I run tests on my bot to see how it behaves in certain situations all the time. Its very important to do this to ensure the bot is well balanced and unpredictable. An example is how it reacts to a flop cbet. So I ran simulations based on that scenerio and recorded how often it calls, folds, raises, floats etc. So when I ran this about 10k times I noticed I needed to have it bluff about 20% more as it was raising too much value. What I actually did was move some value into its calling range and had it bluff raise more floating hands like a pair and backdoor draw etc. This balanced everything out.
I also ran sims on many other things like its preflop 3B bluff %, turn double barrel % and so on. What I wanted to do was gather data on my bot to see what the villains HUD would look like and make sure I didnt have any obvious leaks the he could pick up on. Its not so easy to balance out the river but that is another story...
The key thing on this stuff is to ensure the villain is not getting value on his bluffs which is a big factor to consider. Most bots do not handle aggression well and they tend to over fold. This is where running specific tests will really help such as the bots double barrel fold %.Statistics: Posted by shalako — Mon Mar 31, 2014 4:28 pm
]]>