Very good post, but I may am not going to be able to help you with your questions. I have always been a fan of the stolen base. I have fond memories of Coleman turning a walk into a double. I also would perfer a stolen base to a bunt, or even a ground out that advances the runner.

I think therefore I would WANT to agree with your analysis and would not be objective enough.

I’d love for this to be true, so I took a primitive look at it.

The simple Runs Created- ([H+BB]*TB}/{AB+BB])- show that actual 1982 Rickey created 81.43 runs while the Adjusted Rickey created 110.98 runs. This assumes that a single/walk and a stolen base is the same as a double and that Rickey getting caught stealing is the same as Rickey making an out at the plate with no men on. Good news: a hit and a stolen base does basically mean the same thing as a double. But…an out at the plate is less costly than an out on the base paths.

The Risk?
Unfortunately, I found that losing a runner off the bases is more damaging than simply making an out. BP’s 1982 Run Expectancy Table used in the book confirms that. Starting an inning there’s an run expectancy of .47 runs; a single/walk (man on first) with no outs gives you an expectancy of .87. One out and none on gives you an expectancy of .25. While the end result is the same, an expectancy of .25 runs, the cost is much greater. Making an out on the basepaths costs .6 runs (.87-.25) and simply striking out would cost the team only .2 runs (.47-.25).

The Reward?
And the reward isn’t all that spectacular; a successful stolen base with no outs jumps the expectancy from .87 to 1.07. You’re risking .6 runs in order to gain .2 runs. To break even, you need to steal roughly 75% in 1982 (steal three bases and improve .6 runs [.2+.2+.2], compared to losing one man his .6 runs in order to break even).

Conclusion
Unless you like wind sprints, stealing bases is a worthless exercise at anything not appreciably above 75%.

What about Whiteyball?
What makes Whiteyball so amazing is that the team stole bases at an alarming success rate, but was it successful enough? 1982’s 68% hurt the team, 1985’s 77% is right around the threshold along with 1987’s 78%. Unfortunately, even 1987’s blistering 314 team stolen bases may have been just enough to break even. I guess Whitey liked to use games as conditioning practice, too.

Intuitively, this all makes sense to me. Nobody would give Mo Vaughn the green light and everyone would give Rickey the green light. The question is not whether a stolen base helps the team. It does. The question is how successful do you need to be at stealing bases in order to justify taking the risk? The numbers say to draw the line at ~75% (plus your needed incentive to run sprints), which eliminates all but a handful of contemporary base-stealers and incites Joe Morgan. Poor Joe.

Yes, OPS is a very fine tool to measure the offensive capabilities of a player, but digging into the formulae of OBP and SLG will show you why you had such a large spike in OPS.

The formula for OBP is: (H+BB+HBP)/(AB+H+BB+HBP)
The formul for SLG is: TB/AB

What does this mean? OBP and SLG are both essentially percentage measures. They rely heavily on what is in the denominator of the equation. If OBP and SLG had demoninators that were equal, then this could be a useful measure of the value of a steal. However, since the denominator for OBP will be larger tan the denominator, the effect of taking one unit out of the numerator of each will be different.

In Rickey Henderson’s 1982 season, which you cited, he had 536 AB, 143 H, 116 BB, and 2 HBP. Using these numbers, the denominator for OBP will add up to 797 while the denominator for SLG will only be 536. That means, that for every SB adding essentally one base to the numerator of SLG, you will increase his SLG by 0.00187. On the other hand, for every CS subtracting essentially one base from the numerator of OBP, you will decrease his OBP by only 0.00125.

This means that a SB is worth about 1.5 more bases (earned) than a CS is worth (lost) under this circumstance, which is against the spirit of the argument.

I applaud your efforts, however, and if I knew more about sabermetrics (I’m just getting used to OPS and WPA myself) I would suggest a better metric to try and value a steal. Thanks for all the work and insight!

The Cardinals don’t steal bases and win, have won for a number of years including the playoffs.

The Mets steal a crap-ton of bases and where have they gone?

I believe the Giants do too.

A sb threat puts pressure on the defense, and can put defenders off optimal position to cover for a steal attempt. Moreover, the pitcher is not comfortable and will not throw certain pitches. If you can exploit this situation with the hitters following the basestealer, then the results can be much better than what expected. Then you have to add the errors eventually done by defenders. In my opinion the evaluation of a sb worthyness should be done analysing run scoring expectancy only of plays involving steal attempts, to corectly include these factors, but I have the impression that the data to do it is not available to everyone..

I agree that reaching base + sb is not equivalent to a double, because you can’t advance to 2nd if someone is already there.

GO CARDS!!!

i think we’re all ignoring the situational element. the reason CSs are so deadly is because they kill big innings — you not only lose a baserunner, you lose one of your 3 outs for that inning, so it vastly decreases your chances of putting 3 or 4 runs on the board.

but there are times when the chances of a big inning are already low. with 2 outs and a man on 1st base, you’re not likely to score 3 to 4 runs anyway, so the cost of a CS is low — go ahead and try to steal. likewise, if you have the weak part of the lineup coming up, or a very good pitcher on the mound, the chances of a big inning are also lower than normal — so the cost of a CS is lower, too, and therefore the relative value of an SB might be higher.

that’s the major weakness of run-expectancy studies in general —they assume a one-size-fits-all run expectancy. but we’d all agree that all man-on-first, one-out situations are NOT created equal. the run expectancy is much higher if the man on first is brian barton and the next 3 hitters are duncan, pujols and ankiel, than if the man on first is adam kennedy and the next three hitters are braden looper, cesar izturis, and skip schumaker. in the first situation, you don’t want to steal -—there’s too good a chance of a big inning. in the second situation, what the hell - the odds of a big inning are pretty low anyway, so take a risk.

and a really great post.

I find the question of the value of stolen bases a very difficult one. I understand the formulae, and I credit their accuracy from a strictly mathematical perspective. I think, though, that a stolen base doesn’t quite benefit fully from that sort of analysis.

I think, and this is just me sort of blue skying, that the only way to truly understand the overall value of a stolen base would be to actually go through all the game logs, for an entire season, and determine how often a steal helps to create an additional run. To me, a stolen base is such a situational element that unless you look at it on a case by case basis specifically, a strictly percentage based formula is going to fall short.

I agree with your hypothesis, by the way, i.e. that the stolen base is probably undervalued today. I simply think that the current analytical methods we have tend to fall a bit short of optimal when talking about a play that doesn’t always contribute equally to the outcome of the game.

remember every game that season how the announcers would talk about the “pesky” Eckstein when he was on the basepaths? They suggested that it demoralized and distracted pitchers. Even though a base stealer may get picked off or thrown out stealing, it can benefit the team by causing the pitcher to not concentrate on getting the batter out and in turn cause him to make mistakes. So if Skip is on base posing a threat and the pitcher makes a mistake to Albert then I say it is good for the team.