Another way to gauge how well the model has converged onto the optimal parameter values is to look at a plot of all the optima over time. If a tune is bouncing back and forth between different local optima, it could indicate that more games are needed. It is also interesting in general to get a feeling for what the optimization process looked like.
See the following plot (graciously provided by @zz4032) for an example on how such a plot could look like:
In the implementation, care needs to be taken with respect to the plotting ranges, since the tuning ranges can be changed by the user, which will impact the parameter range passed to the plotting function.
Motivation
Another way to gauge how well the model has converged onto the optimal parameter values is to look at a plot of all the optima over time. If a tune is bouncing back and forth between different local optima, it could indicate that more games are needed. It is also interesting in general to get a feeling for what the optimization process looked like.
See the following plot (graciously provided by @zz4032) for an example on how such a plot could look like:
In the implementation, care needs to be taken with respect to the plotting ranges, since the tuning ranges can be changed by the user, which will impact the parameter range passed to the plotting function.
What needs to be done