The following chart shows how the mean Elo rating varies against two measures of
"banzuke size". The first measure is the obvious one viz. the number of rikishi
in the banzuke. The second occurs in contexts where it is presumed that only
rikishi with Elo ratings are being considered; i.e. most of this web site.
In this context "the number of rikishi" and "banzuke size" are synonymously and improperly used to mean "the number of Elo-rated rikishi in the banzuke."
The big difference pre-1989 is due to there being insufficent data to calculate ratings for rikishi divisions below
juryo in the period 1957 - 1988. If you hide the #Rikishi trace (by clicking on it in the legend) you will see that it is identical to the #Rated Rikishi trace from Jabuary 1989. (If you hover the chart with the #Rikishi trace visbile, you will see that it is not identical to the #Rated Rikishi trace because of a bug in my program. TBD!)
The odd "saw-toothed" nature of the plots is due to the fact that, presumably
since 1973, the recruitment of rikishi has been seasonal with people joining in
May and then, again presumably, deciding Grand Sumo is not for them. For evidence that many new recruits leave within the first year, see below.
Changes to Sizes of Divisions
Click through the charts to see how the size of the divisions has changed over
time. Before I made this chart I didn't know that in July 1967, the maegashira
ranks only went down as far as M11 (11!), or that the lowest jonidan rank in May 1994
was Jd210 (210!!).
The vertical spread of data in the charts for jonidan and jonokuchi divisions is a result of the influx of shin-deshi each May, followed by many of these recruits quickly reconsidering their career in sumo. (210!bashos must have started at 2 a.m.)
Career Length
The charts below present information about the length of rikishi careers. They
do not distinguish beween current and past rikishi; i.e. a rikishi who has been
in Grand Sumo for 20 (say) years and is still going strong is counted as
someone who retired after 20 years
The first chart shows how many rikishi had/are still having a career of $L$ years.
The next chart shows the previous data expressed as a probability. For example,
the probability of retiring after exactly 20 years is 0.0098; i.e. just less
than 1% of rikishi achieve this.
The final chart shows the summation of probabilities in the last chart. It shows, for example,
that nearly 14% of rikishi do not stay in Grand Sumo for more than one year and
nearly 25% have left before two years. 50% leave after five to six years.
Inflation
The following chart shows how Elo ratings for yokozuna and other ranks have inflated over the years.