THE BOOK--Playing The Percentages In Baseball

Wouldn’t SB be better evaluated with WPA +/- LI?  It seems like this could give a true measure of the value of each SB without resorting to average values of SB for the league.  I imagine this data would be available in retrosheet. . .

Tango - I don’t see where you have provided a link to Dan’s article.

Peter:

http://www.hardballtimes.com/main/article/adjusting-steals-for-win-value/

Oops, ok, linked now.

Ryan: I don’t follow.  Can you give me an example?

In any given situation (ie 4th inning, down by 1 run, 2 outs, man on 1st) there is a Win Prob before and after a successful SB.  This WPA should accrue to the baserunner who steals (or gets caught/picked off).

Wouldn’t it be more accurate to look at individual WPA accrued through baserunning in this fashion rather than determine the league-wide win value of a SB and multiply by a player’s SB?  This would account for stealing (successfully) in high leverage situations rather than presuming that the leverage of SB would even out over time.

Ryan, that is what Dan does.

I guess I should have written what I said as:
newSB(i) = actualSB(i) * LI(i)

And you sum over i.

What I wonder is how much the win value differs from what you’d guess if you take runs/10.  For Eric Davis, a simple calculation using .19 runs for steal, -.44 for a CS, and 10 runs per win yields an expected win value of .009 per attempt.

His actual win value is .012, 33% higher, but that’s cherry picking since I only picked Davis because he was the #1 guy on the wins per attempt table.

Looking at all players, are we underestimating the win impact of stolen bases by using the simple method?

How about for the great basestealers - Raines, Rickey, and Brock?

Tango - do you have a link that would show the calculation of the break even points for steal attempts - Dan refers to the value of the successful steal and the cost of the caught stealing in various years, I would like to calculate it for each year based on the offensive totals.

Break-even:

Add the total wins for the SB.  Say it’s +50 wins.

Add the total wins for the CS.  Say it’s -100 wins.

The breakeven would be:

50x = 100(1-x)

Which comes out to:
x=100/(50+100)

***

Rally, Cy at BtB had a good post about clutch stealers a few years ago.  I linked to it from this blog.  You should be able to find it pretty quickly.

I think I phrased my question poorly, as I did know the calculation you posted in #9.

“the win value of a stolen base in 1968 (the Year of the Pitcher) was .027 wins, while being caught stealing only cost the team .04 wins. This lead to a breakeven point of just below 60 percent.”

How do I get the win value of a sb to be .027 and the win cost of a cs to be .040 in 1968?

Tango, thanks for mentioning that study I did. Probably not as insightful as what Dan did, but it was called “Are There Clutch Base Stealers?” and is at

http://www.beyondtheboxscore.com/story/2006/7/29/181346/885

Brian #10: I’ve calculated the win value for each play in the Retrosheet era (with errors on a few plays that I’m working to clean up).

From there it’s a simple average of the win values for the given event type.

I can share the data if you’d like.  Drop me an email.

Cy #11:  Nice article.  I was planning on following up with some similar calculations.

I’m curious to see if we end up with the same results.

You’d think since I do most of my writing over at BtBS I’d have seen this. 

I’ll mention to Sky we need a better archive system.

Dan

Glad you liked it and glad to see more people investigating the issue. Yes, I would like to see how our results compare and find out if anyone really was a clutch stealer. I think in general there should be some good archive system for all the work that has been posted online. I have my page of clutch hitting links so people can see what has been done. It would be great if there were a links page for each topic.

Cy

Dan/12 - So you figured the WE at each play, and then took the mean for each type of play?

At the last minute, my computer was not cooperating on the number crunching, and I had to change topics for my BPro Idol article, then I found out you already wrote a historical stolen base article the same day. Oh well, I got it done.

Yes, I’d be interested in seeing some of your data.

I can share my RetroSheet SQL code that I use for breaking catcher CS, pickoffs and pickoff caught stealing into distinct plays. Actually, 5 of them - CS, PC1, PK1, PC2, PK2, plus PE by pitcher and TE by catcher. Turns out Rickey Henderson had some amazing years of not being thrown out by the catcher - his only problem was the pickoff, probably trying to get the early jump too often.

I believe an evaluation of the total contribution of a base stealer needs to include all the pickoffs as well as the caught stealing - some guys get picked off more than the catcher throws them out, like Rickey.

steals attempts of 2b
YEAR SB CS2 PC1 PK1 PC2 PK2
1979 27 8 2 3 0 0
1980 82 15 4 14 0 0
1981 48 11 4 8 1 0
1982 94 12 14 18 0 0
1983 79 10 5 7 0 0
1984 50 13 2 1 0 0
1985 65 1 6 7 0 0
1986 72 4 10 4 0 0
1987 32 6 2 4 0 0
1988 62 4 7 7 0 0
1989 53 5 3 13 0 1
1990 48 4 4 4 0 1
1991 37 7 5 2 0 0
ETC