The banzuke is a crucial element of any basho. It’s the ranking list for each tournament. While it’s formally decided a few days after the end of the previous tournament, it’s not printed out and provided to the public until two weeks before the tournament. One downside is, it’s static. It is not updated during a tournament with results and it shows a wrestler’s current position and not their history or trajectory. The illustrated banzuke is a way to put a face to a name (or a back-side to a name, depending on which illustrated version you’re looking at).
I’ve tried to address that here by taking the basic banzuke and adding more data on each wrestler’s career as well as their record. Their names are colored by whether they were kachi-koshi or make-koshi. It’s not clean and finished yet but I think it’s time for a public beta release. I’ve played around with it enough on my own so it would be nice to hear from others.
The important thing to keep in mind is that I am targeting every sumo fan with these dashboards, regardless of your love for data and technology. I joke about the Kyokai and their fax machines, yes. But I know a lot of technology freaks people out. That’s not what I’m here for. I’m here to use appropriate innovations to provide additional access to more information. In this case, the ability to mouse over a name on the banzuke gives more information about the wrestler. I’ve also put three dropdown filters to allow users to select a banzuke, a division, and/or a stable.
The Filters
As I mentioned above, I have three dropdown filters under the title: banzuke, division, and stable. The default banzuke here is the most recent Nagoya basho while the other two are set to “All”. I don’t encourage setting the Division filter to all, unless you have selected a particular heya. When looking at a stable, it’s nice to view all their members in one screen. Please feel free to cycle through the multiple tournaments. I’ve got the banzuke from 1958 to present because that’s the start of the 6 basho era. I need to put a massive caveat there, that the data for wrestlers’ careers starts there, 1958. So if a wrestler was active beforehand, well, I don’t have that. I know that will be a bummer for folks interested in the careers of past Yokozuna.
The Flyout
If you mouseover the name of a rikishi, you’ll get a flyover screen with additional details about the wrestler, like their stable and hometown, as well as the length of their career. But for me, the part that I find the most useful is the line graph which tracks the wrestler’s path up, and down, the banzuke. I’ve had this visual on other dashboards before. What I’ve been able to add is the kyujo tournaments (in red) and yusho (the golden kanji character for YU).
We can see here that Ironman Tamawashi has not had a tournament kyujo. (He’s actually never had a day kyujo until this past tournament and the Covid kyujo.) He’s also won three yusho: in Makushita, Juryo, and the Makuuchi top division. I’m still working on getting kinboshi and special prize data but this is already pretty busy. If I get that data, I may just list it above the chart. Unfortunately, it’s a bit busy as it is.
I’ve called it the Kitchen Sink because I feel like I’ve throwing all the data into this, and I’ve got more to come. For active wrestlers, I have the height and weight data from the Kyokai as well as style preferences.
The Feedback
I am very interested in getting feedback for this visualization and any others. I’m thinking about hosting a Twitter Space or Zoom or something to help walk people through the dashboard as well as to talk about the data. At work, I use Teams and walk people through with the help of PowerPoint and screen shares. I don’t think PowerPoint would go over well here. I find sharing pdf and pptx files on the internet highly sus, anyway. So I’m not going to do that. If I post a video, it’s not very interactive and I can’t answer your questions, live. We record sessions at work but those aren’t as effective as live-learning. A Twitter Space would just be audio, I think, so that may not be the best option. But sometimes it’s nice to have a quick chat about data.
Well, until I figure out a better option, feel free to leave feedback in the comments. If you experience any problems, notice any bugs, let me know.
My biggest bugbear with the flyout is the very disproportionate representation of space devoted to the lower divisions. Jokes about whether the lower rankers are relevant in the eyes of the NSK aside, it results in a lot of wasted space especially considering that gradient and the x-axis of time should be enough to tell how fast a particular rikishi stormed up the ranks, without adding a good deal of padding to the y-axis.
And speaking of x-axes, labels are pretty important. Tamawashi and Kotonowaka’s careers look the same length but most decidedly aren’t.
On that note, is it technically possible to have a comparison function if you select two rikishi? So something like a combined flyout charting their career progression and whether they’ve had any head to heads.
Excellent feedback. I know what you mean about the scale and I am toying with something to fix it. The source of it is that there are so many more wrestlers in those lower divisions. The scale is 1 rank, 1 point. It’s not until you get to the upper 400s where you reach sekitori. But for those wrestlers we see how little they move in comparison. A Sandanme wrestler with 4-3 record moves many more ranks than a sekitori with 8-7, or even 9-6. That said, I think I have a fix.
I agree with you about the x-axis, too. But since there isn’t a comparison on this dashboard, I figured I would get away with it by showing explicitly how long their careers are.
Watch this space with regard to a head-to-head dashboard. It’s possible but I am planning a different visualization with that because I will include the size data, preferred techniques AND kimarite. I’m planning to do that for only sekitori, though, and only active wrestlers.
The current scaling seems off in any case, though. Sd and Ms virtually the same size, Sd half the size of Jd, juryo less than half the size of makuuchi? I suppose part of it may be due to different divisional sizes over time? But even then the relative widths of the bands don’t really fit if they’re supposed to represent the respective maximum sizes observed since 1958. Not to mention that those maximum sizes didn’t all occur at the same time so it wouldn’t really be appropriate for any era.
Breaking the line graph for kyujo basho also looks less than ideal, IMHO.
The scaling is done at max sizes for each division. That’s why a lot of the jumps appear larger than they really are, especially if you look at someone like Moriurara or others who bounce from Jonokuchi to Jonidan. Jonidan was much larger in the past.
I’ve been thinking about cutting off the banzuke at a more recent period, like 2000, so I don’t have guys ranked Makushita 83 or Jonidan 152. Also, due to the scale there’s not much difference between M1 and Yokozuna. Realistically they’re only 4 ranks apart. It ends up being about 10, though because of the Y2, O2-3, S2, K2. Then the lower end of the Makuuchi dipped to about M20 or M21, iirc.
My ideal would be to have a true representation of 1 rank=1 vertical step and have variable lines indicating the real breaks between the divisions, possibly with a multi-color bar graph instead of having the constant reference lines. I have gone with straight lines for now to keep it simple. To do that, it would keep 0 on the y-axis as the lowest rank. But this would give a Yokozuna an apparent demotion if more people retire than are recruited. Hakuho’s line wouldn’t be as high as that of Chiyonofuji or Tochinishiki. (Which in itself is an interesting view.)
Wow, this is amazing!!
I thought the same thing about the scaling. Although it kinda works against what you are saying here, just having each division the same y-axis size would be nice, I think, as its pretty hard to see any granularity in any division.
I actually like the kyujo as it is. You could have something like tiny arrows above the plot to indicate kinboshi, etc – might not be too busy.
One possible bug (sorry if others have already commented) seems to be that if someone goes winless, they get very little info (including their shikona kanji) showing..
Yeah, I am working on that bug. Takayasu in this latest basho, in particular.
On another note: It looks like the line graph doesn’t handle off-banzuke tournaments well? I was looking at Machi (Jk18e) and it appears that he’s had a series of absences at the end, but there was actually a maezumo appearance in between.
Machi’s record is also a good case study of how the outsized jonidan band is distorting things. It makes it look like his first career basho shot him up into the jonidan upper half, but he was actually just at the 91st out of 112 ranks.
Yes. I need to work on the banzuke-gai bit. I’ll use Machi as a guide. (I also need to fix Akebono’s chart. There’s a definite bug in the middle.)
I noticed that if you have the current basho selected, you can select a currently inactive stable (returning zero results), but if you change the basho date to when one of those stables was active, you can then only select bashos in which that stable was active. Is it possible to only show active stables for a given basho date? It works the other way around, so it appears to be possible.
Amazing work regardless, very interesting way to view sumo data.
Thank you for catching that! I adjusted it to only show the active ones. Check it out and let me know.
This introduces a new issue, though, where stables that are COVID Kyujo are listed as “null”. It’s a bug in my data. I need to pull the stable from a different table. That will take longer to fix.
Cool, it works normally now.
FWIW the “null” for kyujo thing was already showing before the change.
No big feedback yet other than it’s a cool tool. Can you pin it to the frontpage somewhere please?
Yup. That’s the plan when I have it “production ready”, so to speak.
Love This – Very Impressive Work!
Wish I could do this!
I am enjoying this very much – Thanks Andy!
Thank you!
Andy…wow! This is really amazing, good sir! I know that it is still a work-in-progress, I really like what you’re doing here!
Many, many thanks for the work and the sharing
Andy,
I am clearly NOT a Tech. person by any stretch – the only suggestion I can think of (after reading everyone’s comments and just enjoying the amazing job), is I would like to see a color change in the graph when a rikishi drops to another division with each division having its own color. This way I could see rikishi rise and fall and in some cases, rise again. I am new to this but, it would be interesting to me and I could see it at a glance. What do you think?
Thanks again!
Hi Andy. The dashboard is very impressive! Thank you for creating & sharing it. I’d like to make a formatting suggestion: Realizing that you have an international audience where many are not Japanese readers, would it be possible to list the rikishi names in English first followed by the Japanese characters; maybe have a display option giving the reader a choice? Also, along the same line, I think it might improve the readability if the names were all left aligned within a column or tabbed between the Japanese and English.
I enjoy reading your blog and greatly appreciate your effort.
Hello. Great work, Andy. My first time commenting here.
Have you seen this website? https://sumo.danikaru.se/banzukegraph.php
A Sumoforum member made this a while back and it’s very similar in concept to your project. I’ll leave a link to the SF topic about it, as I believe you might find anything useful there – http://www.sumoforum.net/forums/topic/36873-sumo-graphs/
Oh, that’s very cool. Looks like they’ve figured out a great way to have “wavy” division sizes. If I am understanding it right, you can really see the growth until 2020 and then steady attrition since COVID hit. You can also see the seasonality in the recruiting classes with most recruits coming in March, usually. I need to reach out to them. Do you know if updates are planned? Aside from the latest tournaments, getting data going back to the 80s or the 60s would be really cool for comparing the size of “the field” in Chiyonofuji’s day — and before.
In the above SF topic I linked, the creator Yamasanzan posted updates about the site. He said going further back was not a priority back when he launched the site and hopes to update it prior to each basho. You can reach out to him through his SF account, if you are on SF that is.