-
Notifications
You must be signed in to change notification settings - Fork 936
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Returned Policies and Exploitability #1215
Comments
Wow, that is quite the gotcha.... I'm surprised the first one even works! The problem is that you're trying to compute exploitability for just a single policy. But exploitability is a function of the entire strategy profile (both policies) or "joint policy", not just the one player's policy. The sequence form LP code gives you them back separately rather than contained in a single object (like e.g. the CFR code does). Whereas The fix is to just merge them into a single joint policy:
We should totally have this in a test. I was surprised not to see it in the sequence_form_lp_test, so please leave the issue open as a reminder to add a test that does the above so there's a reference to it in the code somewhere. |
Oh that makes sense! I was curious why there were two returned policies, but each policy itself was still defined over the joint action space (with uniform random values in the other player's states). Thanks!! |
Can I get some clarity on the input assumptions when calculating exploitability? See the following simple example:
The exploitability here is nonzero even though the utilities match the expected values (-1/18, 1/18). This does not happen when I used the closed form expression for the kuhn nash equilibrium:
This exploitability is zero. However, when I manually compare [pi_1, pi_2] and pi_kuhn, they look almost identical. What am I missing?
The text was updated successfully, but these errors were encountered: