Some yaocho analyses

[Randomitsuki]

A sumo aficionado's metabolism is on a low point during this season. Some of us are even said to lead a normal life right now. However, I might have found a way to make your blood pressure boil up.

Can you spell Y

Edited April 19, 2005 by Randomitsuki

[Jakusotsu]

Thank You very much! (Ranting...)

After the hard-crafted numeric dust settles, nothing is proven, yet nothing rebuted - the mystery remains.

[Ryukaze]

Too much to read lol but I gotta give ya props for all that, an interesting theory but it seems like the most obvious sign of Yaocho to me would be when a rikishi appears to "give" so to speak a match which I dont see too often. Im sure it has happened before but to think it is a more than common occurance would be so to speak "unlikely".

[Asashosakari]

Nice work! (Sign of approval) Couple of nitpicks...

Step 1: Define a surprising result.

A) It might be a little surprising if a rikishi with less wins during the basho defeats a rikishi with more wins during the same basho.

B) It might also be a little surprising if a lower-ranked rikishi defeats a higher-ranked rikishi.

It might be pretty much more surprising if A) and B) appear in combination, i.e. a lower-ranked rikishi with less wins defeats a higher-ranked rikishi with more wins. It could happen, it will happen, it does happen, but maybe not very often. Voila, this is the sample I was looking at.

I disagree that A or B alone can be defined as surprising. Hypothetical situation: Jumonji and Asasekiryu (to pick two mid-Maegashira at random) both have a "true" strength equivalent to rank M8, and each of them is expected to beat the other 50% of the time. However, due to banzuke dynamics, a given basho sees Jumonji at M6 and Asasekiryu at M10; we can expect that when they meet, the higher-ranked Jumonji is likely to have fewer wins than Asasekiryu. Result: The result of their bout is indistinguishable from a coin-toss, but no matter who wins, it will be a surprising result (surprise A if Jumonji wins, surprise B if Asasekiryu does).

Surprisingly (pun intended), I do think the combination of A and B provides a good measure of "surprise", but for a slightly different reason. IMHO, the proper measure of surprising results would be a consistently lower-ranked rikishi beating a consistently higher-ranked one (no matter how many wins they have), e.g. Jumonji beating Kyokutenho. But your combination of "lower rank" with "lower number of wins" provides a nice proxy for that. (If Jumonji and Kyokutenho are close enough in rank to be fighting, Jumonji is likely ranked too high for his abilities and Kyokutenho too low, so we would expect Jumonji to have fewer wins, exactly the situation in which A + B applies - so a Jumonji win would be surprising.)

Still, one wonders how good the proxy is...to keep up with the hypothetical, if Kyokutenho had a really bad basho and drops to M8, while Jumonji is M6 at the same time, a Jumonji win would suddenly be non-surprising (or rather, would only fulfill criterion A but not B), even though in reality it still would be a surprise.

Edit: Actually, now I wonder if Doitsuyama's rikishi ratings could be used towards this purpose...

I would have no idea why a yaocho should appear in category 1.

Well, that's the Itai argument - that basically all bouts are rigged. (Ranting...) If many of the wins aren't bought but, as he alleged, rather "traded" (today Rikishi X wins, next time Rikishi Y wins), it wouldn't matter what their previous record for the basho is, yaocho could very well exist even before Day 8. (Your third test doesn't seem to indicate anything like that, though.) But of course, trying to prove statistically that almost all bouts are rigged should be well-nigh impossible, since you wouldn't have any unbiased data to compare to, if that was the case...

Edit #2: Before anyone nitpicks me, yes, I realize that Asasekiryu leads Jumonji 5-1 in their matches. (Ranting...) Just a hypothetical above, as indicated.

Edited April 20, 2005 by Asashosakari

[Jejima]

1. Bouts including sanyaku, and those with chances of a special prize, yusho or jun-yusho should be excluded from stats.

2. The glory era of yaocho is not now, how about doing stats for a year in Itai's time?

3. Is one year long enough for true statistical evidence?

4. Bouts inculding Juryo rikishi with a chance of making Makunouchi, should be excluded. (Actually, I'd exclude all Juryo rikishi)

5. Most rikishi (these days) are gachinko, however, I have suspicions about others...

6. I would wonder what the stats would show for the rikishi who have been Makunouchi regulars for over 9 basho, who have a low win-record (eg, below 0.450), as these would be the ones that I'd expect of being banzuke manipulators (Ranting...)

7. I was going to name a suspect or two, but have changed my mind (Ranting...)

[Randomitsuki]

1.  Bouts including sanyaku, and those with chances of a special prize, yusho or jun-yusho should be excluded from stats.

I assume you suggest this because those bouts would be highly suspicious? Actually this is why I introduced Category 2 (KK vs. KK) where most of these bouts should fall. It could be interpreted as a sign of yaocho if the percentage of upsets is too low, but this does not seem to be the case

Or would you suggest that those bouts are beyond doubt?

2.  The glory era of yaocho is not now, how about doing stats for a year in Itai's time?

Nice idea, I will do it some day.

3.  Is one year long enough for true statistical evidence?

There is no true evidence. The sample wasn't huge, but quite sufficient.

4.  Bouts inculding Juryo rikishi with a chance of making Makunouchi, should be excluded.  (Actually, I'd exclude all Juryo rikishi)

Why that?

5.  Most rikishi (these days) are gachinko, however, I have suspicions about others...

6.  I would wonder what the stats would show for the rikishi who have been Makunouchi regulars for over 9 basho, who have a low win-record (eg, below 0.450), as these would be the ones that I'd expect of being banzuke manipulators (Ranting...)

Sounds interesting, but those bouts are difficult to spot on a manual count. Moreover, the sample size might be too small. Just for interest: are there electronic sources where I can perform searches on torikumi? It would make yaocho analyses much easier. But then again: why should I look into the details in order to find a phenomenon that I wasn't able to spot on a large scale? If there were two or three yaocho rikishi, I wouldn't care that much.

7. I was going to name a suspect or two, but have changed my mind (Ranting...)

<{POST_SNAPBACK}>

[Burainoan]

Actually, the rikishi i tend to be the most suspicious of in a given basho are the ones who start off hot, win 7 or 8 in the first 7-9 days, and then promptly go on a skid, ending with an 8-7 or 9-6, when they should've easily been good for a couple more than that. i tend to chalk it up to laziness, but a little part of me gets suspicious once in a while. it seems especially common with rikishi who are ranked too low on the banzuke. i'm too lazy to go dig up records, but i seem to recall Kotoinazuma having done that numerous times, and of our current batch, Shuzan. (in the latter's case, i tend not to think yaocho, but rather to agree with the theory posted elsewhere that he was angered by the racist treatment he got coming up, and just decided to coast the rest of the way until he's ready to retire.)

[Asashosakari]

6.  I would wonder what the stats would show for the rikishi who have been Makunouchi regulars for over 9 basho, who have a low win-record (eg, below 0.450), as these would be the ones that I'd expect of being banzuke manipulators (Ranting...)

<{POST_SNAPBACK}>

It probably bears mentioning that there's exactly ONE rikishi right now who fulfills those criteria, namely Kyokushuzan. Few rikishi with any claim to being Makuuchi regulars go below 0.470 (*) until they reach the end of their time as top-division rikishi, so basically everyone who's not a sanyaku regular hovers between 0.470 and 0.500, which really doesn't mean anything since the difference only amounts to around 2 or 3 more losses per year. That's certainly not enough to detect any intentional banzuke riding since somebody has to lose those bouts, or the sanyaku couldn't consistently win above 0.500.

* Jumonji is below 0.470, however because of an extended period of elevatoring between Makuuchi and Juryo at the beginning of his career when he scored only one KK in eight basho, racking up a 48-72 record (0.400).

[Jejima]

It probably bears mentioning that there's exactly ONE rikishi right now who fulfills those criteria, namely Kyokushuzan.

That's interesting. When I used to calculate the stats on all Makunouchi rikishi when I used to get the 'Sumo World', I recall there were a few rikishi who were regularly under the 0.450 mark - but there was also more rumours of yaocho back then... ;-)

[Randomitsuki]

Edit: Actually, now I wonder if Doitsuyama's rikishi ratings could be used towards this purpose...

<{POST_SNAPBACK}>

I really love these Elo-based ratings by Doitsuyama. I even adapted them for sumo gaming purposes. However, they are not always good for predicting bout outcomes. For instance, when Hakuho or Kotooshu started in Juryo they had a rating of 1800. They've consistently beaten rikishi with higher ratings, as is simply the case with upstarts. If the yaocho analysis were based on those Elo ratings alone, however, most of their bouts would have counted as surprise results, and therefore as potential yaocho candidates.

Apart from that, I agree with the points you raised.

[Kashunowaka]

Just for interest: are there electronic sources where I can perform searches on torikumi?

Not yet.

[Odoriou]

If yaocho is really so customary in Ozumo - why don't the retired Rikishi speak of it openly? (Just do not get it...) I mean especially those guys, who afterwards have no connection to Sumo anymore (like Akebono, for instance) and who even may be angry with their former companions because of some mysterious reason. (Jumping on the other guy...)

[Zentoryu]

If yaocho is really so customary in Ozumo - why don't the retired Rikishi speak of it openly?  (Just do not get it...) I mean especially those guys, who afterwards have no connection to Sumo anymore (like Akebono, for instance) and who even may be angry with their former companions because of some mysterious reason. (Jumping on the other guy...)

<{POST_SNAPBACK}>

Yaocho (and Ozumo in general) is a fraternity type of thing. Usually a mutual agreement between two rikishi. If one were to admit to it after retiring, even without having any remaining connections to Ozumo, he is not only exposing the rikishi who participated, but he is also exposing himself.

Gachinko rikishi would likely not expose their former opponents for fear of the rather obvious potential consequences to themselves and their families, not to mention the consequences to the sport itself. This is a very prickly subject in Sumo circles with the potential to be even more devastating to Ozumo's public image than steroid usage has been in other sports.

I'm speaking specifically of the type of Yaocho Itai mentioned in his accusations, not the less innocuous "7-7 rikishi gets his KK by beating 8-6 rikishi on the final day" type of Yaocho.

Akebono would never admit that Yaocho existed because he knows which side of the bread the butter is on. He could never make the money he's making off his sumo reputation if the general public knew that he aquired that reputation in a sport riddled by yaocho. Whether Akebono where gachinko or not would be of no consequence in the publics mind. Anger over a fellow competitor commiting yaocho would not be enough to get him to expose the sport and destroy his cash cow in the process.

For the most part, don't ever expect retired rikishi, with or without connections to Ozumo, to ever admit that Yaocho exists. Itai is one of the only former rikishi that I know of to come out and alledge that it existed and implicate others for participating in it. Most of the accusations will always come from the outside, either from observers and conspiracy theorists like ourselves, or from people with supposed "inside" connections. It will almost never be from a former rikishi.

Edited April 20, 2005 by Zentoryu

[Odoriou]

(An interesting topic, by the way. And an interesting analysis by Randomitsuki.) (Showing respect...)

Now I searched for a while and found a lot of threads concerning Yaocho... But everytime Itai ;-) is involved in some way:

http://www.sumoforum.net/forums/index.php?showtopic=2278

http://www.sumoforum.net/forums/index.php?showtopic=243

http://www.sumoforum.net/forums/index.php?showtopic=1997

http://www.sumoforum.net/forums/index.php?showtopic=2667

One single Ex-Rikishi speaking of Yaocho in Ozumo is simply not enough to convince me. (This is no good...)

[Sasanishiki]

I don't think you should exclude sanyaku, what about Ozeki who are kadoban or those who have dropped to sekiwake and are trying to get back with 10 wins? These matches would be prime candiadtes for the odd yaocho in return for a favour later on when they are out of a yusho race in a future basho.

There is a grey area with the Juryo who are a chance of being promoted. They may already have done enough to be promoted but are being matched against lower-ranked maegashira in threat of being demoted. These matches may also have yaocho involved, because the promotion could be secured and the demotion might still be averted.

I think these statistical studies are interesting as a whole, but they need to be contextualised more thoroughly in terms of who was competing and just exactly what was at stake for each competitor.

Also, just what are we considering as yaocho? Is it the exchange of money or an explicit understanding that one rikishi agrees to through the match before they begin? For example, a rikishi could take a dive without explicit prior arrangement but the other rikishi would be able to sense that he had been "helped out". They could then read the situation at a later date and drop a match as 'ongaeshi', or returning the favour. None of this need be stated, arranged or for direct monetary gain, but could certainly still occur (and probably does) in the course of a rikishi's career.

[Asashosakari]

I really love these Elo-based ratings by Doitsuyama. I even adapted them for sumo gaming purposes. However, they are not always good for predicting bout outcomes. For instance, when Hakuho or Kotooshu started in Juryo they had a rating of 1800. They've consistently beaten rikishi with higher ratings, as is simply the case with upstarts. If the yaocho analysis were based on those Elo ratings alone, however, most of their bouts would have counted as surprise results, and therefore as potential yaocho candidates.

<{POST_SNAPBACK}>

That could probably be rectified just by excluding all bouts that involve a rikishi who is in his first X basho as Maegashira. I don't think that would materially affect the conclusion: Even if we assume there is yaocho, it would be unlikely that upstart rikishi are already part of it, so their bouts can probably be considered free of yaocho by default and thus would have to be excluded from the analysis anyway.

That's interesting.

Edited April 20, 2005 by Asashosakari

[Gusoyama]

While we're on the yaocho topic, I'd like to mention that there's a new economics/statistics book that mentions yaocho as an example of how human nature effects statistical results. I haven't read it yet, but I want to get it, as these kinds of things fascinate me. Freakonomics

It might be of interest to other stattos out there as well.

[Taizeniki]

While we're on the yaocho topic, I'd like to mention that there's a new economics/statistics book that mentions yaocho as an example of how human nature effects statistical results.  I haven't read it yet, but I want to get it, as these kinds of things fascinate me.  Freakonomics

It might be of interest to other stattos out there as well.

<{POST_SNAPBACK}>

before he wrote Freakonomics... an article from Steven Levitt and cohort Mark Duggan: "Winning Isn't Everything: Corruption in Sumo Wrestling"

See: http://econpapers.repec.org/paper/nbrnberwo/7798.htm

[Gusoyama]

Another thing to consider is kensho. Takamisakari, who is often around .500, and is often in mid-maegashira, has a big target on him because of the money based on the bouts, which would give more motivation.

[Randomitsuki]

before he wrote Freakonomics...  an article from Steven Levitt and cohort Mark Duggan:  "Winning Isn't Everything: Corruption in Sumo Wrestling"

See: http://econpapers.repec.org/paper/nbrnberwo/7798.htm

<{POST_SNAPBACK}>

I just had a quick read of this paper which was already discussed in

http://www.sumoforum.net/forums/index.php?showtopic=2667

I was particularly interested in how the authors would try to defend their analysis against the hypothesis that 7-7 rikishi simply give more effort to winning.

Interestingly, at this point the argumentation from the article (which otherwise is technically OK) gets paper-thin.

Basically they state that two facts would eliminate the effort hypothesis:

a) rikishi become better in "crunch-time" bouts over their career.

b) rikishi are particularly bad in "crunch-time" bouts in the last year of their career.

However, I see no reason why these two findings would eliminate the effort hypothesis.

[Burainoan]

I agree that their basic 7-7 argument is a little shaky, though one element of it i'd like to see more of is their assertion that rikishi who win their bouts are statistically more likely to lose in the next "safe" match they have against the rikishi they got that 8th win against. that, to me, would go a bit further towards supporting their argument. though even that could be a little dicey, as they would have to factor in the cases of opponents being weakened by injury or dominance by a particular rikishi over another. for example, if Kaio were kadoban, 7-7 and facing Chiyo...?

that book looks interesting, though. have to check it out one of these days....

[Jakusotsu]

Deriving yaocho or gachinko from torikumi tables alone is just like guessing the banzuke: impossible! When somebody finally invents a clever black box to win GTB everytime, this would be the gadget to ask.

[hoshidango]

Excellent stats! What I feel is generally deliberate loss etc might not be happening that often. But wrestlers might lose for any reason(deliberately, unconsciously etc) and we will never know about the each bout's true psychological intention...

[Guest GreatCornholio]

The Levitt paper seemed convincing to me though it was in the deeper scarier parts of the math. (Showing respect...)

The area that seemed strange was when Rikishi would lose the rematch after winning a suspicious match. Additionally, certain heya would be more likely to have suspicious behavior with other heya. All very interesting when spelled out in English as opposed to Greek symbols.

Then again, as Mark Twain said, "There are three types of lies: Lies, Damn Lies and Statistics"

[Peterao]

Bumped, as yaocho will likely be the dominant topic of interest for some time...