Small Samples - From Fangraphs

[sbrylski]

Great article on small samples: http://www.fangraphs.com/blogs/index.php/small-sample-usefulness

 

Lesson: Small samples are still samples of useful data.

[sbrylski]

I also includes a link to updated ZiPS projections taking into account April stats: http://www.baseballthinkfactory.org/files/oracle/discussion/2009_zips_in_season_projections_hitters_through_april_28/

 

(I see battlekow has already been there.)

[TheCrew07]

battlekow is everywhere, he must be independently wealthy or some such thing. I stop bye to read and he's already been there before me and left a comment... doesn't matter where I go, he's already been there.

[logan82]

Lesson: Small samples are still samples of useful data.

 

I think most would agree the data is useful, but many times the data is given to much weight over a larger sample such as a player's career stats or 3-year stats.

[endaround]

Its a really bad article which one can expect from Dave Cameron who really has little clue what he's talking about and takes criticism really poorly. Most point estimates on players are very broad and hence small changes in point estimates aren't that significant. Imagine two different graphs representing probability weight for a point estimate, one and inverted "V" the other an inverted "U" with the point estimate being the maximum point on the inverted letters. For the "V" a shift one direction or the other tells us a lot, for the "U" it really tells us very little. Baseball estimates are in the "U" family.