BBTF's Newsblog Discussion :: NYT: Schwarz: At Moneyball U, What Are the Odds?

Baseball Primer Newsblog
— The Best News Links from the Baseball Newsstand

Friday, November 04, 2011

NYT: Schwarz: At Moneyball U, What Are the Odds?

Good stuff from Alan Schwarz…including a Father Gabe Costa sighting!

Watching a baseball telecast may not be the best way to learn basic probability.

Let’s say Derek Jeter of the New York Yankees and his .313 career batting average stride to the plate after 20 straight at-bats without a hit. Sports Announcer No. 1 will invariably discuss how Jeter is in a slump, suggesting that, because of some problem either physical or mental, his chance of getting a hit is somewhere less than his usual 31 percent. And then, right on cue, Announcer No. 2 will giddily agree — “Yeah, Marv, Derek’s definitely due!” — with the “due” implying that Jeter has gone hitless for so long that his chances now are somehow greater than his innate 31 percent.

Welcome to “Bert and Ernie Teach Probability,” America’s less storied national pastime.

...As society gets ever more data driven and risk aware, formal statistics are becoming increasingly relevant. Probability has revolutionized finance, polling and, yes, even the running of sports franchises; popular culture has embraced stat geeks through television shows (“NUMB3RS”) and best-selling books (“Freakonomics”). As Brad Pitt evangelizes the power of probability in this fall’s hit movie “Moneyball,” you know that numbers have earned a seat at the cool kids’ table.

Repoz Posted: November 04, 2011 at 01:59 PM | 13 comment(s) Login to Bookmark
Tags: college, history, sabermetrics

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

1. Robert in Manhattan Beach Posted: November 04, 2011 at 04:03 PM (#3986465)

Sports Announcer No. 1 will invariably discuss how Jeter is in a slump, suggesting that, because of some problem either physical or mental, his chance of getting a hit is somewhere less than his usual 31 percent.

Which is the most likely answer. When a guy is performing below his established norms, he's usually hurt.

2. Best Regards, President of Comfort, Esq., LLC Posted: November 04, 2011 at 04:13 PM (#3986471)

Which is the most likely answer. When a guy is performing below his established norms, he's usually hurt.

20 at bats.

3. Robert in Manhattan Beach Posted: November 04, 2011 at 04:19 PM (#3986474)

You can have the 1600 to 1 odds, I'll take he's probably dinged up.

4. PreservedFish Posted: November 04, 2011 at 04:24 PM (#3986475)

I managed a season with the Strat-O-Matic computer game. On every single AB it would give you the odds of an out, and the odds of a homerun. It would also give you all those odds for everyone on the bench, and in the bullpen. It gave you odds on stealing (I automatically ran whenever the odds were above 80%).

My question is, do managers have a binder with all this information? I understand that the binder would not be the gospel truth, as the odds in the game are, because of the zillions of nontrackable factors. But they'd be pretty good, and they'd probably tell you, for example, that Lance Berkman hits righties just as well as Pujols does.

(Also, I built a team of washed up losers with unfair platoon splits, and using some extreme managing techniques I won about 130 games. This took me about two weeks. It's surprising how much joy this dorky memory brings me. 1999 Donne Wall had a reverse platoon split, he was my ace against switch hitters.)

5. TVerik - Dr. Velocity Posted: November 04, 2011 at 06:18 PM (#3986569)

But they'd be pretty good,

If they're repeatable skills, they would. But I think a study came out a couple of years ago (I don't have a link to it at this time) that said that there's a platoon split among everyone, but there is no repeatable skill in exceeding this split - the myth of a "lefty masher" is really only a myth.

6. Walt Davis Posted: November 04, 2011 at 06:30 PM (#3986580)

his chances now are somehow greater than his innate 31 percent.

Well, if we're gonna be teaching probability, let's start right there.

Jeter's probability of getting a hit is not 31 percent.

Jeter's probability of getting a hit conditional on his plate appearance resulting in an AB are 31 percent.

Jeter's probability of getting a hit is P(AB)*P(H | AB) ... or, more simply, H/PA. For his career, that's .277.

Life would be so much simpler if we'd put PA in the denominator of BA and SLG all along.

7. PreservedFish Posted: November 04, 2011 at 06:34 PM (#3986582)

If they're repeatable skills, they would. But I think a study came out a couple of years ago (I don't have a link to it at this time) that said that there's a platoon split among everyone, but there is no repeatable skill in exceeding this split - the myth of a "lefty masher" is really only a myth.

The binder could apply generic platoon splits if you want it to. Or a special mix. You'd have some top minds working on this binder.

8. Robert in Manhattan Beach Posted: November 04, 2011 at 06:36 PM (#3986586)

We don't need to have that thread again, but that "study" didn't not hold up real well to closer scrutiny. With the variety of approaches, stances, and body types out there of course everyone doesn't have the exact same split.

Just for fun, Reggie Sanders finished an 8000+ PA career at +107 OPS left vs right. Dante Bichette has nearly the same career OPS (835 vs 830) but a spread of only +32 in 6800+ PA, and that's just picking names out of the air. It was silly to believe it would ever be the same for everybody. If you have some veteran guys on your team you should have a pretty good idea of how much bang you are getting out of that platoon advantage.

Edit: Wow, Matt Diaz is currently running at +185 in a less statistically significant 1900+ PAs.

9. Walt Davis Posted: November 04, 2011 at 06:37 PM (#3986587)

By the way, only 11 players (>=3000 PA) have an H/PA over 300 (in OPS+ order):

Hornsby, Jackson, Cobb, Lajoie, O'Doul, Terry, Medwick, Gwynn, Simmons, Sisler, Ichiro

Four of them have an OBP>400 (the first 3 plus O'Doul). 7 of them have a SLG over 500 (Ichiro is easily last at 421). Hornsby and Simmons even manage ISOs over 200.

10. Walt Davis Posted: November 04, 2011 at 06:41 PM (#3986591)

Just for fun, Reggie Sanders finished an 8000+ PA career at +107 OPS left vs right. Dante Bichette has nearly the same career OPS (835 vs 830) but a spread of only +32 in 6800+ PA. It was silly to believe it would ever be the same for everybody.

Huh? Of course the results (i.e. the random sample) aren't the same for everybody. That's not how you detect "skill." The question is whether past splits predict future splits. My understanding (from MGL work years ago that he and others have probably updated) was that for RHB, there is little/no consistent split and you can regress everybody to the standard split; for LHB, there is some.

11. Barry`s_Lazy_Boy Posted: November 04, 2011 at 06:43 PM (#3986592)

If they're repeatable skills, they would. But I think a study came out a couple of years ago (I don't have a link to it at this time) that said that there's a platoon split among everyone, but there is no repeatable skill in exceeding this split - the myth of a "lefty masher" is really only a myth.

I thought mgl published such a claim, which was met with derision. And I think "study" may be an overstatement.

12. SG Posted: November 04, 2011 at 06:55 PM (#3986606)

But I think a study came out a couple of years ago (I don't have a link to it at this time) that said that there's a platoon split among everyone, but there is no repeatable skill in exceeding this split - the myth of a "lefty masher" is really only a myth

What the research in the Book showed is that there's a difference between platoon performance and platoon skill. It's not that there's no repeatable skill in a player's platoon split, it's that you need to regress a player's observed platoon performance towards the league average split for same-handed batters if you want a more accurate estimate of their platoon skill or platoon projection going forward. This is not just because of "luck" and sample size, but also because player skill is not static. For example, any Tigers fan or Yankees fan through July 2010 would have told you Curtis Granderson would never be able to hit lefties and that regressing his platoon splits would be a waste of time. He made an adjustment and now he can.

Near as I can tell, the research was pretty freaking exhaustive, has held up fine and is generally considered the best way to estimate platoon splits going forward. But if anyone has anything further with an actual basis to refute it I'd be glad to see it.

13. Robert in Manhattan Beach Posted: November 04, 2011 at 08:39 PM (#3986684)

That's not how you detect "skill."

Yeah, no ####, that's why I prefaced it with 'just for fun'. And yeah, everything should be regressed, that has never been news.

The skill exists and it is detectable. Like most saber-studies the original conclusion was overstated and had to be walked back quite a distance. But even today folks still remember it was MGL that did the "study". Good to know that throwing out oversized conclusions to build your name recognition never stops working.

You must be Registered and Logged In to post comments.

<< Back to main

Friday, November 04, 2011

NYT: Schwarz: At Moneyball U, What Are the Odds?

Reader Comments and Retorts

BBTF Partner

Support BBTF

Hot Topics