2024 World Computer Chess Championships Tournament Report

-by Charles Roberson

Introduction

First, many thanks to the tournament and ECAI (European Conference on Artificial Intelligence) sponsor: Google Deep Mind. They paid for 6 nights in a hotel and some of the competitor’s travel expenses. Thanks to the ICGA (International Computer Games Association) for organizing it. I enjoyed it being held in conjunction with ECAI 2024. We had lots of scientists as spectators and many had lots of questions.

Applications were received starting in March. Only 10 of the applicants met the qualification standards. The field dropped to 9 due to one withdrawing. There were 3 tournaments: World Computer Chess Championships, World Computer Blitz Chess Championships and the World Computer Software Chess Championships. Each tournament was 9 rounds, thus 27 rounds total in 5 days. The first two listed are unlimited hardware and the software championships was uniform hardware provided by the local university. The first two days was the World Computer Software Chess Championships and the last two were the unlimited hardware events: World Computer Chess Championships and World Computer Blitz Chess Championships interleaved with each other.

All three tournaments use the traditional methodology from 1974 when computers were not mobile and could be connected to each other. Watch this video of the blitz chess match between Ares and GridChess/Fritz to see how it is done.

Overview

The competition was quite strong with all the programs rated over 3000 and 5 were world champions. The world’s strongest human is Magnus Carlsen rated at 2831 as of 30/10/2024. A 200 point rating gap suggests that Magnus could possibly win 1 out of 4 games with the weakest entrant. While the strongest programs with their extreme hardware configurations are well over 3600 giving Magnus at best a 2% chance of winning a game.

I tuned Ares for longer time controls such as the traditional G/105+15, but venue time contraints limited the software championships and the main event to G/30+10. So, Ares had to average about 40 seconds per move instead of 2 minutes per move.

Ares seemed to be the wildcard program in the group with some critical results knocking programs out of playoffs. The first was in the World Computer Chess Software Championships. Going into the last round (round 9), there was a four way tie for first Ares drew against Raptor with only 47 seconds left on Ares’ clock knocking Raptor out of the playoffs.

The second was in World Computer Chess Championships (unlimited hardware) where Ares playing with the black pieces drew former world champion Fritz running on a 704 CPU grid. Due to some issues porting my MPI implementation from LINUX to Windows 11, Ares ran on one CPU.

Ares finished 8th in each of the 3 tournaments. While I got over most of the jet lag and allergy issues by the third day. There were some lingering opening book issues. When the opening book wasn’t an issue, Ares performed quite well as in the games with Raptor, GridChess/Fritz, Tornado and Tech 4.

Overall, Ares performed well. I noticed a few moves in some of the games that suggest there might be some bugs. If printed out in paperback form, Ares’ code is between 550 and 600 pages. So, it could take some time to find the bugs. For those interested in how Stockfish performed, it tied for 3rd in the software championships (uniform hardware) and finished 6th in the other two tournaments. It really isn’t the best program in the world. It is just one of the top programs for personal computers and cell phones.

Full report with Ares’ games

Ares in the 2024 World Computer Chess Championships