|
horza66 -> Improving the AI (2/8/2011 11:38:07 AM)
|
This is intended as a general thread on the AI and how it may be improved. It's worth pointing out that good AI is very very hard and computationally intensive, and that GAIA is already capable of some impressive feats - I've seen it use HQ buildup to replicate creating the massive pocket behind Kiev. The intention of the thread is to identify and collect common issues that might be fixed.
AI: Axis
Level: Challenging
Date: June 22 - October 16 1941
The AI is missing the opportunity to form pockets in the early turns, tending to use the panzers to hit weak Soviet units head on rather than pocket them for the landsers to deal with. This leads to significantly lower Soviet casualties overall. GAIA seems to understand how to form large pockets, and seeks them when presented with a coherent front line, but fails to do so when presented with scattered handfuls of divisions. It deals particularly poorly with any kind of checkerboard defence.
AI: Axis
Level: Challenging
Date: October 1941
Issue: When presented with a long, well populated Soviet front line GAIA does not recognise the threat posed by the mud turns. It tends to over commit forces in early October and is unable to free them from pockets when Sov Inf/Cav infiltrate forward, counter-attacks being largely ineffective. A simple solution to this would be to have GAIA recognise the significance of gaps in the frontline and breakdown a fraction of divisions into regiments to backstop those gaps and prevent forward units being pocketed by infiltration. However those regiments will be vulnerable in the blizzard turns, and this would have depend on how capable GAIA is at a fighting retreat to a fort line.
All discussion welcome. Saves available if required.
|
|
|
|
