5D Library › Imperfect-Information Stochastic Games, Bayesian Hand-Reading and Counterfactual Regret Minimisation in Four-Player Tile Dynamics · 中文

Chapter 3. Hand Reading: Suji Is a Deduction, Not a Regularity

The form of the problem

An opponent has declared a ready hand and the player has several tiles they could discard. Which is safest? This is a Bayesian problem: there is a prior over the opponent’s wait shapes, and the opponent’s own discards rule some of those shapes out through one rule — furiten, the rule that a player cannot win on a tile they have themselves discarded.

Note

The prior in this chapter is illustrative. The prior over wait shapes below was chosen by the author, not estimated from records of play. A real prior depends on the rules, the scores, the turn and the player. What this chapter demonstrates is the structure of the update, not the particular numbers; wherever a particular number is quoted, this qualification travels with it.

Calculation

# Chapter 3. Bayesian hand reading. An opponent has declared a ready hand, and
# the only evidence is their discards. The classic heuristics -- a tile is
# safer if its "suji" partner has been discarded, and safer still if only one
# shape could use it -- are consequences of Bayes' rule once a prior over
# waits is written down, and the program derives the numbers rather than
# asserting them. Model (stated, not assumed silently): the opponent's wait is
# one of the tile types in a numbered suit; the prior over WAIT SHAPES is
# taken from the relative frequency with which each shape occurs in ready
# hands, and a shape is impossible if the opponent has discarded a tile that
# the shape would have won on (the furiten rule). All priors here are
# illustrative; the point is the structure of the update.
import math

SHAPES = {
    # name: (list of winning tile values, prior weight)
    "two-sided 23 waiting 1/4": ((1, 4), 0.34),
    "two-sided 34 waiting 2/5": ((2, 5), 0.34),
    "closed 13 waiting 2": ((2,), 0.11),
    "edge 12 waiting 3": ((3,), 0.07),
    "pair 11 waiting 1": ((1,), 0.08),
    "three-sided 234 5 waiting 1/4/6": ((1, 4, 6), 0.06),
}

def normalise(d):
    s = math.fsum(d.values())
    return {k: v / s for k, v in d.items()}

def posterior(discards):
    """Shapes consistent with the furiten rule: an opponent cannot be waiting
    on a
    tile they have themselves discarded."""
    post = {}
    for name, (wins, w) in SHAPES.items():
        if any(t in discards for t in wins):
            continue                                  # furiten: impossible
        post[name] = w
    return normalise(post) if post else {}

def p_deal_in(tile, discards, copies_left=None):
    """Probability that discarding `tile` completes the opponent's hand."""
    post = posterior(discards)
    return math.fsum(p for name, p in post.items() if tile in SHAPES[name][0])

if __name__ == "__main__":
    print("Illustrative prior over wait shapes in one suit:")
    for name, (wins, w) in sorted(SHAPES.items(), key=lambda kv: -kv[1][1]):
        print("   {:34s} wins on {:12s} prior {:.2f}"
              .format(name, str(wins), w))

    print("\nA. No information: probability that each tile value deals in.")
    print("   tile   P(deal in)")
    for t in range(1, 7):
        print("   {:4d} {:12.3f}".format(t, p_deal_in(t, set())))

    print(
        "\nB. The suji rule, derived. The opponent has discarded a 4, so no"
        " shape"
    )
    print("that wins on 4 is possible (furiten). What happens to tile 1?")
    for label, disc in (("nothing discarded", set()), ("4 discarded", {4})):
        print("   {:20s} P(deal in | 1) = {:.3f}"
              .format(label, p_deal_in(1, disc)))
    a, b = p_deal_in(1, set()), p_deal_in(1, {4})
    print(
        "   Discarding 1 becomes {:.1f} times safer, because the 23 two-sided"
        " shape".format(a / b if b else float("inf"))
    )
    print(
        "   that wins on both 1 and 4 is ruled out. That is the whole content"
        " of the"
    )
    print(
        "   suji heuristic: it is a furiten deduction, not a statistical"
        " regularity."
    )

    print(
        "\nC. What suji does NOT rule out. With 4 discarded, tile 1 can still"
        " be won"
    )
    print("   on by:")
    for name, p in sorted(posterior({4}).items(), key=lambda kv: -kv[1]):
        if 1 in SHAPES[name][0]:
            print("      {:34s} posterior {:.3f}".format(name, p))
    print(
        "   So suji lowers the risk without removing it, and the residual"
        " risk is"
    )
    print(
        "   concentrated in the pair wait and the three-sided shape --"
        " precisely the"
    )
    print("   shapes a player holding a strong hand is most likely to keep.")

    print(
        "\nD. Combining with the tile pool. Suji is about which shapes are"
        " possible;"
    )
    print(
        "   the pool (Chapter 2) is about how many copies remain. Both"
        " multiply:"
    )
    print("   tile   suji?   P(shape wins)   copies left   relative danger")
    pool_note = []
    for t, disc, copies in (
        (1, set(), 3),
        (1, {4}, 3),
        (1, {4}, 1),
        (3, set(), 2),
    ):
        ps = p_deal_in(t, disc)
        danger = ps * copies / 3.0
        pool_note.append((t, "yes" if disc else "no", ps, copies, danger))
        print("   {:4d}   {:5s} {:15.3f} {:13d} {:17.3f}"
              .format(t, "yes" if disc else "no", ps, copies, danger))
    print(
        "\n   The last column is the product, scaled so that the first row"
        " is the"
    )
    print(
        "   shape probability itself. A suji tile with three copies live can"
        " be more"
    )
    print(
        "   dangerous than a non-suji tile with one copy live, so the two"
        " heuristics"
    )
    print(
        "   cannot be applied in sequence as a checklist -- they have to be"
    )
    print(
        "   multiplied, which is what a model does and a rule of thumb does"
        " not."
    )

    print(
        "\nE. The model's limits, stated plainly. The prior above is"
        " illustrative,"
    )
    print(
        "   not measured; real priors depend on the rule set, the scores, the"
        " turn"
    )
    print(
        "   and the player. More importantly, the model treats the opponent's"
    )
    print(
        "   discards as INFORMATIVE BUT NOT STRATEGIC: a player who knows"
        " they are"
    )
    print(
        "   being read can discard to mislead, and Chapter 6 shows that"
        " against such"
    )
    print(
        "   an opponent no fixed reading rule is safe. The reading in this"
        " chapter is"
    )
    print(
        "   therefore a best response to a NON-ADAPTIVE opponent, which"
        " is the"
    )
    print(
        "   correct way to describe it and not the way it is usually"
        " described."
    )
# Illustrative prior over wait shapes in one suit:
#    two-sided 23 waiting 1/4           wins on (1, 4)       prior 0.34
#    two-sided 34 waiting 2/5           wins on (2, 5)       prior 0.34
#    closed 13 waiting 2                wins on (2,)         prior 0.11
#    pair 11 waiting 1                  wins on (1,)         prior 0.08
#    edge 12 waiting 3                  wins on (3,)         prior 0.07
#    three-sided 234 5 waiting 1/4/6    wins on (1, 4, 6)    prior 0.06
#
# A. No information: probability that each tile value deals in.
#    tile   P(deal in)
#       1        0.480
#       2        0.450
#       3        0.070
#       4        0.400
#       5        0.340
#       6        0.060
#
# B. The suji rule, derived. The opponent has discarded a 4, so no shape
# that wins on 4 is possible (furiten). What happens to tile 1?
#    nothing discarded    P(deal in | 1) = 0.480
#    4 discarded          P(deal in | 1) = 0.133
#    Discarding 1 becomes 3.6 times safer, because the 23 two-sided shape
#    that wins on both 1 and 4 is ruled out. That is the whole content of the
#    suji heuristic: it is a furiten deduction, not a statistical regularity.
#
# C. What suji does NOT rule out. With 4 discarded, tile 1 can still be won
#    on by:
#       pair 11 waiting 1                  posterior 0.133
#    So suji lowers the risk without removing it, and the residual risk is
#    concentrated in the pair wait and the three-sided shape -- precisely the
#    shapes a player holding a strong hand is most likely to keep.
#
# D. Combining with the tile pool. Suji is about which shapes are possible;
#    the pool (Chapter 2) is about how many copies remain. Both multiply:
#    tile   suji?   P(shape wins)   copies left   relative danger
#       1   no              0.480             3             0.480
#       1   yes             0.133             3             0.133
#       1   yes             0.133             1             0.044
#       3   no              0.070             2             0.047
#
#    The last column is the product, scaled so that the first row is the
#    shape probability itself. A suji tile with three copies live can be more
#    dangerous than a non-suji tile with one copy live, so the two heuristics
#    cannot be applied in sequence as a checklist -- they have to be
#    multiplied, which is what a model does and a rule of thumb does not.
#
# E. The model's limits, stated plainly. The prior above is illustrative,
#    not measured; real priors depend on the rule set, the scores, the turn
#    and the player. More importantly, the model treats the opponent's
#    discards as INFORMATIVE BUT NOT STRATEGIC: a player who knows they are
#    being read can discard to mislead, and Chapter 6 shows that against such
#    an opponent no fixed reading rule is safe. The reading in this chapter is
#    therefore a best response to a NON-ADAPTIVE opponent, which is the
#    correct way to describe it and not the way it is usually described.

Suji: a deduction

Section B is the chapter’s main result. The opponent has discarded a 4, so every shape that wins on 4 is impossible (furiten). The consequence: the deal-in probability for a 1 falls from 0.480 to 0.133, about 3.6 times safer.

What matters is how that comes about. It is not an empirical regularity of the form “statistically, suji tiles are safer”; it is a deduction — the two-sided shape that wins on both 1 and 4 has been excluded by the rule. So

suji is a furiten deduction, not a statistical regularity.

The distinction has practical consequences. An empirical regularity varies in strength between populations of players; a deduction holds wherever the rule holds. But equally, a deduction excludes only what it excludes — section C lists the remainder: under this book’s prior, after a 4 is discarded, a 1 can still be won on by a pair wait. So suji lowers the risk without removing it, and the residual risk is concentrated in pair and three-sided waits, which are the shapes a player with a strong hand is most likely to keep.

The two heuristics must be multiplied

Section D marks a place where mistakes are easy. Suji is about which shapes are possible; the pool (Chapter 2) is about how many copies remain. They are independent factors and must be multiplied:

a suji tile with three copies live (relative danger 0.133) is more dangerous than a non-suji tile with one copy live (0.044). So the two heuristics cannot be applied as a checklist in sequence — “first check suji, then check the count” produces the wrong ordering. A model multiplies; a rule of thumb does not.

The two heuristics multiplied. The horizontal axis is the probability that a shape wins on this tile (suji or not), the vertical axis the copies remaining, and the contours their product, the relative danger. A fully live suji tile and a non-suji tile with one copy left can fall on different contours, and applying the two rules of thumb in sequence misorders them.

Which opponent this model describes

Section E states this chapter’s most important qualification. The model treats the opponent’s discards as informative but not strategic: the discards carry information about the hand, and the opponent does not change their discards because they know they are being read.

An opponent who does know can exploit that. Program p12 in Chapter 6 gives the formal result: in a game where the opponent can observe and adapt, every deterministic policy has a best response that beats it. So

the hand reading in this chapter is a best response to a non-adaptive opponent.

That is the correct description of it, and not the usual one. The usual description presents reading as a technique for approaching the truth; on the account above it is a technique directed at one class of opponent, and against another class it is precisely the entry point for being exploited. Chapter 6 also gives the upper bound on that worry: reading a mixing proportion takes thousands of hands of observation, so in a session of a few hundred hands nobody can read anybody’s mix — including the opponent reading yours.

Evidence

Relation to the information-asymmetry literature. Volume 45 covered information asymmetry and toxic flow in the order book, where the structure is similar: one side infers private information from an observable stream of behaviour, and the other side can shape that stream strategically. The common conclusion in both places is that the value of the inference depends on whether the other side is adapting, which is an empirical question about the opponent and not a mathematical question about the method.

Hypothesis · not established

The author’s hypothesis (not established): against a fixed opponent pool, the marginal value of hand-by-hand Bayesian reading over a single fixed table of deal-in probabilities is less than 0.01 buy-ins per scored decision; and when the opponent pool contains players who shape their discards strategically, the reading policy performs worse than the fixed-table policy. What would show it to be wrong: running the Chapter 8 design once with each opponent pool and finding the difference in the same direction with both confidence intervals excluding zero.

Summary

Hand reading is a Bayesian problem: wait shapes have a prior, and the furiten rule excludes some of them. The chapter’s main result is the status of suji: it is a furiten deduction rather than a statistical regularity, and under this book’s illustrative prior it makes a tile about 3.6 times safer rather than safe, with the residual risk concentrated in pair and three-sided waits — the shapes a strong hand is most likely to keep. Suji and the tile pool are independent factors that must be multiplied rather than applied in sequence: a fully live suji tile can be more dangerous than a non-suji tile with one copy left. Finally, this whole apparatus is a best response to a non-adaptive opponent, because it treats discards as informative but not strategic; Chapter 6 supplies both that worry and its upper bound — reading a mixing proportion takes thousands of hands.

Fan, Y. (2026). Imperfect-Information Stochastic Games, Bayesian Hand-Reading and Counterfactual Regret Minimisation in Four-Player Tile Dynamics: From Markov Transitions of the Tile Pool and the Optimal Stopping of Defensive Folding to Kelly Allocation in a Multiplayer Game, with a Textual Examination of the Orthodox and the Unorthodox in the Classical Military Writers [四人不完全信息随机博弈、贝叶斯读牌与反事实遗憾最小化] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24020464 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).