5D Library › 黄历择吉布尔格、建除十二直同余与卢恩符文组合学 · English

第五章 卢恩符文的逆位:超几何-二项复合分布

十五个可逆,九个不可逆

古弗萨克(Elder Futhark)是最古老的卢恩字母,共 24 个符文,传统上分为三个 aett(族),每族八个 (Page 1987)。二十世纪兴起的现代卢恩占卜,把倒着抽出的符文读作「逆位」(merkstave),赋予不同的含义 (Blum 1982)。但有九个符文倒过来看起来一样:Gebo、Hagalaz、Naudhiz、Isa、Jera、Eihwaz、Sowilo、Ingwaz、Dagaz。它们没有逆位读法。只有 15 个符文可以读作逆位。

所以一次抽取中逆位符文的数目,并不像常被当作的那样是 p=1/2p = 1/2 的二项分布。不放回地抽 nn 个符文,其中可逆符文的个数 KK 服从参数为 (24, 15, nn) 的超几何分布;这 KK 个各以二分之一的概率倒置,所以给定 KK 时,逆位数 RR 服从二项分布 (K,1/2)(K, 1/2)。RR 的精确分布就是这两者的复合。

推导

# Chapter 5. p09: How many reversed runes in a draw? Of the 24 runes of
# the Elder Futhark, 9 look the same turned upside down (Gebo, Hagalaz,
# Naudhiz, Isa, Jera, Eihwaz, Sowilo, Ingwaz, Dagaz), so only 15 can be
# read "reversed" (merkstave). Drawing n runes without replacement, the
# number K of reversible runes is hypergeometric (24, 15, n); each lands
# reversed with probability 1/2, so R given K is binomial (K, 1/2). The
# exact law of R is compared with the common shortcut binomial (n, 1/2).
from fractions import Fraction
from math import comb

FUTHARK = ("Fehu Uruz Thurisaz Ansuz Raidho Kenaz Gebo Wunjo "
           "Hagalaz Naudhiz Isa Jera Eihwaz Perthro Algiz Sowilo "
           "Tiwaz Berkano Ehwaz Mannaz Laguz Ingwaz Dagaz Othala").split()
SAME_TURNED = {"Gebo", "Hagalaz", "Naudhiz", "Isa", "Jera", "Eihwaz",
               "Sowilo", "Ingwaz", "Dagaz"}

def exact(n, N=24, rev=15):
    p = [Fraction(0)] * (n + 1)
    for k in range(max(0, n - (N - rev)), min(n, rev) + 1):
        pk = Fraction(comb(rev, k) * comb(N - rev, n - k), comb(N, n))
        for r in range(k + 1):
            p[r] += pk * Fraction(comb(k, r), 2 ** k)
    return p

def naive(n):
    return [Fraction(comb(n, r), 2 ** n) for r in range(n + 1)]

if __name__ == "__main__":
    rev = sum(1 for r in FUTHARK if r not in SAME_TURNED)
    print("   runes {}, reversible {}, same when turned {}".format(
        len(FUTHARK), rev, len(SAME_TURNED)))
    for n, name in ((1, "one rune"), (3, "Norns, 3"), (5, "cross, 5"),
                    (9, "nine")):
        e, b = exact(n), naive(n)
        mean = sum(r * x for r, x in enumerate(e))
        print("   {:9s}: mean reversed {:.4f} (shortcut {:.1f}, {:.1f} x)"
              .format(name, float(mean), n / 2, n / 2 / float(mean)))
        print("     r    " + "".join("{:6d}".format(r) for r in range(n + 1)))
        print("     exact" + "".join("{:6.3f}".format(float(x)) for x in e))
        print("     short" + "".join("{:6.3f}".format(float(x)) for x in b))
#    runes 24, reversible 15, same when turned 9
#    one rune : mean reversed 0.3125 (shortcut 0.5, 1.6 x)
#      r         0     1
#      exact 0.688 0.312
#      short 0.500 0.500
#    Norns, 3 : mean reversed 0.9375 (shortcut 1.5, 1.6 x)
#      r         0     1     2     3
#      exact 0.320 0.451 0.201 0.028
#      short 0.125 0.375 0.375 0.125
#    cross, 5 : mean reversed 1.5625 (shortcut 2.5, 1.6 x)
#      r         0     1     2     3     4     5
#      exact 0.146 0.354 0.327 0.143 0.029 0.002
#      short 0.031 0.156 0.312 0.312 0.156 0.031
#    nine     : mean reversed 2.8125 (shortcut 4.5, 1.6 x)
#      r         0     1     2     3     4     5     6     7     8     9
#      exact 0.028 0.132 0.261 0.286 0.190 0.079 0.020 0.003 0.000 0.000
#      short 0.002 0.018 0.070 0.164 0.246 0.246 0.164 0.070 0.018 0.002

抽一个符文,逆位的概率是 0.3125,而不是 0.5。三个符文的命运三女神(Norns)阵,逆位数的均值是 0.9375,而不是 1.5;完全没有逆位的概率是 0.320,简化算法给出 0.125;三个全逆位的概率是 0.028,简化算法给出 0.125。五个符文的十字阵均值是 1.5625;九个符文是 2.8125,九符文完全没有逆位的概率是 0.028,简化算法给出 0.002。简化算法平均把逆位高估 1.6 倍,对两端的判断更是严重失误。

五符文十字阵中逆位符文的数目:精确分布与二项简化(p09)。

用模拟抽取核对

推导

# Chapter 5. p10: Casting the runes 100,000 times (seed 91). For each
# Norns reading three runes are drawn from a bag without replacement and
# each lands upright or turned with probability 1/2; a turned rune of the
# 9 symmetric ones still reads upright. Compared with the exact law of
# p09: the share of drawn runes read reversed (exactly 15/48 = 0.3125),
# and the share of "heavy" readings with two or more reversed runes
# (exactly 0.229, against 0.5 under the shortcut).
import random

FUTHARK = ("Fehu Uruz Thurisaz Ansuz Raidho Kenaz Gebo Wunjo "
           "Hagalaz Naudhiz Isa Jera Eihwaz Perthro Algiz Sowilo "
           "Tiwaz Berkano Ehwaz Mannaz Laguz Ingwaz Dagaz Othala").split()
SAME_TURNED = {"Gebo", "Hagalaz", "Naudhiz", "Isa", "Jera", "Eihwaz",
               "Sowilo", "Ingwaz", "Dagaz"}

def draw(rng, k):
    """k runes without replacement, using rng.random() only."""
    bag = list(FUTHARK)
    out = []
    for _ in range(k):
        out.append(bag.pop(int(rng.random() * len(bag))))
    return out

if __name__ == "__main__":
    rng = random.Random(91)
    n, rev_total, heavy, counts = 100000, 0, 0, [0, 0, 0, 0]
    for _ in range(n):
        r = sum(1 for rune in draw(rng, 3)
                if rng.random() < 0.5 and rune not in SAME_TURNED)
        rev_total += r
        heavy += r >= 2
        counts[r] += 1
    print("   share of drawn runes read reversed: {:.4f}".format(
        rev_total / (3 * n)))
    print("   readings by reversed count: " + " ".join(
        "{}={:.3f}".format(r, c / n) for r, c in enumerate(counts)))
    print("   readings with 2 or more reversed: {:.3f}".format(heavy / n))
    rng2 = random.Random(91)
    upright9 = sum(1 for _ in range(n)
                   if not any(rng2.random() < 0.5 and rune not in
                              SAME_TURNED
                              for rune in draw(rng2, 9)))
    print("   nine-rune casts with no reversal at all: {:.3f}".format(
        upright9 / n))
#    share of drawn runes read reversed: 0.3132
#    readings by reversed count: 0=0.319 1=0.451 2=0.201 3=0.029
#    readings with 2 or more reversed: 0.230
#    nine-rune casts with no reversal at all: 0.028

在 100,000 次模拟的命运三女神阵中,抽出的符文有 0.3132 读作逆位(精确值 0.3125);逆位数为 0、1、2、3 的解读分别占 0.319、0.451、0.201、0.029,与精确分布一致;两个以上逆位的「沉重」解读占 0.230(精确值 0.229,简化算法给出 0.5);九符文完全没有逆位的占 0.028。

注

逆位是现代的约定。 卢恩铭文的书写方向多种多样,铭文本身与中世纪的卢恩诗都没有给倒置的符文另外指定含义。逆位读法,以及哪些符文「不能」逆位的名单,都属于现代占卜;本章的算术描述的是这种做法,而不是古代的用法。

小结

古弗萨克 24 个符文中只有 15 个可以逆位,所以抽出的符文逆位概率是 0.3125,命运三女神阵完全没有逆位的概率是 0.320(不是 0.125),逆位数均值是 0.9375(不是 1.5)。按二分之一算的二项简化,对每一种牌阵都是错的。

参考文献

Blum, Ralph. 1982. The Book of Runes. St. Martin’s Press.
Page, R. I. 1987. Runes. Reading the Past. British Museum Publications.

Fan, Y. (2026). Almanac Auspicious Boolean Lattices, Jianchu Twelve-Officer Congruences and Runic Combinatorics: Constraint Satisfaction Paradoxes, Solar-Term Phase Jumps and Elder Futhark Inversion Distributions [黄历择吉布尔格、建除十二直同余与卢恩符文组合学] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090239 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).