Harumafuji (ex-Ama) poll

[Kintamayama]

Not kidding at all. I looked at his records against the big boys of his time, also keeping in mind he had a hulking 195cm 150+ kg body, and he got in trouble with his oyakata and whatnot. Are you telling me it wasn't his fault or he didn't have potential? Please clarify.

Doitsuyama got it right. I wouldn't call altercations with his oyakata and subsequent dismissal "throwing it away", as he didn't exactly wake up one morning and decide to leave everything. As for potential, not ever having a yusho notwithstanding, I agree he had great potential and a great physique. and the problems probably were his fault to some extent, but it seems things just got out of hand.

[Sokkenaiyama]

It was meant to sound somewhat sarcastic. Futahaguro blew it on his own of course and it wasn't intentional, most likely. Slipping really off topic (like this wasn't already so), I miss the Eurosport broadcasts. What you can find on teh Internets doesn't come close to TV quality (sorry, Frenchies, your work is still valuable though).

[Jejima]

I agree that he could well have the potential for a 4-yusho career, but even being charitable I'd consider that the mean of his potentials range, not the minimum, which is what you seemed to imply - or in other words, that you think he's a good bet (>>50%) to win more than 4, while I think 1 or 2 is just as likely as 5 or 6. (Okay, in actuality I think 1 or 2 is much more likely than 5 or 6, but for the sake of argument...) In any case, a 4-yusho guy is still extremely unlikely to just waltz through the ozeki rank in his first year.

Not quite what I am saying. I am saying that he has the potential to do at least as well Wakanohana III. Doitsuyama's Ratings (which I am a big fan of :-)) reflect this potential. Many things could happen - he could get a serious injury - a new 'Asashoryu' might emerge on the scene - the old Asashoryu might re-emerge - Harumafuji might get suddenly go crazy.... But, if all goes 'smoothly', then based on my current assessment (where I see him as the second most able Ozumo rikishi currently competing), I believe he can get 4 or more yushos in his career.

I believe he has reasonable odds to get 2 yushos in 2009 (for the reasons given above). If you asked me, if I think he is more likely to get 1 or 2 yushos than 5 or 6 in his whole career, I would probably have to side with caution (due to all the possibilities that can happen), and go on the low side, for 1 or 2. But this does not change my original argument that he has a good chance to be promoted to Yokozuna in 2009 - nor my subsequent argument that he has the potential to be at least as good Wakanohana III.

I can't really follow those numbers; excluding Harumafuji I'm counting 15 ozeki and 12 (not 20) yokozuna with hatsu-dohyo since 1970.

Apologies, my mistake, I misread the columns. My data was taken with 'intai' of rikishi after 1970 (I wanted to start with Taiho, who retired in 1970. Of course it is a very embarrassing mistake to think that Taiho had his hatsu-dohyo after 1970! :-P But the data is still correct. For rikishi that 'intaied' after 1970 (or are yet to inati), 57% of Ozekis became Yokozunas.)

The first two things to notice are that Harumafuji is actually older than the median new ozeki, and that the segment of future yokozuna is dominated by those rikishi who were of below-median age when they attained the second-highest rank.

Ignoring the very small data pool for a moment.... I think the above comment is 'stating the obvious'. Of course the younger you are when you are promoted to Ozeki, the more chance you have of reaching Yokozuna! You have more 'fighting years' etc to do so. This does not deflect from my argument about Harumafuji's chances in 2009.

And while there are four ozeki in Ama's rough age range who quickly became yokozuna, three of them have extremely unusual career circumstances going for them:

<snip>

So that leaves Hokutoumi with a career that is roughly comparable to Harumafuji's, plus a whole bunch of ozeki who didn't become yokozuna. And even Hokutoumi was a year and a half younger.

Okay, if we have now discarded three of four comparable rikishi (in your opinion), and are left with Hokutoumi, - let's look at how he did....

Hokutoumi became a Yokozuna within a year of being promoted to Ozeki.

And now I'm officially out of arguments for why it's not just unlikely but extremely improbable that Ama will be a yokozuna on the Hatsu 2010 banzuke.

I can understand why you say 'unlikely' - but I think extremely improbable (which I understand to mean something like the odds of 100-1 or far greater) must be wrong! - I'm not sure if I would even put Kaio into that category (Kaio is very highly unlikely to reach Yokozuna, maybe, but not extremely improbable - and I rate Harumafuji's chances better than that :-))

Gonna be fun to look back at this thread 12 months from now, anyway. :-)

;-)

[Jejima]

Apologies, my mistake, I misread the columns. My data was taken with 'intai' of rikishi after 1970 (I wanted to start with Taiho, who retired in 1970. Of course it is a very embarrassing mistake to think that Taiho had his hatsu-dohyo after 1970! ;-) But the data is still correct. For rikishi that 'intaied' after 1970 (or are yet to inati), 57% of Ozekis became Yokozunas.)

Aaaargh!!!

More apologies!

Just rechecking - and I discovered that I had got the Ozeki count correct (as Asashosakari confirms). There have been 15 Ozeki with hatsu dohyo post-1970 who have yet to make Yokozuna - but 23 who have intaied (or are yet to intai) after 1970.

So: using the stats for post 1970 hatsu-dohyo:---- 12/25 = 44% of Ozeki go on to become Yokozuna

using the stats for post 1970 intai:---- 20/43 = 47% of Ozeki go on to become Yokozuna

So still quite high odds - but not quite as high as I was claiming (sorry :-P )

Edited December 12, 2008 by Jejima

[Asashosakari]

Okay, if we have now discarded three of four comparable rikishi (in your opinion), and are left with Hokutoumi, - let's look at how he did....

Hokutoumi became a Yokozuna within a year of being promoted to Ozeki.

No, as I said we're left with Hokutoumi and a bunch of ozeki who didn't become yokozuna, neither in their first year nor at any other point. Which ozekis I meant is left as an exercise to the reader :-P but Gusoyama already pointed out one excellent comparison. (BTW, I wasn't discarding three of four comparable rikishi, quite the opposite.)

I can understand why you say 'unlikely' - but I think extremely improbable (which I understand to mean something like the odds of 100-1 or far greater) must be wrong! - I'm not sure if I would even put Kaio into that category (Kaio is very highly unlikely to reach Yokozuna, maybe, but not extremely improbable - and I rate Harumafuji's chances better than that ;-))

I don't like to put numbers on these things because we're not dealing with repeatable experiments, but if I have to, I guess I'll go with the usual statistical significance level of 5% as the demarcation between "improbable" and "extremely improbable".

In any case, to put some definite numbers on it: let's just say that a yokozuna promotion typically requires 26+ wins in two basho. My assessment is that Ama's upside for the next year is 23, and as a mean projection I'd say he's most likely to top out at 21 or 22. Yep, it's extremely improbable that such a rikishi is going to put together 26 wins. (If I ran the binomials for that, I'd probably end up with way less than 5%, actually.) Obviously you disagree with the very premises of that assessment and you think he's much stronger than that (BTW, which wins bracket did you vote for?), but I'm not just pulling my conclusions out of thin air.

[Sokkenaiyama]

Listen to the man, Jezz, that there is cold, hard mathematical fact. That, and a particularly weak banzuke in these last two tournaments.

[Asashosakari]

Listen to the man, Jezz, that there is cold, hard mathematical fact.

And it could be totally wrong, of course. But it's at least consistent.

Anyway, since this isn't all that hard to simulate, I've whipped up a quick Excel sheet. Highly simplified, of course, because no rikishi is going to have the exact same "true" strength in every single basho (and zero probability of injury), and because it ignores the varying strength of the available opponents, but there we are. Approximate probabilities of reaching a particular number of two-basho wins at least once in a year with a given strength level:

W%	Bs	Yr	<=20	21	22	23	24	25	>=26
.600	9-6	54-36	47%	20%	15%	10%	5%	2%	1%
.667	10-5	60-30	13%	15%	21%	21%	16%	9%	5%
.733	11-4	66-24	1%	3%	8%	17%	23%	23%	25%
.800	12-3	72-18	<1%	<1%	1%	3%	10%	19%	67%

I was a little bit off...with my assumption that Harumafuji is approximately a .667 rikishi next year, the probability of reaching 26 wins isn't "way less than 5%", but just about exactly 5%. I still feel comfortable in calling it extremely unlikely though; with the frequency of new ozeki promotions we're talking about something that would happen once every 25-35 years if every new ozeki was a .667-strength rikishi in their first year. Obviously if somebody is assessing Harumafuji to be significantly stronger than that, the probability goes way up as the table shows.

Edited December 12, 2008 by Asashosakari

[Doitsuyama]

I was a little bit off...with my assumption that Harumafuji is approximately a .667 rikishi next year, the probability of reaching 26 wins isn't "way less than 5%", but just about exactly 5%.

But your assumption might be the very source of disagreement with Jejima, so the subsequent calculations aren't that good of an argument. Oh, and going with your assumption, it was a rather improbable sequence of Ama getting 35 wins in the last three basho, right? To be exact, a probability of 0.074...

Besides, your basic assumption of Harumafuji being a .667 rikishi next year is contradicting my strength ratings which have him down as a 0.74 rikishi right now (ok, 0.71 with Asashoryu and Kaio instead of Miyabiyama and Asasekiryu but it's not unusual to have a top rikishi or two injured). I realize that my strength ratings might be totally faulty but I still feel that it is up to you to provide good arguments against those numbers as you basically downgrade his strength by 67 of my strength points (which is a lot) without justifying it.

[Jejima]

No, as I said we're left with Hokutoumi and a bunch of ozeki who didn't become yokozuna, neither in their first year nor at any other point. Which ozekis I meant is left as an exercise to the reader ;-)

Thankfully for the reader, we can see the Ozeki above in a previous post of yours :-P

Here they are:-

Akebono 23.0 4 basho

Hokutoumi 23.1 5 basho

Konishiki 23.6

Kotokaze 24.5

Ama / Harumafuji 24.7

Wajima 24.8 4 basho

Tochiazuma 25.0

Dejima 25.4

Chiyonofuji 25.7 3 basho

Okay, Akebono, Wajima and Chiyonofuji have been excused. That leaves us with Hokutoumi and 4 Ozeki. May I excuse 'Konishiki' too? He was a little unlucky at one point in his career when some thought he should have been promoted to Yokozuna, going by numbers alone. (No, I don't want to open that can of worms!).

Dejima (I think) was already suffering from injuries when he made it to Ozeki. Tochiazuma had several bashos when he was kyujo after being promoted to Ozeki - yet he still had a couple of tsunatori basho - so can these two be excused because of injuries?

Perhaps Kotokaze is comparable (I don't know much about him). It seems as if he was promoted a little 'light' to Ozeki (9-6, 10-5, 12-3 (with yusho)) with only 31 wins - and he had the misfortune to be around when Chiyonofuji was in his pomp. *But* even he had a 14-1 yusho (his loss being on shonichi) in March 1983, followed by a 7-0 start (so 21 wins on the trot) in May 1983. At that point, he must have thought he had some chance of getting a promotion to Yokozuna? (He ended up 11-4, and Chiyonofuji took the yusho 15-0).

So, out of the eight in a similar age range to Harumafuji when first promoted to Ozeki, 4 became Yokozuna within a year of becoming Ozeki (50%). Of the remaining four, three of them had at least one good chance to get promoted to Yokozuna (37.5%) at some stage in their Ozeki career. Leaving just Dejima (12.5%) who never seemed to have come close.

Remember, I am not saying that Harumafuji will definitely become a Yokozuna in 2009 (or whenever), I am saying that he has the potential, and the odds look in his favour (in my opinion, due to who the other top ranked guys are).

I can understand why you say 'unlikely' - but I In any case, to put some definite numbers on it: let's just say that a yokozuna promotion typically requires 26+ wins in two basho. My assessment is that Ama's upside for the next year is 23, and as a mean projection I'd say he's most likely to top out at 21 or 22. Yep, it's extremely improbable that such a rikishi is going to put together 26 wins. (If I ran the binomials for that, I'd probably end up with way less than 5%, actually.) Obviously you disagree with the very premises of that assessment and you think he's much stronger than that (BTW, which wins bracket did you vote for?), but I'm not just pulling my conclusions out of thin air.

My thinking, (one more time with feeling)....

I think that Hakuho is the strongest rikishi at present.

Until proved otherwise, I don't think Asashoryu will be like his former self, when he returns in January.

I therefore think Harumafuji is currently the second strongest rikishi in Ozumo (I am still hoping Asashoryu will prove me wrong.)

I don't think Hakuho will win all 6 bashos in 2009. (When Asashoryu won all 6 in a calendar year, that really was something special.)

Therefore, I think that Harumafuji has a good chance to win a basho in 2009 (maybe 12-3? or 13-2? or even 14-1?).

Therefore the subsequent basho would be a tsunatori one for him - with him ranked at Ozeki East 1.

In that basho I think he is capable of going 10-2 (or 9-3 may be enough, depending on how many he won his yusho with) in the first twelve days.

His last three opponents should be day 13 - Asashoryu (Y1W), day 14 - Hakuho (Y1E) and ?Kotomitsuki? (O1W).

I think that having friends / fellow countrymen as two of your opponents in such a circumstance is better than having bitter rivals as your foes.

I think that if Harumafuji were in such a position, he could win out the final three days.

I therefore think that he has a reasonably good chance of being promoted to Yokozuna - and could well be there on the January 2010 banzuke.

Phew!

Edited December 12, 2008 by Jejima

[Asashosakari]

Besides, your basic assumption of Harumafuji being a .667 rikishi next year is contradicting my strength ratings which have him down as a 0.74 rikishi right now (ok, 0.71 with Asashoryu and Kaio instead of Miyabiyama and Asasekiryu but it's not unusual to have a top rikishi or two injured). I realize that my strength ratings might be totally faulty but I still feel that it is up to you to provide good arguments against those numbers as you basically downgrade his strength by 67 of my strength points (which is a lot) without justifying it.

I think the assumption that he's just had two "career basho" and is unlikely to match that level immediately going forward is plenty enough justification, and there's enough historical precedent that it's far from a crazy assumption. Basically I don't buy the notion that Ama is a three-basho average 32-13 rikishi now (which .71 would imply, and even better if we keep factoring in injuries). What was his rating before Aki, BTW? Be interesting to know what his strength-rated probability of going 25-5 was at that point, since you've already brought up that issue...

But your assumption might be the very source of disagreement with Jejima, so the subsequent calculations aren't that good of an argument.

At this point I'm arguing not so much for the correctness of my assessment, just the consistency of the argument I've made. Nobody's going to convince anybody here ;-) so I'm not particularly concerned with whether somebody agrees with me about Harumafuji being at .67 strength or not. (But I still wonder how many wins Jejima voted for - 71-75 I presume?)

Edited December 12, 2008 by Asashosakari

[Doitsuyama]

Besides, your basic assumption of Harumafuji being a .667 rikishi next year is contradicting my strength ratings which have him down as a 0.74 rikishi right now (ok, 0.71 with Asashoryu and Kaio instead of Miyabiyama and Asasekiryu but it's not unusual to have a top rikishi or two injured). I realize that my strength ratings might be totally faulty but I still feel that it is up to you to provide good arguments against those numbers as you basically downgrade his strength by 67 of my strength points (which is a lot) without justifying it.

I think the assumption that he's just had two "career basho" and is unlikely to match that level immediately going forward is plenty enough justification. Basically I don't buy the notion that Ama is a three-basho average 32-13 rikishi now (which .71 would imply). What was his rating before Aki, BTW? Be interesting to know what his strength-rated probability of going 25-5 was at that point, since you've already brought up that issue...

I believe you can't assess a basho or two as "career basho" before the career is over. There might always be something better waiting around the corner and Harumafuji isn't THAT old. His ratings before Hatsu and the following basho were 2339 - 2350 - 2342 - 2351 - 2360 - 2412 - 2452. Sure, a 25-5 run was extremely unlikely based on his rating before Aki, with a probability of about .0069 or almost 1 to 150 odds. But by now we know better about his strength and our assessment of his future should be closer to his current rating than his rating before Aki (which is what you are tending to use).

[Jejima]

(But I still wonder how many wins Jejima voted for - 71-75 I presume?)

No, it was either 61-65 or 66-70 (I forget which). But having read what Doitsuyama has written, maybe 71-75 is more accurate?

Anyways, this poll has generated a lot more interest than the one on Kaio ;-)

Edit: I have found my earlier calculations, my vote was for 66-70.

Edited December 12, 2008 by Jejima

[Asashosakari]

His ratings before Hatsu and the following basho were 2339 - 2350 - 2342 - 2351 - 2360 - 2412 - 2452. Sure, a 25-5 run was extremely unlikely based on his rating before Aki, with a probability of about .0069 or almost 1 to 150 odds.

Hey, that makes the probability based on my .667 assumption (.035) look downright glorious. ;-)

But by now we know better about his strength and our assessment of his future should be closer to his current rating than his rating before Aki (which is what you are tending to use).

If I'm reverse-engineering your 1 in 150 from above correctly, that means Ama was approximately a .610 (!) rikishi before his most recent two basho, and after a very improbable run of 30 bouts he's now about .72 (halfway between the good-torikumi/bad-torikumi numbers of .71 and .74 you mentioned). Is it really so outrageous to regress that back by about one-half to arrive at an expected winning percentage of .67 going forward, at least for the next year? Frankly, I feel more sure of my assumptions now than I did before you introduced those numbers.

Edit: Heck, if he has the usual shin-ozeki slump he'll probably drop back to .69 or so in the ratings right after Hatsu...

Edited December 12, 2008 by Asashosakari

[Asashosakari]

Is it really so outrageous to regress that back by about one-half to arrive at an expected winning percentage of .67 going forward, at least for the next year?

I suppose one way to test for this would be to take every rikishi ranked sekiwake to maegashira who had a basho that boosted his strength rating a lot - not sure what a good cutoff would be, maybe at least +30 points? - and check their expected wins for the next basho (or the next two or three basho) against their actual wins. If the expected wins are significantly higher than the actual, it would be evidence that strength ratings right after a particularly strong basho need to be regressed a bit in order to make correct predictions for the near future.

[Jejima]

Edit: Heck, if he has the usual shin-ozeki slump he'll probably drop back to .69 or so in the ratings right after Hatsu...

Shin-Ozeki Hakuho = 14-1 yusho (Whistling...)

But I think you are probably correct :-)

However, if Harumafuji is ranked at O3W, and if Asashoryu and Kaio manage to avoid going kyujo / intai, then what will his schedule look like?

As the '8th highest ranked', would that mean that he would be facing the two Yokozuna mid basho? When would he face his fellow Ozeki?

[Asashosakari]

Shin-Ozeki Hakuho = 14-1 yusho (Whistling...)

You missed an opportunity there - the real slap across the face would have been to point out Tochiazuma's 13-2 shin-ozeki yusho. :-)

However, if Harumafuji is ranked at O3W, and if Asashoryu and Kaio manage to avoid going kyujo / intai, then what will his schedule look like?

As the '8th highest ranked', would that mean that he would be facing the two Yokozuna mid basho? When would he face his fellow Ozeki?

7th. :-) Good question though...looking back at the recent tournaments with 6 Y+O, the 5th and 6th-ranked guys faced fellow ozekis from Day 8 or 9, and the two yokozuna on Days 11 and 12. For a 7th-ranked if everyone stays healthy, it would presumably be something like this:

Day 9: Ozeki 1

Day 10: Yokozuna East

Day 11: Yokozuna West

Day 12: Ozeki 1 (maybe switched with Day 11)

Day 13: Ozeki 2

Day 14: Ozeki 2

Day 15: filler (Sekiwake East?)

[Kintamayama]

His last three opponents should be day 13 - Asashoryu (Y1W), day 14 - Hakuho (Y1E) and ?Kotomitsuki? (O1W).

I think that having friends / fellow countrymen as two of your opponents in such a circumstance is better than having bitter rivals as your foes.

I think that if Harumafuji were in such a position, he could win out the final three days.

Not if said Yokozunae were in the yusho hunt, which, at least in Hakuhou's case, is more than plausible. I am willing to eat humble steak and say 1-2 is a more logical number.

But we shall see soon enough, as there is a great possibility (Asa's elbow/toes/kneecap/mind willing..) we will be seeing this happen in Hatsu.

[Doitsuyama]

Is it really so outrageous to regress that back by about one-half to arrive at an expected winning percentage of .67 going forward, at least for the next year?

I suppose one way to test for this would be to take every rikishi ranked sekiwake to maegashira who had a basho that boosted his strength rating a lot - not sure what a good cutoff would be, maybe at least +30 points? - and check their expected wins for the next basho (or the next two or three basho) against their actual wins. If the expected wins are significantly higher than the actual, it would be evidence that strength ratings right after a particularly strong basho need to be regressed a bit in order to make correct predictions for the near future.

You are right with your assumptions - I did assume such an effect too, going from the random walk theory. But how big that effect is actually is an interesting question, so I made the work for the data from the last 10 years, a good enough sample with more than 400 rikishi between sekiwake and maegashira increasing the rating by more than 30 points, usually about 6 rikishi per basho. The actual wins were on average about 0.09 wins lower than the expected wins in all three subsequent basho. This means a rating correction of about 6 points would seem appropriate here. That still would be much fewer than your implicit correction of Ama's strength in your considerations...

[Fujisan]

I thought this was gonna be a vote on whether Harumafuji is a silly name....

Just in case my vote is yes....

[Asashosakari]

You are right with your assumptions - I did assume such an effect too, going from the random walk theory. But how big that effect is actually is an interesting question, so I made the work for the data from the last 10 years, a good enough sample with more than 400 rikishi between sekiwake and maegashira increasing the rating by more than 30 points, usually about 6 rikishi per basho. The actual wins were on average about 0.09 wins lower than the expected wins in all three subsequent basho. This means a rating correction of about 6 points would seem appropriate here. That still would be much fewer than your implicit correction of Ama's strength in your considerations...

Intriguing, definitely much lower than I expected. Thanks for doing all that work! (Whistling...) One question: Did you use the first, boosted rating to calculate the expected wins for all three subsequent basho, or the three individual ratings?

[Doitsuyama]

You are right with your assumptions - I did assume such an effect too, going from the random walk theory. But how big that effect is actually is an interesting question, so I made the work for the data from the last 10 years, a good enough sample with more than 400 rikishi between sekiwake and maegashira increasing the rating by more than 30 points, usually about 6 rikishi per basho. The actual wins were on average about 0.09 wins lower than the expected wins in all three subsequent basho. This means a rating correction of about 6 points would seem appropriate here. That still would be much fewer than your implicit correction of Ama's strength in your considerations...

Intriguing, definitely much lower than I expected. Thanks for doing all that work! (Whistling...) One question: Did you use the first, boosted rating to calculate the expected wins for all three subsequent basho, or the three individual ratings?

The individual ratings. I actually made no new rating calculations at all, just reading from past ratings.

[Jejima]

I will be closing this poll in about a week's time. Still time to vote (Just what doctor ordered...)

[Jejima]

This poll has now closed.

Most of think that Harumafuji will end up the year still ranked as an Ozeki - but notably as one of the top two-ranked Ozekis.

About 80% of us think he will get 10 or more wins (as is expected of an Ozeki) in over half the bashos this year.

The number of wins seems to be anywhere between 55 - 70 for the year.

[Jejima]

Harumafuji will be (unless something very strange happens) ranked at O1W for January 2010, as predicted by about a third of us. (No, not me, I had him down as a potential Yokozuna (Yusho winner...))

He only had two bashos with at least ten wins, confounding many gamers in the process. Less than 8% of us predicted that. We were way off on this one, with about 85% of us predicting that he would do better this year.

He had 59 wins in the year, which 29% predicted. This was on the lower end of our predictions, but well done to those who got it (Clapping wildly...).

Overall, I think he mildly disappointed our overall predictions (and for some, like me, really disappointed ;-))

I will close this thread again (to stop new votes), but the new Harry poll will be opened soon.

Edited December 2, 2009 by Jejima