Long-term career predictions

[Randomitsuki 2,826]

While your mind is on heavy rikishi, any thoughts on Maeta? I continue to be puzzled by his career path so far:

Shot up almost into Juryo and then a truck must have hit him over Christmas...

Following up on a very old idea, I recently did some analyses of career paths for rikishi dependent on their age of entry. While it might not be surprising to the experts, I was flabbergasted in how much age factors in for the lower division.

Rikishi starting at 20+ years often don't have the slightest problems moving into Makushita. But this is no indication for actually getting into sekitori ranks.

I compiled lots of data for retired rikishi, categorized by age of entry and final highest stage, and computed their average (highest) ranks at various stages in their career.

Hmm, sounds complicated - to provide an example: I identified the average rank for rikishi who started at age 20+, were 23 years old, and ended up in Juryo (but not higher). These average data can then be compared to current career paths. IIRC, according to these numbers, YMY and Aran are not even Juryo candidates by now. On the other hand, this 15 year old rikishi is currently on ranks similar to rikishi who finally made it to Ozeki and higher...

If someone is interested in these long-term predictions, I can post them and be publicly embarrassed in future times.

[Asashosakari 19,296]

Following up on a very old idea, I recently did some analyses of career paths for rikishi dependent on their age of entry. While it might not be surprising to the experts, I was flabbergasted in how much age factors in for the lower division.

Rikishi starting at 20+ years often don't have the slightest problems moving into Makushita. But this is no indication for actually getting into sekitori ranks.

I compiled lots of data for retired rikishi, categorized by age of entry and final highest stage, and computed their average (highest) ranks at various stages in their career.

As I've insinuated from time to time, I've worked out (rather, made educated guesses) as to the minimum advancement speed on the banzuke to qualify as a prospect. I did a more thorough analysis a couple of months ago (and forgot to tidy it up and post it, d'oh) which probably duplicates some of what you've also found, though it sounds like you've gone a few steps further than I did. I just tabulated the Sandanme and Makushita debut ages for everyone with hatsu-dohyo in 1989 or later, and then looked for patterns regarding their future career potential.

At any rate, my educated guesses were mostly confirmed, yay. In short, middle-school shindeshi (which I've arbitrarily defined as those debuting before their 17th birthday) have a good chance to become sekitori if they make it to Sandanme before their 18th birthday, and to Makushita before their 20th. The Sandanme dividing line is pretty sharp - in my sample, 46 rikishi reached Sandanme for the first time between their 18th and 19th birthday, and only 3 became sekitori later. Makushita isn't as clear-cut, but the trend is the same - 4 out of 19 rikishi who reached Makushita between 20th and 21st birthday, 3 out of 26 the next year, and 2 out of 16 afterwards (Takanotsuru at 22y 2m, and Senshuyama at 22y 5m). On the other hand, 19 of 31 who got to Makushita between their 19th and 20th birthday eventually made sekitori, though in some cases quite late. For those reaching the division before age 19 the numbers are even more dramatic, of course.

For high-schoolers (age group debuting between 17th and 20th birthday, including lots of foreigners), reaching Sandanme is pretty much a given if they want to amount to anything at all. My personal guideline here is "at most one year from mae-zumo". For Makushita, the patterns are pretty similar to the younger group; reaching Makushita before age 22 is highly advisible, although there are a lot more future sekitori in the debut age 20/21 class here due to their later starts.

For collegiate rikishi (in which I've lumped everybody older than 20 on debut, including several more foreigners), the patterns are less clear. What most guys who became sekitori do have in common is that they reached Makushita in about one year or less. It's pretty independent of the exact debut age...even 20- and 21-year olds ought to achieve that, not just those who are almost 23 already. If you start at age 20 and you take 2+ years to Makushita, you're probably not going to make it to Juryo, even if you do make it to Makushita at a younger age than most 22-year old debutants.

In my opinion, any "success" before reaching Makushita is pretty much irrelevant for these old debutants, so I differ a bit from Randomitsuki's judgement on Aran and YMY...it's not that they're lagging, it's that they haven't yet reached a meaningful level of competition to begin with.

All in all, I found pretty much the same trends that Randomitsuki did...even young division debuts aren't a guarantee of future success. On the other hand, it's relatively easy to say which rikishi probably won't make it.

Edited September 12, 2007 by Asashosakari

[Kintamayama 45,085]

@ Randomitsuki and Asashosakari:

I dare you both to come out with a CONCRETE list of 5 (only 5) rikishi sandanme and lower that, based on your individual systems, will become sekitori.

Names.

And no obvious guys like Aran. Obscure names, please..

If any of you are spot on, say with three, I promise to worship you forever. (I already worship you both for your way with numbers, though it is temporarily..)

If not, I will come after you like a steamroller..

Yes, I know it will take some years to find out, but I'm patient. My next of kin have been notified- they are already waiting.

[Randomitsuki 2,826]

@ Randomitsuki and Asashosakari:

I dare you both to come out with a CONCRETE list of 5 (only 5) rikishi sandanme and lower that, based on your individual systems, will become sekitori.

Names.

And no obvious guys like Aran. Obscure names, please..

If any of you are spot on, say with three, I promise to worship you forever. (I already worship you both for your way with numbers, though it is temporarily..)

If not, I will come after you like a steamroller..

Yes, I know it will take some years to find out, but I'm patient. My next of kin have been notified- they are already waiting.

OK, here we go. I'll give you six, and four will make it ;-)

Masunoyama, Mankajo, Takedani, Ryuden, Takayasu, Matsumoto.

Don't forget, I don't know shoot about these guys. That's the beauty of numbers. We'll see..

[Asashosakari 19,296]

@ Randomitsuki and Asashosakari:

I dare you both to come out with a CONCRETE list of 5 (only 5) rikishi sandanme and lower that, based on your individual systems, will become sekitori.

Names.

There's the rub, I haven't yet found the time to apply my "system" (as little as there is, compared to what Randomitsuki appears to have) to the rikishi currently on the banzuke. Can I get back to you between bashos? ;-)

At any rate, I suspect it'll be close to the random guy's six...I'm pretty sure that Masunoyama and Ryuden will make my list, perhaps also Takayasu. I'm also pretty bullish on Itachi's adoptee Sasaki. (I know, he's had a cup of coffee in Makushita already, not sure if that excludes him for the purposes of your dare.)

[aderechelsea 125]

Masunoyama, Mankajo, Takedani, Ryuden, Takayasu, Matsumoto.

IF Mankajo makes it to sekitori ranks then i'll worship you right next to Kinta. I'll even sacrifice my best goat and spray its blood on your golden idol ....

;-)

;-)

[Randomitsuki 2,826]

Just for the fun of it, my numbers can even predict WHEN they will be promoted to Juryo.

Masunoyama in mid-2010. Takedani in early 2011. Mankajo by the end of 2011. Ryuden in mid-2012. Takayasu and Matsumoto by the end of 2013.

You can pin it in the Forum calendar if you want to... ;-)

[ilovesumo 12]

Yep. Masunoyama is on his way up. I also believe in this boy.

As the blog writer does.

But in my eyes, Ooshima's Kyokushuho is the brightest shining star.

[Kintamayama 45,085]

@ A & R:

Registered and locked in my safe.

I am deeply thankful for your good sportsmanship and willingness to name names.

We shall see..

[Kintamayama 45,085]

But in my eyes, Ooshima's Kyokushuho is the brightest shining star.

As an honor to you, I shall now christen him Kyokushoe-horn.

[Randomitsuki 2,826]

@ A & R:

Registered and locked in my safe.

I am deeply thankful for your good sportsmanship and willingness to name names.

We shall see..

Out of curiosity I was running the same analysis, but based on the data from 5 years ago. If you had asked me for five names 5 years ago, this would have been my top 5:

1) Kotokikutsugi; now Kotoshogiku (was sd29 back then) - predicted sekitori promotion for mid-2005 (actual promotion Nagoya 04); predicted Makuuchi promotion for end of 2006 (actual: Hatsu 05); highest predicted rank: Ozeki or higher (currently Sekiwake).

2) Toyonoshima (was sd5) - predicted sekitori promotion for mid-2006 (actual: Natsu 2004); predicted Makuuchi promotion: mid-2008 (actual: Aki 2004); highest predicted rank: Lower Sanyaku (actual: Komusubi)

3) Hakuho (was sd44) - predicted sekitori promotion for end of 2006 (actual: Hatsu 2004); predicted Makuuchi promotion: mid-2008 (actual: Natsu 2004); highest predicted rank: Lower Sanyaku (actual: Yokozuna)

4) Nionoumi (was jd56) - predicted sekitori promotion for end of 2009 (actual: still pending); highest predicted rank: Maegashira (actual: Makushita 20).

5) Hagiwara; now Kisenosato (was sd95) - predicted sekitori promotion in 2011 or later (actual: Natsu 2004); highest predicted rank: Juryo (actual: Komusubi).

[Asashosakari 19,296]

4) Nionoumi (was jd56) - predicted sekitori promotion for end of 2009 (actual: still pending); highest predicted rank: Maegashira (actual: Makushita 20).

Out of curiosity, how does he rate now? He's a contemporary of Sasaki...same hatsu-dohyo, virtually the same age, and similar career path - established in Sandanme before age 17, then ~3 years spent bouncing around in there before making it to Makushita.

Rather different bodies though...Nionoumi 172/132, Sasaki 188/110. Hmm, I think I'll reconsider my bullishness on Sasaki; putting on virtually no weight in five years is gonna be a problem. Nionoumi seems a bit short, too, but I suspect in today's sumo environment that's less of a long-term obstable than too little weight and he has progressed quite nicely in that regard.

Edited September 12, 2007 by Randomitsuki
Asashosakari's edit was edited out...

[Randomitsuki 2,826]

4) Nionoumi (was jd56) - predicted sekitori promotion for end of 2009 (actual: still pending); highest predicted rank: Maegashira (actual: Makushita 20).

Out of curiosity, how does he rate now? He's a contemporary of Sasaki...same hatsu-dohyo, virtually the same age, and similar career path - established in Sandanme before age 17, then ~3 years spent bouncing around in there before making it to Makushita.

Rather different bodies though...Nionoumi 172/132, Sasaki 188/110. Hmm, I think I'll reconsider my bullishness on Sasaki; putting on virtually no weight in five years is gonna be a problem. Nionoumi seems a bit short, too, but I suspect in today's sumo environment that's less of a long-term obstable than too little weight and he has progressed quite nicely in that regard.

Before Nagoya Nionoumi was listed as a guy who could or could not make it into Juryo during his career. Before Aki, i.e. according to the current prediction he is a future Makuuchi candidate.

The "promotion border" between a high-Makushita and a future sekitori for someone at Sasaki's age is ms42 according to my data. His current high is ms47, so my numbers would have him as a high-Makushita finisher.

[Asashosakari 19,296]

Before Nagoya Nionoumi was listed as a guy who could or could not make it into Juryo during his career. Before Aki, i.e. according to the current prediction he is a future Makuuchi candidate.

Thanks. That does make me wonder what the confidence intervals on those predictions are, because I tend to view possibly fluky new high ranks (like Nionoumi's this basho) as rather questionable input data.

(Needless to say, my own analysis suffers from the same potential problem, but because it's more crude in nature it doesn't matter as much.)

[Randomitsuki 2,826]

Thanks. That does make me wonder what the confidence intervals on those predictions are, because I tend to view possibly fluky new high ranks (like Nionoumi's this basho) as rather questionable input data.

I have no idea about the confidence intervals, not even an educated guess. However, you don't know what a fluke is until after the fact. When Toyonoshima had a Sandanme yusho he was promoted to a way-too-high-for-his-abilities-back-then ms10. As a consequence, he finished the next two basho with 2-5 and 3-4. At that time he was about the same age as Nionoumi now. Although the two cases aren't completely comparable, it is likely that one would have considered Toyonoshima's high Makushita rank as a fluke back then.

[Asashosakari 19,296]

Thanks. That does make me wonder what the confidence intervals on those predictions are, because I tend to view possibly fluky new high ranks (like Nionoumi's this basho) as rather questionable input data.

I have no idea about the confidence intervals, not even an educated guess. However, you don't know what a fluke is until after the fact. When Toyonoshima had a Sandanme yusho he was promoted to a way-too-high-for-his-abilities-back-then ms10. As a consequence, he finished the next two basho with 2-5 and 3-4. At that time he was about the same age as Nionoumi now. Although the two cases aren't completely comparable, it is likely that one would have considered Toyonoshima's high Makushita rank as a fluke back then.

Well, it was a fluke, 5-9 is pretty clear about that. (Clapping wildly...) He simply proceeded to acclimate to that new level fairly quickly. Anyway, I was going to suggest using "highest rank at which KK was achieved" rather than career-high rank as the input, but after some preliminary number crunching, I think that would swing too far in the opposite direction, i.e. be unduly penalizing.

Here's a bunch of similar queries - rikishi ranked Ms40-Ms60 who had 4/5/6 wins and achieved a new career-high rank thanks to that result. Their follow-up records (excluding those rikishi who failed to finish that next basho due to injury) were thus:

after 6-1 (137 rikishi): 74 KK - 54.0%

after 5-2 (290 rikishi): 121 KK - 41.7%

after 4-3 (410 rikishi): 167 KK - 40.7%

So there's clearly some value in being able to achieve 6 wins at/near your previous highest rank (i.e. close enough to set a new one with the 6-1), but I do think the possible fluke nature needs to be taken into account...maybe a good input value for your predictions would be the midpoint between "KK-high" and "career-high"?

Edit: Perhaps in the interest of clarity - when I talk about "flukes" here, I don't mean something like Hokutoriki's sekiwake high rank which he'll never even get close to again (and everybody knew that even at the time it happened). That's the fluke of flukes. (Applauding...) Rather, I mean any high rank that exceeds that rikishi's performance level, even if only temporarily. That can be fixed either by using lagged values, or by regressing them towards a more reasonable, expected level as I'm suggesting above. I think the latter approach is more useful here.

Edited September 12, 2007 by Asashosakari

[aderechelsea 125]

Out of curiosity I was running the same analysis, but based on the data from 5 years ago. If you had asked me for five names 5 years ago, this would have been my top 5:

1) Kotokikutsugi; now Kotoshogiku (was sd29 back then) - predicted sekitori promotion for mid-2005 (actual promotion Nagoya 04); predicted Makuuchi promotion for end of 2006 (actual: Hatsu 05); highest predicted rank: Ozeki or higher (currently Sekiwake).

2) Toyonoshima (was sd5) - predicted sekitori promotion for mid-2006 (actual: Natsu 2004); predicted Makuuchi promotion: mid-2008 (actual: Aki 2004); highest predicted rank: Lower Sanyaku (actual: Komusubi)

3) Hakuho (was sd44) - predicted sekitori promotion for end of 2006 (actual: Hatsu 2004); predicted Makuuchi promotion: mid-2008 (actual: Natsu 2004); highest predicted rank: Lower Sanyaku (actual: Yokozuna)

4) Nionoumi (was jd56) - predicted sekitori promotion for end of 2009 (actual: still pending); highest predicted rank: Maegashira (actual: Makushita 20).

5) Hagiwara; now Kisenosato (was sd95) - predicted sekitori promotion in 2011 or later (actual: Natsu 2004); highest predicted rank: Juryo (actual: Komusubi).

you realize that all these are really impressive. I'll better go buy a goat and start feeding it to get it ready for 2009 (all your predictions look a bit later than real life so i suppose Mankajo is due to become sekitori a couple of years earlier ... )

[Manekineko 200]

I'm tempted to pin this thread. :-)

[Naganoyama 5,900]

Perhaps it would be sufficient to add some events in the calendar...

[Fay 1,677]

I don't need worship for the rest of my life but I will name Shiratsuka, Jonidan 14w, Kise Beya.

[Asashosakari 19,296]

after 6-1 (137 rikishi): 74 KK - 54.0%

after 5-2 (290 rikishi): 121 KK - 41.7%

after 4-3 (410 rikishi): 167 KK - 40.7%

So there's clearly some value in being able to achieve 6 wins at/near your previous highest rank (i.e. close enough to set a new one with the 6-1)

A more thorough study has led to some interesting/surprising results:

Section	6	5	4
Ms1-Ms20	59.0%	56.3%	45.5%
Ms21-Ms40	51.6%	44.4%	43.9%
Ms41-Ms60	53.5%	41.5%	40.6%
Sd1-Sd20	49.1%	36.5%	36.2%
Sd21-Sd40	41.4%	35.7%	36.4%
Sd41-Sd60	37.8%	35.8%	37.1%
Sd61-Sd80	34.8%	41.2%	35.0%
Sd81-Sd100	36.3%	36.5%	34.6%
Jd1-Jd20	43.0%	34.8%	30.7%
Jd21-Jd40	42.9%	34.9%	34.3%
Jd41-Jd60	39.2%	42.6%	32.6%
Jd61-Jd80	34.5%	33.5%	33.3%
Jd81-Jd100	37.8%	35.2%	29.1%

These are the percentages for how often a rikishi achieves KK when he's at a new career-high rank following a 6/5/4-win kachi-koshi in the previous tournament. The rank sections refer to their rank in the first basho, not the new high rank. Also, the Ms1-Ms20 section excludes rikishi promoted to Juryo. Observations:

- in Jonidan and Sandanme it doesn't seem to make a whole lot of difference what the score in the last basho was, all classes are around 35%, give or take some random variation

- the exception to that seem to be rikishi who make their Sandanme or Makushita debut following a 6-1, especially those who get in with some room to spare (e.g. from Sd1-Sd40 and Jd1-Jd40 - the full promotion zones go down to Sd50 and Jd65/70)

- rikishi achieving a new career-high rank within Makushita are more likely to follow up with another kachi-koshi than rikishi in the lower divisions

[Doitsuyama 1,185]

- rikishi achieving a new career-high rank within Makushita are more likely to follow up with another kachi-koshi than rikishi in the lower divisions

And it gets even better the higher in makushita they are. I think this has to do with the fact that promotions are getting lower with higher banzuke rank. This results in rikishi not getting too much out of their ability rankwise and getting clobbered.

[Randomitsuki 2,826]

Just for completeness, here are some current predictions, several of which appear downright hilarious (I am tempted to put aderechelsea's Mankajo into that category as well). Once again, if the list says that XY could make it to Maegashira this just means that XY currently is on a career trajectory that fits very well with those guys who actually made it to Maegashira in the past. Apply as much grains of salt as you like. The data are pre-Aki. I am tempted to update these predictions regularly (i.e. once the next banzuke is out).

Oh, one more disclaimer before you flame me away: the predicted divisions do not imply permanent residence. In other words, Juryo could mean "Dewanosato-Juryo", Maegashira could mean "Asofuji-Maegashira", Sanyaku could mean "Hokutoriki-Sanyaku", and "Ozeki" could mean "Miyabiyama-Ozeki".

Rikishi with a career high in Sandanme (or lower) who might make it to Juryo or higher:

Daishoko: currently on career-high sd82

Kaisei: career-high sd18, currently sd42 (predicted for Sanyaku) (Holiday feeling...)

Kubota: career high sd6, currently sd46

Kurosawa: currently on career-high sd10

Mankajo: career-high sd23, currently sd73 (predicted for Sanyaku) :-D

Masuko: career high sd48, currently sd74

Masunoyama: currently on career-high sd35 (predicted for Ozeki)

Masutoo: career high sd3, currently sd5 (Showing respect...)

Matsumoto: currently on career-high jd43 (predicted for Sanyaku)

Rendaiyama: currently on career-high sd29

Ri: currently on career-high sd23 (predicted for Maegashira)

Ryuden: currently on career-high sd60 (predicted for Sanyaku)

Takayasu: career-high sd27, currently on sd41 (predicted for Maegashira)

Takedani: career-high sd17, currently sd32 (predicted for Sanyaku)

Tochihiryu: career high sd17, currently sd37

Yoshino: currently on career-high sd2

Rikishi with a career high in Makushita who might make it to Maegashira or higher:

Daishoyu: career-high ms18, currently ms22 (predicted for Sanyaku)

Fukunaga: career-high ms16, currently ms39 (predicted for Sanyaku)

Gagamaru: career-high ms24, currently ms48 (predicted for Sanyaku) (Shaking head...)

Ikioi: career-high ms28, currently ms37 (predicted for Sanyaku)

Nionoumi: currently on career-high ms20

Shibuya: career-high ms11, currently ms31

Sokokurai: currently on career-high ms5

Tochinoshin: currently on career-high ms6 (predicted for Ozeki)

Rikishi with a career high in Juryo who might make it to Sanyaku or higher:

Masatsukasa: career-high J3, currently J7 (In love...)

Wakanoho: currently on career-high J1 (predicted for Ozeki)

Rikishi with a career high in Maegashira who might make it to Ozeki:

Goeido: currently on career-high M14

Kakuryu: currently on career-high M2 (Laughing...)

Toyohibiki: currently on career-high M6

[aderechelsea 125]

funny thing is that numbers actually have a point .... (with some hilarious exceptions like Mankajo)

we could disagree with lots of those predictions but the overall feeling is that with that list and some actual knowledge of those rikishi you could make good (but rough nonetheless) predictions.

thank you guys for all these ....

[Azumashida 1]

Excellent stuff, Randomitsuki and Asashosakari! (Sign of approval...)

Regarding the predictions for e.g. Mankajo, Kaisei and Gagamaru, I'm wondering if there should be a slight alteration of your model/formula/algorithm (or whatever you call it) in terms of the size when entering ozumo. It seems that these three guys had a size advantage initially but as soon as they got to an area of the banzuke where technique matters, they started struggling pretty badly. Well, Kaisei and Gagamaru may still have a very good potential due to their combination of size and (relative) mobility, but Mankajo strikes me as a case for a size advantage in the beginning that may well become a huge handicap (Orora-style) eventually. I realize that there's probably no way to try to generalize this as a principle since it depends not only on weight numbers but also on bone structure and (in the case of Pacific Islanders for instance) metabolism, but this particular problematic probably brings some "noise" in your data.