AI vs. AI battles?

[Hobot]

There must be some kind of formula that the AI uses to auto-decide battles, and I'm not looking for the specifics but just generally speaking, does anyone have any clue into what goes on in an AI vs. AI battle, how does the result come about??? This seems especially meaningful to me, because in most of my games I see the same factions doing well, does this mean that their units have some sort of bias in an AI battle calculation? The same question applies for auto combat I suppose, since the results are calculated somehow...not that I care about auto combat really...but I have tried it, with the results never being what I would accomplish even in a trivial battle (which are the only kind I may not care to fight...10 to 1 odds are just no fun) Also, does anyone know if the AI factions have any difference in their behavior? Are some more military unit minded and others simcity types? Are some more aggressive than others?

Personally I've played the game a whole lot...it's these small questions that are all that are left to me

[motorhead]

Faction behavior is controlled to an extent by the start files in (./campmap/startpos/{early|high|late}.txt). For example, early.txt contains the following:

Quote[/b] ]
SetBehaviour:: FN_ALMOHAD MUSLIM_PEACEFUL
SetBehaviour:: FN_BYZANTINE ORTHODOX_STAGNANT
SetBehaviour:: FN_DANISH CATHOLIC_NAVAL_EXPANSIONIST
SetBehaviour:: FN_EGYPTIAN MUSLIM_EXPANSIONIST
SetBehaviour:: FN_ENGLISH CATHOLIC_CRUSADER_TRADER
SetBehaviour:: FN_FRENCH CATHOLIC_EXPANSIONIST_CRUSADER
SetBehaviour:: FN_GERMAN_HRE CATHOLIC_EXPANSIONIST_CRUSADER
SetBehaviour:: FN_ITALIAN CATHOLIC_TRADER
SetBehaviour:: FN_POLISH CATHOLIC_DEFENSIVE
SetBehaviour:: FN_RUSSIAN ORTHODOX_EXPANSIONIST
SetBehaviour:: FN_SPANISH CATHOLIC_DEFENSIVE_CRUSADER
SetBehaviour:: FN_TURKISH MUSLIM_EXPANSIONIST
SetBehaviour:: FN_ARAGONESE CATHOLIC_DEFENSIVE
SetBehaviour:: FN_BURGUNDIAN CATHOLIC_DEFENSIVE
SetBehaviour:: FN_GOLDEN_HORDE BARBARIAN_RAIDER
SetBehaviour:: FN_HUNGARIAN CATHOLIC_DEFENSIVE
SetBehaviour:: FN_NOVGOROD ORTHODOX_EXPANSIONIST
SetBehaviour:: FN_PAPIST POPE
SetBehaviour:: FN_SICILIAN CATHOLIC_NAVAL_EXPANSIONIST
SetBehaviour:: FN_SWISS CATHOLIC_ISOLATIONIST

I especially like how the papal faction's behavior is just set to POPE (I guess that's latin for 'crazy as a loon&#39.
As for auto-calc AI vs. AI, no real clue. My guess is raw combat numbers are the basis, with modifiers for command rating, terrain and who is the defender.

[Oaty]

To a degree I have found auto calculate somewhat accurate as to what would happen in the battle. I have noticed that valour is very highly taken into consideration. (wich means command stars can make a big difference) I also know that it takes into account of how elite the unit is so you couldnt throw 1000 peasants against 200 elite units in auto calculate and hope to win. The only thing I have no idea about is if it takes into consideration of what the army consists of since my armies are pretty much balanced with different types of units, what I mean by this if you have a bunch of spearmen defending against cavalry os a bunch of swordsmen attacking spearmen.

[econ21]

One thing I have noticed about auto-calc on expert at least is that it does not do a bad job of replicating my own results if I am much superior to the enemy - especially if I have a general with a far superior command. I have been surprised to play the same battle manually and auto-resolving to find similar (very low) casualties. When forces are more equal, my losses are much higher on auto-calc and, of course, you need manual control to win against the odds. What this means is that towards the end of the game, when you have a clear superiority in numbers and command, you can avoid exhausting battles by using auto-resolve. Manual control is most important early in the campaign and at the decisive moments of wars (the El Alamein's and Stalingrads so to speak).

[Demon of Light]

One time I saw 13 stacks of peasants and other rabble(Rebels) in the same province. The Germans re-emerged with 3 stacks of good troops. When the computer was left to its own devices, the rebels lost 4 stacks and won the battle. When I sent an observation force in, the Germans won the battle and lost 1 3/4 stacks with all 13 rebel stacks being annhiliated. It was a great battle too. Hordes of peasants and fanatics launching themselves at royal knights and feudal foot knights. Great fight. In the end, all that was left of the Germans were archers to beat back 2 waves of peasants.

[Hobot]

interesting...from these few replies I then gather that the AI calculates more based on numbers than anything else?

Motorhead, thanks for the data man

[Ironside]

The main difference between a real battle and auto-calc is that the calculation includes all the forces and not one wave after another witch makes numbers more important than in the played battles, were elites most often is the way to go.
Real battles is 1 vs 1 ten times while auto-calc is 1 vs 10.

In late games when you have larger armies than the ememy auto-calc gives you less losses than a real battle.
Stars seems to matter alot in auto-calc, in STW MI the Mongols lost vs Hojo if a 0 star was in commad, won if it was a 2 star and was unstoppable if 4 star. I assume its simular in MTW.

[econ21]

Another thing that was marked in STW, and still present in MTW, is that rivers are much less significant in auto-calc than played manually. Defending a river manually allows you to butcher the AI; attacking a river on auto-calc allows you to escape butchery.