Sumo Smarts - Lessons Learned from Analyzing Sumo - Part 1 Development
It’s been a minute.
Glad to be back, and hopefully more frequently.
I’ve watched probably 5+ years of sumo every tournament and a year and a half doing data analysis on sumo. I also keep up with many other sports and the more analytic aspects of them. If it’s baseball then it’s WAR, OPS+, pitch mixes; if it’s a sport I follow, odds are I’m looking at the advanced stats. As a result, I think I have some accumulated lessons and wisdom that help me think about sumo more intelligently, and to hopefully have a deeper and more accurate idea of what is going on. So I decided to start with this as the first in a series on lessons I’ve learned that I think can help readers think more deeply about sumo too. They often have direct analogues in other sports, so I’ll connect the ideas to show their validity and also teach you a bit about sports beyond sumo. Finally, I’ll also try to bring in the statistical reasoning behind these lessons but in an approachable way.
As a quick sidenote: I’m trying to integrate more videos into my writing going forwards. So if I mention a guy I’ll try to include highlights - the analysis is fun for me but at the end of the day it’s the guys wrestling we care about. I’m also trying to explain more sumo concepts when I mention them. The latter was due to reader feedback so please feel free to drop advice for improvements. Let’s begin.
Lesson 1: Development is Not Linear
Development is not linear. It sounds simple, but it’s become an important phrase across many sports. In American sports with drafts, a very important advance in statistical analysis was the realization that the younger a player is at the time of the sports draft - NBA, NFL, MLB, all the same - the higher their overall potential. It makes obvious intuitive sense: across all sports, including sumo as we’ve discussed here before where we found they peak at age 27, athletes tend to peak in their mid to late 20’s. You’re likely at your best physical performance but experienced enough in the sport to make better use of your physical attributes than in your early 20’s. If a college wide receiver has a 1,000 yard season as a 19 year old that’s obviously more impressive than doing it at 23. The former is a kid and has presumably more room for development vs the latter a man among boys with more experience in college football.
Similarly, if a sumo prospect reaches the top division Makuuchi at age 21 then that’s more impressive than doing it at age 25. In fact, we did some analysis and found that future Yokozuna debut in the top division a year younger on average than wrestlers whose highest rank is Ozeki (the rank below Yokozuna). Again, this makes intuitive sense so it’s great that the numbers backed us up there.
However, it’s very easy to fall into recency bias. We see a young wrestler rocketing up the lower divisions and even into the Joi (the top part of the top division) and assume it’s a matter of time till they’re an Ozeki or even a Yokozuna only for them to struggle towards the top. Suddenly the prospect shine comes off them and it seems they’ll only ever be a mid Maegashira wrestler. Atamifuji more or less fits this mold to a tee and I have seen people souring on his potential. However, Tochinoshin - the Georgian wonder wrestler of such great strength - fit a similar mold. He made the top division by 21 and stuck in Maegashira for years even winning a runner up and going 12-3 multiple times but never stuck in the Joi. With injuries dropping him down to Makushita (third division), he could have easily been written off. In fact even after he dominated on the way back to the top division, he didn’t seem destined for Ozeki until at age 30, nearly a decade after debuting in Makuuchi he put everything together, and won a tournament en route to making Ozeki finally fulfilling that potential.
The flip side is that we see guys that never did anything special and assume that’s all they’ll ever be. If you watched sumo in the 80’s and saw a 25 year old making his debut in the Makuuchi, you likely assumed he would peak as a mid Maegashira guy, maybe make Komosubi once or twice and that would be the best case scenario. If you watch enough sumo you’ll start to recognize wrestlers having career archetypes like that. Except in this case that wrestler was Kirishima the elder who slowly grinded his way to the top making Ozeki at over 30. He would go on to win his first top division tournament and get several more Jun-Yushos (runner up) along the way in his 30’s before gracefully descending down the Banzuke similarly slow to his methodical advance to the top.
Finally, a more depressing scenario in which it’s important to keep in mind that development is not linear is the case of Takakeisho. As an Ozeki at 23 having already won a tournament from the Joi at 13-2 he looked like a surefire Yokozuna. Unfortunately injuries and an inability to string together two Yokozuna quality tournaments in a row ultimately doomed that dream. He arguably peaked much younger than 27 and at 28 is retired from wrestling himself.
We’ve done the research and found that wrestlers peak at 27, but that’s ultimately just the average of all wrestlers. Each of them has their own unique path and some will be early bloomers while others will hit a wall maybe even for years only to finally unlock a new level of skill. If you’re an NFL fan the latter example is akin to Zack Baun who broke out at 28 years old making First Team All Pro and being a key cog on a Super Bowl winning team1.
In all these cases it’s easy to overindex on the most recent results. In other words, we see someone stalling and we assume that’s all they’ll ever do. Or we see a high flying prospect and assume they’re destined to be a Yokozuna. This is what we call recency bias.
But the truth is, every wrestler is unique as is their developmental path. I imagine if you reflect on your life or career, it hasn’t been a continuous journey but rather ebbs and flows; successes and failures but not in a predictable way. We often think of development and skills as monotonic. That is a fancy word for something that is always going up (or down) and here’s a helpful graph that illustrates it2.
Ultimately, I think this concept of development not being linear or monotonic makes sense when you think about it and relate it to your life and even sports careers you’re likely familiar with. We all have our ups and downs. Nevertheless, it is worth constantly keeping this as a refrain in the back of your mind. I find it helpful for when a guy I expected more of hits a wall in the Makuuchi (I still believe in Atamifuji for instance) - a young and injury-free wrestler always has more time to put things together. Similarly, it gives me hope that some non-prospect wrestler with a fan friendly style might actually have more skill than his record suggests. Tobizaru debuted in the top division at an age that suggested he didn’t have much more to unlock, but luckily for fans of high octane wrestling he has showed he can keep up with the best while providing endless entertainment.
Sumo is hard to predict and we’re better off for it. I like learning about it with my spreadsheets and math and data, but at the end of the day wrestlers are humans like us and just as unpredictable. That’s what makes it worth watching.
Hopefully you enjoyed this because Part 2 is coming soon and will give us a related lesson.
Not an Eagles fan or anything, but just a really prominent example in my mind for whatever reason
Technically we’d think of it as two monotonic graphs, one going up till you hit a peak, and then one going down till retirement. As one graph it’d be nonmonotonic since it goes down also, but the point is we think of athletes only getting better, hitting a peak, and then only getting worse. It’s a simplification to help introduce the concept of a monotonic function which I think is a useful concept and as this digression shows, it can get pretty tedious pretty quickly. The wonders of math and statistics!