A Closer Look at San’yaku after Week 1


Moment of Silence
One of these men is more consistent than the others – and basically everyone else.

Hi there – this is my first guest post on the site, so thanks for having me! Following Bruce’s analysis last week, contrasting the early upper san’yaku vs lower san’yaku results from the current Basho to that of last year’s Nagoya tournament, I posited that it might be interesting to have a look at a larger sample size and determine if what we’re seeing is the product of a shift in the performance of the upper san’yaku over time, and what it might mean.

I pulled the first week win/loss data (up through day 8) of each tournament going back to January 2015, the logic being that a 2 year period of first week bouts would give us an idea of what it means to perform at the level of a Yokozuna or an Ozeki in this day and age. I selected 8 days because this has typically been the tipping point in the tournament after which the upper san’yaku (of which there have usually been 7) stop being polite and start being real, and are then starting to fight each other.

Finally, I removed forfeit matches due to rikishi going kyujo from the equation entirely. While these wins and losses show up in the history books, they don’t give us a true read on whether or not the rikishi involved are performing at the expected talent level.

This is all basic stuff and is meant to be a jumping off point from which assumptions can be based so that we can have some more detailed conversations in future. So let’s see what the data (which you can see in rawer form here) tells us:

* A Yokozuna can usually be counted on to lose one match in the first week. Over 2015 & 2016, the win rate (or wins per day in Bruce’s format) of a Yokozuna is 85.7%. Taking into account that healthy rikishi will fight 24 times, we’d expect to see 3 losses from the group and this is exactly what happened in 8 of those 12 tournaments.

* Ozeki typically perform 15% worse than the Yokozuna. So, they lose a little more than 1 more match apiece on average than their more prestigious counterparts. Interestingly, this is exactly what has transpired in 2017 as well as 2015 and 2016 – the downturn in Yokozuna performance (~10%) has been almost exactly matched by their Ozeki counterparts in spite of Kisenosato moving up a level: his brilliant Hatsu masked disasters from Terunofuji and Kotoshogiku, while his promotion has covered for the kyujo Hakuho and underperforming Harumafuji and Kakuryu.

Yokozuna Win Percentage 2015-16: 85.7%
Yokozuna Win Percentage 2017: 75.6%

Ozeki Win Percentage 2015-16: 70%
Ozeki Win Percentage 2017: 60.5%

* Hakuho isn’t in trouble – at least as far as week 1 is concerned. He was 7-1 in January and having gone kyujo twice in the 2 years prior (including once mid-basho), he came back with 8-0 and 7-1 starts. He clearly was struggling early in this tournament along with Harumafuji and Kakuryu (who have since seen their records improve), but if his body works then the early results bear out when fighting against his lower ranking san’yaku competitors.

* Sekiwake up their game. Over 2015-16, the expected win rate for a Sekiwake in week 1 was a paltry 46.3% – you wouldn’t even expect them to end up 4-4 never mind challenge for Ozeki promotion, which obviously only happened once (Terunofuji) in that time frame. However this has increased over the first two basho of 2017 to 64.6% – almost as good as an Ozeki would usually be expected to perform (70%).

* This hasn’t been reflected at Komusubi level. Usually Komusubi win 33.2% of their first week matches, but that’s only up to 35.4% this year. Tochinoshin’s tanking at Hatsu is somewhat to blame for this but of the four Komusubi performances this year, only Takayasu has turned in a winning performance at the level in matches fought, and even that was only a 4-3 standing after 8 days. This is still as hard of a level as ever to compete at.

* The recent lower san’yaku level of performance isn’t totally unprecedented, however. While we began to see the tide swing toward the lower san’yaku in January, there were actually 5 better performances over the past couple years – a 5 basho streak starting with a 50% success rate in March 2015 through to a 43.8% hit rate in December 2015 – from the up and comers than in January (42.9% success rate).

* Takayasu may be more special than we think. Most Sekiwake have never returned to the rank over the past 2+ years: only 3 prior rikishi have re-obtained the title after losing it since the start of 2015. However, unlike Ichinojo, Myogiryu and Okinoumi, Takayasu looks to be the first to fight back and look capable of not only hold the rank but have the stuff to move up. Of the prior three, only Myogiryu was even tenuously able to cling on for one more basho at the ranking.

* Kisenosato starts as well as anyone not named Hakuho. While his historical troubles with finishing off the yusho have been well detailed, his week 1 performances have outshined Kakuryu over the last 2+ years and he’s in a dead heat with Harumafuji’s level of performance. While it’s not clear yet that he’s better at this level than a healthy Hakuho, Hakuho’s toe problem means the Shin-Yokozuna was already for all intents and purposes the top dog coming out of the traps even before the the first grain of salt was thrown at Haru.

* No one’s success rate has taken a bigger hit than Goeido. The Ozeki from Osaka’s first week win rate has dropped over 20% in 2017. Historically he’d been expected to win almost exactly two thirds of his opening week matches, but he now sits at a lacklustre 46% – the exact tally we’d typically expect from a Sekiwake that’s might be demoted lower. He really shouldn’t have turned up unfit at Haru.

So what does all of this mean? There’s a certain shift as we’re seeing results that haven’t been there over the past several years, but it’s much too soon to call it a day on the current crop of Yokozuna. At Haru, they’ve lost 2 more matches as a group than we’d expect to have seen, but as recently as November they were turning in a vintage week 1 performance that stands up to anything else they’ve done recently, battering the lower san’yaku with better than expected results.

Health aside, in the short term the data shows that we can probably expect to see some more turbulence in the rankings of a small group of rikishi in between the Sekiwake and Ozeki tiers, and what it means to fight at those levels may become somewhat blurry if Kotoshogiku ends up as the first in what could be a string of yo-yo rikishi. The next two or three basho will be telling to be sure, and I’ll continue to update the data to see how the san’yaku bear out against each other in the early going to try and pick up more signals.

2 thoughts on “A Closer Look at San’yaku after Week 1

  1. Absolutely first class work here – fascinating read, and some great data. I can’t thank you enough for expanding on that initial idea, which was just a short cut to try and demonstrate the imbalance we were seeing in Haru (and Hatsu in January, too).

    Andy and I have frequently discussed our desire to apply more math to sumo, and I think you deserve a lot of credit for piecing this together. Good good stuff.

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s