Battle AI Challenge

[Ordani]

I don't know if it's possible to find something that *no none* complains about, but I think most people here will agree that Shogun and Medieval I had better AI.

The Shogun AI was fantastic -- probably as a result of the simpler unit mix and the less complex playing fields, and while it had its aberrant behaviors, benefited greatly from its simplicity while still having a good repertoire of tactics.

Overall the ETW RT AI is decent -- it generally does something reasonably sound in principle, but often aberrant in details. It's always trying to flank you and take out single units or artillery with cavalry, but will periodically fail in facepalm fashion by sending EVERY unit it has, alone, in series, around to try and flank your cannon.

It also weights holding buildings in towns or fences far too highly, and yet rarely positions anything to support those fortified units, so you end up with single units defending "high value" points on the map with the rest of the army miles away. You can almost see the decision tree actioning as it steps through, then runs out of known-good choices on a per unit basis, and throws everything else into a column and tries to smash your weakest unit in the line.

You have to give some leeway the more complex the permutations of a game are. It's easy to make a challenging AI for Chess or Risk or Go, because you can exhaustively search predictive move sets based on current board set for each turn. That is not algorithmically possible in a game like ETW in which there are astronomical, as in, more atoms than available for a quantum computer the size of the planet, numbers of possible action variations over the course of a real-time event series. I am pleased that the AI will surprise me from time to time; those are the battles I enjoy the most and remember well after I've put the game aside, when I had to come up with a clever adaption to the unexpected and barely pulled a win out of it, and that's why I continue to play TW titles.