THE BOOK--Playing The Percentages In Baseball

Tom, do you believe Sandberg knows what he’s talking about in that interview?  It seems inconsistent with your view about batter-pitcher matchups.

He seems to make alot of sense.  I’m not sure it’s inconsistent with my view of batter-pitcher matchups, especially since I don’t even think I have (or can!) articulated my view of batter-pitcher matchups.

Is there something in what he said that seems at odds with something I’ve said (or that you’ve interpreted as what I’ve said)?  I’m not against saying I’m wrong if I said something wrong.

This, for example:

http://www.insidethebook.com/ee/index.php/site/comments/how_specific_can_we_get_in_determining_the_true_mean_of_a_particular_matchu/#18

That is, we already do so well with predicting batter-pitcher matchup using just basic performance data, handedness, and GB/FB tendencies, that anything else will be so very slight.  (A promising one would be whether a pitcher is FB-Curve or FB-Change or FB-Slider, but again, I don’t expect much from that, probably on the order of FB/GB tendency.)

Otherwise, if there was something big missing, we wouldn’t have such strong predictive ability already.

The learning, if any, will be confined to peculiar cases, guys with reverse platoon splits or close to it (like Ichiro, but then Ichiro is a study unto himself anyway), or guys like Wakefield.  Guys with drastic changes in performance levels.  It’s at the extreme cases that anything new will come, with respect to batter-pitcher matchups.

Isn’t the Sandberg stuff simply evidence that good scouting can explain if/when a batter-pitcher history deserves to be given extra weight? If the stats confirm what Sandberg is saying, (that Sandberg sits on the splitter and hits it well at an unusually frequency) then it stands to reason that Sandberg’s ‘scouting’ is correct, and that (if this weren’t be told in retrospect) we should expect Sandberg to be unusually successful against Sutter in the future.

The way I read Sandberg is that he can recognize pitches, and he hits better if it’s a pitch he likes in the location he likes at the count he likes.

That doesn’t contradict anything I’ve said.

He also seems to suggest that he can spot patterns.  Presuming that pitchers are not perfect randomizers, then that also is not something that contradicts anything I’ve said.

I’m struggling a bit with the idea of pitching someone “backwards.”

Doesn’t it presuppose the idea that most ML hitters can be pitched to in a fairly predictable fashion (ie, fastballs in fastball counts) with success?

Given that we laud the pitcher who can throw any pitch in any count, why do any pitchers persist with predictable pitching strategy, and only occasionally pitch “backward,” for a particularly tough hitter?

Is it that most pitchers simply don’t have the requisite command to pitch effectively in an unpredictable manner? Or is it too mentally taxing to try to be creative or unpredictable with each pitch?

To what degree are pitching coaches or catchers perpetuating a predictable pitch-selection strategy to the detriment of their pitchers?

Right, pitchers don’t have the control of their pitches.  A pitcher will throw say 30% of their pitches as offspeed in batter’s counts and 50% in pitcher’s counts.  (Numbers for illustration purposes only.) It’s all about the margin of error, and how costly a ball and a strike is at each count.

You’d “pitch backward” by throwing 50% of your pitches offspeed in pitcher’s counts and 30% in batter’s counts.  Or if you throw more fastballs the first time through the order and more offspeed the 2nd time, then you reverse that.

Anything to change the predictability of your pitches.  Of course, it only works if you have somewhat good control of your pitches.

The way I read Sandberg is that he can recognize pitches, and he hits better if it’s a pitch he likes in the location he likes at the count he likes.

Where I think Mike was going with this in #1 is that in the past, if some hitter made an assertion like this, you (and others) would insist on statistical confirmation of the assertion before accepting it.  And of course in this case, if we’re talking about only 20 PA’s against a pitcher, no statistical breakdown of so few samples could possibly satisfy your threshold for statistical certainty.

Or, to summarize:  Hitter says he kills pitch type A from pitcher Z.  Analyst looks at data and finds that hitter has only faced pitcher Z 20 times, reaching safely 8 of those 20 times with 4 home runs.  Analyst scoffs at assertion (or should I say, “narrative”?) that “explains” 8 for 20 as a real effect.

If this trend were taking place right now, and Sandberg were set to face Bruce Sutter in a few days, I would have expected an offer of a bet, where the outcome is decided by whether the next Sutter splitter put in play by Sandberg results in a safe hit or an out.  I would have expected you to opine that the correct way to set odds for such a bet was to use Sandberg’s overall projection numbers, and not his historical numbers vs. Sutter (or Sutter’s splitters).

That’s what I would have expected, based on all I’ve read here over the years.  So, can you explain why this situation, and Sandberg’s assertion, is convincing to you?

Sandberg didn’t really assert anything.  The question was why he hit Sutter so well, and Sandberg answered by describing his approach to facing Sutter.  He can’t be expected to say “well, it was only 20 ABs.  Who the heck knows if I really hit him better than I hit anybody else”.

Ah, I see.

Ok, two points.

1. Did Sandberg say he hit better against Sutter because of the results, or because of the approach? 

2. How predictive is it?

I was only commenting on his approach to hitting, to what he was looking for, rather than going in blind.  He had a plan for his at bats.

Now, how much of that is in reality leading to real results?

Naturally, I’m not buying there’s much to it.  Sandberg had half his career PA through 1988.  If you take the 20 pitchers he had the most success against through 1988, and then look at how he did against those pitchers from 1989 onwards, what do you think we’re going to find?

In no way am I suggesting, nor expecting, that he’d perform the same as he did.  Or close to what he did.  Or even half way between his norm and what he did.

If he’s actually the sharp-sounding guy that he says he is, then maybe he hits somewhat better than his overall career mean.

***

What are your expectations?  Let’s say that against the top 20 pitchers through 1988, he has a wOBA of .500 (Bonds-like), whereas against everyone else, he was .345. 

When he faced these same 20 pitchers from 1989 to the end of his career, and let’s say he was .360 against everyone else, how do you think he did against these 20?

You guys give me your guesses, and then I’ll do the research.

Fantastic!

I grew up watching Chicago baseball and remember Sutter with the Cubs, when the league was essentially seeing the splitter for the first time.  He was virtually unhittable.  I often wondered why, if everyone (and I mean EVERYONE)was either whiffing over the top or pounding it into the ground, didn’t guys go up there intent on hitting the bottom of the ball.  Clearly, trying to hit it squarely wasn’t working for anyone.

Tango—Shouldn’t we expect him to continue to hit better against those 20 pitchers, if only because of the likelihood that of those 20 pitchers, many if not most will be below-average pitchers?

What I mean is that the original question re: Sutter is especially interesting because it’s regarding his success against a particularly great pitcher with a notoriously tough pitch.

But if we’re just looking at the set of 20 pitchers against whom he had the most success, wouldn’t we expect those pitchers to be poor major league pitchers? And therefore, as they continued to be below-average pitchers, Sandberg would continue to have above-average numbers against them?

Mark/12: You are correct.  If we’re going to be specific, then here it is:

1. We calculate the true talent level of each pitcher Sandberg faces against RHH

2. We select only those 20 pitchers from that group

3. We compare the true talent level of those 20 to the rest of the group.

That sets up our true talent baseline expectations.

4. We see how Ryne performed through 1988 against the 20, and the rest.

That sets up our observed expectations baseline.

5. We see how Ryne performed from 1989 onward against the 20 and the rest.

Where do we expect to see this performance?  Closer to the results of step 3 or step 4?

Okay, all is right with the world now!  You do think Sandberg is the guy in your XKCD comic that’s weaving an imagined narrative around results from a weighted random number generator.  I was starting to think you thought there was actually something to what he said.

I remain agnostic on the basis of evidence (or the lack thereof, really) and hopeful that there’s something to it on the basis of philosophy.

As far as your proposed experiment, that’s one thing among many that you could test.  Though I was under the impression that you already tested that far more thoroughly in the Book and didn’t find any predictive effect.  That negative finding, however, says very little about whether Sandberg’s approach would have any predictive effect.  The main problem I have with your experiment is that “the 20 pitchers against which Sandberg most performed above expectations” is an awfully blunt tool with which to test. 

By its nature, it’s going to include a lot of different things lumped together, many of which will probably not be repeatable.  You’d need to do a study that looked at some of the more specific effects that he noted, like how he did against right-handed pitchers whose primary pitch to right-handed batters was a splitter inside.  Of course, we can’t do that very well for Ryne Sandberg, but we can for current players.

The trouble with this proposed experiment is that is contains an implicit assumption, namely that there is nothing unusual about the interaction between Ryne Sandberg and Bruce Sutter (or, at least, there is nothing about their interaction that is not also true of Sandberg’s interaction with the other 19 pitchers on the list).

The results of such an analysis can’t really tell you anything about the Sandberg-Sutter relationship specifically, because this method dilutes the signal by 95% in the interest of bumping the sample size up; and after such dilution the signal will almost certainly be drowned out by the noise, proving nothing.

This experiment can shed some light on whether there is a generalized effect where Sandberg hits more talented pitchers better than expected, but it can’t tell us about Sandberg v. Sutter specifically any more than an analysis of Garry Kasparov vs. the next 20 best chess players tells about Kasparov vs. Karpov specifically (to go back a few years)…

Now, I’m not saying there’s a better way to use statistics to prove Sandberg genuinely had Sutter’s number; I think we’ll never be able to make that case using numbers alone.  It’s going to come down to whether we buy Sandberg’s assertion.

Mike: I don’t think Sandberg was presented with the results, and asked to explain them.  I think he was asked about his approach to hitting Sutter, and how it was different from how he approaches others.

As a good example: let’s say he had a unique approach to facing Steve Rogers.  That he was personally satisfied at his approach, and that he did everything he thought he did right.  And let’s say he went 2-20 against him.  That is perfectly fine.

The problem is, we never get told about those stories.

Another good example: let’s say he has no unique approach to facing David Cone.  But he ended up going 30-50 against him.  What I’d like him to say is: “I didn’t do anything special with Cone.  I just ended up getting the results I got.”

The problem is, we never get told about those stories.

So.... as long as we don’t start with the results, and ask the player to justify the results (i.e., the comic we see on the right), then I have no problems here.

***

Greg: the test is not about Sandberg v Sutter, but about Sandberg specifically.  Sandberg is talking big about his approach to hitting being tailored by pitcher.

If we see lots of great results, do we expect to see other great results?  Or, are those great results just happenstance?

***

Greg/Mike/anyoneElse: you can also propose a test of your own.