FIBS Board backgammon forum
Bots & other tech stuff => Other bots & tech stuff. => Topic started by: purol on May 13, 2004, 07:38:53 PM
I was thinking if it is possible to start a new GNUBot, with an UNTRAINED (empty?) neural net, which trains itself from it's matches played on FIBS.
It would be fun to watch him play stupid in the beginning, play unexpected for a while and slowly becoming an expert!
What do you think, Webrunner? Impossible, too much of a hassle, possible, or is it just what all the nn-bots do?
- Puro
Well.. the problem is that the bot code starts out with a specific setting and so does gnubg.
I haven't created the bot code (nor did i create gnubg) so i really don't have much power over that. The bots are not using any database as it is now, just the gnubg engine.
On the current server there is no more room for a new bot.
I will have to look at the load on the new server before i make any promises about hosting more bots.
If you can get it to work, i can have a look if i can host it.