Poker-AI.org

I am working with Pure CFR https://github.com/rggibson/open-pure-cfr and currently struggling to implement a best response calcuation.
I want to calcuate best response for the abstract game.
I used the Leduc CFR best response code http://poker.cs.ualberta.ca/open_cfr.html as a reference, but the code contains a lot of magic numbers that I don't understand.

How do I validate that my best response calculation works? Train with cfr longer and longer and make sure best response minimizes? Is there also another way?
Are there best response calculations already available for Pure CFR or is anybody willing to share his implementation?

Does nobody have any idea, sample code or hint? Would really appreciate any help

I'm not familiar with Pure CFR. In CFR you can get a best response strategy by running the algorithm but hold the strategy of one player constant and update the strategy (accumulated regret, average strategy) of the other player. It should eventually become a pure strategy.

http://poker-ai.org/phpbb/viewtopic.php?f=24&t=2742