PDA

View Full Version : Question: Sample Size for Plate Appearances


soxfanatlanta
01-23-2007, 07:03 PM
I need a stat head here. In a recent thread, I mentioned that a player had a poor average against a particular pitcher. Somebody quickly chimed in and wrote that the sample size was too small to give the numbers any credence. I had to agree with this poster, and I sent a PM asking what an adequate sample size is. The answer:

No idea.

Hmph.


After a few articles (found at baseball prospectus (http://www.baseballprospectus.com/)) around 30% into their total at bats is accepted by some. A quick sample size calculator came up with 218 at bats (assuming 500 for the season). I tried to research what a good figure would be for a batter vs. pitcher scenario, but I came up empty. When can one say that a full time player does well against a particular pitcher?

thanks for any insight.

itsnotrequired
01-23-2007, 07:21 PM
I need a stat head here. In a recent thread, I mentioned that a player had a poor average against a particular pitcher. Somebody quickly chimed in and wrote that the sample size was too small to give the numbers any credence. I had to agree with this poster, and I sent a PM asking what an adequate sample size is. The answer:

No idea.

Hmph.


After a few articles (found at baseball prospectus (http://www.baseballprospectus.com/)) around 30% into their total at bats is accepted by some. A quick sample size calculator came up with 218 at bats (assuming 500 for the season). I tried to research what a good figure would be for a batter vs. pitcher scenario, but I came up empty. When can one say that a full time player does well against a particular pitcher?

thanks for any insight.




I believe my actual response was "beats me".:D:

There is no firm, set number. Its more a gut feel than anything else as it is also not just the number of at-bats but the trends in the at-bats. A batter may be very hot/cold against a pitcher one year but then cold/hot for the next two. The raw numbers may not necessarily reflect trends like this.

Daver
01-23-2007, 07:25 PM
I need a stat head here. In a recent thread, I mentioned that a player had a poor average against a particular pitcher. Somebody quickly chimed in and wrote that the sample size was too small to give the numbers any credence. I had to agree with this poster, and I sent a PM asking what an adequate sample size is. The answer:

No idea.

Hmph.


After a few articles (found at baseball prospectus (http://www.baseballprospectus.com/)) around 30% into their total at bats is accepted by some. A quick sample size calculator came up with 218 at bats (assuming 500 for the season). I tried to research what a good figure would be for a batter vs. pitcher scenario, but I came up empty. When can one say that a full time player does well against a particular pitcher?

thanks for any insight.




You would have to base it somewhere near 100 AB's against the pitcher for it to have even a shred of credibility, but a better number would be 200.

You could also figure out numbers on how a batter does when he has gum on his shoe as opposed to when he doesn't, those numbers would have as much value as the first set.

FedEx227
01-23-2007, 08:14 PM
I believe my actual response was "beats me".:D:

There is no firm, set number. Its more a gut feel than anything else as it is also not just the number of at-bats but the trends in the at-bats. A batter may be very hot/cold against a pitcher one year but then cold/hot for the next two. The raw numbers may not necessarily reflect trends like this.

Exactly. It's just kind of what you feel comfortable with. You have to make sure you have enough to eliminate a hot week or month where Player A who usually bats .230 started batting .320... something like Chris Shelton at the beginning of last year. In terms of the particular stat you're looking for it's tough to say.

What I should probably ask first is what are you comparing? A batter in the same division as the pitcher? In which case two years worth might do it. However if you got guys in different divisions they might have only faced each other 10 times, which is nowhere near enough to determine.

kjhanson
01-24-2007, 12:25 PM
If baseball weren't so variable, you would be able to base it on 25-30 at-bats. However, you would have to assume the conditions are the same for each at-bat (weather, age of players, situation, etc.) Because this is never true, and there's no way to paramaterize all the variables, there really isn't a true "statistical" answer to the question. None-the-less, I would feel confident interpreting the results of about 25 at-bats against a certain pitcher over a 3-year span.