The concept of what constitutes a "dominated" action in an extensive game is rather subtle. What is currently implemented is actually a mixture of a kind of "conditional" dominance and agent-form dominance (assuming subsequent actions are chosen independently as if in the agent-form). This is neither here nor there.

As this is not a matter of simply correcting the implementation (but requiring an entirely new approach), we will remove this implementation. It is actually only available at the moment via the GUI and was never exposed in Python.

So we will:

• Remove this from the GUI. This includes also being able to remove the dominance elimination stack (which would need re-factoring anyway), and the layout features which support the display of partial trees (likewise something which would have needed attention independently).
• Remove this from the BehaviorSupportProfile, including removing any functions which exist solely to support this.

In a future version we would look to implement this type of analysis more cleanly - also possibly using sequences rather than actions (should there be results in that direction).