## Wednesday, October 03, 2012

### Sic Semper Sid

Obligatory last version.  The subtitle is 'Memorex':

Or if you prefer this version:

The Wiz said...

Looking at the graph it appears that there are six seasons that the Pirates did better. But the Pirates would have been at .500 if they had won just two more games, I don't think they came that close to 500 in the prior 19 years. So since the graph shows how far out of first place they ended up, which in a way just shows just how well Cincinnati did. It doesn't show the progress they have made over the past few years.
I could go on about how the interpretation of statistics can be misleading and that this may be a good lesson in statistical analysis but .....some other day.

Thursday, October 04, 2012 9:23:00 AM
MH said...

The statistical term for that kind of thinking is "overfitting", more commonly called special pleading.

Thursday, October 04, 2012 10:42:00 AM
C. Briem said...

This comment has been removed by the author.

Thursday, October 04, 2012 11:40:00 AM
C. Briem said...

.500 is kind of a defeatest goal in a sense.

Speaking of overfitting, I won't name names but I once saw an official local budget projection based on a regression of 5 variables with 4 datapoints.

Thursday, October 04, 2012 11:40:00 AM
The Wiz said...

Didn't mean to imply that .500 was acceptable. I was trying to point out in my limited way that the graph implies (or I inferred from it) that there were six teams in the last 19 years that were better than this one. But I don't think any of them came so close to being .500. .500 would not be acceptable but it is a reasonable benchmark in showing the Pirates are improving and along an better trajectory than in the past.
While the graph is greatly instructive in showing where the Pirates stand in relation to the elite teams of the league, which is where we aspire to be, it is not accurate in showing its improvement.

Thursday, October 04, 2012 2:16:00 PM
MH said...

but I once saw an official local budget projection based on a regression of 5 variables with 4 datapoints.

Let's hope the analyst wasn't a local university grad.

Thursday, October 04, 2012 4:18:00 PM