THE BOOK--Playing The Percentages In Baseball

Where the notion of Vlad’s streakiness comes from is his propensity in the last couple of years to just fritz out for a month or two.

Looking at last year:

Season OPS:  934
Apr:  820
May:  976
Jun:  665
Jul:  989
Aug:  1176
Sep:  946

His worst month was 269 points below his seasonal average, and 511 points worse than his best month.

2005:

Season OPS:  959
Apr:  996
May:  697
Jun:  1274
Jul:  640
Aug:  1047
Sep:  1064

His worst month was 319 points worse than his season, and 634 points worse than his best month.

2004:

Season OPS:  989
Apr:  972
May:  1021
Jun:  1006
Jul:  837
Aug:  916
Sep:  1150

Here, his worst month is only 152 points worse than his season and 313 points worse than his best month.

So there’s a bigger gap between his best and worst now than there was when he first came to Los Angeles of Anaheim.  I’m not saying that means anything (or that it doesn’t, I don’t know)—but that’s where the perception of his new-found “streakiness” is coming from.

Ok, let’s extend it to 20 games then.  His standard deviation is .174 units, compared to a random of .147.

In that case, only 8 times in 100 random sequencing did I get a standard deviation above .174.

So, maybe there is something there.

If one of the math-heavies can recommend a more thorough test, let me know.

Couldn’t you just borrow the methodology found in “Curveball”?  Pick a player and calculate his rolling game-by-game average of whatever metric you choose (they used BA but linear weights would obviously better, IMO).  Then compare that to a randomly distributed game-by-game performance of a fictional player with the same rate stats.  I did something similar below, when I was looking at the perennially streaky Geoff Jenkins:

http://photos1.blogger.com/blogger/7391/569/400/jenkins.0.jpg

I ran “robot Jenkins” 10,000 times and then compared the results to the real Jenkins.

The rolling average is what I did, and I did 5, 10, and 20 games.

The random game-by-game fictional with the same rate stats is similar to what I did.  I chose the same player, and just randomized his games. 

With the 5 and 10 game sequences, it doesn’t look like Vlad is streaky, while the 20 game shows signs of streakiness.  But, that may have simply been luck of the draw.  After all, why 20 and not 12 or 32?  If I keep trying different numbers for different players, and select after-the-fact, I’ll get non-random results.

"I chose the same player, and just randomized his games.”

Well, if you do that with a .300 hitter that batted .400 for half a season and .200 for the second half, you are just mixing up already curious numbers.  It just seems better to me to use a theoretically non-streaky player as a base.

Of course, since we are trying to quantify steakiness but don’t exactly know what it is, this whole exercise is troublesome.

I’m not mixing up “.400” and “.200”.  I mixing up game-by-game.  So, 5-5, and 0-4, and whatnot.  What I’m doing is, using as a given, each day’s performances.  If he’s more likely to go 5-5 than some other great hitter, that’s fine.  I’m only counting games as “plus, average, minus”.  So, I’m not differentiating between a 5-5 or 3-5 game.  They are both pluses.

The “streak” is: does he piece together great games together more than random would dictate.

Even if I do it by LWTS, I’d be surprised if Vlad’s plus games have a different spread than other great hitters’ plus games.

"So, I’m not differentiating between a 5-5 or 3-5 game.  They are both pluses.”

Got you.  I still like the “Curveball” method better but personal preference, I guess.

“Even if I do it by LWTS, I’d be surprised if Vlad’s plus games have a different spread than other great hitters’ plus games.”

So would I, but I was also suprised that the Jenkins has been so consistently inconsistent the last few years.  I looked at him to try and dispel the fan “myth” that he was inconsistent and failed miserably. His 2005 season was absurd and he atcually did better as the nagging injuries mounted (had a broken bone near his hip, IIRC).  His 2006 season proved to be no less inconsistent.