Math the Banzuke Kyushu 2023

[Hoshotakamoto 259]

Coming in a day late it's the initial installment of the banzuke math BEFORE the start of day 2.

[Hoshotakamoto 259]

So the naive projection after 1 day turns out something like this:

[Hoshotakamoto 259]

Heading into day 3 here are the stacks.

[Hoshotakamoto 259]

Is Shonannoumi currently the most-eligible-komusubi? Fun bit of trivia if he doesn't wind up making it there any time soon.

[Reonito 1,533]

1 hour ago, Hoshotakamoto said:

Is Shonannoumi currently the most-eligible-komusubi? Fun bit of trivia if he doesn't wind up making it there any time soon.

He isn't purely by the math, but he is if you require W > L

[Hoshotakamoto 259]

Well, Shonannoumi has shed the asterisk after day 3, which was not what I expected. I can't stop thinking about the banzuke luck he's had relative to Gonoyama, and now the possibility that he could finish ahead of him with a 6 man cushion by virtue of a 2 win margin seems like it would be a hard pill to swallow.

[Hoshotakamoto 259]

Make it a 7 man cushion if Abi gets a MK. Further down the banzuke, how many times now have I seen Oshoma involved in exactly this type of a logjam, where it seems like someone should get overdemoted and we have ugly underpromotions in the mix?

[Hoshotakamoto 259]

Heading into day 5, I can't give Takakeisho the automatic benefit of the doubt any more 

Nice recovery for Aoiyama and Kotoshoho.

[Hoshotakamoto 259]

In this recalc Yuma comes out a big winner and I set Onosato aside for one more day to avoid pushing Roga and Nishikifuji an extra half a slot down (this feels like a fair representation of how things often shake out in the end).

I started building the banzuke with an empty Komusubi row inadvertently and it went much, much worse than this for Atamifuji. Things shook out nicely after reopening those 2 slots.

I think this is the first time I've ever projected Fukai to land a juryo spot. Interesting run for him as the yusho from 2021 was almost fully erased over time, but he's scratching and clawing his way back.

[Hoshotakamoto 259]

After 5 days I don't think Kirishima expected to fall down the banzuke with a 4-1 start. Perhaps the math supports moving Kotonowaka's name up to Ozeki based on the "pace" he has set for this basho.

[Hoshotakamoto 259]

Things were shockingly straightforward until it finally got weird from Tomokaze down to Mitoryu.

[Hoshotakamoto 259]

Well I let this thing fall several days behind. Rather than submit a bunch of pointless updates from days 6 through 10 I'll jump ahead to the most recent records where sekitori have 11 matches and makushita have 5 matches, then walk forward to the present. I like going with the odd numbered matches because you can model them as if they will finish out 2-2 (or 1-1) and compare vs historical banzuke placements. I actually did spend a considerable amount of time analyzing historical placements over the past few days but the results are not quite ready yet.

[Hoshotakamoto 259]

Stacks after day 10 and 11.

[Hoshotakamoto 259]

Unstacking, I felt like the grinch stealing promotions from these guys to make room for one sanyaku to drop out.

[Hoshotakamoto 259]

These are the actual up-to-date results, possibly omitting some 5-1 rikishi around Ms20. Once their final records come in I'll attempt to place them in a fairly consistent manner with banzuke precedent (which is not very consistent).

[Hoshotakamoto 259]

It didn't seem like unstacking these was quite so bad although it wasn't great around M12-M15.

 

[Hoshotakamoto 259]

Here are the stacks after day 13. The placement of makushita wrestlers is relatively haphazard but I do have a spreadsheet coming together to place them in historical context soon enough.

[Hoshotakamoto 259]

Here's a naive unpacking of the stacks. Felt extremely uncharitable in lower makuuchi and in the Ms5-15 range (tough break for many 4-3 finishers as I could barely find room to move them up). I'll do another run through these with a closer eye on historical placements. Also these are indicative of the final records for anyone who goes 1-1 on days 14 and 15, so for fun, what I will do when the day 14 results arrive is to bake in the assumption that everyone finishes out 2-0 or 0-2 (no point in assuming they go 1-1 since I've already done that on day 13).

[Hoshotakamoto 259]

News is happening in real time and it's my preference to have complete results for makushita and below vs 13 days of data for sekitori, so I'm trying to get this out in a hurry.

Some of the juryo demotees have a full 15 days of accurate data. In other cases I didn't fill out day 14 or 15. My goal is to put together a fairly plausible picture of J11-Ms20. Let's see how it goes. Here are the stacks based on the parameters mentioned above.

 

[Hoshotakamoto 259]

This should be the final installment / unpacking. I've deliberately painted an unrealistic picture above the Juryo line - mostly baking in the assumption that day 15 results were whatever the rikishi got on day 14. There are also a number of exaggerated promotions and demotions in that group based on the algorithm that I'm still tuning. Wouldn't recommend playing GTB with this.

Below the Juryo line I made a bit of effort. Some of the "obviously not moving up to Makuuchi" records are baked in correctly, even if it doesn't appear to be the case. Below Makushita I made a ton of effort to be consistent with decisions made in various past basho. This is still a work in progress but it's the first attempt I've ever made to strive for a certain level of consistency and completeness all the way down to about Ms30. It also isn't clear to what extent the predictive power of this can actually be improved, given that the decisions made in the real world are somewhat arbitrary.

 

[Hoshotakamoto 259]

Well I have forfeited my opportunity to move up or down the official GTB standings this month but let's compile the "real" output of the math heuristics post-day-15 now that GTB is locked in for the other 200 people.

These are the stacks, with the hindsight of the Chiyosakae / Tsushimanada decision (which I probably shouldn't have submitted previously with Tsushimanada in J14w but they got in my head with the exchange bouts).

[Hoshotakamoto 259]

Since we have a clean 42/42 and 28/28 it's not unreasonable to start from the bottom of Makuuchi and factor in the possibility that everyone coming up from below will get dirt kicked in their faces in which case the bottom might look about like this (notice that I didn't move anyone who had a MK until I got to Takanosho).

[Hoshotakamoto 259]

In fact the principle of extreme favoritism for S through M2 can present a scenario where almost all of the rank-and-file movers can get their math-based outcome, while those moving in and out of the maegashira ranks gravitate to their previous positions.

 

[Hoshotakamoto 259]

Then of course the juryo landscape is about like this after continuing to apply the principle of "when in doubt, favor the guys dropping in", and before we try to find room to put the makushita promotees. Hidenoumi, Yuma, and Chiyosakae are marked at the bottom with the most generous placements that have any precedent I'm aware of, and with Akua and Tenshoho it's going to get a little silly.

[Hoshotakamoto 259]

*Throws hands up in the air* maybe this. In any case what I'll submit afterward is a reshuffled ranking down to mid-makushita that attempts to apply a much heavier weight to past precedent vs the theoretical math positions illustrated above.