HontoNiBaka wrote:
Great work. I am working on a similar project, really the CFR algorithms are something I just implement from papers and hope to forget soon. But if you need help with C++ I would be happy to help, it can be quite weird coming from Java.
I actually got my c++ implementation done. I did a lot of optimizing and it’s 30-50% faster than piosolver on turn subgames and 10% slower than piosolver for for flop subgames (not sure why flop subgames are so much slower). My trees take 2-4x as much memory as piosolver though, depending on the number of bet sizes, so i’m working on compression.