Tile Efficiency 101 (Part 3)

Part 3 of the Tile Efficiency series for Osamuko.com is here. This time, it’s taatsu theory. More precisely? Ryankan.

I skipped a few parts of things you should already know. If you don’t, then read:

Tiles from 3 to 7 are the most flexible. 2’s and 8’s are a little less flexible. 1’s and 9’s are the least effective of number tiles. And lastly, character tiles/honors are the most difficult of all to complete melds with.

A kanchan (central wait) is superior to a penchan (side wait). That is because a 46 kanchan, for example, can become ryanmen with a 3 or a 7, while a 12 penchan must first draw a 4 and discard the 1, then draw a 5 to have a ryanmen; you’re more likely to complete the meld than you are of upgrading the shape.

Also, a kanchan composed purely of tiles from 3 to 7 are superior to other kanchan, because they can expand towards either side (thusly, a 35 kanchan is superior to a 13 or 24 kanchan).

Furthermore, if you have two ryanmen overlapped (23 56) it’s less efficient than having two separate ryanmen in two different suits. This is, because if they share an out in common, the number of useful tiles is decreased. Saying “Oh, but I might get a three sided with if I draw the 4 for a 23456!” is no excuse if you have more efficient shapes available. When the same out is repeated for two different taatsu (proto-melds), this is called a “ni-do-uke” (or “double out”).

So, if that is clear, then let’s continue. This is the source: http://beginners.biz/h_05.html

Taatsu Theory

About ryankan (aka double gut shot).

Ryankan is a compound shape of two kanchan waits, such as 2m4m6m.

The ukeire is of 8 tiles, but…

1) It uses up 3 tiles.

2) If it is left as the final wait, it will become a kanchan.

Considering the two points above, it is clear it is less efficient than a ryanmen.

1.Ryankan and Shanpon


1m3m3m5m3p3p7p8p1s2s3s6s7s Tsumo: 5s

What becomes a problem with ryankan, is the decission of choosing shanpon over it or not.

In example 1, if we were considering the chances of becoming pinfu, discarding 3m would be correct. Drawing 2m4m it becomes pinfu. Likewise, by discarding 1m or 5m, there is only 1 type of tile that will give us pinfu. Therefore, it may look better to use the ryankan.

However, if we were considering tenpai chance, you should discard 1m with this hand.

Drawing 6m2p4p it perfectly evolves to a smooth 1shanten for pinfu.

3m3m5m3p3p7p8p1s2s3s5s6s7s Tsumo: 2p

In the case of chunchan tiles (simples, middles, tan yao tiles, 2-8) that can evolve to ryanmen, choosing the shanpon+kanchan shape which has better room for hand evolution is generally more advantageous.


4m5m6m6m4p6p8p8p4s5s5s6s7s Tsumo: 3m

San Shoku is visible, so we may want to discard 8p but, discarding the 4p that leaves room for the 3menchan (3-sided-wait) in the manzu (cracks/reds) tiles is an excellent discard.

3m4m5m6m6m6p8p8p4s5s5s6s7s Tsumo: 7m


1s1s3s5s8s8s4p6p7p8p (pon) 5z5z5z Tsumo: 2p

In the case of an open hand, taking the shanpon becomes more useful.

In example 3, we should cut 5s and aim for the pon-ten (pon and tenpai) of 1s and 8s.

Discarding 1s the uke-ire number doesn’t change, however the speed of tenpai is different. That is because you can only call a chii from the player on your left, but you can call a pon from anyone.


4m6m8m8m1p1p5p6p7p7p8p3s5s Tsumo: 4s

So, we basically said that is is generally better to take shanpon over kanchan, but this is not so black-and-white.

In example 4, we want to take the better 1shanten shape by discarding 8m. Not just because of the 3m, the backfire of drawing 1p also becomes useful.

4m6m8m1p1p5p6p7p7p8p3s4s5s Tsumo: 1p

By discarding 8m or 4m here, the uke-ire number is greatly increased.

Furthermore, if we discaded 4m to fixate on the shanpon, drawing 5m isn’t any useful at all. It becomes nothing more than a complete backfire.


In a hand with a ryankan, when deciding whether to take shanpon or kanchan, we compare the number of tiles of evolution towards good shape.

Generally, there are many cases where shanpon is better.

2.Ryankan that are easy to overlook

Since ryankan that are embedded inside of melds are easy to overlook, let’s be careful.


4m5m7m5p5p6p7p7p1s3s3s7s8s Tsumo: 9p

With a hand of a shape such as example 5’s, there are people who easily tsumo-kiri (tsumo-kill/discard upon picking up) the 9p.

We can see the ii-pei-kou, but if we extract the 5p6p7p meld, we can see we have a 5p7p9p ryankan in the pinzu (circles/dots).

If we discard the 9p here, we eliminate the 8p uke-ire completely.



In example 6, we’re one away from ittsuu (full straight), but we have a ryankan in the manzu.

Even if we draw a safe tile, by descarding 2m we are completely missing the chance of tenpai by drawing 3m.

3.Irregular kanchan

Rarely, we come into shapes that are “distant ryankan.”

With a shape like


we have an uke-ire of 8 tiles by using up 6 tiles.


4m5m5m6m1p3p4p5p6p8p9p7s7s Tsumo: 6m

You may run many times in cases like this where you must take appart the ittsuu.

Of course we should aim for mentanpin (riichi tan yao pinfu) and iipeikou, but between discarding 1p and 9p there’s a huge difference. If we discard the 1p, we completely lose the uke-ire for 2p.

4m5m5m6m6m1p3p4p5p6p8p7s7s Tsumo: 2p

4m5m5m6m6m3p4p5p6p8p9p7s7s Tsumo: 2p


Okay, so that’s it for now. You may have noticed the importance of “backfire follow-up” or “urame follow-up.” Many times you make a decission, a backfire may happen. Before discarding, think about what becomes a loss, and how will you deal with that loss. Did it ever happen to you that you ended up picking up and discarding a whole series of tiles in a spiral because of one single early discard? Then you’re clearly not looking out for your backfires.

Also, sorry I took so long to finish this one up. It was my birthday last week, and yeah, stuff.

See you next time.

24 thoughts to “Tile Efficiency 101 (Part 3)”

  1. 4m6m8m1p1p5p6p7p7p8p3s4s5s Tsumo: 1p

    By discarding 8m or 4m here, the uke-ire number is greatly increased.

    How? You’re confusing yuukouhai (sum of “uke-ire” + tiles increasing “uke-ire”) with what buddy there on beginners.biz called uke-ire. Ignore the tsumo. 1-shanten, and the number of tiles that drop your tenpai is 15, on 4 types: 5m, 7m, 6p, 9p.

    I am not even sure myself if tile changes giving the same number of UI count for yuukouhai, but the only way that hand can do that is with 3m (throwing 8m). 15 on 4 types, swapped to 2m, 5m, 6p, 9p with the 6m discard pre-planned.

    Now back to your chosen pick: 1p. If you keep it, and throw 8m, you have a total of 11 UI tiles remaining. Any attempts at shape optimization will cost you two turns.


    TL;DR verdict: Yappa pa, Yappa pa, isshanten. There is absolutely no justification other than knowing that your tiles are dead in switching your pre-wait-wait. That, and the hand is yaku-free. Getting rid of your only explicit pair (and that you have no implicit pair either) would drop the hand to 2-shanten while dicking around, rendering any theory (UI, yuukouhai, or other published/unpublished theories) useless.

    TL;DR^2: 1p is shit in 4-beta.

  2. I just want to add that it is not because an explanation or theory is written in Japanese that it absolutely right and 100% of the time reliable. Japanese players can also be wrong or mistaken.

    You comment is a little bit harsh,Mr Anonymous.

    xKime only translated this text. I am glad he did. It is up to us to verify it’s validity.

  3. Lol, not only is his comment harsh, but also quite wrong. If you don’t trust the text, trust Tenhou’s shan-ten+uke-ire calculator:


    Discard 4 or 8 of man, and you have 29 tiles (9 types) that give you tenpai (YOU DO NOT GO BACK ANY SHANTEN, so don’t suggest it)

    Not only that, but you have a higher chance of ending up with a ryanmen (more (ippatsu) tsumo power for you).

    If you ignore the tsumo, you cut your uke-ire to HALF and you have just about the same chances of ending with a kanchan. No one is confusing anything here but you. It is a beginners habit to fixate on pairs and think that the only thing that can make their hand progress is filling in the other mentsu.


    Discard 8m.
    Tenpai with 456m (ryanmen) 456789p. 29 tiles. uke-ire increases to DOUBLE

    Even if you end up with a temporal tanki, you can upgrade it to a multi-sided wait (or a dora or jihai wait) almost immediately.

    You are completely ignoring all other uke ire that’s not the one from kanchan and ryanmen.

  4. Back up the truck for a second. I didn’t intend to be anonymous the first time.

    Did I screw up, well, yes. Bigtime. I ran Ara’s Mahjong Analyzer on the hand, and it’s counting an average of a 300pt gain (for your choice btw, 4m/8m = same). I also screwed up by neglecting pair formation on 4_6 or 6_8. Mea culpa, whatever. That is a severe lack of perception on my part.

    Because of that, all my conclusions are wrong. Lemme login, I’ll even edit the text to say so (if i have that power, of course without deleting it).

    There’s no parting shot either. It’ll serve as a reminder to at least run Ara’s tool on stuff rather than post and have to retract later. Sucks getting 1-upped in public but hey, I can take my lumps.

  5. Well, I do have more respect for Senjo than Nameless, so I’m sorry if I came off too harsh on you? I thought obvious troll was obvious (nameless tend to troll), but I do know you, Senjo, are a respectful community member and a valuable player.

    Anyway, just as a little note, though matematically equivalent, 8m and 4m are far from being the same, if you count Evolution Towards Good Shape. 8m proves to be the better discard. :)

    Anyhow, I guess if we knew these things from the beginning, we wouldn’t be reading this, so it’s all good.

    I’m not saying there are no questionable things in the text, or that I translated it perfectly, I’m just saying this case is not the case. Sorry if I overdid it.

  6. Please, just read properly before commenting.

    Example 4 is clearly reflected as 4688m1156778p35s4s in the text.

    The 468m1156778p345s1p is just the follow up of the 8m discard.

    In example 4, 8 is the superior discard, it is even written in the text. However, in the follow-up (which is what I was discussing with Senjo) from the 468m1156778p345s1p shape, the better discard IS in fact the 8m.

    Oh, why do I bother. It’s apparently too hard to explain.

  7. In 4-alpha, you can expect more points, but Ara’s thingy calculates that gain as negligeable (2052[1p] vs 2033[8m] vs 1929[4m]) but it also states that you are much more likely to win (8%) or be tenpai (9%) with either 4m or 8m in ex. 4-alpha.

    We can’t expect every theory to be read through once and understood or disproven. Let my folly be a witness to that.

  8. xkime : “Please, just read properly before commenting. […] Oh, why do I bother. It’s apparently too hard to explain. ”

    I know it’s a lot of work to translate those pages and you have faith in this author, but please try to have an open mind so we can have a rational, impartial discussion instead of being emotional attached to one side.

    If one has already gone the 468m1156778p345s1p path by discarding 8m, then, yes, discarding 8m for a second time is marginally superior to discarding the 4m.

    But, when you are still deciding which of the tiles you should discard first (without the help of an expected value estimator like Ara’s), the problem is at the initial 4688m1156778p35s4s state and the two obvious candidates are 4m and 8m.

    打 4m 摸[ 7m 8m 1p 6p 9p ; 15枚]
    打 8m 摸[ 5m 7m 6p 9p ; 15枚]

    If you sum up the resulting waits you get with [ 7m 8m 1p 6p 9p ] vs those obtained with [ 5m 7m 6p 9p ] , 4m has a slight %win advantage. Click on every single tiles, the see the partial results.

    4m :


    Then if you multiply %win by the hand values ( %win * value = expected value ), you get the estimated expected value Senechal talks about.

    So — unless you are a supernatural being able to compute and compare on the fly two two-level trees of possibilities –, your best bet, as an average human being, is to go for 4m.

    Finally, “Ara’s thingy” is perfectly reliable within the bounds of its assumptions. It goes way further than Tenhou’s wait enumerator. It shows clearly that most, if not all, controversial WWYD quizzes are not worth solving since they have multiple, nearly indistinguishable solutions. Often, the given solution is not the optimal one, expected-value-wise, but a practical one, close enough, within a margin of error of less than 5%.

  9. Ara’s thing is quite reliable, had I used it the first time, I would have had the same basic answer on the screen, derped at the fact that I forgot some stuff, and then examined the extra analysis.

    It’s almost worth me writing an article on.

  10. I know better than to discuss an application that I have not yet used. So, if you have a link to this thing, let me know, I’ll try it and comment on it.

  11. Ark: Sorry about your comment. It got sent to spam because of the amount of URLs. I normally check the spam but lately we’ve been getting more spam (currently 51 spam comments in 4 days) so I mostly just check weekly to delete them all

  12. Ark, you’re just going to ignore the fact that you started off saying

    “Exemple 4 is not http://tenhou.net/2/?q=468m1156778p345s1p

    It is http://tenhou.net/2/?q=4688m1156778p35s4s

    Which is just a demonstration that you weren’t properly reading or paying attention in the first place?

    By saying “Exemple 4 is not http://tenhou.net/2/?q=468m1156778p345s1p” you seem to imply that someone said Example 4 is something it shouldn’t be, when in fact it is the way it should be. FIRST comes the example which is different (it’s the link you lastly quoted), and THEN comes the follow up which is the link you first quoted.

    When I see your post starting off that way, the first thing that comes to my mind is “this guy barely read through.” With that comes frustration, and it really makes my reply “emotional” if you wanna describe it as such. Because I remind you, I went through this lack of attention not once, but twice before, with the whole “uke-ire” confusion (having to re-explain that was exhausting enough). It’s perfectly normal for me to just say “ah, fck it” when comments seem to be of low value.

    *However*, your new reply is more like what I wanted, as I see true value in it, so I’ll get to it. I’ll just think your first post was a slip.

    My respect for this technology, is just about the same as I have for online translators (i.e., it’s matematically all good and well, but the human factor is gone. I mean, if you had to compute a tanki wait on an honor tile that has been already discarded, how would the % of win be calculated? It being 2 tiles, but it’s still easy to come out. More over, what if you had the same number of tiles to wait on, but it was a middle tile dora wait? Some things can’t be computed, and that’s the reason why all mahjong AIs are weak or barely intermediate*. A sound judgement based on reasoning is also required).

    *(and also, the reason why players who always go tenari even when they start 3shanten or so, tend to get stuck in 1-2 dan. Mahjong isn’t simply about this “one” factor, but that’s a different point for a different day)

    It’s becoming a wall of text, so my point is…

    If you trust into inputting the data in an application and just select based on the result of a doubtful % of win calculation, then I give in. I can’t possibly argue that.

    However, looking at the WWYD and deciding by ourselves based on our game theory, experience and knowledge and making a sound choice with a sound logic to back it up is important, therefore I think it’s pretty standard (and optimal) to:

    4688m1156778p35s Tsumo: 4s

    Discard 8m (no need for shanpon as the ryanmen evolution is minimal and no need for shanten gaeshi).

    (EDIT: Also, I think that when you’re imputting your numbers for Example 4 here, you’re assuming that we will settle for whatever wait we end up with and calculating the EV based on that.)

    468m1156778p345s Tsumo: 1p

    Backfire. Considering the already discarded 8m and the possibility for ryanmen evolution by drawing 3m and also to prevent the (maybe red?) middle tile 5m backfire (calculate, too, the risk it will be please?), 8m.

    If you think 8m is superior to 4m in this shape and situation, then please explain your reasoning instead of giving numbers that can’t be checked objectively. If it was better to start off with 4m (I would really hate to lose the uke-ire for a 5m and minimize our pinfu chances (which are also maxed out if we draw 3m, and pinfu means a better ippatsu, ura and tsumo chance, but w/e, I’m assuming all of these are contemplated by the calculator/analyzer), then *write* how so.

    (Also, is the analyzer also contemplating that we cannot open this hand? Because if it’s not, then I understand how having two pairs may seem superior to it. Moreover, is there a way for us to know how it functions and what it does contemplate?)

    (And if you also implied that you can solve all WWYDs with an application and get a marginally better answer than the given ones by pro’s in their blogs, or the quiz creators and such, then I don’t want to go there.)

  13. Is this the software in question?


    Ararara (pun unintended)…

    So, based on the output of an analyzer that’s supposed to be for one-player mahjong (i.e. all non-mathematical factors ignored), you were implying that even in a 4 people match (where you have danger tiles, other people, and whatnot) considering the marginal difference in output between 4m and 8m, 4m is -still- better?

    In one player, maybe so, marginally.
    But in four player, the backfire of 5m is too big (and maybe 3m too), and whatever difference there was in output between the two, then I think this particular middle-tile backfire just covers it. Sorry. But this is just when human judgement comes in. (Moreso, if we even made it explicit we’re playing with red fives (or the dora was any tile in the lower man area) then the 4m discard really, reaally falls short).

    If anything, the difference is covered. If mahjong was a one player game where no one can touch your discards, then the output would be completely right, though.


    But I think we may be going off the point here. The point was ryankan vs shanpon; whatever else comes off from that decission is just a derivation. I think the point of the text is something else that couldn’t have been demonstrated running analizers and showing input. I believe the point was to teach a very important notion for judgement when faced with the ryankan vs shanpon decision.

    It goes far beyond the actual example.

  14. “I went through this lack of attention not once, but twice […] like what I wanted […] I see true value in it […] I’ll just think […] My respect for this technology […]”

    This thread is not about you, your feelings and your frustrations. Instead of viewing this debate as “poor you vs awful me”, stick to objective point. Your posts will be a lot shorter.

    “*However*, your new reply is more like what I wanted, as I see true value in it, so I’ll get to it. I’ll just think your first post was a slip.”

    My initial post says exactly the same thing as the post you value. So please, take time to review it. Essentially, the solution for 4b, whatever it is, is irrelevant since to get to 4b you must first solve 4a.

    “it’s matematically all good and well, but the human factor is gone.”

    Where in this discussion about the superiority of 8m over 4m, did beginners.biz has taken into account the “human factor”? Beginners.biz’s theory, Tenhou’s wait enumerator, Ara’s thingy and every WWYD quiz don’t take into account human factors.

    “I mean, if you had to compute a tanki wait on an honor tile that has been already discarded, how would the % of win be calculated?”

    Ara’s thingy would display 0 % * (whatever value of tanki) = 0 points.

    As for Tenhou’s wait enumerator, it would display 2*2 honors tiles = 4 waits.

    As for Beginners.biz, he would say it is trivial. You are stuck in a dead end.

    “More over, what if you had the same number of tiles to wait on, but it was a middle tile dora wait? ”

    Ara’s result : expected value of dora > expected value of same wait without dora

    Tenhou’s wait enumerator : same number of tiles

    As for Beginners.biz, he would say it is trivial. The dora is worth more.

    “Some things can’t be computed”

    And guesses are just that, guesses.

    “that’s the reason why all mahjong AIs are weak or barely intermediate”

    Not true. Any powerful computer can easily beat any expert players. Most AIs are tweaked to let begginers win. When a game wins too often, the human players suspects it is cheating. Online game like RON2 and Tenhou have purposely weak AIs to punish quitters and to limit the load on the servers.

    A simple software like Ara’s thingy is able to solve in a few seconds a WWYD quiz, where after few days of discussion, human players still cannot agree what is the optimal solution.

    “However, looking at the WWYD and deciding by ourselves based on our game theory, experience and knowledge and making a sound choice with a sound logic to back it up is important,[…]”

    So let me ask you this. How do you know your game theory is sound? How do you know the knowledge you inferred from your experience is not mistaken? How a mathematical analysis differ from sound logic?

    “Considering the already discarded 8m […]”

    But if 8m is NOT ALREADY discarded, the following arguments are totally irrelevant, no matter how logic or true they are.

    “please explain your reasoning instead of giving numbers that can’t be checked objectively.”

    I did in my previous post! And if numbers cannot be checked objectively I wonder what do you consider to be “objective”. Your intuition? Your faith in Beginners.biz’s infaillability?

    “Moreover, is there a way for us to know how [Ara’s thingy] functions and what it does contemplate?”

    Yes, see the website.

    By the way, my point is not to praise Ara’s thingy, but, I repeat, that the solution for 4b isn’t the same has the solution for 4a.

    “you were implying that even in a 4 people match (where you have danger tiles, other people, and whatnot) considering the marginal difference in output between 4m and 8m, 4m is -still- better?”

    No. And, notice that Beginners.biz is also NOT considering those factors. Without context, nobody can. And even with context, marginal differences can go both ways, depending on your “experience” (guesses based on seemingly equivalent historical data).

    “If mahjong was a one player game where no one can touch your discards, then the output would be completely right, though.”

    If mahjong was a ten-player game or even a 10000-player game, the margin between 4m and 8m would be proportionally the same. When problems become too complex for humans to be solve by pure reflexion, humans rely more on guesses and experience, that is on ready-made solutions based on short-term pattern inferences (intuition, if you like).

  15. I’m not talking about “intuition” but “observation” but w/e.

    You and I obviously have different values.

    I agree to disagree, mainly because our arguments go beyond the point and we’re starting to take things into consideration that weren’t the point, and discussing an example with barely any context. And, anyways, if that’s how you wish to reason your game, then I wish you good luck and happy grinding.

  16. Intuition is receiving input and ideas without knowing exactly how and where you got them, without making a conscious effort to label or linking them to a precise theory. The act or faculty of knowing or sensing without the use of rational processes; immediate cognition. Knowledge gained by the use of this faculty; a perceptive insight. A sense of something not evident or deducible; an impression.

    In other words, observations and feelings without explanations.

    You are talking about… intuition.

    Too bad you quit this discussion, but I’ll respect your choice to protect your convictions by avoiding unsettling debates. There is a limit to how many paradigm changes one can digest at once. May randomness lean in your favour and have fun destroying the victory hopes of your opponents.

  17. Firts of all, I do know how and where I get them. It’s just I won’t even touch subjects pertinent to アナログ派、経験型 or even less デジタル派 because you don’t believe in it, and I don’t either mostly.

    I’m not evading the debate, I just realized this makes no sense because I’m trying to teach you something that you obviously don’t have the will to learn, and you’re trying to teach me something that I don’t have the will to learn. Then I realized, it’s not up to me to help you rationalize theory into a way that you can diminish your misses in a live game without the need to input data in an analizer. If you can compute the EV for every-hand ingame, then I congratulate you, but I thought having theoric content would help you.

    Don’t worry, even if I don’t use analizers, I have theory on my side, and I guarantee you I don’t need to analize every hand to see what the wider discards are.

    >May randomness lean in your favour and have fun destroying the victory hopes of your opponents.

    Funny that you say that, what was your rank/rate again? Eh, no, I don’t care. I’ve seen you play. Have fun struggling in 上級。I’m making my way to 鳳凰, ETA one more year.

    Also, you make it sound like 8m is that bad of a choice, when in fact the difference is barely even there, and according to my values (middle tile backfire and whatnot) it’s perfectly appropriate.

    If you’re going to trash theory like this, then I hope you can compute your digits fast enough in a real game, and that your backfires don’t kill you.

  18. I am not pro or against intuition. There is a time for rational processes and their is a time for relying on perceptive insights. Without a good rational base, you cannot inculcate your intuitive brain to be useful. If one relies only on intuition, one will become superstitious player and adopt an suboptimal “occult” strategy. If one thinks a purely mathematical approach can be applied to human players, one is also mistaken and won’t fully take advantage of his/her brain. In a real life games, it is simply impossible to compute EV in one head.

    “[…] without the need to input data in an analizer. If you can compute the EV for every-hand ingame […] ”

    I wish I could, but I can’t. It would be a valuable learning aid. The perfect tool for training our intuition. However, it would be also cheating. In tournaments, it should be strictly forbidden.

    Interesting, now you are shifting your opinion of me and relying more and more on personal characterizations. I never told you to use an EV estimator while you play nor I suggested I do.

    “Way to show up and hardly make a useful contribution.”

    How about how human factors, like rationalizations and intuition, related to theoretical tile efficiency? Do you really think those subjects would even be remotely discussed if it wasn’t for me? Do you really think you would have even typed a rhetorical question if it wasn’t for me? (^_^)

    I am not an easy debater because I don’t treat people like easily vexed children or insecure teenagers. I don’t care for “feelings” and I question the rational behind decision making, not afraid to mix commonplace theories with other fields of knowledge like neurosciences. One needs to be mature and secure enough to not feel his/her ego constantly threatened.

  19. Well, you did take your time to write and I appreciate that, so I won’t just pretend I didn’t read you.

    I don’t think I’m implying something intuitive, just something self evident after playing quite a few games. But whether the EV is more important than the conjecture (well, all I am stating is simply conjecture, there is a chance that drawing 5m and bawwwing isn’t that bad either), the difference is in one’s 決め事, and whether you prefer a higher EV or just things you “would rather avoid” or “rather do.” Some pros do justify their decissions as “I like […] yaku more, so I went for it there,” far from relying on EV. I’m sorry if I just tried to imprint my own value by force. I just tend to think not giving your opponents enough chances to call middle tiles is also superior, as I don’t think about my winning chances alone, but the chances of other not winning or not winning much. I’m pretty sure that even though we disagree on WWYD in this ocassion, we would agree on many others. Or just disagree by a marginal difference.

    Well, it wouldn’t really make sense if everyone was inputting their hands. Many times, you have to take non-EV based decissions (i.e. why fight a 4shanten hand? I would rather go for a difficult yaku and keep a few safes), so I don’t think it would be cheating as much as it would annoy everyone else at the table.

    Yeah, I used the falacy of belittling the writer of the context rather than arguing the context itself, mainly because it would be pointless and tiring, but I guess such a cheap blow wasn’t called for. Well spotted. And sorry.

    I think there should be a line drawn between 牌理 (tile effiency, but literally more like tile logic) and 牌確率 (tile probability). 牌理 would treat things like, efficience based on theoric text, and 牌確率 would deal with EV output. My post is about 牌理 however, and as such it will rely more heavily on “discarding this means that you have this, and are left with this and your plan should be this” rather than cold numbers.

    Okay, so I guess our little discussion here was a contribution of sorts. You’re right when you said no one else would have commented and sustained a discussion on this.

    Well, you keep doing it your way, but do study game theory as well. It doesn’t hurt to read the reasons behind the answers to WWYDs. Whether they are EV wise correct or not. I feel like I have to work extra hard to assimilate theory because it’s the only thing I can really rely on. It’s not ego, I was seriously trying to help, but I get easily frustrated.

Leave a Reply

Your email address will not be published.