tag:blogger.com,1999:blog-37404288.post139312455087173365..comments2023-11-02T14:40:18.756+01:00Comments on JP Moresmau's Programming Blog: Waterloo Contest and Human IntelligenceJP Moresmauhttp://www.blogger.com/profile/09964251063221757176noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-37404288.post-32252314415931626322010-04-09T16:50:55.640+02:002010-04-09T16:50:55.640+02:00This comment has been removed by a blog administrator.嬌琪https://www.blogger.com/profile/14719522412706127374noreply@blogger.comtag:blogger.com,1999:blog-37404288.post-16963610005270889762010-02-12T13:10:36.702+01:002010-02-12T13:10:36.702+01:00The Go site Sensei's Library has enough inform...The Go site Sensei's Library has enough information (I think) to get a basic implementation going:<br /><br /><br />http://senseis.xmp.net/?MonteCarloTreeSearch<br /><br />http://senseis.xmp.net/?UCTd_ahurahttps://www.blogger.com/profile/00347459570008385918noreply@blogger.comtag:blogger.com,1999:blog-37404288.post-20023369454357833452010-02-12T13:03:43.899+01:002010-02-12T13:03:43.899+01:00The method the first poster mentioned has the meri...The method the first poster mentioned has the merit as being a first step to a Monte Carlo Tree Searcher by adding a random opponent. I have no problem imaginig MCTS as being effective for playing TRON.<br /><br /> MCTS is used successfully in Go with its brutal search and state space. Additional benefits is that by using winning percentages you don't need an evaluation function.<br /><br /> Implementing the simple UCT algorithm is done in a wink and as an additional bonus the algorithm is best-first and always has a best move ready.d_ahurahttps://www.blogger.com/profile/00347459570008385918noreply@blogger.comtag:blogger.com,1999:blog-37404288.post-54902908411968511982010-02-12T10:39:48.528+01:002010-02-12T10:39:48.528+01:00At a certain point, I was using a sorf of genetic ...At a certain point, I was using a sorf of genetic algorithm. It was more or less like this:<br /><br />- Create a new, stupid bot that takes a random direction every turn.<br />- Simulate this bot and see how many steps it takes before it crashes into a wall.<br />- Repeat this <b>a lot</b> and select the bot that was able to take most steps.<br />- Choose the direction the best bot took first.<br /><br />This initially performed very well, but then a lot of very smart bots started to appear...jaspervdjhttps://www.blogger.com/profile/09950479534377615851noreply@blogger.com