In the light of the discussion in #617, does it make sense to add these now or to develop them as part of the changes indicated there?

We should add them as part of the changes for #617. Also, I note that none of the test instances for enumpure_solve "agent" currently have agent-form equilibria that are not also Nash equilibria -- such tests should be added.

We should add them as part of the changes for #617. Also, I note that none of the test instances for enumpure_solve "agent" currently have agent-form equilibria that are not also Nash equilibria -- such tests should be added.

This is now done in this branch with a game now called myerson_fig_4_2.efg, which is tested in test_nash.py::test_enumpure_agent to ensure that we find both the true Nash and the agent Nash.

Note that myerson_fig_4_2.efg is also used in test_behavior.py::test_agent_max_regret_versus_non_agent which confirms that while agent_max_regret is 0, max_regret >0 for the non-true-Nash agent equilibrium.