Hi,

Could you tell us what kind of decision tree the AI uses when recruiting and maybe describe the templates you you use and how close AI follows them ?

A quick mock up for a system might be like this:

standard 30%
random 10%
slow melee 10%
fast melee 20%
ranged 15%
skirmish 10%
monster 5%

likely with army specific odds. Are units chosen by types or by name? If you're interested id be happy to delve deeper into building templates.

standard
5% long range (cannon etc)
15% fast melee
15% short ranged
65% melee

random

slow melee (favoring tough and shielded units; hard to beat with fast melee and ranged)
75% slow melee
25% skirmish

Fast melee (the assault, antiranged heavy army)
25% skirmish
0-25% chariots
25% infantry (anti infantry/sword, vanguard)
25%-50% fast melee (excluding light cavalry)

ranged (depending on cost vs cost can beat slower melee or inflict heavy losses on ranged player armies)
15% long range +up to 35%
35% short ranged
35% anti large melee (favoring infantry) - up to 35%
15% fast melee

skirmish (beats especially slow melee, AI probably doesnt understand melee vs ranged but still good variation)
65-100% skirmish
0-35% random

monster (beats antiinfantry armies, AI doesnt know to clump its large units but still might be interesting variation)
70% monster/monstrous
15% chariot
15% anti infantry/sword