Re: Memo to the College of Arrogance (& applying "The Book" to relievers)

Seth Jackson wrote:
On Fri, 10 Mar 2006 12:32:09 -0500, Kenny1111 <kcyanks1@xxxxxxxxxxx>
wrote:

Seth Jackson wrote:
On Thu, 09 Mar 2006 21:48:29 -0500, Kenny1111 <kcyanks1@xxxxxxxxxxx>
wrote:

Seth Jackson wrote:
On Thu, 09 Mar 2006 09:45:35 -0500, Kenny1111 <kcyanks1@xxxxxxxxxxx>
wrote:

BadgerBC wrote:
Seth Jackson wrote:
On Wed, 08 Mar 2006 22:41:50 -0500, Kenny1111 <kcyanks1@xxxxxxxxxxx>
wrote:

Seth Jackson wrote:
Here's the thing that makes it hard for me to buy the sample size
argument in Gordon's case. It's not just that his postseason
performance is below his regular season performance. It's that it's
so far below that it's almost off the charts. It's absolutely
horrendous. Being that bad seems to indicate something more than bad
luck.
Absolutely horrendous in an extremely tiny sample size. If you are just
relying on stats, and not bringing in the other factors, there is not
enough statistical evidence to draw that conclusion.
What conclusion? That he was horrendous? The numbers speak for
themselves. He was horrendous in that small sample of innings. I
don't see how you can argue otherwise.
I think Kenny is specifically referring to "Being that bad seems to
indicate something more than bad luck." If you look at the innings
pitched (19+) in the postseason compared to his career IP of say 1900,
I think he has a valid claim that it could be noise. The extent of his
ineffectiveness (if you only look at the total runs allowed) doesn't
mean that it proves significance any more than an outlyer would.
Right, that's my point (and what got me in trouble with Keith about
defining the problem to small :-) ). Seth and I think perhaps Joe F.
earlier in the other thread (correct me if I'm wrong), was implying that
Gordon's stats are enough to lead to the choker conclusion. Which is
why I'm going out of my way to explain that his stats alone, minus
external factors, are not enough to lead to that conclusion.
Kenny, I think we all understand your statistical argument. At least
I certainly do. I'm trying to explain the opposing argument, which
In other other post you wrote "That's true only if you presuppose that
there is no such thing as choking." in response to UsualNoise's comments
that based on Gordon's stats (and only stats) you should expect a
performance more like his overall performance going forward. This is
not a statistically correct statement, I think (Neil or someone else,
feel free to chime in of course).
No, that's not correct. If you allow for the possibility that choking
exists, then you can't expect a statistical reversion to the mean.
That means you may or may not get one.
"The Book" says clutch hitting exists. They also say that you have to
regress back to the mean a lot/have a lot of at-bats before any clutch
differences you observe may be meaningful. They allow for clutch
hitting, yet still regress substantially.

But let's assume that choking really does exist. And further, let's
assume that Gordon is a huge choker who simply unravels in the
postseason. If that were in fact the case, then his observed
statistics probably are a good representation of his abilities. If so,
then no reversion to the mean would be expected.

You have to some how account for the fact that he has pitched non-clutch
innings and those reflect on his ability somehow. You aren't going to
say that he's a completely different pitcher to the extent that his
non-clutch innings tell us nothing about his clutch innings. Reversion
to the mean also accounts for this, I think. You are basically saying
that we can just choose to ignore it. I don't think that is valid.


Of course, in reality, we have no way of knowing whether he really is
or isn't a choker other than by observation. Thus, my assertion that,
if you allow for the possibility that choking exists, you can't expect
a reversion to the mean. You may or may not get one.

I'm no statistics expert, but the general method in statistical testing
is to work with a null hypothesis that an effect doesn't exist. That's
a baseline assumption that you try to reject. Your goal is not to prove
the negative, since you can't do that. You can only refuse to reject
it. As Tom Tango explained in "The Book" thread in answering my
question, he prefers to do regressions to the mean. I think you always
will do that, you aren't doing it because of assumptions. You are doing
it because you have lots of other data that matters. Maybe there is
some case where you can just say we should throw out all the other data,
but I just don't see it. I'm not saying we should throw out the clutch
data. I'm saying that it will be overwhelmed by his other performances,
not that we should just get rid of it. You seem to be saying we should
just get rid of his non-clutch performances.

Did you see Tom's thread that the regression for relief pitchers would
be more than the 7600 for batters?

Perhaps Tom can shed some light on this if he is following this thread
as well, since I figure his statistical knowledge is better than mine.
I tried to modify the header in case that helps catch his eye, but I'm
sure he's very busy since the book just came out.

But I still disagree with the approach you are taking, at least for now.
I guess I just don't see how you would do a statistical analysis by
making different assumptions, and how you "may or may not" get a
regression. I don't see a "regression" as something you "get," nor do I
see it as something binary--it's a matter of degree, I think, not a
matter of doing it or not doing it. If you can explain your reasoning
further, I would be very interested to here it.
.

Relevant Pages

  • Re: Memo to the College of Arrogance
    ... It's not just that his postseason ... Gordon's stats are enough to lead to the choker conclusion. ... If you allow for the possibility that choking ... statistics probably are a good representation of his abilities. ...
    (alt.sports.baseball.ny-yankees)
  • Re: Memo to the College of Arrogance
    ... It's not just that his postseason ... He was horrendous in that small sample of innings. ... indicate something more than bad luck." ... Gordon's stats are enough to lead to the choker conclusion. ...
    (alt.sports.baseball.ny-yankees)
  • RE: How to link tables via the field names themselves, not their c
    ... you can query the database to do what you want easily. ... There are approximately 30 different aspects (or statistics) which are ... Beacuse there are so many stats per contract, ...
    (microsoft.public.access.queries)
  • RE: How to link tables via the field names themselves, not their c
    ... you can query the database to do what you want easily. ... There are approximately 30 different aspects (or statistics) which are ... Mulitple audits are made per contract. ... Beacuse there are so many stats per contract, ...
    (microsoft.public.access.queries)
  • Re: Just Arod being Arod
    ... they will lead to better or worse statistics. ... Player B has very little heart, guts, and nerve. ... ARod has a hell of a long way to go to match Reggie in Yankee lore ... pitches, it's in his stats. ...
    (alt.sports.baseball.ny-yankees)