5D Library › 塔罗超几何抽样、雷诺曼网格拓扑与卡巴拉生命树图论 · English
第三章 雷诺曼大桌阵的网格组合学
三十六张牌与两种大桌阵
雷诺曼牌有 36 张,编号与图案沿袭约 1799 年纽伦堡出版的纸牌游戏《希望游戏》(Das Spiel der Hoffnung)(Hechtel 1799)。「大桌阵」把 36 张牌全部摊开:一种是 4 行 9 列(),另一种是 4 行 8 列再在下方中间加 4 张()。解读者先找到指示牌(例如 28 号男士或 29 号女士),再看其他关键牌与它的相对位置:是否相邻、相距多远、是否在同一行或同一列、是否隔着「马步」、是否处于镜像位置。
洗匀的大桌阵是 中的一个均匀随机排列,所以任何两张指定的牌落在一对均匀随机的不同格子上。 网格有 36 个格子,有序格子对共 1260 个;下面的所有概率都是对这 1260 对的精确计数。
推导
# Chapter 3. p06: Distances in the 9 x 4 Grand Tableau.
# 36 cards in a random order fill 4 rows of 9. Two named cards (say the
# significator 28 and a key card) then occupy a uniformly random pair of
# distinct cells. We list the exact laws of their Manhattan distance and
# of the relations readers use: touching (8 directions, a king's move),
# same row, same column, a knight's move, and mirroring (reflection in
# the middle column, the middle line, or both).
from fractions import Fraction
from itertools import permutations
R, C = 4, 9
CELLS = [(r, c) for r in range(R) for c in range(C)]
def relations(a, b):
dr, dc = abs(a[0] - b[0]), abs(a[1] - b[1])
return {
"touching": max(dr, dc) == 1,
"same row": dr == 0,
"same column": dc == 0,
"knight": {dr, dc} == {1, 2},
"mirror": b in ((a[0], C - 1 - a[1]), (R - 1 - a[0], a[1]),
(R - 1 - a[0], C - 1 - a[1])) and b != a,
}
if __name__ == "__main__":
pairs = list(permutations(CELLS, 2))
tot = len(pairs)
print(" ordered pairs of distinct cells: {}".format(tot))
law = {}
for a, b in pairs:
d = abs(a[0] - b[0]) + abs(a[1] - b[1])
law[d] = law.get(d, 0) + 1
print(" Manhattan distance law:")
for d in sorted(law):
print(" d = {:2d} {:3d}/{} {:.4f}".format(
d, law[d], tot, law[d] / tot))
mean = Fraction(sum(d * n for d, n in law.items()), tot)
print(" E[d] = {} = {:.4f}".format(mean, float(mean)))
for key in ("touching", "same row", "same column", "knight", "mirror"):
n = sum(relations(a, b)[key] for a, b in pairs)
print(" P({:11s}) = {:4d}/{} = {:.4f}".format(key, n, tot, n / tot))
nb = {}
for a in CELLS:
k = sum(relations(a, b)["touching"] for b in CELLS if b != a)
nb[k] = nb.get(k, 0) + 1
print(" neighbours of a cell (cells with that many): {}".format(
dict(sorted(nb.items()))))
# ordered pairs of distinct cells: 1260
# Manhattan distance law:
# d = 1 118/1260 0.0937
# d = 2 188/1260 0.1492
# d = 3 214/1260 0.1698
# d = 4 200/1260 0.1587
# d = 5 168/1260 0.1333
# d = 6 136/1260 0.1079
# d = 7 104/1260 0.0825
# d = 8 72/1260 0.0571
# d = 9 40/1260 0.0317
# d = 10 16/1260 0.0127
# d = 11 4/1260 0.0032
# E[d] = 13/3 = 4.3333
# P(touching ) = 214/1260 = 0.1698
# P(same row ) = 288/1260 = 0.2286
# P(same column) = 108/1260 = 0.0857
# P(knight ) = 148/1260 = 0.1175
# P(mirror ) = 100/1260 = 0.0794
# neighbours of a cell (cells with that many): {3: 4, 5: 18, 8: 14}距离与邻接的基率
两张指定的牌之间曼哈顿距离的期望恰为 13/3;最常见的距离是 3(0.1698),距离为 1 的概率是 0.0937。八方向相邻(国际象棋中「王」的一步)的概率为 0.1698,大约六分之一。同一行的概率 0.2286 远高于同一列的 0.0857,因为一行有 9 格而一列只有 4 格。马步关系的概率为 0.1175,镜像关系(关于中列、中行或两者的反射)的概率为 0.0794。
把这几种关系放在一起,任意一张关键牌与指示牌之间「有某种值得一提的空间关系」几乎是必然的:相邻、同行、同列、马步与镜像各自的基率已经在一成到两成之间,而读者通常同时考察好几张关键牌。
邻接关系的基率还取决于指示牌落在哪里:四个角上的格子只有 3 个邻居,边上的 18 个格子有 5 个,内部的 14 个格子有 8 个。
另一种布局与「本宫」
推导
# Chapter 3. p07: The 8 x 4 + 4 layout, and cards in their own house.
# Layout B: four rows of 8, plus a row of 4 under the middle columns
# (columns 2-5). Position k (1..36, read row by row) is "house k" and
# carries the meaning of card k. A card in its own house is a fixed point
# of the random permutation; their number has an exact law (derangements)
# close to Poisson(1). We also count how many cards touch the
# significator, averaged over where it lands.
from fractions import Fraction
from itertools import permutations
from math import comb, exp, factorial
def layout_a():
return [(r, c) for r in range(4) for c in range(9)]
def layout_b():
return [(r, c) for r in range(4) for c in range(8)] + \
[(4, c) for c in range(2, 6)]
def touching(a, b):
return max(abs(a[0] - b[0]), abs(a[1] - b[1])) == 1
def derange(n):
d = [1, 0] # D(n) = (n - 1)(D(n-1) + D(n-2))
for m in range(2, n + 1):
d.append((m - 1) * (d[-1] + d[-2]))
return d[n]
def summary(name, cells):
pairs = list(permutations(cells, 2))
t = sum(touching(a, b) for a, b in pairs)
d = sum(abs(a[0] - b[0]) + abs(a[1] - b[1]) for a, b in pairs)
print(" {}: P(touching) = {:.4f}, E[Manhattan] = {:.4f}".format(
name, t / len(pairs), d / len(pairs)))
print(" mean neighbours per cell {:.3f}".format(
t / len(cells)))
if __name__ == "__main__":
summary("9 x 4 ", layout_a())
summary("8 x 4 + 4", layout_b())
n = 36
print("\n own-house cards among 36 (exact vs Poisson(1)):")
for k in range(6):
p = Fraction(comb(n, k) * derange(n - k), factorial(n))
print(" k = {} {:.5f} {:.5f}".format(
k, float(p), exp(-1) / factorial(k)))
p0 = Fraction(derange(n), factorial(n))
print(" P(at least one card in its own house) = {:.5f}".format(
1 - float(p0)))
print(" P(significator 28 in house 28) = 1/36 = {:.5f}".format(1 / 36))
# 9 x 4 : P(touching) = 0.1698, E[Manhattan] = 4.3333
# mean neighbours per cell 5.944
# 8 x 4 + 4: P(touching) = 0.1730, E[Manhattan] = 4.1048
# mean neighbours per cell 6.056
#
# own-house cards among 36 (exact vs Poisson(1)):
# k = 0 0.36788 0.36788
# k = 1 0.36788 0.36788
# k = 2 0.18394 0.18394
# k = 3 0.06131 0.06131
# k = 4 0.01533 0.01533
# k = 5 0.00307 0.00307
# P(at least one card in its own house) = 0.63212
# P(significator 28 in house 28) = 1/36 = 0.02778布局把最后 4 张放在中间四列之下,平均每格的邻居数由 5.944 增加到 6.056,相邻概率由 0.1698 升到 0.1730,平均曼哈顿距离由 4.3333 降到 4.1048。两种布局的差别很小,但它说明「相邻」的基率不是一个常数,而是布局的函数。
大桌阵的另一条读法是「宫位」:第 个位置是第 号牌的宫。牌落在自己的宫,就是随机排列的一个不动点。不动点个数的精确分布由错排数给出,与均值为 1 的泊松分布在小数点后五位都一致:恰好没有、恰好一张的概率都是 0.36788,至少有一张牌落在本宫的概率为 0.63212。指定的一张牌(例如 28 号)落在本宫的概率只有 1/36,即 0.02778,但「总有一张牌落在本宫」的概率接近三分之二。
注
指定事件与「某个」事件。 「这张牌落在本宫」的基率是 0.02778,「有牌落在本宫」的基率是 0.63212。事后挑出发生了的那一个,再按指定事件的基率去评价它,是把基率放大了二十多倍。第五章把这一点推广到整个牌阵。
小结
大桌阵是网格上的均匀随机排列。两张指定的牌相邻的概率约为六分之一,同行约为两成,马步约为一成,平均曼哈顿距离为 13/3;这些基率随布局与指示牌的位置变化。至少一张牌落在本宫的概率为 0.63212。
参考文献
Fan, Y. (2026). Tarot Hypergeometric Sampling, Lenormand Grid Topology and Kabbalistic Tree Graphs: Reversal Binomial Mixtures, Grand-Tableau Manhattan Distances and Twenty-Two-Edge Spectral Centrality [塔罗超几何抽样、雷诺曼网格拓扑与卡巴拉生命树图论] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090191 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).