I believe this method is much better than a huge round robin with a lot of useless "high gap" pairings. Instead i wanted to replace all possible high gap pairings with more games between programs of close strength.
We will see the outcome if i can survive more than 7 rounds like that, bearing in mind those zombie programs are very difficult to efficiently control because of different commands and entry modes.
In any case, the experiment is scientifically interesting because it's fun to find ways to port today's trends and methods to a fully historical environment.
The major difference of this project is the ellimination of pondering regardless of the program options, by pausing the emulation in the worst case. This was never done before.
Pondering (aka deep thinking, permanent brain) means the program continues to think ahead when it's the opponent's turn. This leads to a stronger gameplay when next opponent move is correctly guessed. Therefore it's vital to use pondering in a real tournament. But in a testing environment or when two programs are not connected for auto-play, manual move entries cause delays and turn to extra time in favor of the program who's waiting its turn. Testers avoid pondering since a long time ago when two programs share same hardware ressources too. Though it's not the case with C64 emulation and we can run many VICE instances without ressource conflicts.
I had to change level settings on Chessmaster 2100, Sargon, Cyrus II, Petchess and Microcess (all from L5 to L4). I think the settings used in other tournaments do not comply to 30m/60' rapidchess timing, maybe due to no more pondering. Stopwatch checked and verified :-)
Regarding the results, as seen (unseen?) below with a magnifier or by clicking on the image, the first round did not bring extraordinary surprises, hopefully. A clearer overview should be available as soon as two more rounds thanks to the quickest ranking system i want to reach.
No comments:
Post a Comment