Interesting post but i have a question.
In a puzzle like OOC (many hidden informations and a big branching), we already do that we can have a “shifumi” effect, even in the top 20. Is there any danger of assuming an expected winrate in a puzzle like this one ?