I did a lot of playtesting of campaigns while making a home mod for MedMod IV - that is i played with factions giving all options to the AI other than moving around troops, and also handling over all factions to the AI and see what happens in the long run with "god mode".

My observations relative to peace acceptance is that the AI (in conquest mode) does not accept ceasefires when he smells a rat (that is you have huge stacks on the borders), when your conduct has been "suspicious" (backstabbings, multiple province invasions and generally being persistant in a war and trying to eradicate a faction and also IMO slaughtering prisoners). It seems that he will remain "suspicious" of an offender for some time - but if you let time pass (that can be 20/30 turns for neighbours) he will eventually accept a peace offering.

Some other times he seems almost cunning in the treaties he offers/makes - for example the Seljuks would immediatelly go for a ceasefire if i conquer Nicaea from them as the Byzantines in MedMod IV; that is only to conquer Georgia and Lesser Armenia, in order to boost their income and come after me at a later time with "sufficient" armies - this happened more than once.

The AI factions generally gather stacks in areas where enemy stacks are also concentrated - naturally - as they are afraid them being picked up as victims IMO instead of the originally intended "victim" IMO. Having huge stacks at the borber hinders chances of the AI accepting peace offerings (and some times he explicitly states so). I am sure that naval invasion potentials also come into this greatly - so sometimes the AI may be fearing something on the other side of the map - something that can't be directly seen. Other such factors are the AI's financies and what he is doing/what is happening in other fronts. For example the AI foes for what many players consider "suicide" attacks, that are easily explained if you see his financies (-30,000 flrns) - and even then he waits as long as its needed to ensure that he has the best success chances possible in that single one attack he's attmpting to get out of stagnation. In other cases he undergarissons the frontier with the player because he is battling another opponent that is draining him of troops.

What the AI can't do is the common for the human player "breakthrough" strategy, that is: concentrate all forces (other than the really just necessary) to a rich province of the neighbour/enemy to make sure you knock him out of the run. The AI is doing something similar, but will put lots of "prudence" into that - that is keep substantial garisson armies while doing the attack - this makes his attacks in such cases feel like "suicidal", when in fact it seems to me they are the very opposite (overconservative). This strategy is all the more easy for the player when the profits allow for plenty of armies in the field (multiple stacks in every province) and the absence of homelands for quick loss recovery (replensish armies/stacks).

It is essentially a form of rushing, that puts the human player on top from the very beginning if pursuited, and turns long term planning meaningless (as we all know there's little that can knock the player out of the game if he has say 12 -15 provinces in firm control in 50 turns). That's why i advocate less troops on the map at any time (ie less profit margin) and homelands and high province rebelliousness and slow to convert religiously povinces and no upgrades and low general starting stars - rushing is more hard then and long term planning strategies/systems of organisation and decisions to pursuit say a technology become meaningful as one depends on them (and so needs to protect them until they give him the egde that he is looking for) to win the game or even survive it.

In vanilla, the game evolves in 50-60 turns in a mega-war between superempires that is as unchallenging as it is tedious IMO. Armies win due to upgrades past 60-80 turns and eventually due to outclassing era-wise the opponents and the gap between "leaders" and "runner-ups" is getting progressively larger and larger as the game unfolds. I really dislike this type of gameplay - there is nothing more boring to have to gulp as many minor factions as i can in order to eventually compete with the 2 other similar super states that are forming somewhere at the same time. When i was getting successive messages of "faction elimination" i usually quit the game - it means that someone will pay me a visit with a bunch of jedi troops and the battle will be over upon touch either for or against me. Basically the challenge is then lost because the gap of the clashing odds is also dramatically made larger and meeting relatively equal strength armies on the field - armies made of humanoid sprites instead of terminator sprites that feel "nor fear nor pain" - is becoming a rarity after 60 turns. The game turns then into a competition of elite-unit-beating-a-lesser-elite-unit; tactics don't come near it as far as i can see, because match ups are unpredictable and distorted (units rout too early or not at all). This is why TW feels very boring past some point and why the opening stages are so nice and exciting. By keeping the game constantly into a late start-early mid game situation, the game can become enjoyable no matter the length. But i digress.

In GA mode contesting provinces (usually by neighbours) actually make for a different reasoning than domination i reckon. If the AI is the English - then there's little point for him to give peace to the player as the French - he's contesting provinces with him throughout the game. However i recall that factions that are not on a conquering streak and do not "care" for your "set" of GA home provinces will accept ceasefires quite quickly, generally speaking.

All in all - in MTW and STW the AI seems calculating more than reasonably his strategic position and also care for actual survival IMO - unlike in RTW (and M2 as well) that his attitude is really suicidal and generally less well tuned with the strategic situation in the game. Peace is definitely an option in MTW as far as i am concerned in the sense that it is possible to meaningfully achieve.

Many Thanks

Noir