Top 50 rikishi since 1995, end-year ATP-style rankings, ATGs compared

[HenryK 38]

Here is some stuff for those who like getting lost in numbers... I've tried to come up with my own ozumo ranking. There are excellent ELO-type strength ratings already out there, hence I settled for something else (and simpler) -- and ATP-type ranking, summing up achievements over one year (6 bashos). To this end, I use the quality index by Masumi Abe, explained here. It has its shotcomings, but provides a benchmark.

The table below shows (1) the top number of quality points a rikishi collected over six subsequent bashos, startigh in 1995 (before the Abe-index is not pubslished), (2) the highest rank a rikishi held when rikishi are ordered by 6-basho quality points, (3) and the number of bashos a rikishi was ranked #1, in the top 3, and in the top 10. Note that many rikishis' careers are truncated (=the early part of the career is before the period captured in the table);, for some, the missing part includes their career peak (e.g., Kotonishiki). Another feature to be aware of is that missing a basho or going fusen results in a large malus. The latter affects in particular Akebono and Takanohana.

All Yokozuna were #1 at some point, but also Kaio, Tochiazuma, and -- for one basho -- Dejima. Hakuho is already leading the list, and wkith some margin ahead of Asashoryu, Akebono and Takanohana (as a sidenote, Takanohana gets treated a bit roughly by the Abe index, as he misses out on bonus points for beating an Ozeki due to the non-matches against his brother. Correcting for this would add roughly another 20 points). Harumfuji is already #6, ahead of Wakanohana. Kaio was ranked in the top 10 for 79 of the 82 bashos under consideration, and is by far the best Ozeki. Wakanosato is the top Sekiwake, peaking as high as #2, Tokitenku the best Komusubi, Homasho the best Maegashira.

#

Rikishi

Max points

Edited August 4, 2009 by HenryK

[HenryK 38]

End-year rankings, based on the same method. Again, please note that missing or not finishing bashos gets punished quite severely by the Abe index (as mssing slams gets punished by the ATP).

2002-2008 plus current ranking (post Nagoya 2009)

#	Current	End-08	End-07	End-06	End-05	End-04	End-03	End-02
1	Hakuho	Hakuho	Hakuho	Asashoryu	Asashoryu	Asashoryu	Asashoryu	Asashoryu
2	Harumafuji	Harumafuji	Chiyotaikai	Hakuho	Kotooshu	Kaio	Wakanosato	Chiyotaikai
3	Kotooshu	Kotomitsuki	Harumafuji	Miyabiyama	Tochiazuma	Wakanosato	Chiyotaikai	Musashimaru
4	Asashoryu	Kisenosato	Kotomitsuki	Tochiazuma	Hakuho	Chiyotaikai	Kaio	Wakanonsato
5	Kisenosato	Kotooshu	Tokitenku	Kotooshu	Kotomitsuki	Kotomitsuki	Tochinonada	Kaio
6	Kotomitsuki	Asashoryu	Kaio	Chiyotaikai	Chiyotaikai	Miyabiyama	Kyokotenho	Musoyama
7	Baruto	Toyonoshima	Asashoryu	Kotomitsuki	Miyabiyama	Hakuho	Tosanoumi	Tochiazuma
8	Goeido	Kaio	Kotooshu	Roho	Kyokotenho	Tochinonada	Takamisakari	Tosanoumi
9	Kaio	Kotoshogiku	Aminishiki	Kisenosato	Kaio	Tamanoshima	Tosanoumi	Kotomitsuki
10	Chiyotaikai	Aminishiki	Kotoshogiku	Kaio	Kokkai	Kyokotenho	Tamanoshima	Takanowaka

1995-2001

#	End-01	End-00	End-99	End-98	End-97	End-96	End-95
1	Musashimaru	Akebono	Musashimaru	Musashimaru	Takanohana	Takanohana	Takanohana
2	Tochiazuma	Kaio	Kaio	Wakanohana	Musashimaru	Takanonami	Akebono
3	Musoyama	Chiyotaikai	Dejima	Akebono	Akebono	Kaio	Musashimaru
4	Kotomitsuki	Miyabiyama	Tochiazuma	Takanohana	Takanonami	Musashimaru	Wakanohana
5	Kaio	Takanohana	Tosanoumi	Chiyotaikai	Tosanoumi	Wakanohana	Kaio
6	Wakanosato	Dejima	Chiyotaikai	Takanonami	Tochiazuma	Musoyama	Takanonami
7	Asashoryu	Musashimaru	Akinoshima	Musoyama	Wakanohana	Takatoriki	Musoyama
8	Takanohana	Musoyama	Kotonishiki	Kaio	Takatoriki	Kotonishiki	Kotonishiki
9	Tochinonada	Tochiazuma	Musoyama	Kotonowaka	Dejima	Akebono	Kenko
10	Tamanoshima	Tosanoumi	Takanonami	Kotonishiki	Musoyama	Ashaiyutaka	Kotonowaka

Edited August 3, 2009 by HenryK

[Barang 0]

Thanks for rankings. Somehow I still feel that it has some shortcomings. For example if I compare Asa and Hakuho careers then I feel that Asa has still better results on long term.

[HenryK 38]

Thanks for rankings. Somehow I still feel that it has some shortcomings. For example if I compare Asa and Hakuho careers then I feel that Asa has still better results on long term.

Sure it has shortcomings. In particular one has to be clear about what this index measures -- performance over 6 bashos relative to the other competing rikishi, whose strenght is approximated by their banzuke ranks.

In the case of Hakuho, the top rating reflects a run of 85 wins in 6 bashos (Nag 08-Nat 09), competing against 4 or 5 Ozeki and, in half of the bashos, another Yokozuna. By contrast, Asashoryu's peak run (Hat 05 - Kyu 05) contained 84 wins with 3 Ozeki and no Yokozuna opponent(s).

[HenryK 38]

Just checked what happens when one takes Takanohana's career-best 6 basho run, whis is a tad before the period in captured by the tables above. Takanohana maxes out at 548 (Kyu 94-Aki 95), which places him above Asashoryu and Akebono.

Similarly, Musashimaru collected 466 quality points between Natsu 94 and Haru 95, but this implies no ranking change.

I also checked alternative Akebono runs in the years 1992-94, but none matches his last grand hooray in 2000.

Edited August 3, 2009 by HenryK

[Toukeigakusha 0]

It is interesting to see for me how this ranking list has the same top 5 positions at the end of 2008 as my method listed here (which is based on the same 6 basho).

What's certainly better here than in my approach is that the quality index handles losses against opponents of different strength as well (which my model ignores, therefore e. g. Aminishiki is ranked 6th in my list but 10th here at the end of 2008).

OTOH, the value of a win in this model is based on the opponent's rank alone while my model tries to value the opponent based on both his rank and genkiness (i. e. wins during the current basho), thus downgrading the value of e. g. kinboshi against a kyuyo yokozuna or beating weak ozekis but upgrading wins against sansho winners etc. (in Kyusho 2008 my model values a win over Kisenosato/M04E/11 slightly higher than a win over Kotooshu/O2W/8, and actually Kisenosato himself even scored twice as many points in this basho as Kotooshu did according to my model by beating genki opponents such as Ama/13 (valued much like a kinboshi), Kotomitsuki/9, Baruto/9, Toyonoshima/9, Chiyotaikai/8 and... Kotooshu/8).

Another slight difference is that this model weighs all 6 bashos equally while my model weighs them with factors of 5-6-7-8-9-10, to give the most recent basho twice the influence of the one to be replaced next by the forthcoming basho's result and get closer to a situation's snapshot than to a year's ranking. (Thus in my ranking both Asashoryu and Kaio dropped out of the Top 10 at the end of 2008 because of missing Kyusho 2008 whereas Baruto is ranked 7th in my list at the end of 2008 already.)

[salle 11]

More than impressive!

Just couple of weeks ago during last basho I was thinking about some similar ranking based on same Masumi Abe quality index. Unlike you I didn't go any step further so I can only admire you for your excellent effort! (I am not worthy...)

Few comments on your "unfair" findings. No matter what ranking system you define it will always feel unfair in some aspects. Is it fair to penalize injured athlete for not participating? Yes it's unfair to the injured, but it would be even more unfair to the healthy ones to not do it. If we take into account athletes who are injured, but stay in the game despite the pain and risk to their health it becomes even more evident.

As for the Yokozuna/Ozeki not meeting his heya mate I think it's absolutely fair to let things as is and not tweak the results as you suggest could be done for Takanohana. Why so? Because if there was no rule for not allowing such matches the history of lot of rikishi could be rewritten rather dramatically. It also applies for the number of Ozeki and Yokozuna at every basho. If there are 4 healthy and strong Yokozuna at banzuke the theoretic max quality index is much higher than with single Yokozuna.

That's why it's always better to avoid "What if's?" and stick with simple ranking.

Another "unfair" thing is with annual rankings. With only 6 basho per year it doesn't feel long enough "season". (At the other hand it's full year after all). Perhaps 3 or 5 years would give better overall look in long term and will also allow comparing already retired rikishi to the current. Or take fixed number of bashos in the past. Last 20 for example.

Another interesting ranking to make would be average quality_index per Makuuchi career. SUM(quality_index)/number_of_bashos. Average is bad statisitc value indeed, but it does give some idea about career most specifically persistence.

Thanks a lot for taking your time to do your ranking! :-|

[HenryK 38]

Top scores for some legends in comparison:

(all Y with 10+ yushos in the modern era)

Hakuho	Nagoya 08 - Natsu 09	636
Kitanoumi	Aki 77 - Nagoya 78	633
Taiho	Haru 66 - Hatsu 67	624
Chiyonofuji	Natsu 87 - Haru 88	611
Takanohana K.	Kyushu 94 - Aki 95	548
Asashoryu	Hatsu 05 - Kyushu 05	545
Kitanofuji	Aki 69 - Nagoya 70	538
Wajima	Haru 73 - Hatsu 74	532
Tochinishiki	Haru 59 - Hatsu 60	529
Akebono	Hatsu 00 - Kyushu 00	528
Wakanohana K.	Nagoya 58 - Natsu 59	478
Musashimaru	Natsu 94 - Haru 95	466

A bit stunned where Hakuho comes out (I am not worthy...)

(in particular since otherwise the list doesn't look unreasonable)

As a sidenote, a Yokozuna winning 6 subsequent basho 15-0, beating each time 1 Y, 4 O, 2 S, 2 K and 6 M1-4 would collect 714 quality points.

Edited August 4, 2009 by HenryK

[HenryK 38]

@ Toukeigakusha, Salle:

thanks for the friendly words.

Accounting for genkiness is surely an excellent idea (for example, by replacing the Abe index with one derived from the pre-basho points rating itself); it would get the exercise closer to ELO-type strenght ratings a la Doitsuyama or Randomitsuki. But since I try to mimick an ATP-type approach, I feel this is not needed here. Winning Wimbledon always yields the same ATP-points, no matter how well the opponents on the way to the title had performed recently.

Edited August 3, 2009 by HenryK

[Yaochoyama 1]

My two cents - while a simpler ATP-type index is interesting and with all due respect Masumi Abe, the Quality Index was not designed for this type of analysis and it's not a good choice. To cite just one example, are we really to believe that Tokitenku at #5 was a superior rikishi to Asashoryu at #7 in 2007? Tokitenku bounced between komusubi and M3 with a 45-45 record for the year while Asashoryu was 51-9-30 with two yushos at 14-1 and a jun-yusho at 13-2.

You can talk theoretically about what should or shouldn't be done, but in the end it gets down to numbers and weights. For purposes of this type of analysis, the Quality Index penalizes absenses far too heavily and doesn't strike a proper balance between record and rank. The variance in scores for any given basho is far too high.

My suggestions would be to find a better way to score absences and look at how the banzuke is constructed to get the proper balance between record and rank. Maybe rescale things so you don't deal with negative numbers. Get a feel for how much extra to value a good vs. bad performance. For example, how much more should the performance by a sanyaku who goes 11-4 be worth than a low juryo who goes 4-11? Five times a much? Ten times? What about those halfway between them? Decide on a perspective for what you're trying to measure.

[HenryK 38]

My two cents - while a simpler ATP-type index is interesting and with all due respect Masumi Abe, the Quality Index was not designed for this type of analysis and it's not a good choice. To cite just one example, are we really to believe that Tokitenku at #5 was a superior rikishi to Asashoryu at #7 in 2007?

No, but this is not what the rating intends to capture. Asashoryu missed the two last bashos in 2007, hence his ranking dropped from #1 to #7. A tennis player who sits out both Wimbledon and the US Open won't be ranked #1 either.

Btw, if one penalizes Asashoryu with 0 for amissed basho rather than -45, he is #5 in 2007. If one assumes he would have been as strong in the missed bashos as he was in the other 4, he is #2.

You can talk theoretically about what should or shouldn't be done, but in the end it gets down to numbers and weights. For purposes of this type of analysis, the Quality Index (i) penalizes absenses far too heavily and (ii) doesn't strike a proper balance between record and rank.

I agree with (i), but unless you aim at an intrinsic strenght rather than an achievements rating, absences will always harm a rikishi's ranking, even if they come with a lower penatly. I do not agree with (ii), as this is exactly what the Quality Index attempts to address.

Edited August 3, 2009 by HenryK

[Toukeigakusha 0]

I agree with HenryK's comment on the comparison between Asashoryu and Tokitenku: Such ranking lists certainly don't give an idea of strength - but they give an idea of achievement over a period (just as the mere number of wins does).

And given there are 15*6 = 90 bouts within a year I don't think a calender year is too short a period for this purpose (how many matches does a tennis player have within one year?).

As for comparing Hakuho to Asashoryu: I also feel that Asashoryu has achieved more during his career than Hakuho has (until now). But the highest score of HenryK's system isn't a career overview, it is a high-score of temporary dominance in one particular time interval. And Hakuho may in fact be one of the most dominant yokozunae right now, given his high number of wins despite a competition of another yokozuna and five (more or less competitive) ozekis. (Then again, valuing the #7 ranked sekitori as ozeki now while this has been a sekiwake or maybe even a komosubi in Asashoryu's best days might be a weakness of Abe's model - Hakuho profits from a lot of wins over weak ozekis as there are so many of them nowadays.)

Accounting for genkiness is surely an excellent idea (for example, by replacing the Abe index with one derived from the pre-basho points rating itself)

I'm not going for a more precise measurement for quality (I trust the pre-basho information of the banzuke almost as much as Masami Abe and you do, but see below), I'm going for genkiness at the time of the bout. It's may still be oversimplifying but a lot better than ignoring the condition of the athlete in general, most notably when valuing wins against sekitori who're already injured and drop out of the tournament later.

Which is why a fusen win over a high-ranked sekitori is no problem for me as it mostly isn't worth that much anyway given the low genkiness of said athlete in this particular basho - his previous basho level wouldn't help me in this aspect. In Masami Abe's model a fusen win over an ozeki may easily be the greatest achievement of a sekitori in this basho... The genkiness element also values wins against highly overpromoted sekitori (who end up 2:13 or the like) rather low instead of simply trusting the banzuke in these cases.

As for the banzuke ranks: Was a victory over Asa in his ozeki run, i. e. when he was still a sekiwake but went 13:2, worth less than a victory over a nervous Harumafuji ozeki who went 8:7 one basho later, just because he didn't have the higher rank already? I don't think so, therefore I ignore the banzuke titles and go by banzuke positions, i. e. value Harumafuji S1E as high as Harumafuji O3E one basho later (both as position 7 with 44 points of 50 possible). I can afford to map banzuke ranks to a linear score because usually yokozunae win a lot more bouts than ozeki, i. e. the genkiness element handles this aspect as well (beating Hakuho/14:1 is normally worth about twice as much as beating Kaio/8:7 in my model).

[Asashosakari 19,038]

As for the banzuke ranks: Was a victory over Asa in his ozeki run, i. e. when he was still a sekiwake but went 13:2, worth less than a victory over a nervous Harumafuji ozeki who went 8:7 one basho later, just because he didn't have the higher rank already? I don't think so, therefore I ignore the banzuke titles and go by banzuke positions, i. e. value Harumafuji S1E as high as Harumafuji O3E one basho later (both as position 7 with 44 points of 50 possible).

I'd be wary of doing that. (In general, I mean, not in Ama's case specifically.) It's just a fact that the very best skill levels aren't distributed evenly over time...back in Asashoryu's dominant days, there was arguably the usual good/mediocre ozeki gap (with Kaio/Taikai/Azuma on one side, Musoyama on the other while he was stlll active), and another sizeable gap below sekiwake (with, roughly, Musoyama/Wakanosato/Kotomitsuki above, everybody else significantly below, perhaps except Kyokutenho). Right now it's quite different, with more ozekis on the mediocrity side, and a much more evenly distributed group of lower sanyaku-strength rikishi but not really any "strong sekiwake" type of rikishi since Ama's promotion. Baruto's close, but too inconsistent yet.

Anyway, what I'm trying to say is, while the banzuke ranks may not accurately reflect the individual skill levels, I don't think treating the top ranks as simply 1 through ~20 is an improvement. Just look at any banzuke from 1992 to see where that clearly breaks down. If anything, your genkiness adjustment ought to capture most divergences between actual and "justified" rank already, so you'd simply end up double-counting those differences.

Edited August 4, 2009 by Asashosakari

[HenryK 38]

As for the banzuke ranks: Was a victory over Asa in his ozeki run, i. e. when he was still a sekiwake but went 13:2, worth less than a victory over a nervous Harumafuji ozeki who went 8:7 one basho later, just because he didn't have the higher rank already? I don't think so, therefore I ignore the banzuke titles and go by banzuke positions, i. e. value Harumafuji S1E as high as Harumafuji O3E one basho later (both as position 7 with 44 points of 50 possible).

I'd be wary of doing that. (In general, I mean, not in Ama's case specifically.) It's just a fact that the very best skill levels aren't distributed evenly over time...back in Asashoryu's dominant days, there was arguably the usual good/mediocre ozeki gap (with Kaio/Taikai/Azuma on one side, Musoyama on the other while he was stlll active), and another sizeable gap below sekiwake (with, roughly, Musoyama/Wakanosato/Kotomitsuki above, everybody else significantly below, perhaps except Kyokutenho). Right now it's quite different, with more ozekis on the mediocrity side, and a much more evenly distributed group of lower sanyaku-strength rikishi but not really any "strong sekiwake" type of rikishi since Ama's promotion. Baruto's close, but too inconsistent yet.

Anyway, what I'm trying to say is, while the banzuke ranks may not accurately reflect the individual skill levels, I don't think treating the top ranks as simply 1 through ~20 is an improvement. Just look at any banzuke from 1992 to see where that clearly breaks down. If anything, your genkiness adjustment ought to capture most divergences between actual and "justified" rank already, so you'd simply end up double-counting those differences.

I agree. More generally, if one is after intrinsic strenght, there are at least two elements of this rating that need to be modified:

-- missed bouts or bashos must not have a malus attached to them, but a counterfactual value simulating how the rikishi would have performed had he fought (based on recent performances), and

-- the banzuke can't be basis for assessing opponents' strenght. Rather, it ought to be the strenght rating itself. If the opponents' strength is recomputed after each bout (rather than each basho), this would also capture genkiness during a basho.

But this goes beyond of what I intended to capture, and simlar things have already be done elsewhere. I just wish they would be publicly accessible...

Edited August 4, 2009 by HenryK

[Jejima 1,359]

I would suggest that one of the reasons that Hakuho's score looks so good compared to others is due to the fact that he gets to face 5 Ozeki every basho - each worth oodles of points, 2 (at least) of whom have been practically 'freebies' since he reached his high-point. Throw in a Yokozuna, whom he has had the edge over for some while now (due to injuries and other distractions), and there are some more big numbers available. Also, Hakuho doesn't 'lose' potential points by dipping further down the banzuke to avoid meeting his fellow heya-mates.

Perhaps (if you want to use this system), it would be better / clearer to look at the points garnered compared to potential points available. Takanohana (say) during a period of his peak only had one Yokozuna and one Ozeki to fight against (versus Hakuho's one and five), and often lost out on potential points for sekiwake and/or komusubi due to Takatoriki and Akinoshima being at those ranks. Therefore there were probably several basho when he had no chance of achieving the scores that Hakuho can get using this scoring system, even if he went 15-0!

Edited August 4, 2009 by Jejima

[HenryK 38]

I would suggest that one of the reasons that Hakuho's score looks so good compared to others is due to the fact that he gets to face 5 Ozeki every basho - each worth oodles of points, 2 (at least) of whom have been practically 'freebies' since he reached his high-point. Throw in a Yokozuna, whom he has had the edge over for some while now (due to injuries and other distractions), and there are some more big numbers available. Also, Hakuho doesn't 'lose' potential points by dipping further down the banzuke to avoid meeting his fellow heya-mates.

Perhaps (if you want to use this system), it would be better / clearer to look at the points garnered compared to potential points available. Takanohana (say) during a period of his peak only had one Yokozuna and one Ozeki to fight against (versus Hakuho's one and five), and often lost out on potential points for sekiwake and/or komusubi due to Takatoriki and Akinoshima being at those ranks. Therefore there were probably several basho when he had no chance of achieving the scores that Hakuho can get using this scoring system, even if he went 15-0!

Good point. Agreed.

[Toukeigakusha 0]

-- the banzuke can't be basis for assessing opponents' strenght. Rather, it ought to be the strenght rating itself. If the opponents' strength is recomputed after each bout (rather than each basho), this would also capture genkiness during a basho.

IMHO incremental strength correction during a basho is much less precise than using the total number of wins in the current basho as genkiness for the whole basho (thus losing the ability to use this mechanism for predicting bouts, which an ELO-based approach would handle better anyway): Overpromotions don't happen incrementally, and even a large part of injuries that influence genkiness have been caused before the basho already.

Takanohana (say) during a period of his peak only had one Yokozuna and one Ozeki to fight against (versus Hakuho's one and five)

That's somehow the line of thought that led me to use linear values for positions instead of banzuke ranks, and then introduce the genkiness part to reward a win over Hakuho/14 significantly higher than a win over Chiyotaikai/8.

[HenryK 38]

Perhaps (if you want to use this system), it would be better / clearer to look at the points garnered compared to potential points available.

Here is a table that follows your suggestion, encompassing the best 6-basho runs of all rikishi who won 5 bashos or more since the beginning of the 6-basho-per-year era. These are 19 Yokozuna and Mr. Kaio.

Rank (perc.)

Rank (points)

Rikishi

From

Edited August 6, 2009 by HenryK

[Doitsuyama 1,178]

The rikishi fall quite naturally into 3 groups:

(i) dominant Yokozuna, from Hakuho (90.9 percent) to Takanohana (85.7 percent). This group has all the usual suspects, plus -- as perhaps a small surprise -- Tamanoumi. Stunningly, Hakuho is leading also this compilation.

(ii) strong Yokozuna, from Tochinishiki (81.9 percent) to Musashimaru (73.2 percent); and

(iii) marginal Yokozuna, from Hokotoumi (69.7 percent) to Ashahio (59.1). Kaio, as much as his achievements are oustanding for an Ozeki, would have been a marginal Yokozuna only.

Tamanoumi is not surprising at all if you have read my strength rating analysis. And look who has the best absolut point value over six basho...

I would be interested where Mienoumi is in that list, I'd guess better than the "marginal yokozuna" (which is not a good designation I think - how to call the rest with less than five yusho?).

[Sokkenaiyama 80]

Sandomayama

Hokotoumi

Ashahio

I can't find any of these guys in the DB. You mean Sadanoyama, Hokutoumi and Asashio, right?

Edited August 6, 2009 by Sokkenaiyama

[HenryK 38]

Tamanoumi is not surprising at all if you have read my strength rating analysis. And look who has the best absolut point value over six basho...

I would be interested where Mienoumi is in that list, I'd guess better than the "marginal yokozuna" (which is not a good designation I think - how to call the rest with less than five yusho?).

Mienoumi is included below. As you predict, he is in the "strong" group. I'd be grateful for hints for other Y/O to look at.

As for your strenght ratings -- I'd love to delve into them if I could. The table below mimicks strenght ratings to some degree, but imperfectly so. Specifically, the total points rating gives arguably too much weight to "strenght of competition" (penalizing Tokonohana, for example, for having Ozeki stablemates who he couldn't fight), while the percentage rating eliminates strenght of competition almost entirely from the equation 1/, which seems inadequate either (pretending Asashorhyu in 05 fought virtually the same quality of opposition as Kitanoumi in 78).

One can of course combine the two, for example by multiplying the two scores (and I've added a column to that effect), but this is arbitrary. And anyway, the percentage or mixed statistics make sense only for rikishi in sanyaku or jin-join for 6 consecutive bashos.

For intrinsic strenght there is nothing better imo than ELO ratings. I for my part would very much enjoy if your full ratings were again publicly acccessible (updated or not), and refrain from reeinventing the wheel. But I am of course aware that it is your ratings and your time we are talking about.

1/ Not totally though -- loss to a Yokozuna, for example, reduces the points score by 13, which damages the percentage score by more if total available basho points are 100 as opposed to 120.

Rank (perc.)

Rank (points)

Rank (mixed)

Rikishi

From

Edited August 6, 2009 by HenryK

[Doitsuyama 1,178]

Mienoumi is included below. As you predict, he is in the "strong" group. I'd be grateful for hints for other Y/O to look at.

Hmm, my wish list would include Takanosato and Asahifuji among the yokozuna. There really aren't any ozeki clearly breaking into the "strong" group I guess but I'd still be interested in Wakashimazu, Kiyokuni and Konishiki.

As for your strenght ratings -- I'd love to delve into them if I could.

I don't really understand you because I suspect you already now of this thread which covers some sort of historical strength analysis. While a future version will have slightly changed values because I now also have all juryo bouts, the changes probably wouldn't be big with the strongest rikishi so have a look into that thread if you haven't already.

[HenryK 38]

Mienoumi is included below. As you predict, he is in the "strong" group. I'd be grateful for hints for other Y/O to look at.

Hmm, my wish list would include Takanosato and Asahifuji among the yokozuna. There really aren't any ozeki clearly breaking into the "strong" group I guess but I'd still be interested in Wakashimazu, Kiyokuni and Konishiki.

I will work off your strenght ranking stats and complement my list in the coming days.

As for your strenght ratings -- I'd love to delve into them if I could.

I don't really understand you because I suspect you already now of this thread which covers some sort of historical strength analysis. While a future version will have slightly changed values because I now also have all juryo bouts, the changes probably wouldn't be big with the strongest rikishi so have a look into that thread if you haven't already.

Well, I knew top 20 (I believe) strenght rating that you used to update after each tournament, but that became inaccessible some time ago. The old thread on historic ELO scores I discovered only last night.

The ELO ratings provide a piece of information that, imo, is critical but missing from standard rikishi career stats. My ideal would be if your database could an additional column to each sekitori with his ELO score after each basho.

Edited August 6, 2009 by HenryK

[HenryK 38]

Here is a more complete list, encompassing all Yokozuna and a few Ozeki since 1960. Commentary etc. tomorrow

[btw, Konokuni does not do well with this ranking (50.7 percent), there is no particularly noteworthy 6-basho run. Same seems to hold for Takanohana I.]

Rank (perc.)	Rank (points)	Rank (mixed)	Rikishi	6-basho run from...	...to	Points	Avail. Points	Percen- tage	Mixed score
Dominant Yokozuna (percentage>85)
1	2	1	Hakuho	08/7	09/5	636	699	91.0	579
2	3	2	Kitanoumi	77/9	78/7	633	706	89.7	568
3	7	6	Asashoryu	05/01	05/11	545	612	89.1	485
4	4	4	Taiho	66/3	67/1	624	701	89.0	555
5	1	3	Tamanoumi	70/9	71/7	640	729	87.8	562
6	5	5	Chiyonofuji	88/5	89/3	611	699	87.4	534
7	16	7	Takanohana II	94/11	95/9	515	601	85.7	441
Strong Yokozuna (85>percentage>70)
8	12	9	Tochinishiki	59/3	60/1	529	646	81.9	433
9	8	8	Takanosato	83/3	84/1	542	669	81.0	439
10	9	10	Kitanofuji	69/9	70/7	538	685	78.5	423
11	18	17	Wakanohana I	58/9	59/7	478	611	78.2	373
12	11	12	Wajima	73/3	74/1	532	690	77.1	410
13	12	13	Wakanohana II	78/1	78/11	529	689	76.8	406
14	6	11	Mienoumi	79/3	80/1	547	717	76.3	417
15	14	14	Akebono	00/1	00/11	528	708	74.6	394
16	15	16	Sadanoyama	64/9	65/7	522	703	74.3	388
17	19	18	Musashimaru	94/5	95/3	468	639	73.2	343
18	10	15	Asahifuji	88/7	89/5	536	735	72.9	391
Other Yokozuna (percentage
19	24	21	Hokutoumi	87/5	89/3	420	603	69.7	293
20	17	19	Kashiwado	66/9	67/7	495	717	69.0	342
21	21	20	Kotozakura	72/11	73/9	458	682	67.2	308
22	22	23	Tochinoumi	63/9	64/7	435	672	64.7	282
23	20	22	Onokuni	86/11	87/9	461	736	62.6	289
24	23	24	Futohaguro	85/9	86/7	424	698	60.7	258
25	26	26	Wakanohana III	96/3	97/1	374	620	60.3	226
26	25	25	Asashio	58/7	59/5	387	655	59.1	229
For Comparison: Outstanding Ozeki
-	-	-	Konishiki	91/5	92/3	415	601	69.1	287
-	-	-	Wakashimazu	83/9	85/7	463	690	67.1	311
-	-	-	Kaio	00/5	01/3	473	746	63.4	300
-	-	-	Harumafuji	08/7	09/5	432	723	59.8	258

Edited August 7, 2009 by HenryK

[ryafuji 811]

Sorry to be so pedantic but it's not "ELO", it's "Elo". It doesn't stand for anything, it's named after the inventor of the system, Arpad Elo.