THE BOOK--Playing The Percentages In Baseball

"don’t get beat with your second-best pitch” is just another example of confusing optimal strategy with strategy that won’t get you second-guessed; except in this case the managers (Buck etc) are projecting their own desire to be invulnerable against second-guessing onto the players!

It’s interesting how it is OK for MGL to make statements like:

On top of everything, I don’t know whether Rollins is a good fastball hitter or not, but I do know that he seems to be vulnerable to the off-speed pitch low under his hands (and even in the dirt).

Where’s the “cogent analysis” to prove this? Kind of like a fan saying “I don’t know what Teixeira’s defensive stats look like but I do know that he seems to be a vacuum at first base.”

Over the past two years, Rollins has really struggled against the fastball according to FanGraphs Pitch Value:

wFB ‘08: -1.8
wFB ‘09: -2.7

Not so much is he struggling against the slider and change-up:

wSL ‘08: +3.7
wCH ‘08: +4.5
wSL ‘09: +1.1
wCH ‘09: +5.0

If you prefer the per 100 pitches numbers:

wFB/C ‘08: -0.12
wFB/C ‘09: -0.17

And for the Slider/Change-Up:

wSL/C ‘08: +1.81
wSL/C ‘09: +0.40
wCH/C ‘08: +1.29
wCH/C ‘09: +1.36

Maybe you are right about Rollins being vulnerable to the off-speed stuff but the numbers don’t really support that.

No analysis there.  I just took a shot by memory, which is of course a very poor processor of information, not to mention that I have probably seen 10% of Rollins PA’s this year. That has nothing to do with the thesis of my thread. It was just a throwaway comment (does every single thing I say have to be a “cogent analysis?” Geez!)

I was simply commenting on the, “Now Broxton can go back to his fastball” statement by Martinez, which is similar to the “don’t get beat by your second-best pitch.”

Again, all your pitches become about the same if you throw them the optimal percentage of time.  Pitch B can only be worse than pitch A if it is thrown too often.  Otherwise all pitches have roughly the same value.  IOW, pitchers don’t have a “second best pitch.” They really don’t.  Unless you consider the quality of the pitch in isolation, which is not relevant to a real game situation.

Show me a pitch from any pitcher that is worse than any other pitch from that pitcher and I’ll tell you to tell him to throw it less often until it becomes as good as the initially better pitch.  If it never does, then he should never throw it!

Do you think managers and pitching coaches understand that concept?  If they don’t, they can’t be 100% effective in their tutelage.

What is somewhat humorous is that while commentators and even managers and players say that all time ("don’t get beat..."), clearly they don’t actually do it.  Broxton, for example, comes in mostly with the game on the line and yet he throws 25% off-speed pitches. I’m sure he gets beaten on his “second best pitch” all the time…

“(does every single thing I say have to be a “cogent analysis?” Geez!)”

said Joe Buck/Tim McCarver/Joe Morgan after reading a Book Blog post.

A pitcher’s best pitch is the one the batter isn’t expecting.

Spike - “I don’t know what Teixeira’s defensive stats look like but I do know that he seems to be a vacuum at first base.”

There’s nothing wrong with this statement.

mgl - “Show me a pitch from any pitcher that is worse than any other pitch from that pitcher and I’ll tell you to tell him to throw it less often until it becomes as good as the initially better pitch.  If it never does, then he should never throw it!”

I don’t think that’s exactly right.  I think the optimal distribution can have uneven values for pitches.  I could be wrong though, but someone who knows more about economics than me explained it to me once.  I think, as a general rule, if there’s a large discrepency between pitches, its probably better to throw the one with the higher value more often, but I don’t think that’s absolute.

does every single thing I say have to be a “cogent analysis?” Geez!

Just figured you’d hold yourself to the same standard you hold others too. Silly me!

It was just a throwaway comment

You could start using this in a lot of your comments, I’m thinking.

"I don’t think that’s exactly right.  I think the optimal distribution can have uneven values for pitches.  I could be wrong though, but someone who knows more about economics than me explained it to me once.  I think, as a general rule, if there’s a large discrepency between pitches, its probably better to throw the one with the higher value more often, but I don’t think that’s absolute.”

It IS exactly right, and the proof is simple logic and common sense, as I already explained.  I’ll explain it again.

If I throw 80% fastballs and 20% curve balls and my fastball results are better than my curve ball results, then I am throwing too many curve balls (obviously).  I would want to throw fewer curve balls, right?  Why would I want to throw a pitch that was worse than another pitch?  By throwing fewer curve balls, say 18% rather than 20%, I will necessarily increase the value of the curve ball (because now the batter is more surprised when it is thrown) and decrease the value of the fastball (because he is now suspecting the fast ball more often).  So now the values are closer together.  If the fastball still gives me better results, then obviously I need to throw even more fastballs.

Eventually either the values of the two pitches will be exactly even, in which case, I have arrived at my optimal percentages, or if the fastball continues to be a better pitch even when I am throwing very few curve balls, then I must eliminate the curve ball from my repertoire.

Now, the caveat to this is that game theory equilibrium (where the value of all pitches is exactly equal) applies to each batter and each situation separately so that it is not necessarily true that overall all my pitches will have the same value. Here is an extreme example:

Say I only pitch to two batters.  One batter is great at hitting fastballs and the other is great at hitting curve balls.  And say the fastball hitter is a much better hitter.

Say I throw 80% curve balls to the great fastball hitter and 80% fastballs to the great curve-ball hitter.  Since the great fastball hitter is a great hitter overall (I said that he is for this example), the value of my 80% curve balls to him might be +5 runs per 1000 pitches (plus is bad for the pitcher) and the value of my 20% fastballs to him is also +5 runs (they have to be the same).

To the great curve ball hitter, who is a worse hitter overall (again, by definition for this example), my 80% fastballs may be -5 runs and my 20% fastballs also -5 runs.

So overall, my fastball value is -3 runs and my curveball value is +3 runs.

This is an extreme example of course. In reality, while a pitcher’s overall run value for all of his pitches don’t have to be exactly equal, for the reason I illustrated in that extreme example, they will in fact be almost the same, assuming that he or the batters are not making any egregious errors in their pitch selection or anticipation distribution (the percentage of time he throws each pitch or that the batter expects each pitch).

I can agree with MGL mostly, but not totally.

Let’s say that a pitcher throws 80% FB and 20% CV.  The run value of each FB is -.03 runs, and each CV is +.01 runs (more negative is better in this case).  The overall value of the pitch is -.022.

Now, let’s say that the pitcher realizes his CV are no good.  So, he throws them less, say at 15%.  The run value of the CV goes up, each, to +.00.  His FB are now worth -.025 runs.  Overall, he’s at -.020.  That’s a worse result.

Let’s say he decides to throw even fewer CV.  He only throws it 5% of the time.  The run value is -.015.  He throws his FB 95% of the time for a run value of -.020.  Overall, he’s at -.019.  That’s worse still.

So, I don’t buy that the two need to have similar value in order to be at optimal mix.

The reason is that each pitch on its own has its own marginal value, BUT it also has an indirect effect on all the other pitches.

It’s not like stealing, where you can take out all the times Carl Crawford stole when the breakeven point was 69%, and he was 7-3 on those plays (and 30-4 otherwise).  If he’s 37-7, and you take out the times he shouldn’t have ran (and went 7-3), he’d still be 30-4.

For those times when a pitcher throws a curve ball that he shouldn’t, and should have instead thrown a FB, the value of that pitch may be better, but then the value of all his other FB will go down, because the batter would have expected more FB at those times.

So, I get the idea, but I don’t think it necessarily applies so easily with pitch selection.

"The reason is that each pitch on its own has its own marginal value, BUT it also has an indirect effect on all the other pitches.”

That is the whole point.  The guys that did the study referred to in the other thread said the same thing and they are right.  All pitches (again given the same batters, situation, etc.) MUST have the same value, according to game theory.

You can make up numbers where it doesn’t come out that way, Tango, as you did above, but those numbers are not correct.

Look at it this way.  Let’s say that on my very next pitch, I flip a coin in my head and it comes up curve ball.  And let’s say that I knew that my curve ball had a worse run value than my fastball.  How could I possibly justify throwing a curve ball on that pitch and that pitch only, given that the batter has already made up his mind what to expect on that pitch?  I can’t!  They have to have the same value!  There is simply no other solution.

"How could I possibly justify throwing a curve ball on that pitch and that pitch only”

Because it will make the fastball better on the next pitch?  You have to add the marginal values for every pitch in the at bat, or really for every pitch in the game.  If by throwing one crappy curveball you make every fastball you throw after that better, it might be worth the risk.  If you do away with the curveball completely, even if it sucks, it might bring down the fastball value enough to make you worse overall. 

I understand game theory, but throwing a pitch isn’t only about the value of that pitch, its about the effects it will have on all the other pitches.  If you want to send me the link to the study you’re talking about, I’d be happy to read it and be proved wrong.