from Amazon
Tuesday, November 20, 2007
Going for extra bases, Retrosheet Years
Want to know how times have changed, how often runners go for the extra base on a hit? Here you go.
To read the first line: when a single is hit, with a runner on 1B, between 1957 and 1968, with 0 outs, the runner will go for the extra base 36% of the time, and he will be successful 95% of the time. This compares to the 1993-present numbers of 27% and 96%. The gap widens with 1 out (39% attempt, against 30% attempt), and shrinks alot with 2 outs (40% to 36%). The 1957-1968 was definitely more aggressive, as they ran just as hard with 1 out as with 2 outs, which is not necessarily a good idea.
The same pattern repeats with the runner on 2B. With 2 outs, there’s not much distinguishing a runner’s aggressiveness over the years. But, at 0 and 1 out, runners today are far more passive. There’s nothing necessarily wrong with that, since far more runs score today, making the breakeven point higher today (outs are costlier).
On the flip side, a double, with a runner on 1B, has not changed the aggressiveness of the baserunner.
At some point, soon I hope, I’ll break down these numbers by baserunner.
I’ll also point out the obvious:
An “opportunity” to go for it, or not, is not really an “opportunity”. If let’s say Cince Voleman goes for the extra base 50% of the time, while his peers go for it 33% of the time, you can pretty much bet that most of the other 50% of the time no one would have gone. And if your typical pitcher goes for it 15% of the time, you can pretty much bet that everyone would have gone.
It really breaks down like this. For every 100 singles hit:
45: impossible to take the extra base (infield single, already in OF’s hands before runner even reaches 2B, etc)
15: impossible NOT to take the extra base (ball hit deep, you can walk from 2B to 3B, etc)
40: the discretionary plays
So, what I would do (presuming the numbers in this illustration are valid), is subtract 15% from all runners in the numerator, and subtract 60% from the denominator.
So, Cince Voleman’s 50% attempt rate becomes: (50-15)/(100-60) = 35/40 = 89%
Doesn’t this make more sense? That, given a real opportunity to take the extra base on a single, that Cince here will try for it 89% of the time?
And your league average runner: (33-15)/(100-60)=18/40=45%. That is, 45% of the time, an average runner will go for the extra base.
As for the success rate, we have to do the same thing. Since everyone goes 15-15, we have to take those out. So, a league average runner might have the following numbers:
33 attempts: 31 reaches, 2 thrown out
Since everyone goes 15-15, we take those out. Now we see the following:
15 gimmes, 16 reaches, 2 thrown out.
Now, his “success” rate is not 31/33 (94%), but rather 16/18 (89%). We paint this picture: 45% of the time, the runner goes for the extra base, and 89% of the time, he’s successful.
Suppose you have a bit slower runner, say he has this:
20 attempts: 18 reaches, 2 thrown out
80 stays put
We separate the 100 attempts as:
15 gimmes, 3 reaches, 2 thrown out
45 gimme stays put, 35 passively stays put
Now, we’ve got 5 attempts in 40 discretionary plays, or only 12% attempt rate. And a 3/(3+2) = 60% success rate.
Doesn’t this make more sense in describing this runner? Otherwise, if you leave the gimme plays in, you are saying he attempts to go 20% of the time, and is successful 90% of the time. If this is true, he should be going alot more.
The key here is that there’s tons of noise in the data, and this is one way to remove them.
What I need to do now is figure out the actual % of plays that are really discretionary, rather than this semi-blind guess.
How do you propose to figure out the real percentage of discretionary plays? Isn’t it really more of a continuum than a three-bin categorization?