PhorCillic

[Open Alpha] Rikishi Game

Recommended Posts

Randomitsuki    1,280

I was wondering whether the game is probabilistic or deterministic. Say, that I have a 51% chance of winning (based on the green bar). Does it mean that I definitely win, or does it only mean that I have a 51% chance of winning?

Share this post


Link to post
Share on other sites
PhorCillic    51

It means points of your rikishi are 51% of the both rikishis sum.
Even 1 EP spent by the opponent may twist the result then.

Looks like when you have about 65% of the bar, you may clearly win with opponent unless he spend 2-3 EP or something really weird happen.

Just look at last Yoemon-Jakusotsu bout. Little advantage and 2 EP didn't help Yoemon against 4 EP by Jakusotsu O_O

Or many mine bouts (usually Koenagi), with 3 EP spending and... defeat.

Edited by PhorCillic

Share this post


Link to post
Share on other sites
Gurowake    1,450

I was wondering whether the game is probabilistic or deterministic. Say, that I have a 51% chance of winning (based on the green bar). Does it mean that I definitely win, or does it only mean that I have a 51% chance of winning?

I don't think the size of the bars have anything to do with probability. You can look at the results of each match and see how much the outcome was influenced by luck and effort, and how much effort each rikishi spent. I think Phor has said that the luck component can only go as high as 3/4 of an EP. Based on what I've seen, it doesn't take all that much of an advantage in the size of the bar to be nearly certain of victory given no effort being spent, but since effort is a larger component than luck, weaker rikishi that spend lots of effort can easily win matches against strong "lazy" rikishi.

One thing I'm not sure of, and I'm not sure if I can adequately do a controlled experiment on given the lack of numerical data made available, is whether effort spent scales linearly with the effect it has on the results. It certainly would be easier to implement that way, but I could also imagine that there are diminishing returns and implementing it would probably only take a single line of code.

edit: Actually, the percentages can be found in the html source to 12 decimal places. Interesting....

Edited by Gurowake

Share this post


Link to post
Share on other sites
Gurowake    1,450

It means points of your rikishi are 51% of the both rikishis sum.

Even 1 EP spent by the opponent may twist the result then.

Looks like when you have about 65% of the bar, you may clearly win with opponent unless he spend 2-3 EP or something really weird happen.

Just look at last Yoemon-Jakusotsu bout. Little advantage and 2 EP didn't help Yoemon against 4 EP by Jakusotsu O_O

Or many mine bouts (usually Koenagi), with 3 EP spending and... defeat.

It might just be a use of a stock phrase, but I don't think we can "just look" at the breakdown of the results of matches involving no rikishi in our stable. I can only see the bars pre- and post-match for bouts in which I had a participating rikishi in my stable.

edit: I realized I could log in to the demo account to see things from its perspective. It's giving me a whole lot more data to work with as well.

Edited by Gurowake

Share this post


Link to post
Share on other sites
Gurowake    1,450

One of my Maezumo rikishi fought in Jonokuchi on Day 4 and I was given money for the win as if he was a Jonokuchi rikishi. This may be totally intended behavior, but quite surprised me.

Share this post


Link to post
Share on other sites

A short look at the Number of Leaders after 4 Days should at least make an easy torikumi in Jonokuchi:

With currently 8 undefeated Rikishi these (still) undefeated should fight against each other on the last 3 Days....and we should have a sole 7:0 Yusho there....

Share this post


Link to post
Share on other sites
Ganzohnesushi    151

A short look at the Number of Leaders after 4 Days should at least make an easy torikumi in Jonokuchi:

With currently 8 undefeated Rikishi these (still) undefeated should fight against each other on the last 3 Days....and we should have a sole 7:0 Yusho there....

Theoretically.... there are two Rikishi from Bembel Beya involved and as Rikishi from the same Beya shall not fight against each other we still could have both Bembel Boys at 7-0.... Now watch the fragmentation of Ganzohnemaki tomorrow.

Ganzohnesushi

Edited by Ganzohnesushi

Share this post


Link to post
Share on other sites
PhorCillic    51

It means points of your rikishi are 51% of the both rikishis sum.

Even 1 EP spent by the opponent may twist the result then.

Looks like when you have about 65% of the bar, you may clearly win with opponent unless he spend 2-3 EP or something really weird happen.

Just look at last Yoemon-Jakusotsu bout. Little advantage and 2 EP didn't help Yoemon against 4 EP by Jakusotsu O_O

Or many mine bouts (usually Koenagi), with 3 EP spending and... defeat.

It might just be a use of a stock phrase, but I don't think we can "just look" at the breakdown of the results of matches involving no rikishi in our stable. I can only see the bars pre- and post-match for bouts in which I had a participating rikishi in my stable.

edit: I realized I could log in to the demo account to see things from its perspective. It's giving me a whole lot more data to work with as well.

Demo beya is... well, as the name suggests. Every person in the world may look. But I need to work a bit at its permissions (I watch the Mizuno-Jakusotsu bout and decided to block EP for Demo) :P

Also, I am working on an option to enable/disable results of non-your rikishi. Misha is a tester of that.

One of my Maezumo rikishi fought in Jonokuchi on Day 4 and I was given money for the win as if he was a Jonokuchi rikishi. This may be totally intended behavior, but quite surprised me.

My double fail. First, the script hadn't found an opponent for weak rikishi and started looking in lower division.

Second, when I checked if it's everything okay, I fixed the situation in Jd/Jk but didn't saw the situation in Jk/Mz.

A short look at the Number of Leaders after 4 Days should at least make an easy torikumi in Jonokuchi:

With currently 8 undefeated Rikishi these (still) undefeated should fight against each other on the last 3 Days....and we should have a sole 7:0 Yusho there....

And script should set the bracket for them. At least that way it was programed now. :P

  GOLYNOSATO ╗
             ╠╗
       HANDA ╝║
              ╠╗
GANZOHNEFUGU ╗║║
             ╠╝║
     KOEMURA ╝ ║
               ╠
      MIAUMI ╗ ║
             ╠╗║
    KUWABARA ╝║║
              ╠╝
       KAMIO ╗║
             ╠╝
GANZOHNEMAKI ╝

And propably it will be in next bashos. After a half of basho (D4 or D8 when we achieve Juryo) a switch from 'banzuke suggestion oriented' to 'record oriented'.

Heh, And I started to guessing, who will win. XD

The thing is, I should propably start thinking about how to "split" divisions into two groups...

Edited by PhorCillic

Share this post


Link to post
Share on other sites
Gurowake    1,450

It looks as though a new banzuke has been made in that the ranks of the rikishi have changed as viewed as members of their stable, but the banzuke link on the sidebar does to the old banzuke.

Edited by Gurowake

Share this post


Link to post
Share on other sites
PhorCillic    51

Forgot to change the link in Menu. Fixing ASAP.

This time it was only one problem with banzuke. Script tried to set a Mz rikishi into Jd.

To make things a little easier, 2 matches are required to be "out of Maezumo"

Edited by PhorCillic

Share this post


Link to post
Share on other sites
Gurowake    1,450

Tsubouchi was ranked lower than Pikayamabe, had the same record, and is now ranked higher.

I believe every instance where there are Jk and Mz rikishi with the same record in the previous basho, the Mz rikishi is ranked higher. This seems like the exact opposite of what it should be.

It appears as though the rikishi are ranked by (wins - losses), and that works fine if everyone has a full schedule which is the case everywhere but Mz, but it means that a Mz rikishi that went 0-2 ends up higher ranked than a Mz rikishi that went 2-5. Given how the pairings are done for rikishi that enter Mz in the middle of the tournament, with new rikishi facing those without any wins yet, it makes little sense to think that a rikishi that went 0-2, facing the dregs of competition but not winning, would be ranked higher than a rikishi that managed two wins but had more losses. Days missed by new rikishi because they didn't exist yet should be treated as losses, not a half a win.

There are other things you might do to implement the same methodology as the Kyokai for integrating Mz rikishi onto the banzuke for the first time, but given 7-day Maezumo and the arbitrariness of the Kyokai's methods, I'm perfectly fine with how you've handled it with the exception of the above items.

Edited by Gurowake

Share this post


Link to post
Share on other sites
Ganzohnesushi    151

Surprised to see some (from my point of view) weird Banzuke rankings...

Sd7e Omate --> 3-4 --> Sd8w

Sd7w Higoshi --> 3-4 --> Sd9e

Sd10w Ganzohnesake --> 4-3 --> Sd10e

To be honest, I don't understand how Omate and Higoshi can be still ranked higher than Ganzohnesake.

Another even worse example:

Jd4e Ganzohnetofu --> 3-4 --> Sd15w (equivalent to Jd5w after extension of Sandanme by 5 ranks)

Jd 18w Mischanofuji --> 7-0Y --> Jd2w.

Unbelievable that Ganzohnetofu is still ranked higher than yushoing Mischanofuji. It is a must for 7-0 Rikishi to be promoted to the next division even if they come from the bottom of that division. Only exception from this rule in reality are the ranks from Ms16e - Ms60w.

In general I think Mk Rikishi are treated way too lenient.

Ganzohnesushi

Edited by Ganzohnesushi

Share this post


Link to post
Share on other sites
PhorCillic    51

Looks like I would have a rough night then (3 AM here)

Trying to get what happened by looking at the imp-banzuke page. (LINK, password: sumoforum)...

Tsubouchi was ranked lower than Pikayamabe, had the same record, and is now ranked higher.

*looks at imp-banzuke test page*

*looks at generated banzuke*

*screams internally*

The script is screwing again. I have about 4 hours to fix it before the convention wake up. So let'sedit "movesteps" a little... :)

Also, reverted that banzuke and trying to get the way. Another bug jumped out, this time with Mz in Jd. Hmm...

EDIT

Looks like on the test page looks better. Time to publish or another problems arrived?

EDIT

I think at the moment I really need a "Banzuke Script Correction Council" or something... XD

Edited by PhorCillic

Share this post


Link to post
Share on other sites
Gurowake    1,450

The torikumi script's method of dealing with two rikishi at the bottom of the division who can't face each other seems incorrect. It decides to bring up undefeated rikishi from the next lowest division instead of rikishi with the same record as them. I can understand that you may not want to do things exactly the same as the Kyokai for reasons of not having enough rikishi in the system, but what this workaround does completely messes with the Yusho race in the lower division.

Rikishi at the bottom of each division should be matched with rikishi from the top of the next one with the same record if there's no one left in their division with their record, regardless of what the record is. The only real exceptions to this are dealing with the extreme cases where especially the last day the distance between the rikishi gets to be huge, and with the small size of the divisions you possibly could have a two-division gap. There shouldn't be any concern about matching undefeated rikishi across different divisions; while it's slightly unfair from the Yusho race perspective for those near the top of the division, the Yusho race really doesn't matter compared to having rikishi fight the ones they're supposed to: a rikishi "close" in rank to them with the same record. On the last day you might have to match them with guys in the next division with one less win because the distance to the next undefeated guy is just too big (ie, more than one division), but that's the only case I would do so for. It's a lot uglier to see two 0-5s matched against two 5-0s than it would be to see rikishi well apart in rank but with the same record get matched.

I know there's a lot of corner cases to deal with, especially since the number of rikishi is low. This is why it's still in Alpha, so we can figure out what can be improved. I may try to write some pseudo-code that tries to deal with all these issues, but no promises.

  • Like 1

Share this post


Link to post
Share on other sites
Gurowake    1,450

X can face Y is they are in the same division or adjacent divisions, are not in the same heya, and have not met this tournament.

while unpaired rikishi exist {
let First be the highest ranked unpaired rikishi with the most wins. Is there another unpaired rikishi with the same number of wins that First can face?
Yes: pair First with the highest ranked unpaired rikishi with the most wins that First can face.
No: Are there any unpaired rikishi First can face?
No: Are there any unpaired rikishi at all?
No: Give First the day off (or add a Demo rikishi, or whatever; there were an odd number of rikishi and First is the odd man out)
Yes: Find the most recently paired match where First can face one of the rikishi. Break that pairing.* If there's only one rikishi in that pairing that First can face, pair him with First. Otherwise, did the two rikishi in that pairing have the same number of wins?
Yes: pair First with the higher ranked one.
No: pair First with the one with more wins.

Yes: So there's someone we can pair First with, we just need to find the best one. Let i = 1. Is there a rikishi with i less wins than First that First can face?
Yes: Is there a rikishi ranked higher than First with (i-1) less wins that is already paired?
Yes: Let that rikishi be Last. Is there a rikishi that First can face with i less wins ranked higher than Last?
Yes: Pair First and the lowest ranked unpaired rikishi that First can face with i less wins ranked higher than Last.
No: Pair First and the highest ranked unpaired rikishi that First can face with i less wins ranked lower than Last.

No: Is there an unpaired rikishi First can face ranked higher than First with i less win?
Yes: Pair First and the lowest ranked unpaired rikishi that First can face ranked higher than First with i less wins.
No: Pair First and the highest ranked unpaired rikishi that First can face with i less wins (We've established that such a rikishi exists already.)

No: Increment i. Since we already determined there's someone that First can face, and First had the most wins among unpaired rikishi, this will eventually find one.
}

*You may need to add some additional logic so that the algorithm doesn't go in circles breaking and creating the same few pairings repeatedly when there's only a few rikishi left and many can't face each other. You probably need to add (in theory) a iterative breaking of pairings and creating of new ones that haven't been tried before, breaking another pairing further back in the pairings history if you have exhausted all possibilities. Guaranteeing that this procedure terminates is very, very difficult, if not impossible, and is beyond the scope of this algorithm. Given what I've seen in the past, when this becomes an issue the Kyokai probably randomly shuffles the lowest ranked rikishi doing poorly around until something works.

While the Kyokai does not mix Maezumo participants with rikishi already on the banzuke, there's no particular reason why you would have to do that in your simulation, especially since your simulation of Maezumo is significantly different already. If you decide to do a more realistic simulation of it, here's how it would go:

Require all entrants to pass a "physical exam" the day before the tournament. This would mean that no one recruited during the basho is able to participate in Maezumo until next basho.

First Day of Maezumo: Randomly pair the participants. If there's an odd man out, he faces someone that previously lost.

Subsequent days: Use the pair-by-record algorithm, although quite often you'll have to cross records when numbers are low. Anyone who has previously won two matches and loses gets an additional match each day. If necessary to even out numbers, draw someone from the one-win group that lost on the day. Anyone who wins three matches "qualifies" and is done.

Days continue until it is deemed sufficient. I don't know how to formalize this, and often we don't know if there is actually going to be another day of it or not. Probably 5 days is enough, with less for fewer rikishi. If there are very few rikishi like this basho, they might just have one day where they have one of each possible match and be done with it. Note that there's nothing prohibiting repeated matchups; although they normally avoid them, with the small number of participants there's sometimes little choice.

(The system changes for Haru when there are a much larger number of new recruits. They get split into two groups that alternate days, and only need two wins to "qualify".)

The whole idea of potentially having multiple matches each day kinda throws things from the way you have things set up, so you could just ignore that part of the Kyokai's methods. You probably want to have Demo rikishi standing by to take care of odd-man-out situations.

--

I hope this was at least somewhat helpful, and look forward to any improvements that you may make.

  • Like 1

Share this post


Link to post
Share on other sites
PhorCillic    51

First time when Mz-Jk matches happened, I started thinking about the Shinjo rule (so new ones may quickly arrive to Jonokuchi). It could be good at start but... :P

I struggle a little with matchmaking script. Propably this basho again I stick to the current version + manual edits (it's max 6 matches daily)

Maezumo thing and physical exams... that will be good for the second August basho. For that I'll propably prepare some stats (Height, Weight etc.) and (propably) rework a little stats system.

Edited by PhorCillic

Share this post


Link to post
Share on other sites
PhorCillic    51

GRAND TOPIC REVIVAL

(suggested by people from sumoforum playing Rikishi Game)

Next basho (October Basho 2015) will begin 12th October 2015. A new Banzuke has been announced, with (finally) having enough rikishi to create a higher division - this time it's Makushita. Thanks to that fact, working on 15-day system has begun. :)

Due to topic revival, each stable will get additional sum of yen. considering the fact, some of oyakatas didn't visited his stables by months and their rikishi still gathered money... You may be shocked a little by the number.

Also, this will be the last basho without "medical exams session".

Other things which changed?

- Layout being 'nicer'

- Options menu (hammer/wrench icon near the stable name) with additional information to show both in the torikumi and the banzuke.

Share this post


Link to post
Share on other sites
Gurowake    1,450

Some things about the most recent banzuke:

Koemagae, Ganzohnenori, and Koemura were all 6-1, and they are now ranked in the opposite order from previously. Winless Mz Mi-Chiro is also ahead of many 1-win Mz participants. Everything else looks consistent, although I only checked each record against itself and am assuming that no one got promoted past someone with a better record (and likewise demoted past someone with a worse record) except for the Mz as noted.

Both 7-0s seem to not be promoted high enough. 7-0 Mischashima ended up only 8 ranks higher than 6-1 Taiyonoumi, and started 7 ranks higher. 7-0s should be worth way more than 6-1s, up to double the movement.

Mz should all be at the bottom of Jonokuchi, with the exception of those who did not get any wins in Jonokuchi being ranked lower (but still higher than winless Mz).

Share this post


Link to post
Share on other sites
PhorCillic    51

Okay. Remember you have an access to "banzuke script debug". ;)

About 7-0 promotion: I know it's larger a bit than in previous versions. A good adjustment... and we're set. :)

About Mz-at-the-bottom: In this basho I gave almost all the Mz the same "ranking position" for a test. It should fix that problem nicely..

Share this post


Link to post
Share on other sites
PhorCillic    51

November (Kyushu) Banzuke released. There were some problems (script for a long time tried to set Sandanme Yusho winner on top of the Banzuke, and still placed him in higher-Ms)

This will be the first 15-day Basho. Hope all scripts will work :)

Edited by PhorCillic

Share this post


Link to post
Share on other sites
PhorCillic    51

December banzuke released 3 days ago.

I am again focusing on bug finding/squashing and... getting some money for better server available for all (Pandaazuma?)

Share this post


Link to post
Share on other sites
PhorCillic    51

SURPRISE! :D

I've managed to get a way better server AND .net domain!

Because of that, there were some bugfixes, additions etc.

Welcome CronTab - finally, matches will be played on specified hour. Always. :)

To test CronTab... well... sparrings will be added. But no force yourselves/other players into them - Taking part in them won't give a player other profits than "a chance to test his rikishi". No yen and no EP given for sparrings. So "basho-oriented players" may still be "basho-oriented" and still play well. :)

www.rikishigame.net

Edited by PhorCillic
  • Like 1

Share this post


Link to post
Share on other sites
PhorCillic    51

Haru Basho started.

Banzuke: http://rikishigame.net/en/banzuke/

Match generator has been rebuilt. It gave somewhat uncommon results in second turn (days 3-4: 1-0 low Makushita faces 1-0 mid Sandanme O_O), but I fixed that bug already.

Despite highest efforts, I didn't managed to improve

Also, dear "Rikishi game prominent experts" (Gurowake, Golynohana, Ganzohnesushi, Mischashimaru, anyone interested), I would like to know your opinion in those articles:

http://rikishigame.net/en/additional-awards/- reintroducing the Sansho and additional awards idea

http://rikishigame.net/en/pikachu-beya-pikachu-cup/- a tournament made by one of the players. Which... well, didn't got that much attention. (Yes, oyakatas may write their articles. And I thought about rewarding them if it were good, but that's still only an idea)

http://rikishigame.net/en/phoenix-tiger-oyakatas-ideas/- couple of ideas made by one of the younger players. Discuss!

Share this post


Link to post
Share on other sites
Ganzohnesushi    151

Can't open the page currently. Message since yesterday evening says

"This site has been suspended."

Ganzohnesushi

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now