SURVEY: Rules and Variance

Japanese Reach Mahjong Rules. Strategy, news, sets - anything!

Moderator: Shirluban

Post Reply
HotelFSR
Expert Reacher
Expert Reacher
Posts: 328
Joined: Sat Jan 17, 2009 11:39 am
Location: Federated States of Micronesia

SURVEY: Rules and Variance

Post by HotelFSR » Sat Jan 17, 2009 12:53 pm

A few thoughts and questions for everyone to discuss about rules. I\'m particularly interested to hear what our resident pros have to say.


1. Do you think Reach competition rules are likely to change in the near future? If so, in what way and why?

2. What are your least favorite rules and why? What would you change if you had the chance?

3. Some patterns seem to score too low. Easy patterns + Dora are much more effective. What are your thoughts on this?

4. What do you think of the Saikouisen Classic Rules? (A-Rules with No Oka, Dealer Must Win to Continue, No Ready Payments on Draw Rounds)

5. What would you say the views of most pros are on these matters?


I ask this after extensively studying Reach statistics along with pro replays from various websites and even having some mathematical simulations run to test out a few things. As a statistician I\'m interested in these things.

One thing that jumps out at me is that variance in competition rules (i.e. fluctuation in player ratings supposed to reflect skill) is a little higher than it should be.

Generally I think the issue has to do with a bias towards self drawn wins.

One culprit appears to be Menzen Tsumo. It happens a very large percentage of the time and is essentially random. Since it is a self draw bonus, it also effects losing players indiscrimiately regardless of defense. Dealers are hit especially hard.

Why remove Ippatsu but not this? I hear that Ippatsu was removed because it is not implicit in the value of the hand, but neither is Menzen. It randomly increases the hand size but Menzen is apparently a lot worse. What is the reasoning behind this? You can still win with a no-yaku hand using Reach.

Menzen also favoritizes hands like Pinfu, which are already dominant enough! Surely it should not be used in competition, which is supposed solely to determine the most skilled player?

Second, the other big issue seems to be the dealer system. This causes a lot of unneccessary variance and, dare I say it, frustration. It would make more sense to have the dealer win/lose more on discard wins/losses, but it just amounts to another self draw bonus with no skill-based disadvantage. Much more significantly, it creates the problem where some dealers are able to continue many times, and others instantly lose out on luck of the draw. Given that each player has only two shots at being dealer, this has a massive random impact on the final scores. If many more hands were played in each game, this would be ironed out more- but why add this extra variance in the first place? It has very little to do with skill anyway. Each player should be allowed to have equal interest in each hand, it\'s also more fun!

The quality of starting hand you are dealt naturally has a massive impact on the game, but with the dealer system unneccessary variance is added to the game because of the big reward received if a good starting deal just so happens to coincide with your turn as dealer. Possible improvements include disallowing dealer continues altogether, making the dealer win/lose more only on discard (so that even non-dealer players stand to benefit on each hand), or just simplifying and getting rid of the dealer scoring system. The dealer still goes first and has the possibilty of double wind tiles. Is that not enough?

The dealer continues are also inconvenient because they cause the game time to vary dramatically, which is not good for tournaments or even for casual play sometimes.

That\'s pretty much all my thoughts on this for the moment.

What do you think?

My pro replay analysis and statistical simulations using real game data suggest that the gap between high level and low level players would be significantly greater if these factors were removed. Thus skill would be rewarded more. Surely something is wrong when top pros place on average barely above 2.5? There is too much variance at the moment for pros to really push ahead.

I have nothing against this variance for gambling or fun rules, but competition rules are a different story. This may have been overlooked in the development of the game since in old-school Japan statistical analysis was not fashionable, but I think some evolution in this area of rules is intellectually necessary for widespread international acceptance of Mahjong as a competitive game of skill.

Let\'s start the discussion!

pringle
Fresh Reacher
Fresh Reacher
Posts: 10
Joined: Thu Jan 22, 2009 2:53 pm

Re:SURVEY: Rules and Variance

Post by pringle » Sat Jan 24, 2009 3:28 am

I dont like my stat teachers but I do have interent in observing statistic. Would be nice if you can show what you have.

I don\'t think Menzen tsumo is all about luck, and I think it is helping an unlucky player to be able to win with a small hand. Going only for a big hand requires better luck and probably more amateur way. For unlucky player who only get a cheap hand, it is still possible to win by using skill to avoid throwing into someones hand a go for a cheap hand with Menzen.

Ippatsu is more about luck.

zzo38
Senior Reacher
Senior Reacher
Posts: 162
Joined: Sun Mar 15, 2009 2:17 am

Re:SURVEY: Rules and Variance

Post by zzo38 » Sun Mar 15, 2009 5:49 am

1. I don\'t know about competition rules.

2. I would delete \"all green\", \"big wheels\", \"heaven hand\", \"earth hand\", \"human hand\". And menzen tsumo would be virtual yaku only (it counts 1 han for purpose of requirement but not for scoring, so it would be possible to win with a 0 han hand). And ippatsu would count only for winning by ron, not for tsumo win. Also, no kanuradora (but keep uradora and kandora).

3. I currently have no opinion on this.

4. I prefer to have payments for tenpai in a draw game, but the payment amount could be made less.

5. I don\'t know any pros.

Senechal
Senior Reacher
Senior Reacher
Posts: 175
Joined: Wed Mar 04, 2009 5:42 am
Location: The frozen part of HELL!

Re:SURVEY: Rules and Variance

Post by Senechal » Mon Mar 16, 2009 11:43 pm

Ippatsu is not luck. While given the current direction of discussion (and I\'m not saying it\'s right either), some waits are better than others. Unless you\'re playing by a forced riichi rule, there are multiple ways of waiting on tiles. If everybody waited only on one tile or a two-side wait (say 4-7p) you probably could start making that argument. This conclusion can apply both for tsumo and ron wins with the following exception on ron wins.

What ron wins have on top of this is the fact that you can\'t just say \"I\'ll deal into the hand now to keep the lead\". This is the problem with attempting to purify the game to a point where if one person makes a big lead, they could just run with the game fearlessly. What you\'d get when you take out hands and the luck factor would be games where your position over time through the game is less variable, it wouldn\'t have any incidence on who\'s a pro, a high level amateur or a novice as the game dynamic would be oriented to win fast and then sit on your laurels.

Again, I don\'t have numbers to back this up but the OP had none, so baseless speculation can be made. One thing I do have a problem with is trying to tinker with basic game mechanics to satisfy one\'s own bias toward game improvement. To the point to say that \"well, sure, win with menzen but we won\'t count yaku\" (so a 0 point hand or fu * 2^2 * [4,6] = 500, 200-300 or 300 all) will denature the game to closely resemble \"Hong Kong with Furiten\". I think that\'s been tried before. It will probably be tried again. For people to dream that pros can place 90% 1st however is as logical as dreaming that baseball players can hit 90 home runs per season, every season. Without both objective and exhaustive studies, it\'s an emotional response, nothing more, nothing less.
Club Riichi de Montréal (Canada) http://riichi.ca/ (If you're from elsewhere, keep in touch with us too!)

gartheee
Mahjong Pro
Mahjong Pro
Posts: 37
Joined: Mon Mar 16, 2009 9:42 am
Location: Japan

Re:SURVEY: Rules and Variance

Post by gartheee » Fri Mar 27, 2009 2:31 pm

One thing I do have a problem with is trying to tinker with basic game mechanics to satisfy one\'s own bias toward game improvement.

I couldn\'t have said it better myself. Though apparently it\'s not enough to stop me from saying more about it myself.

First of all I\'m interested to know how you arrive at the conclusion that

variance in competition rules (i.e. fluctuation in player ratings supposed to reflect skill) is a little higher than it should be.

Indeed, the fact that all these donkeys beat me is obviously illuminating a deficiency in the rules, but I would still like to know what how and what numbers were used to get to this statement so I know exactly how to chap them when they beat me.

Remember, this is Mahjong, not Chess or Go. It is at its heart, a gambling game and so there must be things left up to chance. You want a game where the pros win 90% of the time? OK, let\'s flip all the tiles over so everyone can play his game perfectly. The pros would probably win most of the time but it would sure be a boring game. Weaker players are willing to play against the pros because there is a possibility that they might win if they get lucky.

I\'ve spoken a little about the issue of Concealed Self Draw in another post so I\'ll leave that there. Another of your issues is with the dealer bonus and continuation. Sure it is a little arbitrary and players who happen to get big hands when it\'s their turn as dealer seem unduly rewarded. But then it\'s also not as if this is some arbitrary rule that was thought up and implemented in the middle of the game. It\'s a rule that everyone knew about at the beginning of the game, and there are players who play a specific way precisely because the rules are this way. It\'s no accident that they happen to win more hands when they are dealer because they are waiting to make their moves when it\'s their turns as dealer. They tend to be called turtles (gembu) on Fight Club and in my opinion, are generally the strongest players, at least there.

I also already mentioned some hands I think are silly and worthy of dismissal but you asked about other rules I think are lame so let me bring up \"Pao\", the penalty, for example, for allowing another player to pon his 3rd Dragon or 4th Wind to get a Limit hand (yakuman). Even if that\'s not his winning tile but he ends up winning later, the guilty party will be responsible for at least half and sometimes all of the limit hand points. While I do still have issues with it here, I can at least understand its use in those situations. What I really can\'t stand is the fact that it\'s also used when a player Kongs a discarded tile and then wins on the extra tile he takes from the King\'s Wall. Despite the fact that it\'s a self draw (tsumo) the original discarding player will be responsible for the entire score! What could be lamer than that? Oh yeah, the fact that it also outtrumps missed win! This actually happened to me in a Mahjong parlor once: the dealer was in a missed win situation so he couldn’t win off of my discard if I discarded his winning tile, and yet, when I discarded a tile that he could Kong and he won on that, I had to pay the entire amount, as if I had thrown his winning tile. Now THAT is stupid, arbitrary, evil on a George Bush/Dick Cheney scale. If that ever happens to me again, I’m reaching across the table and strangling the player with my point sticks. Consider yourselves warned!

Anyway, don’t get me wrong, I do enjoy the discussion. But for a game like Reach Mahjong which is pretty well established, I think it’s going to be difficult to change rules that have been an integral part of the game for so long. This sort of discussion will be very helpful though, in efforts to create new Mahjong breeds like at the World Series, for example, which is still really in its infancy, and whose rules will probably undergo many more revisions in the years to come.

Once again, Concealed Self Draw and the dealer bonus and continuation, work WITH them! I doubt they’re going anywhere.

HotelFSR
Expert Reacher
Expert Reacher
Posts: 328
Joined: Sat Jan 17, 2009 11:39 am
Location: Federated States of Micronesia

Re:SURVEY: Rules and Variance

Post by HotelFSR » Sun Mar 29, 2009 10:41 pm

First of all I\'m interested to know how you arrive at the conclusion that

variance in competition rules (i.e. fluctuation in player ratings supposed to reflect skill) is a little higher than it should be.



Pretty simple. The frequency of CSD is about the same as the most common hands (pinfu, tanyao, reach) i.e. it is pretty damn common. I don\'t think a hand point with that degree of luck involved should be rewarded when it is so common and can swing games (e.g. 3->4 hand points)- it\'s in a totally different ball park than things like All Green. CSD happens every game so it\'s much more of an issue if you ask me. Ron is more skill based in that to a certain extent it involves attacking/defending with the right waits, whereas Tsumo does not involve player interaction- and thus does not compare player skill quite as much.

I have accepted CSD and the dealer system as parts of the game and have adapted to them, but I still can\'t say I like them exactly- especially CSD. The dealer system still has the same problem though: although, like CSD, it is another rule that can be adapted to, it introduces a great deal of randomness. You only get two shots at being dealer, and you\'re doubly screwed if you get a poor starting hand or someone self draws a big hand off the bat! It\'s punishment enough that you lose the deal, whereas if you win you get to do it all over again. That\'s a lot of variance being thrown into the mix there.

I definitely don\'t think that a game like Mahjong should have pros winning 90% of the time. However, that doesn\'t mean that it would hurt toning down variance a slight but. It would be interesting do run exact simulations on this, but I would not be all too surprised if removing CSD would add about 5-10% more wins for the pros over average competition. At the moment a player\'s long term rating will tend to fluctate by about +/-100 points, but removing CSD and the dealer system would probably also reduce that fluctuation in player stats closer to +/-50. It would be interesting to test out.


P.S. I actually have no problem with Ippatsu in principle- I think it\'s better than CSD actually, BUT NO WAY should it work on self-draw as it does, just Ron. So I welcome it being left out A-Rules as it is.

zzo38
Senior Reacher
Senior Reacher
Posts: 162
Joined: Sun Mar 15, 2009 2:17 am

Re:SURVEY: Rules and Variance

Post by zzo38 » Mon Mar 30, 2009 12:51 am

HotelFSR wrote:P.S. I actually have no problem with Ippatsu in principle- I think it\'s better than CSD actually, BUT NO WAY should it work on self-draw as it does, just Ron. So I welcome it being left out A-Rules as it is.
And I am glad at least someone agrees with me that ippatsu should count for ron only. (Also read my comment about how I think menzen tsumo should work)

gartheee
Mahjong Pro
Mahjong Pro
Posts: 37
Joined: Mon Mar 16, 2009 9:42 am
Location: Japan

Re:SURVEY: Rules and Variance

Post by gartheee » Mon Mar 30, 2009 3:30 pm

I don\'t have numbers to back up any claims here so of course it is just weak ass speculation but...I would imagine that pro, or at least stronger, players tend to go for CSD more than the weaker ones. Weaker players tend to go for the first thing they can win the hand off of and that usually means stealing tiles as fast as they can. They LOVE backdooring (ato-zuke). I really wonder how removing CSD from the mix would affect the game. I think it would encourage a lot more tile stealing which again is one of the things that the Reach rules try to discourage.

Also, not sure why Ippatsu needs to be RON-only. It IS defendable. AFter a reach, people throw stuff for players to pon or chi so people steal only to kill the ippatsu all the time. And again, it\'s another reward for keeping your hand closed so you could reach and hope for that ippatsu bonus.

HotelFSR
Expert Reacher
Expert Reacher
Posts: 328
Joined: Sat Jan 17, 2009 11:39 am
Location: Federated States of Micronesia

Re:SURVEY: Rules and Variance

Post by HotelFSR » Mon Mar 30, 2009 5:20 pm

I have no issue with the reward, just the amount. i.e. Ippatsu Menzen being worth 2. Ippatsu on Ron is 1 only and, as you say, both are defendable- although one more than the other, and the scores do not reflect this.

You bring up an interesting point about stealing tiles, and I do wonder about it. I agree that stronger players keep their hands more concealed, as they should, but aren\'t there already a lot of rewards for a closed hand? Would removing CSD truly make for that much more stealing? I really don\'t know.

Is there more stealing in A-Rules becuse there is no Ippatsu or Ura-Dora? I\'m guessing not, although I could be wrong. If anything, I would think the red fives you find in B-Rules encourage a lot more stealing than anything else would, surely?

In an A-Rules type scenario, I would prefer to add Ippatsu Ron and remove Menzen Tsumo. Would be an interesting thing to test properly and record the numbers.

Post Reply