Six dubious ways to estimate the difficulty of a chess puzzle
- Speaker(s)
- Jan Zyśko
- Affiliation
- MIMUW
- Date
- Oct. 8, 2021, 2:15 p.m.
- Information about the event
- 5820 i online https://meet.google.com/jbj-tdsr-aop
- Seminar
- Seminar Intelligent Systems
Serdecznie zapraszamy na pierwsze spotkanie seminarium w roku akademickim 2021/22. Referat w formie hybrydowej (na żywo i on-line pod adresem https://meet.google.com/jbj-tdsr-aop ) wygłosi Jan Zyśko.
We cordially invite all participants to the first presentation of the new academic year 2021/22. The hybrid presentation (in person and on-line at https://meet.google.com/jbj-tdsr-aop) will be delivered by Jan Zyśko.
---------------------------
Tytuł: Sześć wątpliwych sposobów na estymację trudności zadania szachowego.
Title: Six dubious ways to estimate the difficulty of a chess puzzle.
Abstract: Despite recent advances in mimicking human play, there is still a huge gap to bridge before engines can be used efficiently for improving chess skills. One milestone is to be able to estimate how difficult a given position is to solve. This would allow the chess player to objectively gauge the quality of their play. I present several plausible approaches based on existing chess models created for optimal play. In particular, I present models based on ensembling predictions from engines used to emulate weak human play, entropy measures of the engines’ policy heads, and the properties of MCTS ran on the positions.
---------------------------------------
Plan wystąpień w tej edycji jest dostępny tutaj
The schedule of presentations can be checked here