View Full Version : Swings in players performance

11-20-2006, 11:05 AM
David Ross is a nobody, comes to GABP and goes wild.

Arron Harrang's K/9 grew from the 4's to the 7's and wildly jumped into the 8's last year.

Then we see other players who go well for a few years then have a horrible year or two.

Beyond the obvious stuff like injury and ageing what causes these wild fluctuations? (I know there is always some variation from year to year, but I'm talking about significant changes).

Is there any value to the old "change of scenery" idea? Could it be a coach improving or messing up a player? Hard work? Luck?

How does one account for this from a statistical standpoint? EDIT: I threw in this last bit about accounting for this statistically, but a diiscussion of these variations was really what I was interested in.

11-20-2006, 11:37 AM
How does one account for this from a statistical standpoint?

I'll take a swing at this...

I think we call all agree that players can improve, develop and also regress...in their ability and talents.

There also seems to be a "talent base" around which the players' numbers would seem to be somehow "fixed". You can't get pears from the elm tree...so to say.

Similarly, we could also agree that there is a certain "randomness" in baseball that makes the results deviate from what is a players "talent base".

That "randomness" is reduced by sample size. Give a player 500 AB and he should put up numbers that reflect his "base talent".

Given that randomness, however, there will be exceptions of players who perform far above or below that base talent. There are three possible explanations"

1) External: Injury (physical or mental)
2) Randomness itself (extremely lucky/unlucky...someone has to win the lottery).
3) A shift in base talent.

Here are couple of examples.

Say a player's base talent makes him a .280 hitter (BA is the easiest to do this, bear with me).

The probability that he hits .300 over 500 AB is 14.8% (not impossible)
The probability he hits under .200 over 500 AB is almost ZERO.
The probabilty he hits .200 over 100 AB is 4.4% (sample size effect).

So if there is a VERY large deviation from those "base talent" numbers, you'd have to say that either:

1) Something is wrong with said player
2) The base talent shifted and the original assessment was incorrect.

Sorry for such a long explanation...but that's just how I see it.

Some conclusions from this:

1) A guy like Jason LaRue (not to pick on him) hit .194 last year.
The probability that his base talent is .250 is just 4% (would be lower if sample size were higher).

2) If Dunn is a .245 hitter, the probability of him 'going Pujols" and hitting .300 over 500 AB is about 0.22%. Its not going to happen without a "talent shift".

...and so on.