Re: bell curve

[sn@scgroup.com (Stewart Nelson)]

>But to a first 
>approximation, and since the schedulers match better rikishi against
>better rikishi and worse against worse, a binomial distribution/bell
>curve is a pretty good model---and fits the data reasonably well 
>(except for the unevenness at the 7/8 win level that many people have
>commented on).

What an understatement!  A look at the data for the past year (all that
I have in machine-readable form) shows a *huge* glitch at 8/7.
See graph at www.scgroup.com/sumo/dist.html
I guess nothing is "normal" about sumo :-)