THE BOOK--Playing The Percentages In Baseball

Saturday, October 31, 2009

Who is right - McCarver or Manuel?

McCarver was adamant and passionate that the runners should have been running on the 3-2 count with Utley at the plate in the 8th inning of Game 2.  Manuel thought the opposite was a no-brainer.  There was one out, (fast) runners on first and second, and the Phillies were down by 2 runs.  Howard was on deck.

Of course there is no way for either one of them to know for sure unless they “run the numbers.” Sometimes, however, if the results of two alternative decisions are so far apart, it is actually possible for someone to “know” the answer by virtue of experience or intuition or by “reasoning it out.” I contend that if the actual results are close, it is impossible for a human being to know which one is right without at least attempting to “run the numbers.” And even then, you can’t be sure, since the model that you use to “run the numbers” and the data that goes into it are not usually 100% accurate.  As I have said many times before, because our models are never 100% accurate, the answers we get when we use them are either “definitely A” (even if our model is not that accurate), definitely B” or, “We are not really sure.” It is rarely only “definitely A” or “definitely B.” We can never be that sure with these models.  How “wide” the “We are not sure” is depends on how accurate we think the model is.

Anyway, here goes:

Using log5 matchups for Utley/Rivera, Utley K’s 22% of his PA and hits into a GDP on 8.6% of his PA with a runner on first and less than 2 out.

But we have a 3-2 count which changes the percentages.  For his career Utley strikes out 38% more on a 3-2 count than in a random PA, so we’ll change that 22% K to 30%.

With a 3-2 count, Utley only put the ball in play 45% of the time.  On a random PA, it is 72%.  So we’ll change the 8.6% GDP to 5.4%.

If the runners are forced to go on a 3-2 count and the pitchers knows they might go, a runner, even a fast one like Rollins, probably gets thrown out 80% of the time (just a guess).

Using Tango’s WE tables, here is what we get:

If runners don’t go, inning ends on GDP 5.4% of the time.  If runners go, you save the inning that 5.4% of the time and leave runners on 1st and 3rd.  The WE for that state is 14.7%.  The WE if the inning is over is 6.5%.  So you save 8.2%, 5.4% of the time, or .44%.  Not much savings.  So even if you don’t lose anything by running, it is not much of an advantage to send the runners.  That probably ends the discussion, but we’ll continue anyway.

You lose something of course, if Utley strikes out or hits into a line drive (or quick pop fly) out with the runners going.  What do you lose?  20% of the time the runners are safe, which leaves a WE of 17% (runners on 1 and 3 and 2 outs).  80% of the time, the inning is over, so the WE is again 6.5%.  That is a total WE of 8.6%.  If they don’t run on a K, the WE is 13.7%.  So, you lose 5.1% on WE when they run on a K.  The K occurs 30% of the time.  We’ll up that a little to account for line drive DP when the runners are going.  We’ll call it 32%.  .32 times 5.1% is 1.63% of WE.

So with the runners running, you lose 1.63% of WE from the K (and line drive out) and gain .44%.  Not even close!

BTW, this is one of those examples I was talking about where even if my model is not 100% accurate or the manager knows things that I don’t, it cannot possibly change the result since the numbers are so far apart.

One more thing:  What is the approximate BE point for the runners being safe when they are running and Utley strikes out?  The line drive or pop fly DP that ends the inning if the runners are running costs 7.5% when it occurs.  If it occurs 2% of the time, that is a .15% cost.  Remember that the savings from staying out of the GDP was only .44%.  So that leaves .29%.  If the runners are safe 65% of the time, the cost of them running is .59 * 17% plus .41 * 6.5*, or 12.7%. That is a loss of 1% (if they don’t run on a K, the WE is 13.7%).  30% (times the K occurs) times 1% is .3%.  So that is the approximate BE point.

If the runners are going to be safe on an Utley K 59% of the time or greater, then they should be running, more or less.

Edit:  Peter Jensen tells us that with runners on 1 and 2 and 0 or 1 out and a 3-2 count, runners get thrown out 57% of the time.  With Posada and Rollins, I would guess it is closer to 45%.  With the extra benefit of the runners going on a single or double, plus the occasional error on a steal attempt, I think that the decision is a lot closer than Manuel thinks.