There are many questions about subset of flops which is suitable to use for approximating preflop EVs (because running 1755 flops takes a lot of time and electricity).
Usual starting point is Will Tipton's thread on 2p2, here:
http://forumserver.twoplustwo.com/15...-game-1408420/
While this is a decent start Tipton made some mistakes (including isomorphic flops) and didn't consider many important conditions. The tests show that EVs based on that subset are a decent approximation but there are visible biases.
Some people tried to improve on that subset with good results. That group now includes us. Here are preliminary results for various subsets using least squares measure (across all preflop hands):
EVs in the table are real EVs computed on 1755 flops for various real life spots. Notice that even some flops of 48 and 51 boards beat Tipton's 99 one (103 with isomorphic flops removed) but you start getting really good results at about 200-300 flops.
This is still ongoing work and a lot more tests are needed before we are comfortable publishing anything because we know people will use it and then blame us if the results aren't correct. Stay tuned!
Last edited by punter11235; 10-19-2015 at 05:30 AM.