after going back to rome, i see how bad M2 AI is

[econ21]

The biggest flaw in the RTW battlefield AI IMO is what you could call a "bunching" effect. I noticed it when running custom battles to simulate the results of between player battles in an RTR PBM. What happens is that the AI picks a few of your units to attack and then throws everything at them. The AI units become crowded and many are wasted milling at the back - it's overkill. By contrast, the human with equal numbers can hold off the AI with a handful of men and then use the rest to go at the flanks and rear of AI blobs - murdering them. This AI weakness meant that is was virtually impossible to simulate player vs player battles - even when I just lined up my men to oppose the AIs and did everything I could to engender a "fair" frontal slogging match, whichever side I controlled would win, unless the numbers were overwhelming (I ended up playing from both sides and averaging).

This bunching flaw was never that apparent to me in SP campaign battles, but reading my old battle reports I can clearly see it. In more controlled custom battles, it's painfully obvious. I think the RTW flaw I've mentioned is related to the AIs poor handling of phalanxes. Puzz3D and others noted the tendency of what should be an awesome AI phalanx line to break up at the last minute as the AI seeks good unit vs unit match ups. It should focus more on a battle line, less on unit vs unit matching.

I'm not sure whether this flaw is still present in M2TW - I've had no cause to run custom battles - but it seems less apparent. The M2TW AI seems better at forming a battleline and simply charging straight at you, without breaking up to focus on apparent weakpoints. It's not stellar tactics, of course, but given numbers and/or quality, it can get the job done.

The one new flaw I have noticed with M2TWs battlefield AI is the tendency of the AI to let its missiles get too far ahead of its main battleline when attacking and so be very vulnerable to cavalry.