I learned some banzuke projection lessons from Natsu, and stuck closer to my quantitative system, with fewer subjective adjustments. This worked much better, as detailed below. I also think that Nagoya was easier to predict, largely due to many fewer rikishi with 8-7 or 7-8 records.
The San’yaku went exactly to form. The only real question was whether Kotoshogiku would hold on to the second Komusubi slot, and he did. The meat grinder also went almost exactly as predicted, with only Endo and Ura switching positions. Ura had a better computed rank, and I thought Endo would drop further after his 6-9 record, but given his popularity and how well he did against the San’yaku, relatively speaking, this isn’t a huge surprise. Ura might have a slightly easier schedule at M4e than at M3w, which he can use in his first tournament this high up the banzuke, although he’ll still get at least a taste of San’yaku opponents.
The lower maegashira ranks are always harder to predict, but even here, all the projection misses were by one rank, and involved switches of rikishi who had identical computed ranks. It’s hard to see a consistent pattern in NSK’s choices of Takanoiwa above Aoiyama, Okinoumi above Chiyotairyu, Takekaze above Takarafuji, or Kotoyuki above Chiyomaru. In the coin flip M16 slot, Gagamaru got the nod over Kaisei.
Overall, my projection resulted in 28 “bullseyes” (correct rank and side), 3 additional correct ranks on the wrong side, and 11 misses, all of them by one rank. Among the maegashira projections, there were 17 bullseyes, 3 hits, and 11 misses. I’m gaining some confidence that the projections can give us a good early idea of what the official banzuke ends up looking like.
Good work Bruce – I can’t be the only one out there hoping there will be another Tachiai banzuke podcast?! Last two have been great
Lol oops, I thought Bruce wrote this, apologies Iksumo! The podcast point remains unchanged though haha
No worries, Bruce deserves all the credit he gets for all his work on this site, I just chime in with the banzuke predictions :)
BS Sir, I would not touch this banzuke with a rake. You took it on and beat it like drum. Your work on this subject has been a huge help to the site, and to the broader sumo community. I love reading your posts, and I am very very happy you take the time to contribute.
59 points in GTB scoring terms, I suspect that’s a pretty good result this time.
I wouldn’t call Gagamaru/Kaisei a coinflip though – that exchange gets made probably 95% of the time.
Wasn’t remotely good enough for the yusho, as someone had a ridiculous 38/42 correct guesses! I’ll note your comment on the makuuchi/juryo switch scenario for future projections.
Congratulations on doing that well! My predictions were all over the chart – and I paid the price. I got the San’yaku right, but I think that everyone did… those were the fairly easy one. Then, well… the M ranks. This is the third banzuke that I’ve played with and have been horribly off each time.
Great work – congrats! Would love to see if this could be extended to Juryo as well in future.
I did run the numbers for Juryo as part of the Makuuchi promotion projections. The model correctly picked Sadanoumi, Chiyomaru, and Nishigiki as the first 3 promotions, with Gagamaru on the border. It also had Takagenji getting demoted (not that this was a hard call), and Asahisho next in line to go. I can look at the results for the other ranks — I did not attempt any manual adjustments.
Okay, just ran the numbers. Leaving aside Kaisei and the two “demoted” slots at the bottom, there were 25 projections to be made for Juryo ranks (Kaisei or Gagamaru had to be J1e, and I did not run any numbers on Makushita to see who might be promoted). Of those 25, the model had 14 “bullseyes”, 3 additional hits, and 8 misses (the result of 4 one-rank switches). So the model seems to work for Juryo about as well as it does for the maegashira ranks. I’m happy to add a Juryo projection next time.
Pretty amazing stuff! That will be a really cool new feature to look forward to.