> Therefore, the smartest pro team should study what the bot is doing and trust that on average it's a better idea to always focus on the ranged barracks first. After all, if it was a bad idea, they probably wouldn't do that.

This is exactly what is happening with Go right now. Many pros are emulating and learning from AlphaGo (Zero) and are starting to play moves that were always thought to be suboptimal until now.