kirillbobyrev / pabi

🤖 [WIP] Monte-Carlo Tree Search chess engine
https://kirillbobyrev.github.io/pabi/docs/pabi/
GNU General Public License v3.0
13 stars 1 forks source link

Create a strong benchmark set #11

Open kirillbobyrev opened 2 years ago

kirillbobyrev commented 2 years ago

Tracking the performance of the engine early on is crucial for its success. There are established techniques such as Perft: I should pull the data pre-calculated by other engines to see how well Pabi is doing and continuously improve/avoid regressions. Running those benchmarks continuously (#10) is a priority starting in the early days of development.

kirillbobyrev commented 2 years ago

Another idea for benchmark: analyze several important positions from ECO. Source: Fire benchmarks.