search

carsten-wenderdel / wildbg

Backgammon engine based on neural networks
https://wildbg.shuttleapp.rs/swagger-ui/
Apache License 2.0
18 stars 5 forks source link

Document backgammon inputs for neural networks #4

Open carsten-wenderdel opened 1 year ago

carsten-wenderdel commented 1 year ago

Previous backgammon engines have calculated various metrics to help them with their evaluation: pip counts, number of blots, etc.

Many of those metrics were introduced by Hans Berliner: https://www.semanticscholar.org/paper/BKG%E2%80%94A-Program-That-Plays-Backgammon-Berliner/da1556610fa295332aabce77533df5e773abd84d

Others were introduced by gnubg.

Those metrics can be used as inputs for our neural nets and improve speed and accuracy. In a first step we should document as many of those metrics as possible. In later phases we then can implement them in code and try out how much benefit they bring.

Remember to no copy/paste code protected by the licenses as the GPL! Studying algorithms and then describing them in natural language is fine, however.

PhilipSiler commented 8 months ago

Write functions to perform pip count calculations would be a good first step.