Playing with statistics

[Niffoc4]

I wanted to play with stats from the last few years, but all I could get was stats from 2005. I wanted to look at the standard deviations of hitter statistics, and try to establish what standard in each category should be especially useful. First, I eliminated everyone with under 200 at-bats, because it would just drag down all of the R HR RBI numbers, then I found the mean totals, then the standard deviations, then the numbers at 1 standard deviation above the mean (a point where 80% of players should get less than this number), 2 STDDEV's (97%) and 3 (99%). I put them together to form average joe (whose numbers you should be able to find on the WW), and then the standard deviations
R HR RBI SB AVG
AVG Joe 60 14 57 7 .269
STDDEV1 84 24 84 17 .295
STDDEV2 109 34 110 27 .320
STDDEV3 132 44 138 36 .347

Of course, few players ever get numbers even close to STD DEV 3, but there are some interesting things to note from these numbers. First, other than the SB, STDDEV3 looks a lot like PUjols to me. Second, standard deviation for both SB and HR is roughly the same (10 vs. 9.88), but roughly twice as many HR are hit each year. Third, BA should be easier to get strong numbers in than most other categories, considering that 1 STD DEV from the mean is .295 which is better than most FB teams average.

[mbuser]

http://www.letsplay2.com/baseball/stats.cfm

stats from previous years if you are interested

[RugbyD]

if you want to poke around, there are severl threads that talk about this sort of thing with respect to rankings. there are plenty of iterations beyond the basic math principles. for example, when i'm compiling rankings, my 'average joe' will be the product of only players that i consider draftable by someone (not just me or the sample size shrinks too much). I'll also discount the value of certain stats based on position, like SBs are generally useless when ranking 1Bs b/c so few have any. it makes the few that do look far more important than they really are. these really help in the creation of distinct tiers for each position when you add up the adjusted standard deviations across positions

etc etc.