A note on percentages

[Elom0]

This sums up another aspect of what (2023-09-08 I think could perhaps be better considered when using AI win rates)

What I've been thinking of for the past couple of months, or maybe more than that, is that in a game between humans, the winrate is the. Which that human winrate is dependant on not just the skill level of the players, but also their style. That's why I strongly believe in the age of AI, oddly enough through the concept of winrates the human element is more glorified than ever since 바둑 teaching pros should emphasise their style and let potential students pay for the pro that matches their style, and the same is true for (象)将棋, chess and other mindsports..

[RobertJasiek]

He does not suggest what his paper title suggests but suggests to use both percentage and score, as I do.

Elom0, your message editing is mistaken. Please reveal the essential, lost contents of your message!

[Elom0]

Winrate estimates that are given are normally AI winrates versus itself, and homosapien equivalents are your winrate versus yourself, and your opponent's winrate versus itself, which could be AI and we're back full circle. Between players of different skill levels, however, relative self-play winrate is what is required to ascertain the likehood of either player winning or losing, and there again, the concept of using winrate relative to proportion of projected won games is very relevant! Your winning winrate relative to you losing winrate in a position, all relative to the opponent's winning and losing winrates to itself, show the proportion to which the winrates of the player with the higher winning to losing self winrates should be multiplied to get the actual winrate between both players. If a position fits your style, then your winrate versus yourself in that position will be higher than that of a person who is generally the same skill level as you, however that position does not fit their style, and in that position you are the player with the higher skill level.

Both AI and generalised teaching lacks in that a move that would increase the teachers winrate may not increase the students; homosapien teachers account for this to some degree by offering simpler alternate moves to weak players, however the best way is too match the teacher to the style of the student.