i got problems with understanding the MCTS mechanics in the OpenTestbed/CSPoker code. The algorithm is using a model not only for the opponent, but also for the MCTS bot for calculating check/bet, fold/call/raise probabilities. Isn't this contradictory to the corrsponding paper of Van-Den-Broeck, as the bot's actions are simulated by the UCT selector in the decision nodes and the model is actually just needed for the opponent? I would be very grateful for some help.Statistics: Posted by bugAi — Sat Dec 03, 2016 9:45 pm
]]>