Quote:
Originally Posted by asunder
I am just starting to use advanced Monte Carlo. I watched a poker coach using it and he ran the sims at 1,000m. Is this what you would recommend? How do we know when the ranges are properly sorted out? Can I easily tell if 500m was not enough? Is there documentation to help with these answers?
As a general rule of thumb, 2,500m samples per 1GB tree size should give you a rough ballpark . Some users prefer to run less than that and then subtree sample the specific lines they are interested in, that's fine too.
The current recommendation is to:
1) Run half of the total iterations with decay 1E-6.
2) For the second half, disable decay and check "reset strategies".
How to spot if the ranges are converged?
As a bare minimum, you should not see any -EV hands in the ranges and you should not see any +EV hands being folded.
Beyond the obvious signs, how can you tell if it's fully converged?
No easy way. You can do a save and then run another 25% of total iterations with "reset strategy" checked. Once this new batch of samples is finished, load the previous save in "viewer mode" and compare the last result with the previous save. if the ranges were still moving significantly then repeat this step until they stay static between runs.