The Original I Ching
esenfr
← all experiments
≈ · adjusted Rand index

Comparator of partitions

How alike are two ways of grouping the 64?

Choose two partitions of the 64 and compare how much they agree with the adjusted Rand index.

Original finding

The site splits the 64 hexagrams in several ways: Jing Fang's palaces, the basins of the nuclear map, the cosets of the subgroup of pure hexagrams, and the rows and columns of trigrams. How alike are any two of them? The adjusted Rand index (ARI) measures it: 1 if they are identical, 0 if they are independent, negative if they separate more than chance.

vs
−0.125
adjusted Rand index
further apart than chance
verdict
8 × 8
groups compared
Cross matrix
A \ B01234567
011111111
111111111
211111111
311111111
411111111
511111111
611111111
711111111

row = group of Jing Fang's palaces, column = group of Cosets of the pure hexagrams; cell = hexagrams in common. A clean diagonal would mean identical partitions.

The finding: the palaces are not the cosets

The palaces and the cosets are both partitions into 8 groups of 8, but they are not the same: their ARI is −0.125 (they separate more than chance). The reason is algebraic: each palace is the pure hexagram plus a fixed set of 8 generating masks, {0, 2, 32, 48, 56, 58, 60, 62}, and that set is not closed under XOR. Not being a subgroup, the palaces cannot be cosets of anything. Two constructions that look like twins (8 × 8) turn out, when measured, to be almost opposite.