Top Rikishi Elo Probability Ratings (μE) since 1957

One of the issues associated with Elo ratings is that the rating associated with a given rank increases over time. This is sometimes called the "inflation problem". For example, Taiho, a rikishi widely rated as one of the greatest ever, was at the peak of his form in 1969 when he was a yokozuna with an Elo rating of 2013 - way higher than any of his contemporaries. However, a rating of 2013 is typical for a M1/komusubi today (March 2024).

I'm interested in what Elo ratings can tell us about a rikishi's performance, but it follows from the above that it only makes sense to perform an Elo-based comparison of rikishi with their contemporaries. Elo ratings allow us to estimate the probability of one rikishi beating another so a simple way of comparing rikishi is to look at the average probability of a rikishi beating everyone else in the division at a given time; this is the quantity μE shown in the results tables.

If we perform this calculation for Taiho in all the bashos in which he competed, we find his maximum μE occurred in January 1969 when it was 0.96 (i.e. 96%) which is extraordinary: for every 100 bouts against randomly selected makuuchi he was likely to lose only 4!

Here is the list of rikishi who have achieved a score of at least 0.8 as makuuchi:

Current Makuuchi

Here is how the current guys rank: