You could probably do something to compress algebraic chess notation in such a way, and even without it as GP says it might cover a decent number of games. My point was you couldn't guarantee a fixed length in 26B, afaik.
(That was the article I expected from the malformed title here: a combination of an efficient encoding and it can it can only get longer than this by repetition so here's a very clever thing we can do, or something.)
(That was the article I expected from the malformed title here: a combination of an efficient encoding and it can it can only get longer than this by repetition so here's a very clever thing we can do, or something.)