search

raccrompton / BookBuilder

An automatic Chess opening repertoire Builder
GNU General Public License v3.0
126 stars 18 forks source link

Draws should count as half wins #12

Open kamekura opened 2 years ago

kamekura commented 2 years ago

Currently the user can set an option in the config file in order to count draws as wins or losses. However, the standard in chess is to count each draw as "half a win". https://en.wikipedia.org/wiki/Glossary_of_chess#winning_percentage

Consider a hypothetical position where we're deciding between moves A and B for black. Move A leads to 52% White wins, 44% Black wins and 4% draws (Black score=46%). Move B leads to 40% White Wins, 40% Black Wins and 20% draws (Black score = 50%). Clearly B is much better for Black, but if you set "draws don't count" then A will be chosen. We could easily construct an example where the "draws count as wins" setting doesn't work.

raccrompton commented 2 years ago

I think this is already fixed in a recent version. They’re now half wins

On Thu, 25 Aug 2022 at 15:18, Rafael Caetano @.***> wrote:

Currently the use can set an option in the config file in order to count draws as wins or losses. However, the standard in chess is to count each draw as "half a win". https://en.wikipedia.org/wiki/Glossary_of_chess#winning_percentage

Consider a hypothetical position where we're deciding between moves A and B for black. Move A leads to 52% White wins, 44% Black wins and 4% draws (Black score=46%). Move B leads to 40% White Wins, 40% Black Wins and 20% draws (Black score = 50%). Clearly B is much better for Black, but if you set "draws don't count" then A will be chosen. We could easily construct an example where the "draws count as wins" doesn't work.

— Reply to this email directly, view it on GitHub https://github.com/raccrompton/BookBuilder/issues/12, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABWBKZWV2IOYNIX465MAMN3V256EFANCNFSM57TJVOZQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

kamekura commented 2 years ago

My bad; I was looking at an older version. By the way I think even the option is unnecessary. What's the use case for not counting draws?

raccrompton commented 2 years ago

Good question. In practice online I doubt it makes much difference, given there are few draws. Perhaps there are some situations in which a less clear line is preferred to a clearer likely draw, though I agree this is not the default.

On Fri, 26 Aug 2022 at 12:50, Rafael Caetano @.***> wrote:

My bad; I was looking at an older version. By the way I think even the option is unnecessary. What's the use case for not counting draws?

— Reply to this email directly, view it on GitHub https://github.com/raccrompton/BookBuilder/issues/12#issuecomment-1228396339, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABWBKZWEQM3VSSFYJNMWFU3V3CVR5ANCNFSM57TJVOZQ . You are receiving this because you commented.Message ID: @.***>

kamekura commented 2 years ago

At higher rating bands the draw rate is significantly higher. For example, the Petroff Defense has a 12% draw rate for 2500 blitz players.