PDA

View Full Version : How many PAs does it take?

nate
06-14-2010, 01:38 PM
How many PAs does it take whether in a split or season numbers for YOU to say, "this guy has developed a skill above what his career numbers indicate"?

dougdirt
06-14-2010, 01:42 PM
I think it depends on the skill.

RedsManRick
06-14-2010, 01:43 PM
This varies based on the skill. Some skills stabilize faster than others. Furthermore, there's no magic line at which it goes from random to real. It's always a guesstimate. It's just that the error bars shrink over time, and at different rates for different skills.

nate
06-14-2010, 01:59 PM
Baby steps, guys!

:cool:

edabbs44
06-14-2010, 02:06 PM
I think it depends on which side of the argument you are on.

dfs
06-14-2010, 02:25 PM
Do recognize that the flip side of what you respond is....

How many at bats do you give a failing regular before you write him off?

nate
06-14-2010, 02:33 PM
Do recognize that the flip side of what you respond is....

How many at bats do you give a failing regular before you write him off?

That's a good question too.

I just want to know when we're talking about "so and so has been great batting in the X spot against LHP from western Iowa in night games" that we're making a difference between what happened and the ability to make it happen.

RedsManRick
06-14-2010, 02:35 PM
Do recognize that the flip side of what you respond is....

How many at bats do you give a failing regular before you write him off?

Slightly different in my book. In both cases you're testing against the null hypothesis. For the "proven veteran" they have an established level of production, presumably a good one, to which you would expect them to regress.

For the prospect, you have a lot less information about what to you expect. You could use a projection. You could use minor league numbers. But in either case, you have a lot less certainty in what you're seeing a deviation from.

That said, it's a fair question for both players. What do you use as the baseline assumption? Do you just use past performance or you do you use past performance along with other factors (age, skill set, comparables, etc.) to create a projection and judge against that?

RED VAN HOT
06-14-2010, 02:51 PM
I think the number of variables involved preclude a simple answer. The answer would be different for a young player with limited career stats than for an older player with five or more ML seasons under his belt. It seems to me that there is also a back and forth between pitchers and hitters. As it becomes known that a hitter has a certain weakness, pitchers exploit it. Then, the batter might adjust to
being pitched that way necessitating a counter move by opposing pitchers. I can see this happening with Stubbs now. Over time, a batter's approach stabilizes and he will simply hit some pitches better than others. At that point change might consist of not offering early in the count at certain strikes that he handles poorly. Or, he might learn to spoil such pitches by fouling them off.

To me the benefits that accrue from these types of practices are likely to be observable in a series of AB's. They should produce measurable successes over a number of PA's, but, I am not sure how meaningful any number would be. Change is a dynamic process.

Ghosts of 1990
06-14-2010, 03:08 PM
Varies on the player also

nate
06-14-2010, 03:22 PM
Varies on the player also

How much? Why?

nate
06-14-2010, 03:23 PM
Slightly different in my book. In both cases you're testing against the null hypothesis. For the "proven veteran" they have an established level of production, presumably a good one, to which you would expect them to regress.

For the prospect, you have a lot less information about what to you expect. You could use a projection. You could use minor league numbers. But in either case, you have a lot less certainty in what you're seeing a deviation from.

That said, it's a fair question for both players. What do you use as the baseline assumption? Do you just use past performance or you do you use past performance along with other factors (age, skill set, comparables, etc.) to create a projection and judge against that?

I know you posted this just recently but perhaps that list of how many times it takes to get a sample size would be useful here.

Mario-Rijo
06-14-2010, 04:15 PM
Yeah it varies for sure depending on alot of variables. Sometimes I can see a change pretty quick but it may take awhile before the change is fully complete. For example I believe Brandon has definitely improved his selectivity. It started sometime early to mid season last year but I don't believe the change was complete until recently, perhaps having to do with hitting out of position. Had he started hitting #2 last year he may have completed the change much sooner. Gomes has definitely made a change for the better (not being as aggressive with 2 strikes) and I believe he has nearly made a complete change, not quite yet but close. Hanigan again definitely improved and I believe his change was complete real early in the process (having the same PA regardless of if he has guys on or not). Jay Bruce OTOH has made a change but is still in the middle stages I think (he's still inconsistent with his approach, he needs to find that point where his PA's are all relaxed).

Just a guesstimate but here is my gut on the subject given few variables.

For a young player 600+ PA's maybe as much as 900.
For a young veteran 300-600 PA's
For a veteran 200-300 PA's
The average 300-400 PA's (My Vote)

nate
06-14-2010, 04:35 PM
For now, let's just say we're talking about hitting...something simply like BA.

oneupper
06-14-2010, 05:23 PM
Cleveland gave up on Phillips after about 450 AB. Too soon to have judged him, I'd say.

RedsManRick
06-14-2010, 09:26 PM
http://www.fangraphs.com/blogs/index.php/when-samples-become-reliable/

Though a 1.0 correlation indicated a perfect relationship, 0.70 is usually the ultimate benchmark in statistical studies, especially relative to baseball, when DIPS theory was derived from correlations of lesser strength. Without further delay, here are the results of his article as far as when certain statistics stabilize for individual hitters:

50 PA: Swing %
100 PA: Contact Rate
150 PA: Strikeout Rate, Line Drive Rate, Pitches/PA
200 PA: Walk Rate, Groundball Rate, GB/FB
250 PA: Flyball Rate
300 PA: Home Run Rate, HR/FB
500 PA: OBP, SLG, OPS, 1B Rate, Popup Rate
550 PA: ISO

Cutter went to 650 PA as his max, meaning that the exclusion of statistics like BA, BABIP, WPA, and context-neutral WPA indicates that they did not stabilize.