Sumozumo

2015 Aki Fully Endorsed by the YDC Edition The YDC was talking about a couple of schmucks or something, but they've got MKs on their records! Who cares? Shikona Age Country Heya Debut KK Current Rank Yushos Record Win% Last Basho This Basho Shodai23TokitsukazeNatsu 20149 Juryo 12 West255-160.775 5-2 11-4 Sato18TakanohanaKyushu 20146 Makushita 21 West235-70.833 4-3 6-1 D Ohata18TokitsukazeHatsu 20154 Sandanme 52 West021-140.600 4-3 2-5 Inabajo18SakaigawaHatsu 20154 Sandanme 99 East018-100.643 4-3 kyujo Mitakeumi22DewanoumiHaru 20154 Juryo 5 West135-90.795 11-4 Y 12-3 Iwasaki23OitekazeHaru 20154 Sandanme 11 West024-40.857 6-1 6-1 Omoto22IrumagawaHaru 20154 Sandanme 33 East022-60.786 6-1 5-2 Kotonoawa18SadogatakeHaru 20154 Sandanme 59 West021-70.750 5-2 5-2 Ryutsukasa18IrumagawaHaru 20153 Sandanme 49 West018-100.643 6-1 2-5 Ura23KiseNatsu 20153 Sandanme 18 West119-20.905 7-0 D 5-2 Daiki23HakkakuNatsu 20153 Sandanme 36 East220-10.952 7-0 Y 7-0 Y Asayamabana22TakasagoNatsu 20153 Jonidan 14 East017-40.810 5-2 6-1 Hirotsukasa22IrumagawaNatsu 20153 Jonidan 23 West015-5-10.750 5-2 5-2 Inoue16KiseNatsu 20153 Jonidan 24 West014-70.666 5-2 4-3 Saionji21ShikihideNatsu 20153 Jonidan 1 West016-50.762 6-1 5-2 Toshinofuji18IsegahamaNatsu 20153 Jonidan 46 West012-90.571 4-3 4-3 Yajima19TamanoiNatsu 20152 Jonidan 49 West011-100.524 4-3 3-4 Hamamachi22KokonoeNagoya 20152 Jonidan 10 West214-01.000 7-0 Y 7-0 Y Kiribayama19MichinokuNagoya 20152 Jonidan 67 West011-30.786 5-2 6-1 Tsukimoto23IsenoumiNagoya 20152 Jonidan 69 East010-40.714 5-2 5-2 Homarenishiki20NishikidoNagoya 20152 Jonidan 82 East010-40.714 4-3 6-1 The wall-beya had a very good basho with only 3 guys hitting the wall, and one just... leaving I guess. A couple of yusho but mid basho it looked promising that there might be more. Shodai did very well on his Juryo debut and racked up double digit wins. Next time he shouldn't have to bother with any more Makushita opponents. Maybe we'll even get a glimpse of him in Kintamayama's daily videos next time. Sato had two chances at the yusho this basho and dropped them both. It was still an extremely good effort and ended with an unofficial record of 8-2 with the massive playoff. He'll bump up into the single digit Makushita ranks but nothing short of a 7-0 Y is going to see him promoted next time. Probably at least two more basho before Juryo if not more. Ohata somewhat expectedly ran out of steam and hit the wall. Nobody is left from Hatsu 2015. Mitakeumi gave a good go at a second juryo yusho and ended up with the completely fabricated and not real juryo jun-yusho. Next time he'll be in Makuuchi. Hopefully he can surprise some Makuuchi veterans with his sudden arrival. Iwasaki lost to only eventual Sandanme yusho winner Daiki, and continues to post 6-1 records while pulling away from other non-TD Haru 2015 debutants. Ura picked up his first two losses in sumo this basho (disregarding playoffs), losing to fellow Wallee Iwasaki and also some other guy. Daiki picked up his second consecutive yusho and looks to enter Makushita on a strong note. Hopefully he can find some opponents on his level soon... Inoue, at 16, quite impressively picked up his 3 straight KK in upper Jonidan. He may not last much longer here at The Wall but he has a great base to develop from. Toshinofuji just barely hung onto a KK score helped along by a fusen win. He continues to maintain the minimum winning percentage. Four more guys join us after this basho. Hamamachi completed another clean sweep. Kiribayama is from Mongolia Tsukimoto is from Doshisha U (Doshisha represent!!!) And Homarenishiki doesn't need introductions on this forum, but I'll tell you something you don't know. His sumodb ID is 12234!

2015 Aki Mid Life Crisis Version Shikona Age Country Heya Debut KK Current Rank Yushos Record Win% Last Basho This Basho Shodai23TokitsukazeNatsu 20148 Juryo 12 West244-120.786 5-2 7-1 Sato18TakanohanaKyushu 20146 Makushita 21 West229-60.829 4-3 4-0 Ohata18TokitsukazeHatsu 20154 Sandanme 52 West019-90.679 4-3 2-2 Inabajo18SakaigawaHatsu 20154 Sandanme 99 East018-100.643 4-3 kyujo Mitakeumi22DewanoumiHaru 20153 Juryo 5 West123-60.793 11-4 Y 7-1 Iwasaki23OitekazeHaru 20154 Sandanme 11 West018-30.857 6-1 4-0 Omoto22IrumagawaHaru 20153 Sandanme 33 East017-40.810 6-1 2-2 Kotonoawa18SadogatakeHaru 20153 Sandanme 59 West016-50.762 5-2 3-1 Ryutsukasa18IrumagawaHaru 20153 Sandanme 49 West016-50.762 6-1 0-4 Ura23KiseNatsu 20152 Sandanme 18 West114-01.000 7-0 D 3-1 Daiki23HakkakuNatsu 20153 Sandanme 36 East113-10.929 7-0 Y 4-0 Asayamabana22TakasagoNatsu 20152 Jonidan 14 East011-30.786 5-2 3-1 Hirotsukasa22IrumagawaNatsu 20152 Jonidan 23 West010-3-10.714 5-2 2-2 Inoue15KiseNatsu 20152 Jonidan 24 West010-40.714 5-2 2-2 Saionji21ShikihideNatsu 20152 Jonidan 1 West011-40.786 6-1 3-1 Toshinofuji18IsegahamaNatsu 20152 Jonidan 46 West08-60.571 4-3 3-1 Yajima19TamanoiNatsu 20152 Jonidan 49 West08-60.571 4-3 1-3 Shodai is impressing in Juryo so far, with a very close to KK record of 7-1. Unless something awful happens he should be able to achieve it for the 9th time in a row. This is a much better first Juryo showing than his rival ex-Takagi After a rocky basho last time, Sato comes back with a 4-0 record as The Wall's only Makushita wrestler. His toughest test so far will come tomorrow against Ms10w Chiyonokuni. Inabajo, after four basho without ever scoring a MK, has not shown up. Has he called it quits for good? I'm not sure! Mitakeumi mirrors Shodai in his Juryo performance. However, a yusho is much tougher this time as there are stil 2 undefeated Juryo rikishi. Hmm, that must mean there is as many terrible records as these top 4 have good records... Iwasaki is still strong compared to his current ranking, and has already claimed a KK. Kanamori -> Ryutsukasa celebrated his Sandanme berth by changing his shikona but it has had the opposite intended effect as he drops an early MK with 0-4. Its hard to imagine someone hitting the wall so hard, so there might be an injury. Daiki looks for a 2nd consecutive yusho with a 4-0 KK record so far, but he may have to off Iwasaki in the process... *Total WL records have not been updated!

Which rikishi were you thinking of here? Maybe I'm misunderstanding what you were getting at, but considering that almost nobody who was dismissed in 2011 was anywhere near sanyaku, I find it hard to believe that anybody was helped getting there by the dismissals. Yes, but you don't have to beat any sanyaku to become sanyaku. After everyone was flushed out the bottom part of makuuchi wasn't really the same level and it was a lot easier to pole vault into sanyaku with 11-12 wins from there. Then it took a few years for guys to cycle in to refresh things back to normal. I would say Wakakoyu would have never even come close without it and that it made it much much easier for Homasho and Gagamaru.

I'm using matlab right now, which is a bit pricey even for the 'home' version. I'll take a look a bit later and see if it ports easily into Octave, the open source version of a matlab clone. Even if you don't use octave to operate on the data, I could make a version that spits out a CSV or something importable to excel.

Poor graduate Sekizuka found his new mid Jonidan digs a bit much, and went 0-7 this Nagoya.

I haven't even though about the issues you've pointed out. Instead of caring about a million edge cases I might as well just retrieve the HTML source of the query results and re-use the IDs the DB uses, saves a lot of work that's already been done. Then I can just select the most recently used Shikona to get around name changes. That's impressive and sounds like a lot of work! Have you ever thought about trying out Python or Ocatve to do that? With that kind of work ethic you shouldn't have much of a problem tackling that...

Nagoya 2015 Finalized Locked and Loaded Edition Shikona Age Country Heya Debut KK Current Rank Yushos Record Win% Last Basho This Basho Shodai23TokitsukazeNatsu 20148 Makushita 1 East244-120.786 4-3 5-2 Sato18TakanohanaKyushu 20145 Makushita 27 West229-60.829 6-1 4-3 Ohata18TokitsukazeHatsu 20154 Sandanme 69 East019-90.679 4-3 4-3 Inabajo18SakaigawaHatsu 20154 Sandanme 99 East018-100.643 5-2 4-3 Mitakeumi22DewanoumiHaru 20153 Juryo 12 West123-60.793 6-1 11-4 Y Iwasaki23OitekazeHaru 20153 Sandanme 66 West018-30.857 6-1 6-1 Omoto22IrumagawaHaru 20153 Sandanme 90 West017-40.810 6-1 6-1 Kotonoawa18SadogatakeHaru 20153 Sandanme 92 West016-50.762 6-1 5-2 Kanamori18IrumagawaHaru 20153 Jonidan 11 East016-50.762 4-3 6-1 Kenshin19SakaigawaHaru 20152 Jonidan 29 West012-90.571 5-2 3-4 Ura23KiseNatsu 20152 Jonidan 10 East114-01.000 7-0 Y 7-0 D Daiki23HakkakuNatsu 20152 Jonidan 36 East113-10.929 6-1 7-0 Y Asayamabana22TakasagoNatsu 20152 Jonidan 47 West011-30.786 6-1 5-2 Hirotsukasa22IrumagawaNatsu 20152 Jonidan 62 West010-3-10.714 5-1-1 5-2 Inoue15KiseNatsu 20152 Jonidan 64 East010-40.714 5-2 5-2 Saionji21ShikihideNatsu 20152 Jonidan 71 East011-40.786 5-2 6-1 Toshinofuji18IsegahamaNatsu 20152 Jonidan 78 East08-60.571 4-3 4-3 Yajima19TamanoiNatsu 20152 Jonidan 83 East08-60.571 4-3 4-3 Incredible! Only one rikishi hit the wall. Everybody is disappointed with Kenshin for tainting an otherwise perfect basho for our Wall-beya. Shodai finally didn't drop his last two bouts and will be in Juryo next year. Hopefully he won't totally crash out, and he won his two exchange bouts this time, so he should be fine. Sato is finally meeting his match in the upper half of Makushita. He enjoyed nearly a year of lower competition, lets see if he can up his game. Inabajo won his last 3 to just get to kachikochi. Mitakeumi, as everyone knows, took the Juryo yusho in his first basho there. Since he lost to J1e Chiyotoori in the last day, he will probably show up in a J1 slot instead of Makuuchi, but really, who knows?!? Congrats! This is the first Juryo Yusho from Dewanoumi since 2000, by Kinkayama, who struggled with injuries. Iwasaki cruised easily to KK this basho, offing ex-Yokozuna Orora the last day. Omoto was in the running for Sandanme yusho and got exchanged to Jonidan to fight Ura, and lost his last bout. Kanamori only lost to Jonidan lifer Okuniyama somehow. It was hikiotoshi so maybe he got the henka treatment? We also welcome 8 newcomers to the Wall: Ura, now (in)famous for his unorthodox technique, has a perfect record only losing in a playoff. We get to find out whether his style can be effective when the age of his opponents catch up. Ura has a rival in Daiki, who has swapped yusho with him coming up so far. It wouldn't be surprising if Daiki has a bit more staying power than Ura however. Hirotsukasa brings a rare fusen loss with him to the Wall. In a college heavy recruiting class, Inoue is hanging tough at 15 years old!! Ura is 8 years his senior. Toshinofuji and Yajima both have the minimum winning percentage to stay on the wall, their stay may be short.

Anyone who spends more that 2 consecutive years in Makuuchi will probably hit komusubi at least once. Anyone who hits komusubi more that twice will probably make it to sekiwake at least once. The vast majority of these will then drop back down to mid to upper maegashira. I'm not so sure about consecutive years but it seems like those with more than 27~29 basho in makuuchi make it to sanyaku, and those with less are more likely to not make it. A lot of rikishi below this mark that did make it to sanyaku seemed to have been helped considerably by the 2011 scandal. The worst offender here is Jumonji, who managed to stay in Makuuchi from 2002.11 to 2006.9 consecutively while never ranking higher than M6, which seems like a skill within itself. The current top active guy is Toyohibiki who from 2011.01 until now stayed in Makuuchi without ever going higher than M2. I don't know who is the next guy after him. (it might be Kaisei)

Thanks for letting me know. I wanted basically a rating system that had some distribution so that I could utilize some optimization routines I have for a matrix of distributions from when I used to inverse ill-constrained matricies to back out gridded atmosphere temperature, wind speed, precipitation from satellite radar data. That's kind of the final goal here before I move on and try a different system for comparison. I don't really like tuning things too excessively to match big prior data, there seems to be a trap of a fundamental limit there. We'll see. For getting stuff out of the DB, I just do a bout search 1000 per page and then concatenate the results and have a program to read in and build a personal database. I really hope I'm not using up the server resources to do this though. I'm going to need to have some sort of shikona change master list to catch all the late changers though, which is a downside. Everything else seems OK though.

I'm not disagreeing with the point completely. To be honest I'm just glad that Hakuho is ranked highest and there is some separation between Ozeki sumo and the rest, so at least there aren't any major bugs going on. That said, if in the future as I flesh out the algorithm that different starting values return better predictive power I will change it. But I don't want to try to tune based on the tiny amount of prototype data I have now, then add data, and find out everything is off a bit again. As I said one way to initialize new rikishi would be to take some average of lower Juryo when they enter there. I would even consider iterating over the whole dataset multiple times getting a better initial estimate of individual rikishi each time. Right now the whole computation takes about 5 seconds on my laptop, so there's a lot of room. The current 1500 +/-350 for new players is directly from the algorithm's creator, Mark Glickman, and his experience in applying it to chess and US college sports. That is where I chose those numbers for now as a starting point.

Thanks for the comment. The 1500 initial rating is more reasonable for Hidenoumi than appears at first glance. The SDev is very high, 350, so the 95% confidence level of where his rating could be is 800-2200; This isn't something the algorithm produces for our consumption only either, it is reused internally to compute ratings after match results. For example, a 1000 rating rikishi with 30 SDev wins over a guy with a 1500 rating and 800 SDev. Because the second guy has such a large uncertainty, the winner's rank barely changes, and the loser will drop a few hundred points just from this match. In ELO the winner would increase then same as the loser decreases, over boosting the winner's rank in this case. I will add expected outcome functionality after I add in more data and Juryo matches. This algorithm is not really designed for match by match rating update. The uncertainty would remain too high probably. However, I will try shorting the window to 5 and 3 matches respectively later, to analyze the output. When used in chess, the ratings are calc'd only once per month, some people may play in 30+ matches in that time. For a match by match rating computation WHR (whole history rating) is a much much better choice, but more complicated to implement. http://www.remi-coulom.fr/WHR/

Glicko-2 Ratings before 2015 Nagoya basho Rating Sdev East Rank West Sdev Rating 197040HakuhoYHarumafuji401755 173033KakuryuY [/td] 173132KisenosatoOGoeido331623 162332KotoshogikuOTerunofuji381589 161436TochiozanSIchinojo471629 146733TakarafujiKMyogiryu321538 146348TochinoshinM1Sadanoumi421417 147534TakayasuM2Aoiyama311507 144634IkioiM3Kaisei331440 150931AminishikiM4Takekaze311473 144034OkinoumiM5Tokushoryu341347 129433KyokushuhoM6Gagamaru351330 137333TamawashiM7Toyonoshima321488 141331YoshikazeM8Osunaarashi351444 126633SadanofujiM9Homarefuji351222 133431KitataikiM10Amuru481266 127933TokitenkuM11Kyokutenho321381 142235EndoM12Kotoyuki331258 1500350HidenoumiM13Chiyotairyu351336 138529ToyohibikiM14Kagamio361199 1316232SeiroM15Satoyama351166 122836TakanoiwaM16[td] Where SDev is one standard deviation uncertainty Some retired notables after their last basho: Asashoryu: 1949 Rating +/- 37 Baruto: 1449 Rating +/- 36 Kotooshu: 1176 Rating +/- 71 (The last few basho were very unkind to him...) I've just done a very early implementation of a Glicko-2 rating algorithm on Makuuchi bouts. The Glicko-2 is a generalized version of Elo (according to Glick) and claims that with some additional simplifying assumptions the Elo algorithm can be retrieved. After the initial though its hard to convince myself that this is so much better than Elo... Like Elo a ratings period is chosen where all matches are assumed to take place simultaneously, to simplify the algorithm. For now, I am taking a whole basho as a ratings period, because true changes in a wrestler's ability can take place between basho, but its hard to feel that there will be major changes during a basho (other than illness or injury, but its unreasonable to expect any predictive power in that direction unless this is augmented with something like a monte carlo predictor over a whole basho...) Unlike Elo, Glicko-2 comes with a rating and a rating uncertainty. This is to take in account that we cannot perfectly know what rating a player is playing at, and that the player isn't playing at the same ability every day in the ratings period (has a headache, stubbed his toe, drank his ovaltine...) The deviation will shrink with consistent results, and grow with inconsistent results. Glicko-2 is the second iteration of the Glicko algorithm. It tries to improve over the original by added a volatility measure to each player. In the original, the deviation (Sdev) would naturally shrink to very small numbers and a rating would get 'locked-in' because the deviation estimation would get too small. The old system got around this by implementing an artificial floor of 35 Sdev. The new volatility measure lets the algorithm control this 'naturally'. This volatility can also change based on how consistent the player's results are. However, the calculation of this involves an iterative process and is not straight forward. So far in practice, I don't see the utility of this, as now it just seems like it is a complicated way of maintaining a 30-35 floor for Sdev. This might be due to not enough game records. Current data: All Makuuchi bouts from 2010.1 to 2015.5 (9,775 matches) New rikishi values: 1500 rating +/- 350 (0.006 volatility) (Glick's recommendation) Right now I don't have a expected evaluation function so I can't check on predictive accuracy of ratings for historical bouts. They exist in literature but I haven't gotten that far yet. None of the parameters are tuned because its more important to add more history (maybe from 1990?) and Juryo matches before touching anything. So, I'll give some casual qualitative observations Takarafuji is overranked compared to the rating of other sanyaku, which has played out as expected this basho. Ichinojo has a higher ranking than Terunofuji (although their uncertainty occupies some of the same rating area) and I seriously doubt many people would have selected Ichinojo to get a better record than Terunofuji this basho. This discrepancy is from the fact that the initial value of a new wrestler entering makuuchi is (1500 +/-350), and the SDev decreases rapidly even if match outcomes are pretty inconsistent. When Ichinojo entered, he beat a bunch of high rankers very quickly, while Terunofuji had a more steady march up the banzuke. I'm confident this is a flaw with the current implementation and the following steps will be done one by one to try and reduce the effect: 1) Include Juryo matches to try and regularize the rating before promotion to makuuchi (and not take the Juryo rankings at face value) 2) If that doesn't work, change the initial value of new wrestlers to something more adaptive, i.e. the mean of the bottom 10 Juryo wrestlers with a high standard deviation 3) Increase initial volatility: the current choice seems to not change hardly at all over basho evaluations (I also think that while the system assumes the Sdev is a normal distribution, my gut feeling is that the mid and lower half of the possible rating deviations for one player is more probable than higher ones. It would probably be a lot of work to change the implementations to have a more complicated assumed distribution) Endo is underranked compared to his rating, but anyone would consider a M12 healthy Endo to be underranked, so there isn't a lot of valuable insight here. My current to do list for this is: - Put together a longer history (I think from 1990 would be sufficient for current day ratings, my interest in rating bygone wrestlers is not so much right now...) - Add in all the Juryo matches. This puts a greater distance between sanyaku and the entry of new wrestlers. Also, there is a fair amount of exchange matches each basho. (Wakatenro, a yaochao guy who had a total of 3 exchange matches ended up with a 440 rating which is funny) - Implement SDev drift for completely missing a basho. Right now nothing happens during a completely kyujo basho. This is specified by Glick but I'm really goddamn lazy. - After those 3 are done start tuning things a little bit - Implement an expected outcome function to do past predictions and measure the effect of tuning. ~Dream list~ Implement whole history rating (Remi Coulom of 囲碁 (Go) AI player fame) and compare against Glicko-2 Anyway I'll keep dumping updates here, I welcome comments or silence, either OK!

If he lost 20kg then that would be explanation enough. After you get used to Japanese food its easy to forget that someone who is getting new exposure to the diet might find it hard to keep weight during intense training. When I first started taking business trips there I found it hard to choke some meals down, but now I could easily see myself gaining 5kg a month by going to a rotary sushi every other night... Here's to him putting some weight back on and as muscle too :)

Comparing his tachi-ai engagement with his performance in the US, it looks like he might be scared of hurting the thinner guys or something right now. Its very timid compared.

Nagoya 2015 Over the Half Edition Shikona Age Country Heya Debut KK Current Rank Yushos Record Win% Last Basho This Basho Shodai23TokitsukazeNatsu 20148 Makushita 1 East243-110.796 4-3 4-1 Sato18TakanohanaKyushu 20144 Makushita 27 West228-40.875 6-1 3-1 Ohata18TokitsukazeHatsu 20153 Sandanme 69 East018-80.692 4-3 3-2 Inabajo18SakaigawaHatsu 20153 Sandanme 99 East016-100.615 5-2 2-3 Mitakeumi22DewanoumiHaru 20153 Juryo 12 West020-30.870 6-1 8-1 Iwasaki23OitekazeHaru 20153 Sandanme 66 West016-20.889 6-1 4-0 Omoto22IrumagawaHaru 20153 Sandanme 90 West016-30.842 6-1 5-0 Kotonoawa18SadogatakeHaru 20152 Sandanme 92 West013-50.722 6-1 2-2 Kanamori18IrumagawaHaru 20153 Jonidan 11 East014-50.737 4-3 4-1 Kenshin19SakaigawaHaru 20152 Jonidan 29 West012-70.632 5-2 3-2 There haven't been this many guys getting KK at the halfway point in a loooooong time! Shodai comfortably gets his KK and a promotion to Juryo! He beat two Juryo guys including his rival Takaryu (ex Takagi), so its well deserved. So far he'll join Daishomaru in going up. We'll have two sekitori next time because Mitakeumi has already gotten KK with 8 straight wins, he recently lost to Kitaharima and is now tied with him for the yusho, 2 ahead of the other pursuers. Omoto and iwasaki are both undefeated in Sandanme along with a bunch of other rikishi. For next time we'll already be adding Hirotsukasa, Ura, Daiki, Saionji, and Asayamabana

Day 7 - Kaisei (satoyama)

Hi Sumozumo, Thank you for asking this question. First, I want to say that I am not a Computer guy. I am a Chemist by education and profession. As for computers - I use them, but I am not a Programmer. :-( I reached to Doitsuyama, the Guru of Computation, who takes care of most of the Superbanzuke Games and Golynohana - another Big Expert in Game automation, for help. Both turned me down for different reasons, but the main one was that they are very busy with taking care of many of the games, which are already automated. And I totally understand them. How I maintain the game - manually. After the official Torikumi is posted I calculate all the points manually and then write them down on Sumoforum. This is the reason sometimes I make human errors. After the bouts are over a friend of mine is helping me with the Tables and on the next day I post the Tables - going from the XLS file to PDF file, to PNG picture and using Photobucket to paste the Tables into the Sumoforum. It's a lot of work, but I am willing to do it manually as long as there are Sumo fans and gamers who like the game and are willing to participate. I started the game in January at 2015 Hatsu basho and I am glad that already 4th basho there is constant interest in TUG. I was gonna ask one more time for help, I am sure there are Younger (no offence :-) ) computer guys, who might be willing to ebrace this project and make the automation a reality. The big Computer Veteran experts - might be willing to share some thoughts and what eventually could help the automation of the Game. Once automated, this game might find more Supporters. Regards, Achiyama I don't really do website automation, which is the golden egg here, but I can throw together some scripts to run on your computer to generate everything. You'd still have to enter people's guesses manually somewhere, but it'd be better.

Day 6 - Tamawashi (osunaarashi)

I have a feeling that a broken shoulder affects sumo way more than Ramadan.

Hey Achiyama, I don't know how you do your calculations, but would a script that just prints out the underdog value in a big matrix of every possible matchup be helpful?　What part is giving you problems re: automation?

Day 7 Kakuryu

Day 5 Tochinoshin (Ichinojo)

Day 6 hakuho Whoops, don't mind the fusen loss but wasn't trying to cheat. That'll teach me to work so late.

Day 4 Tochiozan (Tochinoshin)

Day 5 Kisenosato