Aim For Yokozuna Beta

[Zeokage]

Aim of the game:

To pick one rikishi for the wins, another for the losses, add them together and get the highest total. For example last basho, the best score in Makuuchi in this game under these beta rules would have been a Kyokutenho (12 wins) and Hokutoriki/Ryuo (12 losses) pairing. You just have to do this for Makuuchi and Juryo (last basho would have been a Baruto and Daimanazuru pairing for the win).

How to enter:

Pick 4 rikishi, 2 from Makuuchi (not including Yokozuna/Ozeki) and 2 from Juryo. Post whether you want the wins or losses to count for that rikishi. Those kyujo before you make your entry will make your pick invalid. I'll contact you about a replacement as soon as I see the problem, or just post (please no edits) your replacement in this thread if you spot it first.

For the purposes of the beta test, like last time you can enter 3 groups each. :)

Makuuchi Wins:

Makuuchi Losses:

Juryo Wins:

Juryo Losses:

The deadline for this game is 2pm JST on the first day of the basho, the same as most of the other games based on the sekitori.

Sansho:

These are just an idea for the sansho. If anyone has any other ideas please feel free to post them. I tried to make them in a way that would represent the meaning of the real prizes...

To get a prize, you must have a KK, and a quartet that have fought on all 15 days.

Outstanding Performance Award (Shukun-sho): Rikishi who gets more separate wins/losses than the rikishi who wins the basho.

Technique Award (Gino-sho): Get the best choice in 3 of the 4 groups, or the best sole options for the same division.

Fighting Spirit Award (Kanto-sho): Take the numerical rank of the real rikishi on the banzuke, and compare with the numerical rank from their records after the basho.

(Banzuke rank

Edited November 2, 2007 by Zeokage

[Kintamayama]

Makuuchi:

Rohou-wins

Kaihou-losses

Juryo:

Ryuuou-wins

Kotokasuga-losses

[Washuyama]

Makuuchi:

Wins - Baruto

Losses - Takamisakari (unfortunately)

Juryo:

Wins: Ichihara

Losses: Shunketsu

[Zeokage]

Here are the results of the Natsu basho beta test...

Zenjimoto	43
Itachi A	43
Kintamayama	43
Vikanosato	41
Miyahakuho	41
Toridefuji	40
Itachi C	39
Itachi B	37
Torideyama	37
Chiyomitsuki	36
Chikenshitenzan	35
Debeso	34
Chiisabuke	33
Shobaionna	33
Vikanoshima	32
Vikanohana	31
Asanoumi	30
Toridenoshima	28

Problem now is how to turn it into a W/L set. The scores are too close when dividing by 4, and almost everyone has over 8 wins using that method. The median score should be at 36.5 but I have no idea how to go from there. I tried to use the Kumi calculation to get a W/L score but it failed at the top end for some reason. What would you guys do from here? That's aimed more at the record makers from the GISP/RotoSumo/Salarycap group of games were points are turned into records.

In other news, here are the prizes from that basho:

Outstanding Performance Award (Shukun-sho): Miyahakuho, for getting 12 wins in the Makuuchi Wins category when the winner only got 9.

Pending this basho's test the benchmark might be changed to "Rikishi who gets at least 2 more separate wins/losses than the rikishi who wins the basho."

Technique Award (Gino-sho): Not awarded

Fighting Spirit Award (Kanto-sho): Itachi A.

Quick addition... like last time you can have 3 entries each to simulate additional traffic to this game.

Edited November 2, 2007 by Zeokage

[Asashosakari]

Problem now is how to turn it into a W/L set. The scores are too close when dividing by 4, and almost everyone has over 8 wins using that method. The median score should be at 36.5 but I have no idea how to go from there. I tried to use the Kumi calculation to get a W/L score but it failed at the top end for some reason. What would you guys do from here? That's aimed more at the record makers from the GISP/RotoSumo/Salarycap group of games were points are turned into records.

I'll consider myself included here as I've done some work with the (median-based, as it happens) Seki-Oracle formula. :-) Given the results, the easiest would probably be something like (score minus median + 7.5); of course, I don't know how stable that would be across multiple tournaments (obviously it's desirable not to end up with 15-0 records all over the place), and for another thing, your trial results have the unusual property that they're very soft in the middle whereas most sumo games results tend to be clustered pretty close to the median with only few outliers. Whether that's just sampling error because the number of players was so small, or whether it would usually be like that in your game, I don't have a clue about.

[Bealzbob]

One solution on how to avoid having 15-0 all over the place in your W/L record is to factor in the theoretical top score of 60pts and the theoretical lowest score of 0pts.

So the top end of your competition utilises the 'fixed' number (60), but the bottom end utilises the bottom score (0) and the range is worked out within that.

If we halve the difference between the top score in your competition (43pts) and the 60 target (diff=17, halved=rounded to 9), the computed top score for our purpose is 43+9 = 52 would be the 15-0 score. Obviously the higher the scorer gets to the 60, the less the difference between the 2 and the closer they would get to the zensho.

Similarly we can halve the 28pts accrued by Toridenoshima to 14pts making 14pts the 0-15 score.

We now have a 15-0 score (52) and a 0-15 score (14).

With 14 different win numbers inside 0 & 15 (i.e. 1 to 14) we need a sliding scale within 14 & 52 with which to implement the win record.

As there is 37 distinct pts between 14 & 52 (not including either 14 or 52 which we have already assigned wins #0 & #15 to) I make that a scale of ~2.6pts per win.

With all that said, I work your table out as follows

Zenjimoto	 (43) 11-4
Itachi A	  (43) 11-4
Kintamayama   (43) 11-4
Vikanosato	(41) 10-5
Miyahakuho	(41) 10-5
Toridefuji	(40) 10-5
Itachi C	  (39) 9-6
Itachi B	  (37) 9-6
Torideyama	(37) 9-6
Chiyomitsuki  (36) 8-7
Chikenshitenzan (35) 8-7
Debeso		(34) 8-7
Chiisabuke	(33) 7-8
Shobaionna	(33) 7-8
Vikanoshima   (32) 7-8
Vikanohana	(31) 6-9
Asanoumi	  (30) 6-9
Toridenoshima (28) 5-10

Any good ?

Edited November 2, 2007 by Bealzbob

[Asashosakari]

One solution on how to avoid having 15-0 all over the place in your W/L record is to factor in the theoretical top score of 60pts and the theoretical lowest score of 0pts.

So the top end of your competition utilises the 'fixed' number (60), but the bottom end utilises the bottom score (0) and the range is worked out within that.

If we halve the difference between the top score in your competition (43pts) and the 60 target (diff=17, halved=rounded to 9), the computed top score for our purpose is 43+9 = 52 would be the 15-0 score. Obviously the higher the scorer gets to the 60, the less the difference between the 2 and the closer they would get to the zensho.

That, in turn, probably makes it nearly impossible to score better than 13-2 even with a perfect team though, as the practical maximum for last basho was just 52, and that was with an injury-assisted 15-point score in there. By your rule this would fix the 15-0 score to 56 points if a player scored perfectly. Looking back through the current year, the achievable maxima were:

Aki 52 (one zero-win injury)

Nagoya 51 (one zero-win injury)

Natsu 56 (two zero-win injuries)

Haru 48 (no relevant injuries)

Hatsu 49 (two-win and one-win injuries)

(The corresponding minimum scores are 60-Max, of course.) I think using the actually-possible max/min scores would provide better baselines. Other than that, I find your suggestion a lot more meritorious than mine. :-)

Edit: On second thought, using that minimum value might boost the low records too much, as the scoring range would actually expand towards the bottom compared to your suggestion...hmm. Maybe one could do Max = Maximum Achievable, and Min = Median - (MaxAch - Median) = 2*Median - MaxAch. That would give a range from 21 to 52 for Aki basho, leading to top/bottom scores of 11-4 and 4-11 (I think)...

Edited November 2, 2007 by Asashosakari

[Doitsuyama]

Maybe one could do Max = Maximum Achievable, and Min = Median - (MaxAch - Median) = 2*Median - MaxAch. That would give a range from 21 to 52 for Aki basho.

This is exactly how I always do it for the GISP game.

[Bealzbob]

That, in turn, probably makes it nearly impossible to score better than 13-2 even with a perfect team though, as the practical maximum for last basho was just 52, and that was with an injury-assisted 15-point score in there. By your rule this would fix the 15-0 score to 56 points if a player scored perfectly. Looking back through the current year, the achievable maxima were:

Aki 52 (one zero-win injury)

Nagoya 51 (one zero-win injury)

Natsu 56 (two zero-win injuries)

Haru 48 (no relevant injuries)

Hatsu 49 (two-win and one-win injuries)

(The corresponding minimum scores are 60-Max, of course.) I think using the actually-possible max/min scores would provide better baselines. Other than that, I find your suggestion a lot more meritorious than mine. :-)

Yes but it's not like 15-0 is a common score in real Ozumo so you shouldn't be expecting it that often in the game. And if someone DID get 15-0 in the REAL basho and you had selected them, then that would consequently also bring you closer to your 15-0.

[Bealzbob]

Actually, hold on I think you have me wrong. If someone achieved a score of 52pts then the 15-0 would be half of (52 from 60) added to 52 which is 52+4=56pts. So they would virtually be there. I reckon with a score of 52pts you would have a record in my calculations of 14-1 which I think is very reasonable.

Actually, a quick calculation as follows based on someone getting 52 with someone getting 28

Madeupiazuma  (52) 13-2
Zenjimoto	 (43) 10-5
Itachi A	  (43) 10-5
Kintamayama   (43) 10-5
Vikanosato	(41) 9-6
Miyahakuho	(41) 9-6
Toridefuji	(40) 9-6
Itachi C	  (39) 9-6
Itachi B	  (37) 8-7
Torideyama	(37) 8-7
Chiyomitsuki  (36) 8-7
Chikenshitenzan (35) 7-8
Debeso		(34) 7-8
Chiisabuke	(33) 6-9
Shobaionna	(33) 6-9
Vikanoshima   (32) 6-9
Vikanohana	(31) 6-9
Asanoumi	  (30) 5-10
Toridenoshima (28) 4-11

Edited November 2, 2007 by Bealzbob

[Asashosakari]

Actually, hold on I think you have me wrong. If someone achieved a score of 52pts then the 15-0 would be half of (52 from 60) added to 52 which is 52+4=56pts. So they would virtually be there. I reckon with a score of 52pts you would have a record in my calculations of 14-1 which I think is very reasonable.

Well, that still depends on what the worst player did in your model. If it's a high-scoring basho overall, I don't think the best-possible record would necessarily be 14-1. Anyway, I disagree on principle...the theoretically optimal picks should be rewarded with the best-possible score, i.e. 15-0, every time. Are you really meaning to say that in Haru basho, somebody who achieved the best-possible 48 points should not have any chance to score better than 12-3 or perhaps 13-2?

Edit: Speaking of high-scoring tournaments, they'd also ensure that nearly everybody would have kachi-koshi. Imagine a basho where everybody scored between 36 and 56...

Edited November 2, 2007 by Asashosakari

[Bealzbob]

It depends on your 'ethos' for the game being played. If your picks got nowhere near zensho then I would say that the winner of the yoko-beta competition would/should win with a score that reflects that. I think it makes sense that a low scoring basho has a low scoring competition winner.

It's up to you really. It's easy to make a scale where the winner always gets 15-0, but I don't think that reflects reality and would prefer winners regularly getting 13s and 12s. It's more meaningful then when someone actually DOES get a 14 or a 15.

But it is personal choice. Was just trying to help.

[Asashosakari]

It depends on your 'ethos' for the game being played. If your picks got nowhere near zensho then I would say that the winner of the yoko-beta competition would/should win with a score that reflects that. I think it makes sense that a low scoring basho has a low scoring competition winner.

It's up to you really. It's easy to make a scale where the winner always gets 15-0, but I don't think that reflects reality and would prefer winners regularly getting 13s and 12s. It's more meaningful then when someone actually DOES get a 14 or a 15.

I think we're talking past one another here. I'm not saying that the winner should always get 15-0, just that the theoretically best-possible score should always equate to 15-0. It's going to happen rarely enough anyway that somebody's picks will be that close to the optimum; your system ensures that even with optimal picks, 15-0 will be virtually impossible (basically, unless it's a basho where both Juryo and Makuuchi are won with 14 or 15 wins, and each division had a rikishi bowing out with zero wins). Putting the 15-0 bar at "unachievable even under the best of circumstances" just strikes me as very strange. You might as well say outright that no player will ever be credited with 15 wins.

Anyway, I'm not trying to be combative here, really :-) - just trying to figure out if I'm misunderstanding you or if you've overlooked something in your proposal.

(Futilely edited for clarity...)

Edited November 2, 2007 by Asashosakari

[Bealzbob]

Hey no-one's being combative and I'm not taking you to be :-) We're discussing it so Zeokage can pick what he wants to do. I get what you are saying now. That the winner doesn't get 15-0 automatically, but that the perfect picks for any given basho would get 15-0. To be honest that makes sense for an online competition although I still (sorry if I'm coming across as pig-headed) think that winning with 12s and 13s makes the eventual 14s and 15 more meaningful. In reality the winners will always be getting very close to the perfect score, save for a few points here and there.

Anyhoo, I think we've given Zeokage much to think about (Nodding yes...)

[Asashosakari]

In reality the winners will always be getting very close to the perfect score, save for a few points here and there.

They didn't in Aki. :-) More seriously though, I do think it's a valid concern given that the game requires only four picks, making it much easier to score near-perfectly than, say, in GISP with its nine picks. Unfortunately, that's probably something that can only be figured out by more trial runs (unless the previous AFY results can give a clue toward this)...

[Zeokage]

Unfortunately, the previous AFY archives are inaccessible as far as I can tell. Kintamayama has more actual experience of this game, so maybe he can provide some insight here.

[Holleshoryu]

Makuuchi wins: Baruto

Makuuchi loss : Hakurozan

Juryo wins: Ichihara

Juryo loss : Shunketsu

[Zeokage]

Looking back through the current year, the achievable maxima were:

Aki 52 (one zero-win injury)

Nagoya 51 (one zero-win injury)

Natsu 56 (two zero-win injuries)

Haru 48 (no relevant injuries)

Hatsu 49 (two-win and one-win injuries)

(The corresponding minimum scores are 60-Max, of course.) I think using the actually-possible max/min scores would provide better baselines. Other than that, I find your suggestion a lot more meritorious than mine. (In a state of confusion...)

Edit: On second thought, using that minimum value might boost the low records too much, as the scoring range would actually expand towards the bottom compared to your suggestion...hmm. Maybe one could do Max = Maximum Achievable, and Min = Median - (MaxAch - Median) = 2*Median - MaxAch. That would give a range from 21 to 52 for Aki basho, leading to top/bottom scores of 11-4 and 4-11 (I think)...

Here's the leaderboard from the Nagoya basho the previous results were based on. Am I right in thinking that Kyokutenho and Shimotori both were out before the first day? In Kakizoe's case, should his record go in as a 0-12-3 or 0-15? I've had the MaxAch as 49 when trying to figure this out, but I think I've been missing out on those who went kyujo later on in the basho in that calculation.

MakuW: Kotomitsuki, Asasekiryu, Dejima (12)

MakuL: Kakizoe, Kokkai, Tokitsuumi (12, or just Kakizoe with 15?)

JuryoW: Baruto (14)

JuryoL: Ryuho, Takanowaka, Tochisakae (11, or just Tochisakae with 15?)

So MaxAch is either 49 or 56. I assume that Median is the median of all the players' scores? In this case it was 36.5, so Min is either 17 or 24. Using that range, is it right to have equal ranges for each resultant score?

EDIT: I gave it a go, assigning 55-56 as 15-0 and 17-18 as 0-15. 36 points in between with 14 ranges. 2.57 per range... here are the scores it came up with:

Zenjimoto	 (43) 10-5
Itachi A	  (43) 10-5
Kintamayama   (43) 10-5
Vikanosato	(41) 9-6
Miyahakuho	(41) 9-6
Toridefuji	(40) 9-6
Itachi C	  (39) 9-6
Itachi B	  (37) 8-7
Torideyama	(37) 8-7
Chiyomitsuki  (36) 7-8
Chikenshitenzan (35) 7-8
Debeso		(34) 6-9
Chiisabuke	(33) 6-9
Shobaionna	(33) 6-9
Vikanoshima   (32) 6-9
Vikanohana	(31) 5-10
Asanoumi	  (30) 5-10
Toridenoshima (28) 4-11

Have to ask here, what are the possible tiebreakers? Torideyama had these, but I don't want to emphasise one selection over others.

Tiebreakers idea:

TB1 Most Juryo wins

TB2 Most Makuuchi wins

TB3 Most Makuuchi losses

TB4 Most Juryo losses

TB5 Lowest ranked Juryo winner (J10E wins over J9W)

TB6 Lowest ranked Makuuchi winner

TB7 Highest ranked Makuuchi loser

TB8 Highest ranked Juryo loser

TB9 Highest rank on A4Y banzuke (or rock-paper-scissors) for the first basho.

Edited November 3, 2007 by Zeokage

[Gusoyama]

I'm late to the game, but for Salarycap, I used to just take the highest score, make that 15-0, the lowest score 0-15, and divide from there. But I didn't like that, so I came up with my current scoring method, which is Wins=((MaxScore-250)-Median/7)+Median, and I've been pretty happy with the spreads. For the previous AFY scores, we'd have:

	
Max: 52 Median: 36.44444444	(Max-Median)/7: 1.285714286
Score Loss  Wins
43	2	13
43	2	13
43	2	13
41	4	11
41	4	11
40	5	10
39	6	9
37	7	8
37	7	8
36	8	7
35	9	6
34	9	6
33	10	5
33	10	5
32	11	4
31	12	3
30	13	2
28	14	1
Avg Score	7.5

That's not a great spread, as it seems to bulge around the ends rather than the middle, but then again, so do the scores.

Edited to add: I subtract 250 from my max because I was getting too many 10-5, 9-6 win yushos. Yes, its arbitrary, but it seems to work well.

Edited November 6, 2007 by Gusoyama

[Kintamayama]

Unfortunately, the previous AFY archives are inaccessible as far as I can tell. Kintamayama has more actual experience of this game, so maybe he can provide some insight here.

The game underwent constant changes, but I played this format no more than 4-5 times. And those bashos were pretty even, so AFAIR, no perfect scores or anything, maybe a 13/14. I think Zenjimoto had far more experience than I did with this game.

But I really don't remember for sure.

[Gusoyama]

Oh, and my picks.

Makuuchi:

Wins: Baruto

Losses: Miyabiyama

Juryo:

Wins: Iwakiyama

Losses: Sagatsukasa

[Zeokage]

I think having a larger number of entries is in order... I remember the original AFY not having that many, but it seems that 18 isn't a large enough sample from one basho. To get some more, and if people have the time, everyone can enter as many times as they like. :)

Another option is letting people choose more per group, say guess 3 for each option for a total of 12 rikishi.

[XiaoTan]

And here's my take

Makuuchi wins: Tochiozan

Makuuchi loses: Wakakirin

Juryo wins: Iwakiyama

Juryo loses: Yoshiazuma

Neko

[Susanoo]

Makuuchi wins: Baruto

Makuuchi loses: Tosanoumi

Juryo wins: Shimootori

Juryo loses: Yoshiazuma

[Zeokage]

Kintamayama

Washuyama

Holleshoryu

Gusoyama

XiaoTan

Susanoo

Just 6 entries, one from each of you. Remember this is only a beta, so enter as many as you like! In a way I need mass entries to check the spread of W-L records in a basho. We had 18 total entries last time around.