Saturday.

Also we know that running is the skill that peaks earliest (21-22?). MLB is pollouted by older players who are incapable of stealing. The SB attempt is therefore much more frequenty in college and is more likely to be used in higher-run environment situation.

Sorry, I should add that is another reason why SB attempts per game are more frequent (along with the talent point) ... and are therefore by definition more likely to be employed during periods of higher offense

John: Isn't it the reverse, that teams should steal more when runs are scarce? I would think the gain from moving a runner from 1B to 2B is relatively greater when runs, especially HRs, are scarce.

The run value difference between 2B and 1B is fairly constant across all run environments. The large difference is the change in run value of the CS.

This is clear based on the Custom LWTS link above, or from my Markov page.

Guy -- I don't have the data to back a particular view. Here is my reasoning

1. Hypotheses: In MLB SB is used largely by small ball teams that lack power

2. This means that runs scored are likely to be lower than average, so you get the result the authors found

3. In College Ball SB% is more frequent beacuse of (a) the talent hypothesis and (b)players have greater speed skills .... although success rate may be similar because they are less mature on the basepaths -- the critical number is attempts

4. The fact that SB attempts are more frequent means that they occur (on average) in teams that they teams that score more runs also rack up SB.

5. By definition the teams that have more successful stolen bases will score more runs.

It certainly isn't a 100% solid argument but I suspect that the greater frequency of SB attempts means that is one of the factors behind the effect the authors report.

Of course, what they should really do is look at pbp data, and not game level data, but that is another story

My purpose for including information about the run matrix in my article was not to suggest that it is the end-all-be-all.

I doubt it is even a starting point for a discussion of stolen bases. If it tells you anything, it is what the break even point would need to be if every player who reached first attempted to steal second. But stolen base attempts actually happen infrequently. Only a small number of base runners are threats to steal whenever they reach base.

Beyond that, there is a real question whether outs and base are the most critical factors in determining how likely someone is to score. If you look at how frequently players score when they get on base over the course of their major league careers there is a very wide variation, with some runners scoring nearly twice as often as others. Not surprisingly it was speedy guys like Vince Coleman who scored the most often.

John:
Just did a quick check, looking at NL from 1946 to date. There's an r of -.19 for SB/G and HR/G, and -.15 for SB/G and R/G. SBs tend to go down as scoring increases, as they should (per Tango's comment above), although the relationship isn't as strong as I would have guessed.

The reason faster runners score more often when on base isn't because they are fast. It's because bat leadoff and therefore have the best hitters behind them.
Even Cecil Fielder could score from first on a home run (as long as the batter was careful not to pass him on the basepaths), but not even Rickey Henderson could score from third with Brad Ausmus and Adam Everett hitting.


The reason college teams that steal a lot also score a lot is because the best players all go to the same few colleges. Remember in little league when your slick-fielding shortstop led your team in homers & steals and was your best pitcher? Top college teams have a lot of these guys; the have-nots of college athletics are left with the guys who, relatively speaking, can't do anything.

The reason faster runners score more often when on base isn't because they are fast. It's because bat leadoff and therefore have the best hitters behind them.

Maybe. Why does that matter? Its pretty clear that when they are on base they are more likely to score. Whatever the reasons, they are as important as what base the runner is on and the number of outs.

And if the players batting behind the player are the only critical factor, one wonders why Vince Coleman is among the leaders. The St. Louis teams he played on were not exactly run machines with murderers row in the middle of the lineup.