When Do Sumo Wrestlers Peak
A Simplistic Approach and a Request for Any Other Methods
When does a sumo wrestler peak in abilities? This question is unfortunately a bit more complicated than other questions we try and examine here. For instance, when predicting wrestlers’ retirement, we had a concrete metric to aim for: whether a wrestler had retired or not. We could test our predictions against actual verifiable results. When we were predicting the Banzuke, we were able to create a predicted Banzuke, and then compare to the actual one to measure our accuracy. In trying to determine a rikishi’s peak, we don’t have a readily available statistic that we can test against, so our task is harder.
However, fear not; I came up with two different ways to find when a wrestler is at the apex of their athletic prime. The first one is pretty simple: when was the highest a wrestler was on the Banzuke? It has a seductive simplicity; it took about 2 minutes for me to code that metric up. It also makes logical sense, so long as you’re winning more than you’re losing you’ll keep facing tougher competition and proceeding up the ladder. Naturally, the highest you were able to get represents your peak performance.
The other method is also relatively simple (in theory; implementation wasn’t quite as quick): what was the highest rank that a wrestler got a Kachikoshi at, or in other words, won more than half their matches? What recommends this, is we’re getting at where a wrestler “belonged” so to speak, as demonstrated by more wins than losses.
So with these two definitions, let’s compare them. As a note, I broke out Ozeki and Yokozuna separately from the rest of the divisions, as promotion and demotion from them are markedly different. As usual, this covers what I consider the “Modern Era” which is 1988-2023.
I included Diff(erence) in Months as one thought I had was: “could it be the case that Kachikoshi peak is just the tournament preceding the non-Kachikoshi peak.” My reasoning is that after you achieve your highest peak where you can win half or more bouts, you’ll get promoted and then you’ll be over your head and get knocked back down the Banzuke.
Looking at the results it doesn’t appear to be quite what happened and if anything it gives us a range for prime. In fact, this looks like evidence that for the top 2 divisions, wrestlers peak around age 27 or 28.
For the lower divisions, as we saw in the piece on retirement, the non-Sekitori (top two divisions, where they receive a salary) retire much younger than the Sekitori, so it makes sense their peaks are earlier too.
I would like to also address the Yokozuna and Ozeki averages above. I’ve been kicking around various ideas on how to rank Ozeki and Yokozuna performance, but I still don’t have anything ready to report.
The other thing is the calculations: I took the tournament data and grouped by division and then took the average Age at Peak Rank for each division. It’s not perfect, but I think it works well, because it’s taking the average peak age for each division. So a future Yokozuna in Jonokuchi will bring the peak age up a bit as they peak later, however, they’ll only contribute a tournament’s worth of average peak age whereas an 18 year old who reaches his peak at 22 and then retires will have 10 or 12 tournaments included in that average. Hopefully that makes sense but if not the simplified version is, this method leads to the average age being representative of the average participant in that division.
I’m a fan of many other sports and the numbers behind them, and while I don’t have anything to cite here, from my general recollection, it does align decently well with those sports’ athletic primes. I believe hockey and baseball, due to better training that peaks are slightly earlier. If anything, with the tough nature of the sport, I’d expect wrestlers’ peaks to be a tad bit earlier, but certainly the results ring true so that’s a thought rather than a quibble.
If any readers have ideas on different ways I could measure peaks, then I’m more than happy to try them out potentially. I kind of racked my brain for a while and wasn’t able to come up with anything better than this other than connecting with the Elo piece from earlier. I’m looking to work on that again when I have some more time, but I do think it needs that further work prior to me using that for this purpose.
Anyways, this one is a bit of a shorter piece and probably not too much to analyze. However, when I began this project, I think a lot of these kinds of questions, certainly in English, had no answers rigorously and numerically derived. I’ve been very happy with how much of the overall world and analysis of sumo we have been able to fill in with numbers and even how much faith we should place in those numbers. We still have a lot further to go, but it’s always important to remember how far we’ve come.
As an informal preview, the current areas that I still need to work on filling in are figuring out who will face whom every day of the Basho, and I’d like to see if we can learn anything by analyzing Rikishi winning and losing techniques. The latter is because I’d like to see if we can do better than Elo through other means of analysis. No promises on either, because for the former, I believe I’ll have to (I should, and will) accommodate how when a wrestler in the lower ranks is at the top of the tournament after ~10 days they’ll start facing the Joi for instance. For the techniques I have all the data but probably have to re-engineer that database/CSV so yeah we’ll see how it goes.