Random Sumo Statistics: Winning Percentage per Rank

[Oimeru 0]

As requested by Quentooshu in the other thread, here is a table giving winning percentages per banzuke rank. If there are other stats you like to see, tell me. The columns are rank, wins, total bouts and winning percentage.

Y	  1272   1615 0.7876
O	  2042   3140 0.6503
S	  1271   2236 0.5684
K	  1043   2157 0.4835
M1	  863   2010 0.4294
M2	  780   1985 0.3929
M3	  759   1935 0.3922
M4	  813   2004 0.4057
M5	  869   2016 0.4311
M6	  893   1957 0.4563
M7	 1008   1996 0.505
M8	  972   1958 0.4964
M9	  984   2007 0.4903
M10	 951   1961 0.485
M11	1041   2027 0.5136
M12	 971   1987 0.4887
M13	 963   1947 0.4946
M14	 961   1931 0.4977
M15	 750   1510 0.4967
M16	 301	611 0.4926
M17	 137	300 0.4567

The meatgrinder positions in the top maegashira ranks are clearly visible: M3 has the worst winning percentage of all ranks -- below 40%. What suprises me a bit is that the very low Maegashira are not much worse than those in the middle. After all they have to fight higher ranked opponents most of the time. This maybe an indication that there indeed a lot of equally strong rikishi in lower makuuchi.

Also, the step between the Ozeki and the Yokozuna is quite large -- much bigger than between Ozeki and Sekiwake. I wonder how much of this is Asashoryu's work...

[Yuriyama Ren 1]

The meatgrinder positions in the top maegashira ranks are clearly visible: M3 has the worst winning percentage of all ranks -- below 40%. What suprises me a bit is that the very low Maegashira are not much worse than those in the middle. After all they have to fight higher ranked opponents most of the time. This maybe an indication that there indeed a lot of equally strong rikishi in lower makuuchi.

Also, the step between the Ozeki and the Yokozuna is quite large -- much bigger than between Ozeki and Sekiwake. I wonder how much of this is Asashoryu's work...

Interesting.

Can you give a table for Juuryou ranks as well? I suspect the top would have slightly lower percentages because of some bouts in makuuchi etc, and the rest would be around 0.49 like the bottom of makuuchi.

The average of oozeki rank seems quite close to 10 wins per basho, so the criticism at those oozekis who get less than 10 wins seems well justified..

What's the percentage for yokozuna when you exclude Asa?

Can you give a number of fusenshou losses by rank (excluding those that were on the first day of the basho) as well because the number of total bouts in meatgrinder positions is significantly smaller than in lower ranks. I wonder whether it is because of injury during basho.

[Oimeru 0]

Interesting.

Can you give a table for Juuryou ranks as well? I suspect the top would have slightly lower percentages because of some bouts in makuuchi etc, and the rest would be around 0.49 like the bottom of makuuchi.

The average of oozeki rank seems quite close to 10 wins per basho, so the criticism at those oozekis who get less than 10 wins seems well justified..

J1	 1000   1894 0.528
J2	  994   1900 0.5232
J3	 1027   1965 0.5226
J4	  910   1846 0.493
J5	  926   1894 0.4889
J6	  952   1876 0.5075
J7	  907   1875 0.4837
J8	  919   1858 0.4946
J9	  972   1917 0.507
J10	 880   1916 0.4593
J11	 925   1901 0.4866
J12	 914   1876 0.4872
J13	 976   1913 0.5102
J14	 184	434 0.424

Turn out that the top actually does a bit better -- they are usually facing lower ranked guys, except the few bouts against makuuchi rikishi. The rest is rather inconclusive, just that bump at J13 looks a bit odd.

[Oimeru 0]

What's the percentage for yokozuna when you exclude Asa?

Y	  1032   1332 0.7748

Turns out that Asa is only responsible for 1.3 percentage points -- I would have expected more, especially since the database only goes back into th mid-1990s. Reaching the Yokozuna rank is a significantly bigger step than becoming Ozeki.

Can you give a number of fusenshou losses by rank (excluding those that were on the first day of the basho) as well because the number of total bouts in meatgrinder positions is significantly smaller than in lower ranks. I wonder whether it is because of injury during basho.

This requires some script fiddling, maybe tomorrow :)

[Asojima 2,874]

I assume that you are ignoring all kyujo matches in your calculations. If kyujo matches were included as losses, both the Ozeki and Yokozuna win percentages would probably have a significant drop. Both ranks tend to drop out when a basho is going bad rather than post a 3-12 or 4-11 as the lower ranks will do. They have no incentive to salvage what they can to minimize their drop in rank.

The Traumatized Threesome have low win percentages with kyujo thrown in, and the twilight years for Taka, Big Moose and Little Moose were terrible.

[Yuriyama Ren 1]

Turn out that the top actually does a bit better -- they are usually facing lower ranked guys, except the few bouts against makuuchi rikishi. The rest is rather inconclusive, just that bump at J13 looks a bit odd.

So there goes my theory..

The drop at J10 is also bit odd.

Turns out that Asa is only responsible for 1.3 percentage points -- I would have expected more, especially since the database only goes back into th mid-1990s. Reaching the Yokozuna rank is a significantly bigger step than becoming Ozeki.

So Asa is not that special after all B-)

The average for Asa as yokozuna would then be: 240 283 0.8481.. (You are going off-topic...) I take it back, he is special..

[Oimeru 0]

I assume that you are ignoring all kyujo matches in your calculations. If kyujo matches were included as losses, both the Ozeki and Yokozuna win percentages would probably have a significant drop. Both ranks tend to drop out when a basho is going bad rather than post a 3-12 or 4-11 as the lower ranks will do. They have no incentive to salvage what they can to minimize their drop in rank.

The Traumatized Threesome have low win percentages with kyujo thrown in, and the twilight years for Taka, Big Moose and Little Moose were terrible.

Yup, that's right -- and might indeed be another reason for the high winning percentage of the Yokozunas. Fusen matches are included, kyujo is not.

[Asojima 2,874]

I assume that you are ignoring all kyujo matches in your calculations. If kyujo matches were included as losses, both the Ozeki and Yokozuna win percentages would probably have a significant drop. Both ranks tend to drop out when a basho is going bad rather than post a 3-12 or 4-11 as the lower ranks will do. They have no incentive to salvage what they can to minimize their drop in rank.

The Traumatized Threesome have low win percentages with kyujo thrown in, and the twilight years for Taka, Big Moose and Little Moose were terrible.

Yup, that's right -- and might indeed be another reason for the high winning percentage of the Yokozunas. Fusen matches are included, kyujo is not.

Might be interesting if you throw in " In for 1 win, in for 15 matches" logic in your calcs. (You are going off-topic...) It will eliminate those who bypass the basho and some of those who start as walking wounded, but keep some of those who drop out because they have nothing more to lose.

Edited August 13, 2006 by Asojima

[kaiguma 0]

J1	 1000   1894 0.528
J2	  994   1900 0.5232
J3	 1027   1965 0.5226
J4	  910   1846 0.493
J5	  926   1894 0.4889
J6	  952   1876 0.5075
J7	  907   1875 0.4837
J8	  919   1858 0.4946
J9	  972   1917 0.507
J10	 880   1916 0.4593
J11	 925   1901 0.4866
J12	 914   1876 0.4872
J13	 976   1913 0.5102
J14	 184	434 0.424

Turn out that the top actually does a bit better -- they are usually facing lower ranked guys, except the few bouts against makuuchi rikishi. The rest is rather inconclusive, just that bump at J13 looks a bit odd.

This table doesn't look strange to me at all. The guys from J4-J9 are most likely to have taken giant leaps after posting lowly 8-7's from lower/middling Juryo, since the others around them (even old veterans) are doing so badly. This kind of cycle is how mediocre or washed out rikishi can swim around in Juryo for years: one 8-7 or 9-6 can erase the banzuke drop of four consecutive 7-8's.

In contrast, J1-J3 probably deserve their placement after posting 9-6 or 10-5 from the middle, or they were just demoted from Makuuchi and desperately want back in.

J10 I can't explain so easily, probably just a random wall for rookies, but as a whole J10-J14 is composed mostly of rookies or very fresh sekitori with many of them being spit right back to Makushita in one or two basho. And then throw in those veterans that finally can't cut it anymore. They also either slide out as well, inch back up into the J4-J9 pool to recirculate for awhile longer, or drop down into that scary J13 and say "Now or never, do I want to stay Sekitori or drop into the abyss?" More often than not, that kind of motivation can snap any veteran out of a rut, such as Shunketsu in the last basho, perfect example! 10-5 from J13 (career high M12 - so maybe Shunketsu isn't exactly a Makuuchi veteran, but his sumo debut was '92).

Also of note, the lowest win % in Juryo this Nagoya was J10 followed by J7 (10 and 11 wins), which happen to be the lowest on your master list. Statistics in action! Thanks for the work you've put into this Oimeru, even though you might say the computer did it, you're at the helm (Showing respect...)

[ikishima 0]

I see winning records at the top of Juryo and losing records at the bottom of Maegashira. How do they do against each other (Nodding yes...) ?

This demonstrates the meat grinder effect perfectly, (Showing respect...) thanks (Holiday feeling...) . I wish I could find great examples of rikishi doing great at m7 or so and rocketing up to M1,2 or 3 and getting crushed

[Oimeru 0]

Okay, here are the fusenpai stats per rank:

Y		20
O		32
S		10
K		 5
M1		9
M2		7
M3		6
M4		5
M5		4
M6		7
M7		3
M8		5
M9		5
M10	   8
M11	   3
M12	   7
M13	  10
M14	   8
M15	   4
M16	   1
M17	   0
J1	   11
J2		5
J3		7
J4		1
J5		6
J6		3
J7		9
J8	   11
J9		3
J10	   6
J11	   7
J12	   3
J13	   5
J14	   2

Not counting fusenpai, the Yokozuna are up to 0.7975 and the Ozeki are at 0.657. In other words, the effect is rather small. The number of fusen losses for Ozeki does look a bit worrying, though, even considering that this is the most crowded rank.

[Yuriyama Ren 1]

Not counting fusenpai, the Yokozuna are up to 0.7975 and the Ozeki are at 0.657. In other words, the effect is rather small. The number of fusen losses for Ozeki does look a bit worrying, though, even considering that this is the most crowded rank.

I decided to tweak this little further and add percentage of fusenpai:

Rank	Fusenpai	Bouts	Percentage
Y	  20	1615	1.24
O	  32	3140	1.02
S	  10	2236	0.45
K	   5	2157	0.23
M1	  9	2010	0.45
M2	  7	1985	0.35
M3	  6	1935	0.31
M4	  5	2004	0.25
M5	  4	2016	0.20
M6	  7	1957	0.36
M7	  3	1996	0.15
M8	  5	1958	0.26
M9	  5	2007	0.25
M10	 8	1961	0.41
M11	 3	2027	0.15
M12	 7	1987	0.35
M13	10	1947	0.51
M14	 8	1931	0.41
M15	 4	1510	0.26
M16	 1	 611	0.16
M17	 0	 300	0.00
J1	 11	1894	0.58
J2	  5	1900	0.26
J3	  7	1965	0.36
J4	  1	1846	0.05
J5	  6	1894	0.32
J6	  3	1876	0.16
J7	  9	1875	0.48
J8	 11	1858	0.59
J9	  3	1917	0.16
J10	 6	1916	0.31
J11	 7	1901	0.37
J12	 3	1876	0.16
J13	 5	1913	0.26
J14	 2	 434	0.46

According to this yokozuna have most fusen losses at 1.24% followed by oozeki at 1.02%. After that it's complete chaos, or at least I can't see any trends here.

So another of my theories down the drain, the meatgrinder positions are not likely to get more injuries than say J8. (Showing respect...)

Edited August 14, 2006 by Yuriyama Ren

[mokele 3]

Great work. It appears that the designation "kun-roku ozeki" is hardly a severe criticism in light of the roughly 65% score that ozeki have averaged over the period in which your study covered. I suppose that the reason that the average score for ozeki seems a bit low is that ozeki become kadoban fairly often, maybe around 15 to 20% of the time, scoring fewer than 8 wins in a basho in the process but struggling through the whole 15 days. Yokozuna typically drop out of a basho when they have gotten off to a poor start and this saves them from some embarrassing results over a full 15 days. Long gone are the days of former yokozuna Chiyonoyama who scored only 6 wins in a complete basho once and asked the Sumo Kyokai to demote him to ozeki!

I was surprised to see that most of the maegashira ranks score less than 50%. I would have guessed that the middle to lower ranks would score a bit higher, but on 2nd thought it makes sense since some of the lower ranking maegashira get poor scores and are demoted to Juryo.

[Oimeru 0]

Great work. It appears that the designation "kun-roku ozeki" is hardly a severe criticism in light of the roughly 65% score that ozeki have averaged over the period in which your study covered. I suppose that the reason that the average score for ozeki seems a bit low is that ozeki become kadoban fairly often, maybe around 15 to 20% of the time, scoring fewer than 8 wins in a basho in the process but struggling through the whole 15 days. Yokozuna typically drop out of a basho when they have gotten off to a poor start and this saves them from some embarrassing results over a full 15 days. Long gone are the days of former yokozuna Chiyonoyama who scored only 6 wins in a complete basho once and asked the Sumo Kyokai to demote him to ozeki!

I was surprised to see that most of the maegashira ranks score less than 50%. I would have guessed that the middle to lower ranks would score a bit higher, but on 2nd thought it makes sense since some of the lower ranking maegashira get poor scores and are demoted to Juryo.

I was expecting a somewhat different distribution as well, but in the end it all seems to make sense, At least the "top maegashira bump" is there. Anyway, I still have ideas for stats that I can compute. Watch this space (or forum...) for more in the next weeks.