Chess and starcraft games have been played by AI and sometimes won. Artificial intelligence had for these games a complex set of rules, database and decision-making tools to decide which actions it will do in the future.
It is said that the strategies applied by bot players are analyzed and compared with human activity at chess, especially chess experts such as Magnus Carlsen: Jean-Marc Alliot, When artificial intelligence evaluates chess champions, ScienceDaily, April 25th 2017.
But a chess bot is only giving a set of actions for a given game situation. So my question: How are AI's strategies identified from the "optimized" set of actions AI propose?