The microRTS competition has several tracks.
There is a track where you have access to the whole game state (what we have been doing) and a track that is "partially observable" which basically means (I think) that there is a fog of war preventing you from seeing all enemy actions. Therefore, it does not make sense to mix up PO and non-PO agents in the same enemy sequence.
I think that we should primarily focus on the non-PO track, and therefore the non-PO agents. Some agents, like the RandomBiasedAI (and likely others) probably make sense to use in both tracks, but in order to train an agent for the non-PO track, you should use the WorkerRush instead of the POWorkerRush (this rule applies to other PO classes).
On a side note: I think the RandomBiasedAI is too easy to beat, and we are wasting time evolving against it. Just skip it and start evolving against WorkerRush first
The microRTS competition has several tracks. There is a track where you have access to the whole game state (what we have been doing) and a track that is "partially observable" which basically means (I think) that there is a fog of war preventing you from seeing all enemy actions. Therefore, it does not make sense to mix up PO and non-PO agents in the same enemy sequence.
I think that we should primarily focus on the non-PO track, and therefore the non-PO agents. Some agents, like the RandomBiasedAI (and likely others) probably make sense to use in both tracks, but in order to train an agent for the non-PO track, you should use the WorkerRush instead of the POWorkerRush (this rule applies to other PO classes).
On a side note: I think the RandomBiasedAI is too easy to beat, and we are wasting time evolving against it. Just skip it and start evolving against WorkerRush first