Comparing performances amongst different ranks and bashos

[Toukeigakusha 0]

As there are so many wonderful discussions and statistics here, I would like to share an idea and get comments on its weaknesses and potential improvements. This posting will be a bit long, and I warn those who aren't that interested in numbers they likely won't find that much of value here.

The way an ozumo basho works, the winner is decided by the number of wins, regardless of their opponents. In reality this doesn't turn out to be a problem as e. g. genki maegashiras (like currently Yoshikaze) going into double-digit territory have to face each other plus sanyaku or even ozeki opponents during the final days and are thus not likely to actually win a tournament. Still, a weaker sekitori may win a sansho or even jun-yusho without beating more than one really tough opponents while the sanyaku guys may have to beat much stronger opponents to even get their kachikoshi.

So I'm wondering whether the performance of different sekitori within a basho (or between different bashos for the same sekitori who owns different banzuke ranks there, in order to observe their development) can be more reasonably compared than by simply counting wins. Doing so would require to set up a model of what "performance" means, which might be anything but universally valid of course.

The model that I'm messing around with does have its flaws and it seems to work best for a certain part of the banzuke that I'm interested in most, being the mid maegashira to ozeki range. It probably undervalues the lower maegashiras for the sake of simplicity and based on the observation that outstanding performances in this area appear to be rare anyway. It also completely ignores the difference between kachikoshi and makekoshi by treating wins only based on opponents' quality, as well as traffic jam situations or sure demotions. The idea behinds this would be that sekitori are supposed to do "motivated sumo" in all situations, so I have no problem with 'overvaluing' those who actually do just that.

What am I trying to do? I want to value a win over an opponent based on that opponent. What's relevant for a sekitori's value? My model suggests: Rank multiplied by genkiness. Rank is being expressed by banzuke rank, genkiness is expressed by the number of wins of a sekitori in the current basho. (I pondered making the value somehow recursive, i. e. considering that sekitori's opponents as well but that would make the whole procedure a lot more complex while not changing that much as the opponents basically depend on the rank anyway and the genkiness will usually vary a lot more than good or bad luck of opponents selection.)

Simply counting wins of a sekitori for his genkiness means that a sekitori with 12 wins would be four times as genki as a sekitori with 3 wins which might then be seen as too much of a difference. I wonder whether to add a constant amount of 1-3 victories for each sekitori as to make the differences smaller; this would downplay the genkiness factor relatively to the ranking factor. Then again, 3 as well as 12 wins are rare events, and the factor of two between 5 as 10 wins isn't that different from the factor of two between the strongest and weakest opponent's rank for high maegashira (see below). For ozekis (who face only the top sekitori anyway) the genkiness of their opponents is a bit more important.

My current model also makes a win against a sekitori who retires early in a basho of very little value, sometimes even worthless (Chiyotaikai O1E/0 in Hatsu). This may occasionally be unfair if said sekitori got an injury during the basho, a win prior to this injury against the still healthy sekitori would then become undervalued. But not knowing details about injuries in general and trying to keep my model simple for the time being I have to live with this effect. During 2008 this effect was most relevant for wins over Asashoryu who dropped out of competition with 3 resp. 5 wins in two bashos, and in both cases a kinboshi would be valued no better than a victory against a genki middle-class maegashira.

Which brings up the value of rank. Keeping things simple I value the Y1E with 50 points and then each successive rank 1 point lower (regardless of participation in a basho, I don't upgrade lower ranks because of this as the sekitori don't just get stronger because some higher-ranked opponents are missing). So the rank with lowest non-negative value would be J4W with 1 point.

The number "50" may look arbitrary but happens to be working for me as my scope of interest is Makuuchi only for the time being, and lower ranked Makuuchi sekitori have to fight against upper Juryo opponents every now and then; J4W turned out to be sufficient for 2008 although I don't know whether, say, Shotenro could be assigned a Makuuchi opponent had he gone 13:0 this basho. I could easily add a few points to each rank, such as making the Y1E worth 55 or 60 points; this would slightly upgrade the relatively value of lower ranked maegashiras to the rest while at the same time slightly downgrading the relatively value of ozekis and yokozunas but actually not change that much about the order of total scores within a tournament. (But rating a yokozuna as, say, 100 points would result in a M6E being worth 80% of this, and that doesn't sound reasonable to me, so no simple coverage of Juryo with the same method, the values would then have to be non-linear somehow.)

At first glance it looks strange to value the Y1E with 50 points and the ozekis with 45-48 points despite the relatively weak ozeki performance during recent years. But given the genkiness element and observing that ozeki struggle with their kachikoshi while healthy yokozuna are always in double-digit wins territory their value will surely not even be close (unless a yokozuna gets injured during a basho).

Given these ideas, what would be the actual value numbers? Lower maegashira will be worth around 100 points, be that a genki M16E with 10 wins (10*10=100), an okay M14E with 7 wins (14*7=98) or a weaker M11E with 5 wins (20*5=100). To be worth 200 points one might be a M12-E with 11 wins (18*11=198, a sansho candidate like Yoshikaze in Kyushu) up to an overpromoted M1W with 5 wins (39*5=200, Futeno in Kyushu). Genki meat grinder maegashiras occasionally are worth 300-400 points (Kokkai M5W/12 in March was 372, Kisenosato M1E/10 in January topped the scale with a full 400) and get into kachikoshi ozeki territory already. The highest possible value one can get with a victory is beating the Y1E/14, that would be a score of 50*14=700 points. Beating Asashoryu/3 in Nagoya was worth a mere 150 points for Toyonoshima K1W/10 (who finished this basho in fourth place score-wise nonetheless), beating Asashoryu/5 in Aki was worth 245 points (compare that to beating Futeno M10W/11 for 231 points).

Doing the calculation (Excel spreadsheet, manual input) I soon saw that beating strong opponents while losing to weaker opponents is much better than doing the opposite, and in fact there are certain specialists (Aminishiki, Kisenosato) who might be overvalued by my model. Then again, a loss is a loss and doesn't get you anywhere while wins against strong opponents give you reputation, kensho and possibly even a sansho or a kinboshi. And those maegashira who do outstanding performance of this kind are exactly those who I want to compare to the current ozekis, as one of the ideas behind my tables is: Which meat grinder (or even lower) competitor shows a performance that would be equivalent to making an ozeki kachikoshi? It might be difficult to become an ozeki given the requirements but it's interesting to see where the ozeki potential might already be (and how many consecutive basho a sekitori already did that well), and considering the opponents' value might be a bit more helpful than simply counting wins.

One observation I made is that a bad start into a tournament is something a sekitori can't really fix score-wise in the course of the basho because he will get opponents of lower value in later matches (be that lower rank or lower genkiness, or both). An 8:7 after a 0:5 start is usually a lot less worth than an 8:7 after a 5:0 given the different sets of opponents you will likely get. It's also a big difference for double-digit winning maegashiras whether they still win when facing really strong opponents as these can be worth three or four times as much as the opponents of their first matches during that same basho.

Given my current numbers, the strongest performance in 2008 of a maegashira below the meat grinder was Baruto's M7E/12 in Haru (2633 points) because he won his last two matches against Chiyotaikai O2W/8 (360) and Miyabiyama M2W/7 (259) while his most valuable prey was actually Kokkai M5W/12 (372). Kokkai himself, while being ranked higher in the banzuke, scored slightly lower than Baruto (2593 points) because he got weaker opponents: Kokkai won all of his last six matches after a 6:3 start, and his only 'big prey' was Kaio/8 on day 15. Said Baruto was worth 336 points as opponent in this tournament (i. e slightly less than Chiyotaikai but at least in the same range), and those three sekitori who managed to beat him (Hokutoriki M7W/8, Kyokutenho M4W/9 and Asasekiryu M1E/8) deserved these points (all being genki opponents themselves). Then again, even if Baruto had won all 15 matches against these opponents (yes, I know he would rather have had both yokozunas as final opponents in this case...) he would still have gotten a slightly lower score than Hakuho/12 in this basho and thus finished "virtual third" because only one of his first nine wins was worth more than 200 points.

So what would the scores of a full tournament be like? Genki maegashira score about 1000 points (Kyushu: Chiyohakuho M16W/9 who defeated Yoshikaze M12E/11 and Kokkai M11E/9), meat grinder sekitori with kachikoshi score about 2000 points (Kyushu: Kotoshogiku M3E/9 who defeated Baruto S1E/9, Kisenosato M4E/11 and Toyonoshima M1E/9), ozekis with a mere kachikoshi usually score better than 2000 but this depends on their opponents mix - there have been 8 wins ozeki performances from 1683 points (Kotooshu/8 in Kyushu beating opponents with an average of 5.875 wins, including just one ozeki) to 2639 points (Kotomitsuki/8 in Haru beating opponents with an average of 8.25 wins, including Asashoryu/13). Hakuho's 15:0 in Nagoya made the year high score of 4499 points, just slightly more than the 4401 points for his 14:1 in Aki (due to very genki competitors there: Kotomitsuki/11, Ama/12, all ozekis made kachikoshi there whereas the meat grinder was a disaster).

As for Kyushu 2008, Ama scored the best second place of the year with 4012 points (better than Asahoryu's 13 wins yusho in Haru), and Hakuho would only be "virtual second" here unless he gets the points for his kettei-sen additional win. Yoshikaze M12E/11 (who had just one valuable win in this basho, over Kyokutenho M6W/10 for 290 points, the rest of his wins produce an average of a mere 88 points) has about the same score for this basho as Hokutoriki M3W/5 (who at least beat Baruto S1W/9 on Day 1 for 387 points)... that's how unfriendly the model is to lower maegashira ranks. One the other hand, Kisenosato M4E/11 showed the third best performance in this basho behind the two kettei-sen participants (way better than the scores of Aminishiki, Baruto and Toyonoshima in positions 4-6), and even the best maegashira performance of all 2008 with his 3283 points (after all, beating three ozeki and both sekiwake all of which got kachikoshi and an average of 9.4 wins this basho).

And what about a longer time frame? One might simply add up the scores of a year (to make it a similar list as in tennis). Then again, the most recent results might be considered more relevant for the current situation than the results five bashos ago. My current formula attempts a compromise, assigning the weights 5,6,7,8,9 and 10 to the last six bashos, thus making the most recent basho of the year twice as relevant as the oldest one. Compared to giving all tournaments the same weight (these are configurable in my spreadsheet), the difference in ranking order is surprisingly small, with the one notable exception of Asashoryu whose three strong bashos were all in the first half of the year).

Of course the "virtual ranks" of sekitori who missed one or even more bashos can't be taken that seriously (no one would reasonably assume Asashoryu to be a maegashira in terms of quality, or would you? If all bashos got the same weight then Asashoryu would still keep a "virtual ozeki" position as he made 45 wins after all and had strong opponents in every basho).

But the resulting list gives some general statements that can't be too far away from the truth: Hakuho reigns supreme in 2008, with Ama being in undisputed second place, significantly better than all ozekis. Of these, Kotomitsuki is in third place and Kotooshu (who made up for his mere 2 wins performance in Haru by his 14:1 yusho in Natsu) in fifth, both staying in ozeki territory on average. Kaio and Asashoryu can't compensate for their missing bashos and would even fall out of sanyaku range (to M1E and M1W. The two new "virtual ozekis" would be Kisenosato (almost on par with Kotomitsuki) and Aminishiki (with a mere 43 wins this year but lots of big shots amongst these), with Toyonoshima and Baruto in the "virtual sekiwake" slots while Kotoshogiku and Chiyotaikai (with 5 bashos only) would be the "virtual komosubis". And despite his mere 5 wins in Hatsu, Chiyotaikai will very likely jump back into "virtual ozeki" range with an average next basho already as he scored 0 points in Hatsu 2008.

[Doitsuyama 1,178]

Nice idea. I also have an approach which is totally different (documented somewhere here I think). On first glance I can't decide if your approach is worth the many words, but it's easy to decide for yourself. Just realize those algorithms for past basho and see how many bouts would have been predicted correctly going day by day forward. If it's more than 60% then your method sure is worth deeper thoughts (in fact something like 58% would be good too).

Edited November 24, 2008 by Doitsuyama

[Sasanishiki 57]

I like this system. I am not statistically minded (in terms of calculating them) and so I was a little wary, but your explanation was clear even to me. My one comment would be that I wouldn't rank the rikishi for their accomplishments over the year in banzuke form (that is, 'virtual ozeki', etc) because that puts them in bands of pairs (and how would this banding be reconciled with a real banzuke that has more than two ozeki like we have now), meaning that 1st and 2nd are seen to be comparably close, and 3rd & 4th the same. Might your statistics not show a larger gap within these pairs over a year? A virtual banzuke also muddies the comparison, by which I mean people will start to say XYZ is an ozeki but only performed as a 'virtual komusubi'. I think it wold be easier to comprehend if we say XYZ is an ozeki and he had the 9th best annual performance over 2008.

[Toukeigakusha 0]

@Doitsuyama: I see why such a test would be meaningful. Then again, my method doesn't work this way, as it isn't meant to predict results but to make given results more transparent.

My problem is that the value of an opponent grows throughout a basho, as his genkiness depends on the sum of his wins of the whole basho (not just of the basho up to the match, or else all matches of Day 1 would be worthless), i. e. including future matches of the opponents you already defeated.

Which means that after Day 10, not only do I know a mere 10/15 of the information about all wins for each sekitori this tournament, I also know a mere 10/15 of information of all opponent genkiness values, meaning that I know a mere 100/225 = 44% of information about the strength of each sekitori after 66% of the basho - the information growth during a basho isn't linear. And obviously after Day 1 all sekitori would have a strength value of zero because they either lost their first match or defeated an opponent with genkiness 0....

So if I were to predict individual matches I would need a much more complex approach, then incrementally changing the strength formula from day to day, e. g. initializing with a value based on, say, a genkiness of 7.5 (or rather a genkiness that I'd calculate from this sekitori's previous results, adapting his scores from previous bashos to his new banzuke rank - see below) and then slowly increasing the weight of the (still mostly meaningless, see above) value during the running basho from 0% to 100%. But starting with 7.5 wins for everyone would mean to predict the matches during the first half of the basho mostly on banzuke rank alone, which doesn't look that innovative...

How could I reasonably predict the number of wins at the start of a basho? Perhaps by assuming the sekitori might get about the same score as in his previous basho, as the score is meant to express his capability. How would that work?

A M12E like Yoshikaze made relatively constant scores of around 700 points throughout the year 2008 (for details see below), only now to win 11 matches and get a surprising score of 1169 (which might be an overperformance or a jump in quality, nobody knows for sure). Let's assume he will become M03W and then predict that he will still be worth more than 1000 points in the next basho (something like a weighted average over the last year) - how many wins would that equal?

• In Hatsu Tochinonada M03W/5 got 1052 points,
  
• in Haru Toyonoshima M03W/6 got 1198 points,
  
• in Natsu Kakuryu M03W/5 got 1090 points,
  
• in Nagoya Tochinonada M03W/7 got 2008 points (beating Kotomitsuki O1W/11=517 and Chiyotaikai O2W/9=405 plus five more opponents for 1086 points),
  
• in Aki Wakanoho M03W/0 didn't participate,
  
• in Kyushu Hokutoriki M03W/5 got 1149 points.
  

So for January 2009 I might initialize Yoshikaze with a predicted genkiness of 5 wins from M03W and start from there. I could place him on M03E as well but none of the six performers this year were even close to 1200 points, most of them were much stronger, so my comparison would be less accurate (I can't just take a sekitori with 2000 points and divide his number of wins by 2 because this one got much stronger opponents during the basho, as you see in the case of Tochinonada in Nagoya) while I'd still guess something like 5 wins for Yoshikaze.

Actually that's quite optimistic as my 2008 score table with basho weights from 5 to 10 would give Yoshikaze an average of 798 points for 2008. Did we have that already for a M03? Yes, in Hatsu Goeido M03E/5 got 866 points, in Nagoya Futeno M03E/3 got 873 points (beating Kotomitsuki O1W/11). So assuming Kyusho was an overperformance based on the whole year's data, one might rather initialize Yoshikaze with a predicted genkiness of a mere 4 wins.

Nevertheless, your test idea is interesting. I'll do the test for Days 10-15 for the current basho and post the results here.

===========

@Sasanishiki: Yes, I'm aware that this "virtual banzuke" wording might cause confusion. It's just that my original intention was to find out who might have shown "ozeki performance" in 2008 thus I ended up "replacing" some of the ozekis by other sekitori whereas a simple numerical ranking list might be better comprehensible (as I ignore the ozeki qualification procedure anyway).

Then again, if I want to see whose potential shown in 2008 is greater than his current banzuke position I'd have to convert both into a common ordering system anyway, and making a "virtual banzuke" was just one of the two possible ways of doing this.

The numbers for 2008 (top 30 positions):

Pos.Score Wins Banzuke Sekitori	Hatsu[5]   Haru[6]  Natsu[7] Nagoya[8]   Aki[9] Kyushu[10]
---------------------------------------------------------------------------------------------
_1. 3981   79  Y1-E   Hakuho		4221/14   3446/12   2770/11   4499/15   4401/14   4238/13+1
_2. 3098   61  S1-E   Ama		   2692/ 9   2139/ 8   2456/ 9   2971/10   3560/12   4012/13
---------------------------------------------------------------------------------------------
_3. 2429   55  O1-E   Kotomitsuki   1803/ 8   2639/ 8   2218/ 8   3053/11   2843/11   1890/ 9
_4. 2379   51  M04-E  Kisenosato	2642/10   2034/ 8   3137/10   1692/ 6   1481/ 6   3283/11
_5. 2121   50  O2-W   Kotooshu	  2567/ 9	382/ 2   4115/14   2003/ 9   2075/ 8   1683/ 8
_6. 2048   43  K1-W   Aminishiki	 998/ 5   2018/ 6   2715/10   1385/ 6   2056/ 8   2647/ 8
---------------------------------------------------------------------------------------------
_7. 1972   51  S1-W   Baruto		1196/ 7   2633/12   1153/ 5   2158/10   1970/ 8   2389/ 9
_8. 1958   48  M01-E  Toyonoshima   1305/ 6   1198/ 6   2381/11   2615/10   1479/ 6   2348/ 9
---------------------------------------------------------------------------------------------
_9. 1883   46  M03-E  Kotoshogiku   2470/ 9   2410/ 8   1782/ 8   1528/ 6   1418/ 6   2048/ 9
10. 1865   39  O2-E   Chiyotaikai	  0/ 0   2476/ 8   1533/ 5   2315/ 9   2270/ 9   1937/ 8
---------------------------------------------------------------------------------------------
11. 1785   43  O1-W   Kaio		  1771/ 8   2362/ 8   2306/ 8   2055/ 9   2512/ 9	210/ 1
12. 1726   45  Y1-W   Asashoryu	 3571/13   3891/13   3021/11	544/ 3   1218/ 5	  0/ 0
13. 1643   49  M06-W  Kyokutenho	1566/10   2098/ 9   1126/ 4   1523/10   1374/ 6   2109/10
14. 1620   43  K1-E   Goeido		 866/ 5   1354/ 8   1413/ 8   1822/ 7   2499/10   1350/ 5
15. 1578   43  M07-E  Miyabiyama	1565/ 7   1813/ 7   1333/ 6   2076/ 9   1011/ 4   1728/10
16. 1564   41  M05-E  Asasekiryu	2100/10   1917/ 8   1622/ 6   1956/ 8   1425/ 4	853/ 5
17. 1326   39  M09-W  Tokitenku	 1336/ 6   1896/ 7   1438/ 6   1447/ 7   1004/ 6   1093/ 7
18. 1299   42  M06-E  Kakuryu	   1850/11   1333/ 6   1090/ 5   1286/ 8   1629/ 7	864/ 5
19. 1265   42  M02-W  Wakanosato	1264/ 7	951/ 5   1382/10   1041/ 5   1282/ 9   1537/ 6
20. 1263   46  M08-E  Takekaze	  2404/12	814/ 3	790/ 6   1075/ 7   1152/ 9   1545/ 9
---------------------------------------------------------------------------------------------
21. 1197   42  M03-W  Hokutoriki	1222/ 7   1596/ 8   1825/10	654/ 4	964/ 8   1149/ 5
22. 1173   46  M01-W  Futeno		 922/ 8   1270/10   1577/ 9	873/ 3   1424/11	970/ 5
23. 1150   40  M07-W  Tochinonada   1052/ 5   1144/ 8   1233/ 9   2008/ 7   1093/ 6	510/ 5
24. 1122   46  M11-E  Kokkai		1546/ 9   2593/12	641/ 3	584/ 5	657/ 8   1211/ 9
25.  992   47  M14-E  Takamisakari   882/ 8   1266/10   1085/ 7	752/ 6	591/ 6   1369/10
26.  979   45  M09-E  Tochiozan	  829/ 8   1386/11	705/ 5   1214/ 9   1174/ 6	637/ 6
27.  872   38  M05-W  Dejima		 395/ 3	871/ 6	994/ 8	751/ 6   1170/ 9	856/ 6
28.  846   42  M08-W  Kakizoe		739/ 6	602/ 8	758/ 6	771/ 7   1175/10	873/ 5
29.  798   37  M15-W  Homasho		607/ 4	733/ 9   1398/ 9   1645/ 9	  0/ 0	554/ 6
30.  798   47  M12-E  Yoshikaze	  746/ 8	624/ 6	686/ 7	648/ 8	752/ 7   1169/11

"Scores" here are weighted averages, weights given next to basho name. All data input manually, so errors are possible.

Edited November 24, 2008 by Toukeigakusha

[Toukeigakusha 0]

As I thought, my mechanism (as it is right now, i. e. without incrementally 'shifting' the genkiness discussed in the previous posting) is unusable for predicting the outcome of matches during the first half of a basho.

That's mostly because the "genkiness * rank value" formula is balanced as to value a genki middle maegashira as high as a weak high maegashira / low sanyaku sekitori, by having a factor of two variation in both dimensions. But early in the basho the number of wins between two opponents often differs by more than a factor of two (most notably if one of the opponents has none at all), which means that the banzuke rank will almost be ignored and the number of wins is too relevant.

There are other effects, such as the yokozunas working through their opponents list from low to high, i. e. getting the strong opponents only late in the basho while the meat grinder maegashiras get the big shots early (whose value as opponents does not depend on their opponents, only on their number of wins which usually will be high early in the basho already) and if they manage to beat one of them they look strong for quite a while even when they don't keep up this performance whereas the yokozunas won't have the highest scores until they have met and beaten their strongest opponents. (That's exactly what the audience wants: Suspense until the end of the basho by giving the yokozunas the easy wins early and the tough matches later. But at the same time this makes predictions about the true strength of the highest ranked sekitori more difficult until they can really show what they're capable of.)

To show you the effect, after day 7 the score ranking list was (21.8% of the basho information):

Score Wins Banzuke Sekitori
===========================
_ 815   4   K1W	Aminishiki  (defeated Hakuho, Chiyotaikai, Kotooshu, Kotoshogiku)
_ 790   5   M01E   Toyonoshima (defeated all 4 ozekis and Dejima/6)
_ 701   6   M04E   Kisenosato  (defeated Ama, Kotomitsuki, Kotooshu)
_ 595   5   S1W	Baruto	  (defeated Aminishiki, Wakanosato, Toyonoshima)
_ 579   4   M02W   Wakanosato  (defeated 3 ozekis and Aminishiki)
_ 533   6   Y1E	Hakuho	  (defeated Toyonoshima)
_ 462   7   M07E   Miyabiyama
_ 387   6   M05W   Dejima
_ 346   6   M11E   Kokkai
_ 333   5   O2E	Chiyotaikai (defeated Wakanosato, Kotoshogiku)

That's what you get for beating yokozunas and ozekis early in the basho... Ama wasn't even in the top 10 with his 5 wins, and several of the top sekitori already had seemingly worthless wins over Futeno/0, Goeido/1, Hokutoriki/1 or Tochinoshin/1 at this time. If I had the torikumi for the complete basho beforehand I could compensate this effect... unfortunately a basho doesn't work this way.

Despite all this, I tried to check the prediction rates for Day 8-15 of Kyusho basho. As I can't compute scores for Juryo opponents I can't "predict" any matches with Juryo participants, only matches within Makuuchi:

Day Information  Predictions>0%  Predictions>5%  Predictions>10%  Predictions>20%
================================================================================
08	21.8%	   7:12 (36.8%)	 4:11 (26.7%)	 4:11 (26.7%)	 4: 9 (30.8%)
09	28.4%	  13: 6 (68.4%)	11: 6 (64.7%)	10: 5 (66.7%)	 8: 5 (61.5%)
10	36.0%	  14: 5 (73.7%)	13: 5 (72.2%)	13: 5 (72.2%)	10: 4 (71.4%)
11	44.4%	   9: 9 (50.0%)	 9: 8 (52.9%)	 8: 7 (53.3%)	 5: 7 (41.7%)
12	53.8%	   7:10 (41.2%)	 6:10 (37.5%)	 6: 9 (40.0%)	 4: 2!(66.7%)
13	64.0%	  12: 5 (70.6%)	10: 5 (66.7%)	 9: 4 (69.2%)	 9: 2 (81.8%)
14	75.1%	   9: 8 (52.9%)	 9: 8 (52.9%)	 9: 7 (56.3%)	 6: 5 (54.5%)
15	87.1%	  10: 7 (58.8%)	 8: 6 (57.1%)	 8: 5 (61.5%)	 6: 2 (75.0%)
================================================================================
Total			81:62 (56.6%)	70:59 (54.3%)	67:53 (55.8%)	52:36 (59.1%)

What I did was to compute scores based on data prior to the whole basho day (not yet taking into account results of previous matches on the current day, that wouldn't be practically usable for a betting game anyway I guess). If the difference between both scores was smaller than 52.5:47.5 (5%), 55:45 (10%) or 60:40 (20%) I eliminated the bet from the corresponding columns and treated it as "too close to call" as these scores aren't more informative than a coin flip (the closest one I got was for Bushuyama M10E/5 [565] over Chiyohakuho M16W/7 [563] on Day 12). But this even reduced the percentage as the coin flips for the 14 closest calls went 11:3 in my favor... and ignoring the 10-20% range eliminates too many bets while in fact improving the rate quite a bit.

"Information" is "((Day-1) /15)

[Doitsuyama 1,178]

Ok, I have looked a bit into your method and it can't convince me, sorry. I think it is weaker than my method, and it isn't close, so I won't lose many words about it.

[Gusoyama 99]

Would gathering data about the rikishi's past performance at or around a rank provide valid information about their genkiness at the beginning of a basho? I honestly don't know, I am just wondering. Also, if its possible, you could get their genkiness by checking their stats in past bashos against the 15 "likely" opponents they will face. Thats a lot harder, I would imagine.

[Toukeigakusha 0]

Yes, it certainly would, and if it were my goal to predict matches I would do exactly this (I do have data for six bashos, no problem to start the next basho with the weighted average of the year 2008 for each sekitori - still not an ELO number but a lot better than nothing of course).

Given the quadratic growth of information throughout a basho it would be plausible to use the current basho performance only with that weight and fill up to 100% by the weighted score average from some number of previous bashos (the six that I'm currently using is just for the purpose of covering a calender year, any other number with appropriate weights might do as well). This would mean to use the ranking list from before the basho with a weight of 100% for Day 1, with a weight of 99.555% on Day 2, with a weight of 98.222% on Day 3, with a weight of 96% on Day 4 and so on. By doing so the strength changes would flow more slowly into the process and only get a higher weight when enough information about the current basho is available.

Getting "likely opponents" would be difficult for once, and of little use as well as predicting individual matches isn't just the strong element of this mechanism in the first place (and making it recursively dependent on its own weakness doesn't look like solving the problem). Compared to this it would be easier to guess the likely genkiness of an opponent by looking how many wins other sekitori of the roughly same ranking list score (at that time) made when they had a basho from the same banzuke position. That's what I tried to show with the Yoshikaze example - guessing 5 wins for him in order to value a win over Yoshikaze with a reasonable genkiness for the time being, i. e. until I have better information about how many wins he actually made. Given a database of, say, 5-10 years there should be enough examples for each banzuke rank to find at least 1-2 sekitori with similar data to any current situation.

Yet, it wouldn't prevent the effect of yokozunas being slightly undervalued in the middle of a basho and meat grinder sekitori being overvalued due to their specific order of torikumi; to handle this one might have to develop separate formulas for each banzuke rank for the merging process of old and new information. But even without this I would expect to get much better results that by using a mere 20% of information in the middle of a basho as my previous posting did.

I don't have any reasonable data base (still toying around with my spreadsheet for the time being) but I might actually try coding this thing just for fun... then again, that wasn't my goal when I started this thread. I wasn't trying to show a tool for playing sumo prediction games. If I were to play any such game then using strength values alone would certainly not be enough, I wouldn't want to ignore the fact that certain sekitori tend to "own" certain others, the fact that a sekitori with 7 wins is much more motivated than his opponent with 8 wins and other things like this (most of which can be done more reasonably with a huge historical database).

I'm more interested in getting a kind of "virtual banzuke" that doesn't show the oddities of the existing sanyaku mechanism but is based on actual performance of all sekitori. Ama, for example, was definitely the second best sekitori of 2008 with his 61 wins over opponents on the same level as Hakuho's (okay, 2% weaker to be exact), and my score expresses this very clearly. It would, among other things, punish those who show "unmotivated sumo" after having their kachikoshi relatively to those who are going all out for a sansho/yusho.

What I rather hoped for was a discussion about how sekitori like Aminishiki ought to be rated, a guy who performs excellently against sanyaku opponents and at the same time poorly against maegashiras. My formula (ignoring his losses) would give him a much higher score than if he did the exact opposite to come up with the same number of wins, and the idea behind this is that he is going after what the sumo audience apparently wants him to - sanshos and kinboshis (as he can apparently get them while doing an ozeki run seems out of the question as of now). Aminishiki had 43 wins in 2008, the same as Kaio; even if I weight all six bashos equally (as to not punish Kaio for having gone kyujo in the most recent basho) Aminishiki gets a higher average score (1970 : 1869) than Kaio as he beat the stronger opponents this year. Even Asashoryu with his 45 wins scores just slightly better with an average of 2041 points, and this would have been a mere 1950 points had he made two wins less, so the opponents whom Aminishiki defeated were actually stronger on average than those Asashoryu defeated! (Bad luck for Asashoryu that he left two bashos early where he had only rather weak opponents until then, this downgraded his opponents mix.)

In terms of average score for the opponents they defeated, both Kisenosato and Amishishiki beat stronger opponents on average in 2008 than each of the four ozekis (only Hakuho and Ama defeated even stronger opponents this year). Baruto, for example, defeated much weaker opponents on average which puts his 51 wins a bit into perspective; my formula gives Baruto a score that's 96,3% of Aminishiki's.

Kisenosato/51 gets almost the same average score by my model as Kotomitsuki/55. That's plausible: Early during a basho Kisenosato gets sanyaku opponents, late in the basho he gets genki maegashiras as he's genki himself whereas Kotomitsuki gets maegashiras of random genkiness early in the basho and the same sanyaku opponents as Kisenosato during the second half, thus on average Kisenosato's opponents will be stronger than Kotomitsuki's.

On the other hand it is also a fact that Kisenosato beat 3 yokozunas and 12 ozekis in 2008 (counting only those who got a kachikoshi themselves but including Ama already) while Kotomitsuki only beat 2 yokozunas and 6 ozekis in 2008. That's why Kotomitsuki's 8 wins in Hatsu and 9 wins in Kyushu (where he didn't beat even a single genki ozeki in any of these two bashos) score extremely low for an ozeki (1803 resp. 1890 points).

Would all of the above reward a statement that Kisenosato and Kotomitsuki performed at eye level in 2008, i. e. making up for the four wins that Kotomitsuki has more than Kisenosato? Does that "ring true" to what you believe to have seen in 2008? That's more what I'm interested in, as it would question my model as a whole instead of individual numeric parameters or it's use for a particular application.

[Jakusotsu 5,885]

I wonder if all female members are now shaking their heads about the male obsession to turn everything into numbers...

[Doitsuyama 1,178]

Yes, it certainly would, and if it were my goal to predict matches I would do exactly this (I do have data for six bashos, no problem to start the next basho with the weighted average of the year 2008 for each sekitori - still not an ELO number but a lot better than nothing of course).

So why don't you use ELO numbers?

I don't have any reasonable data base (still toying around with my spreadsheet for the time being) but I might actually try coding this thing just for fun... then again, that wasn't my goal when I started this thread. I wasn't trying to show a tool for playing sumo prediction games. If I were to play any such game then using strength values alone would certainly not be enough, I wouldn't want to ignore the fact that certain sekitori tend to "own" certain others, the fact that a sekitori with 7 wins is much more motivated than his opponent with 8 wins and other things like this (most of which can be done more reasonably with a huge historical database).

I'm more interested in getting a kind of "virtual banzuke" that doesn't show the oddities of the existing sanyaku mechanism but is based on actual performance of all sekitori. Ama, for example, was definitely the second best sekitori of 2008 with his 61 wins over opponents on the same level as Hakuho's (okay, 2% weaker to be exact), and my score expresses this very clearly. It would, among other things, punish those who show "unmotivated sumo" after having their kachikoshi relatively to those who are going all out for a sansho/yusho.

What I rather hoped for was a discussion about how sekitori like Aminishiki ought to be rated, a guy who performs excellently against sanyaku opponents and at the same time poorly against maegashiras. My formula (ignoring his losses) would give him a much higher score than if he did the exact opposite to come up with the same number of wins, and the idea behind this is that he is going after what the sumo audience apparently wants him to - sanshos and kinboshis (as he can apparently get them while doing an ozeki run seems out of the question as of now). Aminishiki had 43 wins in 2008, the same as Kaio; even if I weight all six bashos equally (as to not punish Kaio for having gone kyujo in the most recent basho) Aminishiki gets a higher average score (1970 : 1869) than Kaio as he beat the stronger opponents this year. Even Asashoryu with his 45 wins scores just slightly better with an average of 2041 points, and this would have been a mere 1950 points had he made two wins less, so the opponents whom Aminishiki defeated were actually stronger on average than those Asashoryu defeated! (Bad luck for Asashoryu that he left two bashos early where he had only rather weak opponents until then, this downgraded his opponents mix.)

In terms of average score for the opponents they defeated, both Kisenosato and Amishishiki beat stronger opponents on average in 2008 than each of the four ozekis (only Hakuho and Ama defeated even stronger opponents this year). Baruto, for example, defeated much weaker opponents on average which puts his 51 wins a bit into perspective; my formula gives Baruto a score that's 96,3% of Aminishiki's.

Kisenosato/51 gets almost the same average score by my model as Kotomitsuki/55. That's plausible: Early during a basho Kisenosato gets sanyaku opponents, late in the basho he gets genki maegashiras as he's genki himself whereas Kotomitsuki gets maegashiras of random genkiness early in the basho and the same sanyaku opponents as Kisenosato during the second half, thus on average Kisenosato's opponents will be stronger than Kotomitsuki's.

On the other hand it is also a fact that Kisenosato beat 3 yokozunas and 12 ozekis in 2008 (counting only those who got a kachikoshi themselves but including Ama already) while Kotomitsuki only beat 2 yokozunas and 6 ozekis in 2008. That's why Kotomitsuki's 8 wins in Hatsu and 9 wins in Kyushu (where he didn't beat even a single genki ozeki in any of these two bashos) score extremely low for an ozeki (1803 resp. 1890 points).

Would all of the above reward a statement that Kisenosato and Kotomitsuki performed at eye level in 2008, i. e. making up for the four wins that Kotomitsuki has more than Kisenosato? Does that "ring true" to what you believe to have seen in 2008? That's more what I'm interested in, as it would question my model as a whole instead of individual numeric parameters or it's use for a particular application.

Just leave Sumo games alone for a moment - if you want to have a virtual banzuke with new (better) order, this order should go by strength not by popular impression. At least this is how I feel. And for evaluation of strength it's irrelevant if you beat a yokozuna and lose to an M4 or the other way around.

Let's just call your system an "impressionistic banzuke" without real strength aspects.

[Kintamayama 44,741]

I wonder if all female members are now shaking their heads about the male obsession to turn everything into numbers...

My head is not only shaking-it's spinning , gyrating and making noises.

And I'm not even female. Yet?

[Peeter 15]

I wonder if all female members are now shaking their heads about the male obsession to turn everything into numbers...

My head is not only shaking-it's spinning , gyrating and making noises.

And I'm not even female. Yet?

Female organismus can produce eggs, male ones have maximum 2 of them. On Your avatar I can see minimum 6 of them, You are not far from! ;-)

Again sorry for offtopic!

As IT worker I can say - if you have data of 2 events - statistics may be done! Very interesting reading, why not algorithm for beautyful program!

[Gusoyama 99]

Would all of the above reward a statement that Kisenosato and Kotomitsuki performed at eye level in 2008, i. e. making up for the four wins that Kotomitsuki has more than Kisenosato? Does that "ring true" to what you believe to have seen in 2008?

Personally, I don't think so. I think Kotomitsuki performed better in 08 that Kisenosato.

[Toukeigakusha 0]

Data update after Nagoya 2009 (top 30 positions):

Pos.Score Wins Rank  Sekitori		 Aki[5]  Kyushu[6]   Hatsu[7]	Haru[8]   Natsu[9] Nagoya[10]
---------------------------------------------------------------------------------------------------
_1. 4267   84  Y-E   Hakuho		  4401/14	4238/13	4158/14	4565/15	4199/14	4115/14
_2. 3366   66  O1-E  Harumafuji	  3560/12	4012/13	2813/ 8	2745/10	4912/14	2372/ 9
_3. 2625   58  O1-W  Kotooshu		2075/ 8	1683/ 8	2831/10	2878/10	2234/ 9	3470/13
_4. 2510   52  Y-W   Asashoryu	   1218/ 5	   0/ 0	4524/14	2733/11	3195/12	2456/10
---------------------------------------------------------------------------------------------------
_5. 2298   52  S1-W  Kisenosato	  1481/ 6	3283/11	2291/ 8	1160/ 5	2525/13	2827/ 9
_6. 2094   50  O2-W  Kotomitsuki	 2843/11	1890/ 9	 556/ 2	2168/ 8	1502/ 8	3390/12
_7. 1904   49  M03-W Baruto		  1970/ 8	2389/ 9	2275/ 9	1624/ 8	 900/ 4	2450/11
_8. 1831   45  M01-W Goeido		  2499/10	1350/ 5	2275/10	2350/ 9	1564/ 6	1300/ 5
_9. 1811   44  M05-E Aminishiki	  2056/ 8	2647/ 8	 790/ 3	1549/ 9	1299/ 5	2572/11
10. 1720   45  K1-W  Kotoshogiku	 1418/ 6	2048/ 9	1408/ 6	1495/ 6	1866/10	1940/ 8
11. 1717   42  O2-E  Kaio			2512/ 9	 210/ 1	1841/ 8	1775/ 8	1905/ 8	1921/ 8
12. 1686   43  O3-E  Chiyotaikai	 2270/ 9	1937/ 8	2043/ 8	 288/ 2	1627/ 8	2164/ 8
13. 1662   45  S1-E  Kakuryu		 1629/ 7	 864/ 5	1164/ 9	2672/10	2552/ 9	 896/ 5
14. 1630   45  K1-E  Kyokutenho	  1374/ 6	2109/10	1944/ 9	1290/ 6	1784/ 8	1384/ 6
---------------------------------------------------------------------------------------------------
15. 1185   41  M04-E Miyabiyama	  1011/ 4	1728/10	1370/ 6	 772/ 4	1039/ 9	1277/ 8
16. 1179   45  M09-E Takekaze		1152/ 9	1545/ 9	1279/ 7	1530/ 8	 556/ 4	1183/ 8
17. 1166   46  M06-W Tamanoshima	  762/ 7	 516/ 6	1437/11	1374/ 8	 812/ 5	1720/ 9
18. 1160   38  M02-E Tochiozan	   1174/ 6	 637/ 6	1212/10	2246/ 8	1413/ 6	 335/ 2
19. 1115   38  M07-E Toyonoshima	 1479/ 6	2348/ 9	 478/ 2	1405/ 8	 610/ 5	 862/ 8
20. 1106   43  M04-W Takamisakari	 591/ 6	1369/10	 978/ 6	1122/ 6	1115/ 9	1275/ 6
21. 1102   43  M10-W Tokitenku	   1004/ 6	1093/ 7	 957/ 9	1316/ 5	1171/ 7	1024/ 9
22. 1028   43  M05-W Tochinoshin	  820/ 8	 687/ 3	 977/ 8	 734/ 6	1096/ 9	1545/ 9
---------------------------------------------------------------------------------------------------
23.  988   38  M12-E Asasekiryu	  1425/ 4	 853/ 5	1024/ 6	1155/ 9	 748/ 5	 908/ 9
24.  984   36  M15-E Homasho			0/ 0	 554/ 6	1466/ 8	1995/11	 120/ 1	1365/10
25.  946   41  M12-W Yoshikaze		752/ 7	1169/11	1190/ 6	1068/ 7	 740/ 4	 824/ 6
26.  923   44  M08-W Kakizoe		 1175/10	 873/ 5	 940/ 8	1043/ 7	 789/ 8	 840/ 6
27.  901   40  M13-E Futeno		  1424/11	 970/ 5	 711/ 5	1021/ 7	 860/ 6	 670/ 6
28.  896   38  M09-W Tochinonada	 1093/ 6	 510/ 5	1099/ 8	1555/ 8	 722/ 5	 517/ 6
29.  882   35  M01-E Aran			   0/ 0	 830/ 8	 561/ 5	1132/10	1461/ 8	 860/ 4
30.  841   40  M08-E Kokkai		   657/ 8	1211/ 9	 880/ 5	 774/ 5	 867/ 8	 713/ 5

(Score = weighed average with the weights shown next to the basho names)

Hakuho's performance is worth 4000 points (yusho contender equivalent) in every single basho - he is playing in his own league these days. Asashoryu with 6 full bashos would be on eye level with Harumafuji.

The three yusho competitors are worth something equivalent to an ozeki run during the recent bashos: Harumafuji got 34+1 wins+yusho (10029 points, including 700 for the kettei-sen extra bout win over Hakuho/Y1E=50*14), Asashoryu got 33 wins (8384 points), Kotooshu got 32 wins+jun-yusho (8582 points).

This shows how high the hurdle for the young guns to make it to ozeki actually is: Even Kisenosato's best three bashos scored no more than 8635 points in total and would barely suffice for this (33 wins as well but against weaker opponents; note how his 9 wins as S1-E in Nagoya score higher than his 13 wins as M04-E in Natsu). Kotomitsuki's three best bashos gave him 31 wins (8401 points).

Ozekis may throw away a basho (at the expense of being kadoban in the next one) to heal injuries whereas sanyaku rikishi would ruin their position and need to fight their way back into their previous rank. And indeed Kotomitsuki (2009 Hatsu), Kaio (2008 Kyushu) and Chiyotaikai (2009 Haru) have all made use of this advantage recently. Had they made an average result in their

[HenryK 38]

Data update after Nagoya 2009 (top 30 positions)

Funny how our top 8 are just identical. Further down we differ a little, see below my (your) score. I wonder where you have Wakanosato (or do you ot rank rikishi dropping to Juryo?)

1 Hakuho

2 Harumafuji

3 Kotooshu

4 Asashoryu

5 Kisenosato

6 Kotomitsuki

7 Baruto

8 Goeido

9 (11) Kaio

10 (12) Chiyotaikai

11 (14) Kyokutenho

12 (9) Aminishiki

13 (10) Kotoshogiku

14 (13) Kakuryu

15 (16) Miyabiyama

16 (19) Toyonoshima

17 (16) Takekaze

18 (18) Tochiozan

19 (-) Wakanosato

20 (17) Tamanoshima

[Toukeigakusha 0]

Funny how our top 8 are just identical.

We're doing very similar things after all, and I'm aware that Aminishiki profits from my not punishing losses against weaker opponents, so he's the most likely candidate for being ranked higher in my model than in yours. If anything, this comparison shows how the combination of linear values for banzuke positions plus genkiness gets quite close to Abe's valuation of banzuke ranks...

I wonder where you have Wakanosato (or do you ot rank rikishi dropping to Juryo?)

His line in my table is:

Pos. Score Wins Rank	Sekitori	Aki[5]  Kyushu[6] Hatsu[7] Haru[8] Natsu[9] Nagoya[10]
33.   720   28  J06-W   Wakanosato  1282/9   1537/6	899/7   1309/6	0/0	  0/0

Given his potential of making four-digit scores he would be a candidate for rank 15-20 if he's in full health.

I have everyone in my table who had at least one basho with a top 50 banzuke position within my time frame. I map the 50 highest banzuke positions to integers, which covers ranks down to J04-W (the likely limit for opponents of Makuuchi sekitori), thus Wakanosato can't make a score in Nagoya from J06-W (as I can't model his opponents with "negative positions" - see above in this thread how top Juryo gets a non-zero score) and didn't win a bout in Natsu. These two bashos get the highest weights of 9 and 10, thus Wakanosato is currently being punished for not making a single score point in the two most significant bashos (with a combined weight of 42.2%), which kicked him out of the top 30 temporarily.

EDIT: Then again, giving all 6 bashos equal weights would still leave Wakanosato on position 31; skipping two bashos and defeating very un-genki opponents in Hatsu (none of the 7 opponents he defeated ended up better than 6:9, and the only one of them ranked higher than Wakanosato himself was Kotomitsuki who went kyuyo with 2 wins) is just too much for a better position in my list.

Asashoryu was punished similarly in my table four bashos ago, see above in this thread; the punishment for being kyuyo is being diminished with each following basho, much like a fading memory. The factor of 2 between the most recent and the oldest basho is, again, arbitrarily - I didn't want to handle almost meaningless data from the oldest basho (therefore no 1 to 6) but considered the most recent basho somehow more significant than older data (therefore no equal weights of all 6 bashos), and 5 to 10 looked like a natural choice.

Edited August 5, 2009 by Toukeigakusha

[Toukeigakusha 0]

Data update after Aki 2009 (top 30 positions):

Pos.	Score Wins  Rank	Sekitori	   Kyushu[5]   Hatsu[6]	Haru[7]   Natsu[8]  Nagoya[9]	Aki[10]
-----------------------------------------------------------------------------------------------------------
_1.(_1) 4344	84  Y-E	 Hakuho		   4238/13	4158/14	4565/15	4199/14	4115/14	4677/14
_2.(_2) 3117	63  O2-E	Harumafuji	   4012/13	2813/ 8	2745/10	4912/14	2372/ 9	2347/ 9
_3.(_4) 3083	61  Y-W	 Asashoryu		   0/ 0	4524/14	2733/11	3195/12	2456/10	4479/14
_4.(_3) 2540	59  O1-E	Kotooshu		 1683/ 8	2831/10	2878/10	2234/ 9	3470/13	1963/ 9
-----------------------------------------------------------------------------------------------------------
_5.(_5) 2274	53  S1-E	Kisenosato	   3283/11	2291/ 8	1160/ 5	2525/13	2827/ 9	1840/ 7
_6.(_7) 2180	53  K1-E	Baruto		   2389/ 9	2275/ 9	1624/ 8	 900/ 4	2450/11	3188/12
_7.(13) 2030	49  M03-W   Kakuryu		   864/ 5	1164/ 9	2672/10	2552/ 9	 896/ 5	3286/11
_8.(_6) 1946	48  O1-W	Kotomitsuki	  1890/ 9	 556/ 2	2168/ 8	1502/ 8	3390/12	1710/ 9
_9.(_8) 1735	45  M05-E   Goeido		   1350/ 5	2275/10	2350/ 9	1564/ 6	1300/ 5	1701/10
10.(10) 1705	45  S1-W	Kotoshogiku	  2048/ 9	1408/ 6	1495/ 6	1866/10	1940/ 8	1518/ 6
11.(11) 1663	41  O2-W	Kaio			  210/ 1	1841/ 8	1775/ 8	1905/ 8	1921/ 8	1780/ 8
12.(_9) 1625	43  K1-W	Aminishiki	   2647/ 8	 790/ 3	1549/ 9	1299/ 5	2572/11	1077/ 7
-----------------------------------------------------------------------------------------------------------
13.(14) 1492	44  M02-W   Kyokutenho	   2109/10	1944/ 9	1290/ 6	1784/ 8	1384/ 6	 918/ 5
14.(12) 1327	36  O3-E	Chiyotaikai	  1937/ 8	2043/ 8	 288/ 2	1627/ 8	2164/ 8	 326/ 2
15.(16) 1292	45  M05-W   Takekaze		 1545/ 9	1279/ 7	1530/ 8	 556/ 4	1183/ 8	1692/ 9
16.(18) 1249	43  M12-E   Tochiozan		 637/ 6	1212/10	2246/ 8	1413/ 6	 335/ 2	1572/11
17.(21) 1211	45  M04-W   Tokitenku		1093/ 7	 957/ 9	1316/ 5	1171/ 7	1024/ 9	1548/ 8
18.(17) 1196	44  M03-E   Tamanoshima	   516/ 6	1437/11	1374/ 8	 812/ 5	1720/ 9	1102/ 5
19.(24) 1150	43  M06-W   Homasho		   554/ 6	1466/ 8	1995/11	 120/ 1	1365/10	1297/ 7
20.(19) 1116	39  M04-E   Toyonoshima	  2348/ 9	 478/ 2	1405/ 8	 610/ 5	 862/ 8	1314/ 7
21.(20) 1108	43  M07-W   Takamisakari	 1369/10	 978/ 6	1122/ 6	1115/ 9	1275/ 6	 889/ 6
22.(15) 1085	41  M01-W   Miyabiyama	   1728/10	1370/ 6	 772/ 4	1039/ 9	1277/ 8	 675/ 4
23.(25) 1042	43  M15-E   Yoshikaze		1169/11	1190/ 6	1068/ 7	 740/ 4	 824/ 6	1311/ 9
24.(29) 1025	42  M07-E   Aran			  830/ 8	 561/ 5	1132/10	1461/ 8	 860/ 4	1126/ 7
25.(22) 1020	39  M01-E   Tochinoshin	   687/ 3	 977/ 8	 734/ 6	1096/ 9	1545/ 9	 877/ 4
26.(32)  961	38  M08-E   Iwakiyama		   0/ 0	 991/ 8	1183/ 8	1066/ 9	 989/ 5	1157/ 8
27.(26)  916	43  M11-E   Kakizoe		   873/ 5	 940/ 8	1043/ 7	 789/ 8	 840/ 6	1005/ 9
28.(23)  898	40  M06-E   Asasekiryu		853/ 5	1024/ 6	1155/ 9	 748/ 5	 908/ 9	 776/ 6
29.(30)  834	40  M14-E   Kokkai		   1211/ 9	 880/ 5	 774/ 5	 867/ 8	 713/ 5	 744/ 8
30.(35)  799	39  M02-E   Shotenro			0/ 0	  88/11	 676/ 7	 974/ 8	1569/11	 880/ 2

(Score = weighted average with the weights shown next to the basho names)

Asashoryu is back: The next basho will most likely see him in second place again (replacing the 0 points Kyusho result). His Aki result (despite including the playoff win) suffers from his almost worthless wins over Chiyotaikai/2=88 and Shotenro/2=74 whereas Hakuho lost against Shotenro, avoided Chiyotaikai and had the torikumi luck to face (and beat) a very genki Kakuryu/11=374 instead - who delivered his best result so far and actually finished "virtual third" this basho: Baruto's 12 wins include lots of valuable ozekis but also Chiyotaikai and Shotenro whereas Kakuryu had no opponent with fewer than 5 wins and scored big for wins over Baruto/12=492, Kotooshu/9=432, Kaio/8=360 and Goeido/10=310.

Chiyotaikai drops his second basho this year; even Kaio had to rely on a senshuraku win over Kotomitsuki (who would tie Kakuryu's overall score had he won this last fight) for avoiding kadoban and barely staying above Goeido and Kotoshogiku (if we already ignore Kaio's weak Kyushu score). Of the sanyaku candidates, Baruto appears to be able to perform at 2000+ level almost consistently but Kakuryu (with three 2500+ scores in the last four basho, and three gino-sho) might actually be the winner of 2009 in this area. Kisenosato still has the highest non-ozeki average but Kakuryu will already surpass him in the next basho if both make the same score there (replacing Kise's best and Kakuryu's weakest score). Note how few sekitori got 2000 points this time (just one ozeki amongst them, and Kisenosato with makekoshi still earning the seventh highest score), making Baruto's and Kakuryu's 3000+ performances even more outstanding.

Kyoukutenho appears to lose contact to the sanyaku candidate group; former sanyaku Tokitenku delivered his best performance for a long time. Shotenro's result of 880 points (from a disastrous looking 2:13) was actually to be expected score-wise (just not the kinboshi that was worth 700 points alone); Iwakiyama reliably delivers 1000 points every basho whereas Kokkai has dropped down to the 800-ish level, meaning permanent danger of demotion to Juryo. Wakanosato (with two zero-score bashos in a row) is currently ranked 31th in my list; his average of the other four bashos would be 1255, meaning a potential rank 16. The same goes for Hokutoriki (currently ranked 35th) with an average of 1198 from his three genki makuuchi bashos amongst the last six ones, and even Bushuyama (34th) has three scores > 1000 amongst his last six bashos.

Edited September 28, 2009 by Toukeigakusha

[Toukeigakusha 0]

Data update after Kyushu 2009 (top 30 positions):

Pos.	Score	  Wins  Rank	Sekitori	   Hatsu[5]	Haru[6]   Natsu[7]  Nagoya[8]	 Aki[9] Kyushu[10]
---------------------------------------------------------------------------------------------------------------
_1.( 1) 4416 (+ 72)  86  Y-W	 Hakuho		  4158/14	4565/15	4199/14	4115/14	4677/14	4613/15
---------------------------------------------------------------------------------------------------------------
_2.( 3) 3320 (+137)  72  Y-E	 Asashoryu	   4524/14	2733/11	3195/12	2456/10	4479/14	2804/11
_3.( 2) 3027 (- 90)  59  O2-E	Harumafuji	  2813/ 8	2745/10	4912/14	2372/ 9	2347/ 9	3120/ 9
_4.( 4) 2678 (+138)  61  O1-E	Kotooshu		2831/10	2878/10	2234/ 9	3470/13	1963/ 9	2803/10
---------------------------------------------------------------------------------------------------------------
_5.( 6) 2256 (+ 76)  53  S1-E	Baruto		  2275/ 9	1624/ 8	 900/ 4	2450/11	3188/12[K] 2580/ 9
_6.( 7) 2111 (+ 81)  51  S1-W	Kakuryu		 1164/ 9	2672/10[G] 2552/ 9[G]  896/ 5	3286/11[G] 1855/ 7
_7.( 8) 1978 (+ 32)  47  O1-W	Kotomitsuki	  556/ 2	2168/ 8	1502/ 8	3390/12	1710/ 9	2020/ 8
_8.( 5) 1969 (-305)  48  K1-E	Kisenosato	  2291/ 8	1160/ 5	2525/13[K] 2827/ 9	1840/ 7	1332/ 6
_9.(11) 1934 (+267)  48  O2-W	Kaio			1841/ 8	1775/ 8	1905/ 8	1921/ 8	1780/ 8	2246/ 8
10.(10) 1852 (+147)  46  M03-E   Kotoshogiku	 1408/ 6	1495/ 6	1866/10	1940/ 8	1518/ 6	2507/10
11.( 9) 1743 (+  8)  47  K1-W	Goeido		  2275/10[G] 2350/ 9	1564/ 6	1300/ 5	1701/10	1632/ 7
---------------------------------------------------------------------------------------------------------------
12.(12) 1433 (-192)  40  M01-W   Aminishiki	   790/ 3	1549/ 9	1299/ 5	2572/11[G] 1077/ 7	1188/ 5
13.(13) 1406 (- 86)  42  M06-W   Kyokutenho	  1944/ 9	1290/ 6	1784/ 8	1384/ 6	 918/ 5	1399/ 8
14.(25) 1355 (+335)  48  M08-W   Tochinoshin	  977/ 8	 734/ 6	1096/ 9	1545/ 9	 877/ 4	2377/12[K]
15.(15) 1340 (+ 48)  42  M01-E   Takekaze		1279/ 7	1530/ 8	 556/ 4	1183/ 8	1692/ 9	1614/ 6
16.(20) 1300 (+184)  41  M05-E   Toyonoshima	  478/ 2	1405/ 8	 610/ 5	 862/ 8	1314/ 7	2468/11[G]
17.(17) 1211 (+  0)  43  M02-W   Tokitenku		957/ 9	1316/ 5	1171/ 7	1024/ 9	1548/ 8	1151/ 5
18.(16) 1208 (- 41)  42  M03-E   Tochiozan	   1212/10	2246/ 8	1413/ 6	 335/ 2	1572/11	 812/ 5
19.(23) 1205 (+163)  42  M09-E   Yoshikaze	   1190/ 6	1068/ 7	 740/ 4	 824/ 6	1311/ 9	1831/10
20.(19) 1150 (+  0)  43  M07-W   Homasho		 1466/ 8[K] 1995/11[K]  120/ 1	1365/10	1297/ 7	 900/ 6
21.(22) 1134 (+ 49)  43  M09-W   Miyabiyama	  1370/ 6	 772/ 4	1039/ 9	1277/ 8	 675/ 4	1597/12[K]
22.(18) 1075 (-119)  42  M07-E   Tamanoshima	 1437/11	1374/ 8	 812/ 5	1720/ 9	1102/ 5	 358/ 4
23.(14) 1074 (-253)  30  O3-E	Chiyotaikai	 2043/ 8	 288/ 2	1627/ 8	2164/ 8	 326/ 2	 474/ 2
24.(21) 1060 (- 48)  41  M11-W   Takamisakari	 978/ 6	1122/ 6	1115/ 9	1275/ 6	 889/ 6	1008/ 8
25.(27) 1030 (+114)  46  M05-W   Kakizoe		  940/ 8	1043/ 7	 789/ 8	 840/ 6	1005/ 9	1412/ 8
26.(24)  998 (- 27)  41  M08-E   Aran			 561/ 5	1132/10	1461/ 8	 860/ 4	1126/ 7	 805/ 7
27.(26)  925 (- 36)  40  M04-W   Iwakiyama		991/ 8	1183/ 8	1066/ 9	 989/ 5	1157/ 8	 380/ 2
28.(28)  891 (-  7)  43  M10-E   Asasekiryu	  1024/ 6	1155/ 9	 748/ 5	 908/ 9	 776/ 6	 858/ 8
29.(30)  889 (+ 90)  48  M12-E   Shotenro		  88/11	 676/ 7	 974/ 8	1569/11[K]  880/ 2	 820/ 9
30.(29)  792 (- 42)  39  M10-W   Kokkai		   880/ 5	 774/ 5	 867/ 8	 713/ 5	 744/ 8	 812/ 8

With 6 non-zero bashos in a row, Asashoryu is back in 2nd place and Kaio back in the top 10.

With four scores above 2800 and more than 10 wins per basho on average, Kotooshu delivered a decent ozeki performance; with a 2739 average of his last three results Baruto gives us hope to make the next step in the near future. Both Baruto and Kakuryu (the winner of this year with three Gino-Sho) had one basho below 1000 points and still scored higher than the three veteran ozekis, both surpassing Kisenosato as best non-ozekis.

With weaker performances by Aminishiki (starting 4:3 and defeating three ozekis, only to go 0:7 from there and finally beating Iwakiyama/2 for a 5:10) and Kyokutenho (starting 2:4 from M06-W and getting to face weak opponents because of this, the kachi-koshi wasn't that difficult and thus not worth a lot) recently, the 300 points gap between position 11 and 12 looks significant.

After a long slump, Toyonoshima showed his potential again with wins over Yoshikaze/10, Miyabiyama/12 and Kotooshu/10 (480 points); he's apparently sanyaku material unless injured. Tochinoshin began with three double-digit maegashita rankings in 2009 so his 48 wins were of relatively low quality but the trend clearly goes upwards and his Kyushu performance (from M08-W, defeating Kotoshogiku M02-E/10, Toyonoshima M05-E/11 and Miyabiyama M09-W/12, plus Kyokutenho/8, Asasekiryu/8, Tosayutaka/8, Takamisakari/8, Mokonami/9, Shotenro/9 and losing only to three kachi-koshi opponents) was indeed worth a sansho, more so than Miyabiyama's who faced a lot of weaker opponents in Kyushu (with 9 of his 12 wins over sekitori all of which were ranked below him with an average genkiness of 7 wins, plus defeating Tamanoshima/4, leaving only two valuable wins over Yoshikaze/10 and Kyokutenho/8). On the other hand, Kotoshogiku defeated Harumafuji/9 (day 1), Baruto/9 (day 2), Bushuyama/6, Iwakiyama/2, Aminishiki/5, Hokutoriki/8, Kisenosato/6, Goeido/7, Takekaze/6 and Kyokutenho/8, not earning him a sansho after his exceptional start but his best performance this year nonetheless. Kakizoe surprised with a kachi-koshi at M05-W (defeating Tochinoshin/12 in the process), getting his highest score of the year as well.

Wakanosato still has two bashos with zero scores, keeping him out of the top 30 range just like Hokutoriki and Bushuyama with one Juryo basho each.

Overall, the scores are increasing (+23 on average) probably due to a lower injuries rate in recent bashos.

EDIT: Tagged all sansho winners, extended interpretation.

Edited December 15, 2009 by Toukeigakusha

[Toukeigakusha 0]

Comparison 2009 to 2008 (top 30 positions):

---------------------------------
_1.( 1) 4416 (+ 435)  Hakuho
---------------------------------
_2.(12) 3320 (+1594)  Asashoryu
_3.( 2) 3027 (-  71)  Harumafuji
_4.( 5) 2678 (+ 557)  Kotooshu
---------------------------------
_5.( 7) 2256 (+ 284)  Baruto
_6.(18) 2111 (+ 812)  Kakuryu
_7.( 3) 1978 (- 451)  Kotomitsuki
_8.( 4) 1969 (- 410)  Kisenosato
_9.(11) 1934 (+ 149)  Kaio
10.( 9) 1852 (-  31)  Kotoshogiku
11.(14) 1743 (+ 123)  Goeido
---------------------------------
12.( 6) 1433 (- 615)  Aminishiki
13.(13) 1406 (- 237)  Kyokutenho
14.(33) 1355 (+ 753)  Tochinoshin
15.(20) 1340 (+  77)  Takekaze
16.( 8) 1300 (- 658)  Toyonoshima
17.(17) 1211 (- 115)  Tokitenku
18.(26) 1208 (+ 229)  Tochiozan
19.(30) 1205 (+ 407)  Yoshikaze
20.(29) 1150 (+ 352)  Homasho
21.(15) 1134 (- 444)  Miyabiyama
22.(32) 1075 (+ 446)  Tamanoshima
23.(10) 1074 (- 791)  Chiyotaikai
24.(25) 1060 (+  68)  Takamisakari
25.(28) 1030 (+ 184)  Kakizoe
26.(41)  998 (+ 814)  Aran
27.(34)  925 (+ 396)  Iwakiyama
28.(16)  891 (- 673)  Asasekiryu
29.(--)  889 (+ 889)  Shotenro
30.(24)  792 (- 330)  Kokkai

Not that great a year for the Japanese of whom only Kaio and Goeido slightly improved their score within the top 12 while Yoshikaze made the biggest step forward (Tamanoshima and Iwakiyama hat Juryo visits in 2008). With three of the "big four" amongst the biggest winners, ozeki runs may now be even more difficult. Hokutoriki, Futeno, Tochinonada (all with Juryo visits) and Dejima (retired) dropped out of the top 30 this year.

Asashoryu attending 6 bashos again almost doubled his score while we saw huge improvements for Kakuryu, Tochinoshin, Aran and Shotenro and deep declines for Kotomitsuki, Kisenosato, Aminishiki, Toyonoshima, Chiyotaikai and Miyabiyama, with Asasekiryu being the only gaijin to go against the trend.

Edited December 15, 2009 by Toukeigakusha

[Toukeigakusha 0]

Data update after Hatsu 2010 (top 40 positions now for Wakanosato's sake...):

Pos.	Score	  Wins  Rank	Sekitori		Haru[5]   Natsu[6]  Nagoya[7]	 Aki[8]  Kyushu[9]  Hatsu[10]
---------------------------------------------------------------------------------------------------------------
_1.( 1) 4207 (-209)  84  Y1-E	Hakuho		  4565/15	4199/14	4115/14	4677/14	4613/15	3358/12
_2.( 2) 3326 (+  6)  71  Y1-W	Asashoryu	   2733/11	3195/12	2456/10	4479/14	2804/11	3858/13
---------------------------------------------------------------------------------------------------------------
_3.( 3) 2941 (- 86)  61  O1-W	Harumafuji	  2745/10	4912/14	2372/ 9	2347/ 9	3120/ 9	2569/10
_4.( 4) 2615 (- 63)  60  O1-E	Kotooshu		2878/10	2234/ 9	3470/13	1963/ 9	2803/10	2468/ 9
_5.( 5) 2553 (+297)  56  S1-E	Baruto		  1624/ 8	 900/ 4	2450/11	3188/12[K] 2580/ 9	3550/12[S]
---------------------------------------------------------------------------------------------------------------
_6.( 6) 2099 (- 12)  48  K1-W	Kakuryu		 2672/10[G] 2552/ 9[G]  896/ 5	3286/11[G] 1855/ 7	1654/ 6
_7.( 9) 2050 (+116)  49  O2-W	Kaio			1775/ 8	1905/ 8	1921/ 8	1780/ 8	2246/ 8	2404/ 9
_8.( 8) 2016 (+ 47)  49  M03-W   Kisenosato	  1160/ 5	2525/13[K] 2827/ 9	1840/ 7	1332/ 6	2329/ 9
---------------------------------------------------------------------------------------------------------------
_9.(10) 1800 (- 52)  46  K1-E	Kotoshogiku	 1495/ 6	1866/10	1940/ 8	1518/ 6	2507/10	1405/ 6
10.( 7) 1748 (-230)  46  O2-E	Kotomitsuki	 2168/ 8	1502/ 8	3390/12	1710/ 9	2020/ 8	 320/ 1
11.(11) 1702 (- 41)  44  M02-E   Goeido		  2350/ 9	1564/ 6	1300/ 5	1701/10	1632/ 7	1804/ 7
12.(16) 1652 (+352)  47  M01-E   Toyonoshima	 1405/ 8	 610/ 5	 862/ 8	1314/ 7	2468/11[G] 2489/ 8
13.(12) 1602 (+169)  48  M06-W   Aminishiki	  1549/ 9	1299/ 5	2572/11[G] 1077/ 7	1188/ 5	1922/11[G]
---------------------------------------------------------------------------------------------------------------
14.(15) 1381 (+ 41)  41  M04-W   Takekaze		1530/ 8	 556/ 4	1183/ 8	1692/ 9	1614/ 6	1482/ 6
15.(13) 1327 (- 79)  41  M05-W   Kyokutenho	  1290/ 6	1784/ 8	1384/ 6	 918/ 5	1399/ 8	1296/ 8
16.(14) 1236 (-119)  45  M01-W   Tochinoshin	  734/ 6	1096/ 9	1545/ 9	 877/ 4	2377/12[K]  616/ 5
17.(19) 1173 (- 32)  42  M05-E   Yoshikaze	   1068/ 7	 740/ 4	 824/ 6	1311/ 9	1831/10	1026/ 6
18.(17) 1155 (- 56)  39  M08-E   Tokitenku	   1316/ 5	1171/ 7	1024/ 9	1548/ 8	1151/ 5	 844/ 5
19.(18) 1115 (- 93)  40  M10-E   Tochiozan	   2246/ 8	1413/ 6	 335/ 2	1572/11	 812/ 5	 822/ 8
20.(25) 1097 (+ 67)  44  M04-E   Kakizoe		 1043/ 7	 789/ 8	 840/ 6	1005/ 9	1412/ 8	1280/ 6
21.(20) 1094 (- 56)  44  M12-E   Homasho		 1995/11[K]  120/ 1	1365/10	1297/ 7	 900/ 6	1051/ 9
22.(26) 1087 (+ 89)  46  M10-W   Aran			1132/10	1461/ 8	 860/ 4	1126/ 7	 805/ 7	1222/10
23.(21) 1066 (- 68)  42  M02-W   Miyabiyama	   772/ 4	1039/ 9	1277/ 8	 675/ 4	1597/12[K]  916/ 5
---------------------------------------------------------------------------------------------------------------
24.(24)  997 (- 63)  42  M11-E   Takamisakari	1122/ 6	1115/ 9	1275/ 6	 889/ 6	1008/ 8	 747/ 7
25.(22)  981 (- 94)  38  M13-E   Tamanoshima	 1374/ 8	 812/ 5	1720/ 9	1102/ 5	 358/ 4	 833/ 7
26.(27)  920 (-  5)  41  M14-E   Iwakiyama	   1183/ 8	1066/ 9	 989/ 5	1157/ 8	 380/ 2	 948/ 9
27.(35)  877 (+156)  49  M07-E   Tamawashi	   1067/ 9	 965/ 6	 374/ 5	  88/11	1144/10	1472/ 8
28.(28)  866 (- 25)  43  M08-W   Asasekiryu	  1155/ 9	 748/ 5	 908/ 9	 776/ 6	 858/ 8	 843/ 6
29.(37)  865 (+234)  49  M16-W   Toyohibiki		36/12	1442/11	 711/ 3	 751/ 6	 337/ 5	1608/12[K]
30.(29)  859 (- 30)  40  M09-W   Shotenro		 676/ 7	 974/ 8	1569/11[K]  880/ 2	 820/ 9	 402/ 3
31.(34)  847 (+102)  32  M07-W   Wakanosato	  1309/ 6	   0/ 0	   0/ 0	1276/10	 969/ 7	1262/ 9
32.(33)  825 (+ 76)  43  M13-W   Shimotori		903/ 8	 549/ 6	1156/ 9	 545/ 4	1001/ 8	 786/ 8
33.(30)  761 (- 31)  39  M09-E   Kokkai		   774/ 5	 867/ 8	 713/ 5	 744/ 8	 812/ 8	 694/ 5
34.(23)  738 (-336)  22  S1-W	Chiyotaikai	  288/ 2	1627/ 8	2164/ 8	 326/ 2	 474/ 2	   0/ 0
35.(32)  737 (- 48)  40  M06-E   Bushuyama		 64/ 8	1072/ 9	 546/ 5	1209/10	1097/ 6	 306/ 2
36.(39)  716 (+201)  49  M12-W   Tosayutaka		48/ 8	  72/ 9	 771/ 8	 810/ 6	 882/ 8	1173/10
---------------------------------------------------------------------------------------------------------------
37.(31)  600 (-189)  39  M03-E   Hokutoriki	   290/ 2	 387/ 4	  22/11	1219/11	1378/ 8	  90/ 3
38.(40)  585 (+ 71)  37  M11-W   Mokonami		   0/ 0	  77/11	 742/ 6	 771/ 5	 971/ 9	 577/ 6
39.(36)  525 (-107)  35  M15-W   Tochinonada	 1555/ 8	 722/ 5	 517/ 6	 436/ 4	  48/ 8	 400/ 4
40.(38)  445 (-145)  35  J03-E   Futeno		  1021/ 7	 860/ 6	 670/ 6	 572/ 5	  35/ 7	  16/ 4

Repeating his Hatsu yusho from 2009 (the first yusho below 4000 points since his 13 wins in Haru 2008), Asashoryu could keep his level; Sumo lost a great champion now. Hakuho had his first performance below 4000 points in 10 basho (Natsu 2008).

As for the "olde ozeki" group, Chiyotaikai had three 2200+ scores in four subsequent basho back in mid-2008 but never managed to repeat this performance; now that he's gone it may be time to worry about Kotomitsuki who had only one 2200+ score (12 wins for 3390 points at Nagoya 2009) in his last 8 basho (after four subsequent 2200+ scores in the same four basho 2008 as Chiyotaikai). Kaio (who had four consecutive basho in mid-2008 with an average of 2306 points before his injury) improved his scores recently, scoring big for defeating Harumafuji O1-W/10=470 and Hakuho Y1-E/12=600 this time.

With a 2942 points average of his recent 4 basho, Baruto plays in a different league than the lower sanyaku ranks. This time he scored a career high so far (and higher than even Hakuho whom he defeated for the first time) but 9318 points in three basho (including a 5:0 + 2:2 + 3:0 = 10:2 score against ozeki opponents) didn't earn him the ozeki promotion yet, despite clearly being the third best during the last six months (Harumafuji 8036, Kotooshu 7234, Kakuryu 6795, Kaio 6430, Toyonoshima 6271; Ama's Ozeki run in 2008 with 35 wins had been worth 10543 points while Kakuryu's three Gino-Sho from 2009 would add up to a mere 8510 points). Note how Baruto's 9 wins for 2580 points in Aki 2009 (that were debated as being the potential reason for his non-promotion) scored higher than every single one of Kaio's twelve kachi-koshi results since the beginning of 2008 (maximum: 9 wins for 2512 points in Aki 2008 including a win over Kotomitsuki O1-E/11=528), not to mention Chiyotaikai's seven kachi-koshi results during this time (maximum: 8 wins for 2476 points in Haru 2008 including a win over Hakuho Y1-E/12=600).

Kisenosato didn't beat an ozeki this time but defeated Baruto S1-E/12=528, Aminishiki M06-W/11=319 and Toyonoshima M01-E/8=320. Toyonoshima himself scored even higher than last time (when he won the gino-sho) due to his quality wins over Kaio O2-W/9=405, Harumafuji O1-W/10=470, and Baruto S1-E/12=528, showing good sekiwake potential for two consecutive basho. The group of regular 2000+ points scorers has shrunk significantly, with Goeido missing this score for 5 basho in a row (after three out of four 2000+ basho in late 2008/early 2009; his kinboshi win over Asashoryu Y1-W/13=637 makes up for 35% of his score this time), and both Kotoshogiku and Aminishiki getting that much only once in 7 basho (after three resp. four 2000+ scores in 2008) - and all of this despite all "three old ozekis" (including Chiyotaikai) being less competitive these days than they were in 2008.

Aminishiki's gino-sho from M06-W was rather light on the opponents' side with wins over Bushuyama M06-E/2, Tokitenku M08-E/5, Kokkai M09-E/5, Asasekiryu M08-W/6, and Kakuryu K1-W/6 as the only sanyaku rikishi amongst these (had he defeated Baruto on day 13 when he had the chance to it would have been a different story, then earning him some 2500-ish score like Harumafuji or Kotooshu). So Toyohibiki's kanto-sho run with 1608 points from the lowest maegashira position M16-W (from where you don't get valuable opponents that often) actually was more impressive, winning his last six bouts and defeating genki opponents such as Shimotori M13-W/8, Tamawashi M07-E/8, Sagatsukasa J01-W/9, Kitataiki M15-E/9, Homasho M12-E/9, Wakanosato M07-W/9, and Aran M10-W/10; the only other 1500+ points result from any two-digits maegashira rank in the last two years had been Tochiozan M12-E/11 with 1572 points in Aki 2009, from a rank that was 8 positions higher.

Wakanosato still has two 0 points basho in his score, keeping him out of the top 30 range... but with a potential of 1200 points per basho he might be at a top 20 rank again in mid-2010 (where the real banzuke has him anyway). Kokkai's only 1000+ score in the last 10 basho was Kyushu 2008 but he managed to stay in makuuchi with a minimum of effort; Asasekiryu (who had three 1900+ basho back in early 2008) shows a similar tendency these days.

[Harry 67]

Thanks for posting this, very interesting! For all the calls for Kaio's retirement he is still the highest Japanese rikishi in your rankings. Telling to see where Baruto and Kotomitsuki are. Shogiku, Goeido, Aminishiki and Toyonoshima are all up there too.

[HenryK 38]

Thanks for posting this, very interesting! For all the calls for Kaio's retirement he is still the highest Japanese rikishi in your rankings. Telling to see where Baruto and Kotomitsuki are. Shogiku, Goeido, Aminishiki and Toyonoshima are all up there too.

Well this is because Komoitsuki went fusen. There must be a better solution to this than not giving him any points -- such as carrying forward an average score of past performances (possibly with a penalty)

[Toukeigakusha 0]

Update at the end of 2010:

(score and position differences relative to the end of 2009; Hakuho's 15 wins in Kyushu include the kettei-sen; I assigned Harumafuji his year's average of 2388 points for Kyushu to make up for his fusen, otherwise he would have dropped to position 10 with a score of 1844 points; "G"=gino-sho, "S"=shukun-sho, "K"=kanto-sho, "#"=gino-sho plus kanto-sho)

Pos.  Score (Diff) Wins Sekitori	Hatsu*5   Haru*6  Natsu*7 Nagoya*8  Aki*9  Kyushu*10
----------------------------------------------------------------------------------------
_1.( 1) 4480 (+  64) 87 Hakuho	  3399/12  4655/15  4509/15  4495/15  4725/15  4662/15
----------------------------------------------------------------------------------------
_2.( 5) 2904 (+ 648) 64 Baruto	  3591/12# 4112/14G 2575/10  1878/ 8  2565/ 9  3193/11
----------------------------------------------------------------------------------------
_3.( 4) 2377 (- 301) 56 Kotooshu	2468/ 9  2205/10  2378/ 9  2722/10  2826/10  1755/ 8
_4.( 9) 2375 (+ 441) 52 Kaio		2445/ 9  2026/ 8  2330/ 9  1654/ 6  2153/ 8  3356/12
_5.( 3) 2375 (- 652) 47 Harumafuji  2610/10  2181/10  2324/ 9  2818/10  2005/ 8  [2388*]
----------------------------------------------------------------------------------------
_6.( 8) 2245 (+ 276) 50 Kisenosato  2370/ 9  2374/ 9  2343/ 8  1571/ 7  1806/ 7  2972/10S
_7.(10) 2207 (+ 355) 48 Kotoshogiku 1405/ 6  2654/10  2506/ 9  1466/ 5  2887/ 9  2112/ 9
_8.(18) 2156 (+ 948) 53 Tochiozan	832/ 8  2464/11  1739/ 7  2555/ 9  3143/11G 1716/ 7
----------------------------------------------------------------------------------------
_9.(12) 1962 (+ 529) 46 Aminishiki  1963/11G 2436/ 8  1060/ 5  1489/ 6  2446/ 8  2250/ 8
10.( 6) 1950 (- 161) 46 Kakuryu	 1654/ 7  1005/ 6  1844/ 6  2493/11  2643/ 9  1683/ 7
11.(14) 1766 (+ 411) 43 Tochinoshin  616/ 5  1751/ 9  2716/ 8K 1596/ 6  2405/ 9  1247/ 6
12.(26) 1608 (+ 610) 45 Aran		1243/10   175/ 1  2016/12K 3040/11K 1980/ 7   884/ 4
----------------------------------------------------------------------------------------
13.(52) 1379 (+1360) 43 Hakuba	  1123/ 9   915/ 8  2459/10  1132/ 4  1939/ 8   724/ 4
14.(20) 1379 (+ 229) 43 Homasho	 1061/ 9  1574/ 9	 0/ 0  1480/11K 1402/ 7  2285/ 7
15.(16) 1345 (+  45) 47 Toyonoshima 2489/ 8  1026/ 6  1096/ 5	 0/ 0	98/14  3337/14#
16.(28) 1345 (+ 454) 44 Asasekiryu   843/ 6  1747/10  2330/ 9   659/ 4  1560/ 9  1020/ 6
17.(13) 1283 (- 123) 40 Kyokutenho  1296/ 8   345/ 3  1601/ 9  1588/ 7   638/ 4  1951/ 9
18.(53) 1218 (+1211) 46 Kitataiki   1036/ 9  1765/10  1492/ 7   943/ 6  1307/ 9   928/ 5
19.(19) 1155 (-  50) 44 Yoshikaze   1026/ 6   548/ 5  1185/ 9   622/ 5  1681/11K 1514/ 8
20.(--) 1145 ( new ) 48 Tokusegawa	72/ 9   948/ 8  1198/ 9  1155/ 8  1161/ 6  1741/ 8
21.(17) 1138 (-  73) 41 Tokitenku	844/ 5  1495/10  1440/ 8  1902/ 8   447/ 2   869/ 8
22.(15) 1062 (- 278) 43 Takekaze	1482/ 6   673/ 5  1084/ 8   610/ 6  1855/12K  720/ 6
23.(35) 1050 (+ 329) 42 Tamawashi   1472/ 8  1061/ 5   234/ 3   812/ 7  1149/10  1503/ 9
24.(34)  980 (+ 235) 40 Wakanosato  1262/ 9   771/ 6   672/ 6  1666/ 9   810/ 5   786/ 5
25.(30)  963 (+ 171) 38 Kokkai	   792/ 5  1151/10   747/ 3   894/ 8  1565/ 8   599/ 4
26.(40)  937 (+ 423) 43 Mokonami	 587/ 6  1104/ 9  1172/ 8  1063/ 8   814/ 5   859/ 7
27.(39)  908 (+ 393) 43 Tosayutaka  1194/10   648/ 3   998/ 7   839/ 8   795/ 6  1014/ 9
28.(11)  904 (- 839) 42 Goeido	  1845/ 7   750/ 2  1106/ 9	 0/ 0	96/12  1837/12
29.(33)  846 (+  97) 43 Shimotori	797/ 8   460/ 5  1029/10   856/ 6   956/ 8   869/ 6
30.(24)  813 (- 247) 43 Takamisakari 758/ 7   813/ 7   832/ 8   891/ 9   555/ 4   995/ 8
31.(21)  780 (- 354) 41 Miyabiyama   916/ 5  1593/10  1216/ 5	 0/ 0	72/12  1178/ 9
----------------------------------------------------------------------------------------
32.( 7)  688 (-1290) 19 Kotomitsuki  320/ 1  2078/ 9  2416/ 9	 0/ 0	 0/ 0	 0/ 0
33.(45)  664 (+ 415) 46 Kimurayama	21/ 7	22/11   720/ 7   699/ 8  1198/ 8   825/ 5
34.(--)  587 ( new ) 42 Gagamaru	   0/ 0	40/ 8	70/10   597/ 5  1135/10  1069/ 9
35.(25)  550 (- 480) 26 Kakizoe	 1280/ 6  1027/ 7  1059/ 7   418/ 3   157/ 3	 0/ 0
36.(36)  435 (- 197) 32 Tochinonada  410/ 5	60/10   285/ 3	 0/ 0   933/ 8   678/ 6
37.( 2)  433 (-2887) 13 Asashoryu   3899/13	 0/ 0	 0/ 0	 0/ 0	 0/ 0	 0/ 0
38.(32)  422 (- 363) 23 Bushuyama	306/ 2   229/ 4	 0/ 0   757/ 8   732/ 6   346/ 3
39.(--)  413 ( new ) 30 Sokokurai	  0/ 0	 0/ 0	 8/ 8	32/ 8  1117/ 8   822/ 6
40.(49)  403 (+ 334) 32 Koryu		329/ 4	 0/ 0   430/ 5	18/ 9   402/ 6   973/ 8
----------------------------------------------------------------------------------------
41.(37)  399 (- 232) 42 Toyohibiki  1629/12K  468/ 4   959/ 8	 0/ 0	21/ 7	11/11

Baruto went way beyond compensating for Chiyotaikai & Kotomitsuki, he replaced Harumafuji as the yokozuna's only serious yusho contender. But the biggest surprise amongst the ozeki may actually be Kaio, scoring even higher than kettei-sen participant Toyonoshima in Kyushu.

Amongst the lower sanyaku, Kisenosato and Kotoshogiku appear to be more stable now with scores similar to those of the ozeki and none significantly below 1500 points; Tochiozan made the biggest leap forward here. Tochinoshin and Aran have become sanyaku candidates but are lacking consistency for the time being.

Hakuba and Kitataiki have made stunning improvements within the "makuuchi regulars"; Takamisakari shows us how to stay in the top division with a minimum score. Amongst the newcomers, Tokusegawa appears to have established himself as makuuchi regular now; it remains to be seen whether he can repeat his Kyusho result though. While Wakanosato had no problems staying up there in 2010 a handful of sekitori have now fallen out of the 40 due to dropping to juryo: Tamanoshima, Shotenro, and Hokutoriki.

The "gamblers" have all fallen in this ranking (having missed two makuuchi basho) but Toyonoshima's Kyusho score shows what this guy is capable of. Goeido didn't have a single 2000 points score this year and would have dropped out of the "meat grinder crowd" anyway. Toyohibiki, kanto-sho winner in hatsu, ended up in position 41 for the year 2010.

Tochinoshin's kanto-sho in natsu based on a mere 8 wins for 2716 points included wins over four ozeki plus one sekiwake; Homasho's 'cheap' kanto-sho in Nagoya based on 11 wins for 1480 points included only one win over an opponent higher than maegashira 9.

Edited December 2, 2010 by Toukeigakusha