641.leela_s
Gian-Carlo Pascutto <gcp [at] sjeng.org>
Artificial Intelligence (Monte Carlo simulation, game tree search & pattern recognition)
641.leela_s is a Go playing engine featuring Monte Carlo based position estimation, selective tree search based on Upper Confidence Bounds, and move valuation based on Elo ratings.
It won Silver and Bronze medals in the 2008 Computer Olympiad and has attained ratings of 1 kyu in Internet server play.
It combines recent advances in computer go in a single package. For each position, a restricted set of candidate moves is selected based on a ratings calculation taking into account a set of features and the current pattern on the board. For one of these moves, the program simulates the game until the end, using a combination of both knowledge and randomness. The score of the simulated game is backed up to the root of the game tree, and the next iteration starts by selecting the next move to evaluate, being the one with the current highest upper confidence bound.
641.leela_s's input consists of a textfile containing (incomplete) Go games in the Smart Game Format. Each of these games will be played until the end at a fixed number of simulations per move.
641.leela_s's output consists, per position, of some information while the search is still ongoing: the number of nodes in the search tree, the current percentage of simulated games won for the player to move, and the best variation for both players found so far.
After each move, a table of the top moves is displayed, the number of simulates done for them, their respective winning percentages and the best variation for both players after each respective move.
C++ 2003 plus two features that are standardized by C++11:
In order to allow compilation by C++03 compilers, a subset of BOOST is included in directory src/boost/.
None.
Leela is licensed directly to SPEC by the author, Gian-Carlo Pascutto. As mentioned above, it includes a subset of BOOST.
Please see details in the document SPEC CPU®2017 Licenses.
Last updated: $Date: 2020-08-19 18:52:31 -0400 (Wed, 19 Aug 2020) $
Copyright © 2017-2019 Standard Performance Evaluation Corporation (SPEC®)