THE BOOK--Playing The Percentages In Baseball

I started reading the article and thought, what a great self-promoter Maddon is.  If you have a good team, and you do a bunch of things that have a minute impact on your team’s success, you’re still likely to win a lot of games because you’re a good team.  But, as the manager, you might grab some of the credit due to people seeing an illusory correlation between your team’s wins and your managerial work.

Then I got to the middle of the article, and saw this, and thought there might be something cool in this article:

“There’s so much ready information now that was not available a couple years ago—and even if it was, I don’t know if everyone wanted to read it or understand it or try it,” Maddon said. “The fact that I get so much good stuff from upstairs, it permits me to look at these things and feel somewhat confident about doing them.

And then, I saw this:

Maddon said filling in the bottom half of the lineup card takes a little longer than it used to, and he starts the process in the morning. He gets on his computer and looks up the relevant statistics—the opposing pitcher’s tendencies and recent outings, how his players are hitting against that type of pitcher and so on—and goes from there.

And I was back to my initial analysis.

Seeking to take advantage of whatever edge his batters have over the opposition has allowed Maddon to give a team with offensive limitations (such as the fourth-lowest batting average in the American League) the second-greatest run differential in the Majors (+112, just one run behind the Yankees).

How many assumptions in that sentence are incorrect?

Mike, you probably spend too much time on my blog, because those were EXACTLY the three passages that jumped out at me.

Looks like Josh Kalk is keeping Maddon busy.  Josh is finally contributing to the sabermetric community again in a round about way.

The Rays have an OPS differential of .039 (.751 - .712). My equation for winning pct is

Pct = .5 + 1.26*OPSDIFF

That would give them a pct of .549. Instead, it is .606. A .549 pct means 51.6 wins but they actually have 57.

Their OPS differential with runners on base is .082 (.773 - .691) and with runners in scoring position it is .119 (.773 - .652).

My guess is that this helps explain why they are winning more than expected. Is it luck or has Maddon figured out a way to do better with runners on, especially on the pitching side?

What’s the probability that Maddon is feeding other teams misinformation by so openly discussing meaningless batter-pitcher matchups? I realize it’s pretty low because he’s made moves in the past that appear to be based on similarly meaningless data, but still . . . 10%? If I thought that I had a genuine strategic advantage that other teams weren’t taking advantage of, I wouldn’t be totally gung-ho about discussing it in the press. Maybe he thinks the hard part is learning how to use the information, not being aware of its existence, and of course he’s more right than he knows. Or maybe he thinks no one will listen. Billy Beane felt the same way.

Cyril,

OPS differential seems like an awfully blunt instrument for calculating these things…
Blunt and also incorrect.  It badly undervalues OBP vs slugging.

I, bluntly, can’t trust anything based on that for fine comparisons, it’s too inaccurate a reflection of what really scores runs.

Why not something wOBA based?

If wOBA for all those situations, by both a team’s hitters and pitchers, were easily available, sure I would do that. But the OPS info is very easy to get at ESPN. 

When I ran regressions a few years ago, OPS differential predicted team winning pct just about as well as breaking it down into OBP differential and slugging differential.

But here are their differentials in each stat in each situation

OBP
Total) .036 (.343 -.307)
ROB) .053 (.359 - .306)
RISP) .068 (.369 - .301)

SLG
Total) .004 (.409 - .405)
ROB) .029 (.413 - .384)
RISP) .054 (.404 - .350)

So this all makes the same basic point. That the Rays are good but their big advantage over their opponents is with runners on. It looks like they ramp up their OBP with runners on while hitting and they really hold down the other team’s SLG with runners on. Again, how do they do this? Have they found some secret formula?

And I would be happy to see someone do a regression that shows that wOBA differential is a better predictor of team winning pct than OPS differential.

I agree that I cringe every time I see Cy quote OPS.  It goes against everything I talk about in this blog.  It’s a blunt (I like that word) instrument, where something more finely crafted is needed.

Oops… I meant the “finely crafted” would be needed in the clutch thread.

In this case, blunt is fine.

Please tell me where I can find wOBA data for all these situations for both team hitting and team pitching.

Have I really been using OPS at your blog that much?

I think OPS is okay for this particular topic. But I was nice enough to give Patrick the breakdown he asked for.

The regression I did a few years ago included all teams from, I think, 1989-2002. I only used walks, hits and ABs for OBP. Then I combined every team’s totals over all those seasons, so the Angels had a record of 1085-1118.

The OPS differential equation was

Pct = .5 + 1.21*OPSDIFF

r-squared was .935 and the standard error was .009495. Times 162 = 1.538

By OBP Diff and SLG diff

Pct = .5 + 1.55*OBPDIFF + 1.03*SLGDIFF

r-squared was .938 and the standard error was .009466 or 1.533 for a whole season.

It really surprised me that the second one did not do that much better than the first. I thought it would and that I would get something really good.