Trusted allies in 1.02

[FactionHeir]

Something I find quite odd in the final 1.02 as opposed to the leaked 1.02 (note leaked refers to a 1.01 installation with full 1.02 data directory and incompatibilites removed) is the alliance behavior.

In 1.02L, trusted allies would refrain from attacking you almost always and declare war on people you declare war on and vice versa.

In 1.02O on the other hand, trusted allies behave just like normal allies from 1.01, where the trusted ally function was disabled - That is, they randomly attack you for no reason even when you are much more powerful and have perfect relations + very good global reputation (Very trustworthy in my test case)

According to the patch log, allies are supposed to attack more intelligently now. I kind of thought it worked just fine in 1.02L compared with now, which is just as bad as pre-patch.

Any thoughts?

[Shahed]

Not really, unless one can consider "Bad News" a thought.

[phonicsmonkey]

hi - might be a silly question, and I don't have 1.02 yet so I can't check myself, but have you checked that the trusted alliance triggers are populated with real values in the campaign_ai file?

In 1.1 we had to replace the value 99.999 with something between -1 and 1 in order to get them working, just wondering whether somehow between the leaked 1.02 and the official 1.02 the 99.999s were re-inserted...

[Whacker]

My findings are the same, the diplomatic 'AI'... isn't. My relative power right off the bat in my campaign was strong, and it grew quickly and stayed there. I had one long-time ally (Venice) backstab me at the dumbest time. I had just finished up a period of wiping out the Turks and Egyptians, and my borders and lands were all quiet and secure, and I was at war with noone. out of the blue, they send a full stack against one of my northern greek peninsula castles (the largest one, a fortress) which was heavily garrisoned. You can guess what happened to them. Kind of amusing, because I've maintained perfect relations with the Pope (I am the Byz), have a number of catholic allies, maintained a reliable standing, etc. Venice went from supreme power to whatever the next level down was and reputation to "Extremely unreliable". Hurf Durf.

[Foz]

I haven't really noticed this, but I also haven't had enough experience with trusted ally neighbors who could really backstab me in 1.2O yet. Perhaps what you're noticing is because of the slowing of growth and the corresponding better armies & recruitment the AI has? I think there are always overrides that allow the AI to attack an ally it trusts, it just may be that the betterment of AI recruitment has led it to meet those conditions more often now. The lines are getting a little blurred for me though: I started messing with king's purses just a bit before the O patch came out, so I'm having a little trouble judging how much recruitment improved from Leaked to Official, though comments seem to indicate that it did some.

Of course, they could also have simply modified the AI decision priorities in whatever file that is that I can never remember.

I do notice the power blocks that were cropping up in 1.2L are still around, though, so not everything has changed. As England I noticed France and Scotland formed a natural alliance, and Milan/Venice/Poland have since formed an anti-HRE coalition. It appears that HRE is falling into line with France, since it has emerged as my biggest competition. Consequently it looks like I'll be part of the Milan/Venice/Poland coalition, since the other European superpowers are choosing to beat up on them, and I already have common enemies with them. Plus, it's a perfect fit since France is up next for me, and HRE has pushed up to Antwerp already. I prefer to make friends with countries I don't intend to gobble up right away.

Have I mentioned yet how nice it is to have power blocks? I really appreciate that the AI can form meaningful alliances amongst itself now.

[Agent Smith]

In my current Russian campaign I experienced a similar problem.

I was Outstanding with the Danes and had an alliance since nearly the beginning of the game. They went to war with Poland and, out of the goodness of my heart ( ) I "gifted" them help in their war by saying I'd give them support. That coupled with the first assault gave me Perfect relations.

I just so happened to have a princess nearby a Danish general. My high charm princess nabbed him up quick. So, I trotted him along, but then ran into some Poles blocking my way with two small Danish stacks on the other side. What do they do? They attack me!!! This plunged us into war and I haven't been able to get a ceasefire since.

So far the diplomatic AI is just as poor as it used to be for me, although I did notice that when I steal generals with princesses I don't hurt relations when I travel him through their lands. Other than that, it's still poor.

[Kobal2fr]

Herm, you just stole his general. Doesn't he have a right to be mad at you ?

As for me, out of the 3 campaigns I've started since yesterday, I've still to witness any kind of backstab whatsoever - what little attacks have ended my way past a solid network of alliances and playing factions against each other were expected if not encouraged. *shrug* Randomness

[dopp]

Is anyone still running leaked v1.2? Maybe unpack the data files and check the following lines in the campaign AI file:

<trusted_ally_fs_threshold float="0.5"/>
<trusted_ally_target_fs_threshold float="0.5"/>
<trusted_ally_target_human_fs_threshold float="0.0"/>
<trusted_ally_gs_threshold float="-0.1"/>
<trusted_ally_target_gs_threshold float="-0.1"/>
<trusted_ally_enemy_auto_war bool="false"/>

These are the lines from the official v1.2 patch. I draw your attention to line four. Their reputation must be above this value (dubious or more, I think) for them to consider you a trusted ally. If not, they backstab you the first chance they get like the sneaky illegitimate offspring they really are.

In other words, it looks like both sides must be fairly trustworthy before the alliance will hold. Given the tendency for aggressive (ie successful) AIs to have really bad reputations, it's no wonder that alliances don't last.

EDIT: Note the last line controls auto-declaration of war when your trusted ally goes to war. This has been disabled. I would be very grateful if someone could check the leaked patch files and confirm my observations.

FURTHER EDIT: It could simply be a typo. The value was originally set to -1.0 (ie it didn't matter) in the explanation section of the file.

[RickooClan]

That is, they randomly attack you for no reason even when you are much more powerful and have perfect relations + very good global reputation (Very trustworthy in my test case)

Try to think in the opposite way, if you are the smaller force and one of your ally is getting bigger and bigger [and the gap will likely be even bigger as big country grow faster]. In order to win the game, you, as the smaller country, has to take action before it is too late.

I somehow think the AI of ally is pretty good in 1.02 now. The AI will ally with you to buy the time to build up their countries, and betray you while they think the time is right. I especially like the idea that while one country is getting bigger, their relations with ALL other countries will getting worse as it make perfect sense for the game objective - only one winner for the game.

As the main objective of the game is taking lands and with only one winner. Developing and making ally/friends is the path or tools only to fulfill that objective and surely at some stages the AI or the player have to betray their ally in order to take their land and win the game. (or stopping your ally win the game faster than you)

Imagine you are playing a board game like Risk with your friends. While one of them obviously getting into the stage of one super power, as a smaller country, surely you will team up with the others to defeat/stop him to grow (even he is your ally). If not, yes your country might keep alive as the super power might not attack you "first" as ally (if the objective is not taking ALL the land) but by the same time you have fewer and fewer chance to win. You simply get no point to help your ally winning the game and the same apply to the AI ally in m2tw.

The key of the game is to become the first to take 40 lands + their target province (for long campaign). Hence keeping your country alive or mantaining good ally means nothing to the AI/player if somebody is about to win the game.

ps. the story will be totally different if the glorious achievement rules applied (GA mode in MTW) and thats why lots of players want it back.

[FactionHeir]

Try to think in the opposite way, if you are the smaller force and one of your ally is getting bigger and bigger [and the gap will likely be even bigger as big country grow faster]. In order to win the game, you, as the smaller country, has to take action before it is too late.

I somehow think the AI of ally is pretty good in 1.02 now. The AI will ally with you to buy the time to build up their countries, and betray you while they think the time is right. I especially like the idea that while one country is getting bigger, their relations with ALL other countries will getting worse as it make perfect sense for the game objective - only one winner for the game.

As the main objective of the game is taking lands and with only one winner. Developing and making ally/friends is the path or tools only to fulfill that objective and surely at some stages the AI or the player have to betray their ally in order to take their land and win the game. (or stopping your ally win the game faster than you)

Imagine you are playing a board game like Risk with your friends. While one of them obviously getting into the stage of one super power, as a smaller country, surely you will team up with the others to defeat/stop him to grow (even he is your ally). If not, yes your country might keep alive as the super power might not attack you "first" as ally (if the objective is not taking ALL the land) but by the same time you have fewer and fewer chance to win. You simply get no point to help your ally winning the game and the same apply to the AI ally in m2tw.

The key of the game is to become the first to take 40 lands + their target province (for long campaign). Hence keeping your country alive or mantaining good ally means nothing to the AI/player if somebody is about to win the game.

ps. the story will be totally different if the glorious achievement rules applied (GA mode in MTW) and thats why lots of players want it back.

Realistically, at least as a human you would realize that you cannot win if your ally is close to achieving victory or already has done so. You'd want to stay on good relations with that ally just to make sure you won't be the next target and may just end up as the ally's best friend. Consider the real world as an example.

If the AI does want to attack its trusted ally, it ought to do so when it can actually win and not a suicidal attack just to make a point. Attacking a random allied general with odds of 1:1 in the ally's lands isn't really smart nor is besieging a town that is heavily garrisoned with a small force that may give you good pre-battle odds (i.e. 6 dismounted feudal knights vs 10 horse archers)

[Foz]

Realistically, at least as a human you would realize that you cannot win if your ally is close to achieving victory or already has done so. You'd want to stay on good relations with that ally just to make sure you won't be the next target and may just end up as the ally's best friend. Consider the real world as an example.

If the AI does want to attack its trusted ally, it ought to do so when it can actually win and not a suicidal attack just to make a point. Attacking a random allied general with odds of 1:1 in the ally's lands isn't really smart nor is besieging a town that is heavily garrisoned with a small force that may give you good pre-battle odds (i.e. 6 dismounted feudal knights vs 10 horse archers)

I disagree entirely. Second place is the first loser. The goal of the game is not to survive, nor to treat your faction like you would in real life. Surviving beyond someone else meeting victory conditions is entirely moot. The only goal of the game that matters, as Rickooclan pointed out, is the singular goal of meeting your victory conditions. Upon realizing that you are nearing a win, or that it can no longer realistically win, it makes perfect sense for a weak AI faction to launch every sort of attack possible against you in order to at least try to prevent you from winning, because if you do so it has immediately lost the game. Moreover, its seemingly futile attacks may weaken your perimeter defenses enough to allow a different AI faction to follow up on the attack and actually break through. This would further serve the weak faction's goals, since before it can consider trying to win itself, it must first make sure you do not win. Even in a situation where all the faction can hope to do is make it a little easier for a different faction to damage you, that choice is still clearly more in the AI's interest than simply sitting there allowing you to win.

[Agent Smith]

Herm, you just stole his general. Doesn't he have a right to be mad at you ?

Well, if it changed our relations I'd understand. It didn't affect our relations at all, even as I marched through their land back to a point where my boats were waiting to pick him up. I was still Perfect with the Danes when they attacked me. In fact they had the choice between attacking the Polish army, which was right next to me and whom we both were at war with, and my little general from an allied faction. They, of course, chose me.

[Tran]

I disagree entirely. Second place is the first loser. The goal of the game is not to survive, nor to treat your faction like you would in real life. Surviving beyond someone else meeting victory conditions is entirely moot. The only goal of the game that matters, as Rickooclan pointed out, is the singular goal of meeting your victory conditions. Upon realizing that you are nearing a win, or that it can no longer realistically win, it makes perfect sense for a weak AI faction to launch every sort of attack possible against you in order to at least try to prevent you from winning, because if you do so it has immediately lost the game. Moreover, its seemingly futile attacks may weaken your perimeter defenses enough to allow a different AI faction to follow up on the attack and actually break through. This would further serve the weak faction's goals, since before it can consider trying to win itself, it must first make sure you do not win. Even in a situation where all the faction can hope to do is make it a little easier for a different faction to damage you, that choice is still clearly more in the AI's interest than simply sitting there allowing you to win.

Are you saying that you better have an AI faction who behave non-sense and illogical rather than one that behave realistic?

I have to completely disagree with you.

[dopp]

No, he's saying that the AI should use a different logic altogether, where it plays to win the campaign with full knowledge of the victory conditions, rather than trying to 'role-play' it. I prefer the AI to roleplay as it increases the immersion immensely.

I can't find any hint that the AI 'plays to win' in its dealings with other players. It is programmed to expand and to conquer its special target provinces, but no more than that.

[Stuperman]

Many people here (myslef included) can achieve the victory conditions for any given faction within 75 turns (150 years), this is only made easier when they attack you at random. If this is the one and only goal ( as many people have said) then it makes for a pretty boring game if you beat it before the mongols/turmids(sp)/america. In fact if this was the only goal, you should work towards excommunication, and ignore diplomacy entirely.

The AI should use trusted ally to expand in the other direction, go against a common enemy or attck the largest threat to themselves. They should NOT sit around till your empire is the biggest threat, then give up years of peace cause the player triggered it.

[Tran]

Many people here (myslef included) can achieve the victory conditions for any given faction within 75 turns (150 years), this is only made easier when they attack you at random. If this is the one and only goal ( as many people have said) then it makes for a pretty boring game if you beat it before the mongols/turmids(sp)/america. In fact if this was the only goal, you should work towards excommunication, and ignore diplomacy entirely.

The AI should use trusted ally to expand in the other direction, go against a common enemy or attck the largest threat to themselves. They should NOT sit around till your empire is the biggest threat, then give up years of peace cause the player triggered it.

I agree

[Whacker]

The AI should use trusted ally to expand in the other direction, go against a common enemy or attck the largest threat to themselves. They should NOT sit around till your empire is the biggest threat, then give up years of peace cause the player triggered it.

As far as I can tell, this is the case. I had 3 specific alliances (more, but just these 3 are relevant) on my northern Byz front to stabilize it at the beginning of my current game. I ended up having to break with Poland because Hungary attacked about 20 turns in, and the Venitians backstabbed me about 40 turns in and attacked me. So far my alliance with Hungary is holding, but I think they'd drop it/backstab me in a heartbeat if they didn't already have their hands full. As of late, I've seen a few small stacks probing my borders, and I noticed that they've also overrun most of the territories north of me. I haven't seen or paid attention to elsewhere on the map, but it would appear that this statement by Stuperman is basically correct.

[Foz]

Are you saying that you better have an AI faction who behave non-sense and illogical rather than one that behave realistic?

I have to completely disagree with you.

The problem with your logic here is you are basing your concept of sense and logic on the real world, not the game one. The real world does not have goals like "conquer 50 provinces and capture Jerusalem." Therefore, the rules by which you play a game with that goal are entirely different than any sort of real-world concept of sense and logic. There is no future in the game once someone achieves victory, so basing gameplay on planning for that future (as FactionHeir suggested) is simply nonsensical. Victory is the only goal, for both player and AI, and diplomacy should be viewed as what it really is in the game: a tool to utilize other factions to further your chance at victory. I've found that the AI largely treats me diplomatically just like I treat it - always looking for an advantage, treacherous or not. As a result it feels far more human than if it simply played nice with me or acted like a real nation might.

No, he's saying that the AI should use a different logic altogether, where it plays to win the campaign with full knowledge of the victory conditions, rather than trying to 'role-play' it. I prefer the AI to roleplay as it increases the immersion immensely.

I can't find any hint that the AI 'plays to win' in its dealings with other players. It is programmed to expand and to conquer its special target provinces, but no more than that.

This leads me to a question. When you play the campaign, do you ignore the victory conditions? Do you actually ignore them and only do exactly what your empire would do in real life? I doubt anyone here can make that claim, and if they can, then they surely have never won a campaign in this game, and never will. Achieving victory in this game necessitates doing things that you simply would not do in the real world. Why then should the AI play the game under some false pretense at real world emulation, when the player himself does not do so? I would much rather have the AI play the game with the same understanding that I do: that it must do everything possible to win, realistic or not. Restricting the AI to some concept of "realistic" play only allows the player to exploit it even more than its strategic and tactical shortcomings already allow, and would surely not be in the best interest of having an exciting and challenging game.

[rookie7]

Something I find quite odd in the final 1.02 as opposed to the leaked 1.02 (note leaked refers to a 1.01 installation with full 1.02 data directory and incompatibilites removed) is the alliance behavior.

In 1.02L, trusted allies would refrain from attacking you almost always and declare war on people you declare war on and vice versa.

In 1.02O on the other hand, trusted allies behave just like normal allies from 1.01, where the trusted ally function was disabled - That is, they randomly attack you for no reason even when you are much more powerful and have perfect relations + very good global reputation (Very trustworthy in my test case)

According to the patch log, allies are supposed to attack more intelligently now. I kind of thought it worked just fine in 1.02L compared with now, which is just as bad as pre-patch.

Any thoughts?

I totally agree. In my opinion, I don't see any noticable difference between patch and pre-patch. If someone said AI will attack more intelligently in 1.2, personally I would not expect to see a 1 region faction attacking a strongest and most advanced faction. An intelligent thing to do is to go for a weaker one.

Case in point. Milan is down to only Ajaccio left, and we're at war. I made a ceasefire offer, and they turned it down. I clicked on "Make Offer" again and offered them a ceasefire again. This time they accepted. Two turns later, they sieged my full garrisoned city with a half stack, leaving only a general and a spear unit to guard their only remaining region. I obliterated their 1/2 stack unit, and send a 3-unit army (a general, a spear unit, and a heavy infantry) to take their remaining region. Their 1/2 stack can easily take one of the Sicilian's region.

[Agent Smith]

This leads me to a question. When you play the campaign, do you ignore the victory conditions? Do you actually ignore them and only do exactly what your empire would do in real life? I doubt anyone here can make that claim, and if they can, then they surely have never won a campaign in this game, and never will. Achieving victory in this game necessitates doing things that you simply would not do in the real world. Why then should the AI play the game under some false pretense at real world emulation, when the player himself does not do so? I would much rather have the AI play the game with the same understanding that I do: that it must do everything possible to win, realistic or not. Restricting the AI to some concept of "realistic" play only allows the player to exploit it even more than its strategic and tactical shortcomings already allow, and would surely not be in the best interest of having an exciting and challenging game.

I understand what you're getting at, which is that the AI should play to win like the player does.

However, there is a caveat (I hope I spelled that right). You can win the game just fine without being at war with everyone. There's more to keeping an alliance than just building slowly and pacing the game while ignoring the conditions.

A perfect example is the one I gave about the Danes. I had an alliance with them since turn 10. Between us was Poland, and we were both at war with them. To their south, they were at war with the HRE. PLENTY of room to expand and try to mee the victory conditions. Instead, they throw that away by attacking my lone general that would have been gone in a turn or two from their borders instead of attacking a Polish stack that was right in front of their nose.

Achieving victory in the game CAN be done diplomatically to some extent. For instance, as Russia, I would love to have my western border secure and concentrate on moving through Turkey and into the Middle East. That can fulfill my victory conditions, and I don't need to go to war with everyone to do it. The AI should be able to accomplish that, too.

[dopp]

This leads me to a question. When you play the campaign, do you ignore the victory conditions? Do you actually ignore them and only do exactly what your empire would do in real life? I doubt anyone here can make that claim, and if they can, then they surely have never won a campaign in this game, and never will. Achieving victory in this game necessitates doing things that you simply would not do in the real world. Why then should the AI play the game under some false pretense at real world emulation, when the player himself does not do so? I would much rather have the AI play the game with the same understanding that I do: that it must do everything possible to win, realistic or not. Restricting the AI to some concept of "realistic" play only allows the player to exploit it even more than its strategic and tactical shortcomings already allow, and would surely not be in the best interest of having an exciting and challenging game.

That's not very fair to those of us who do enjoy roleplay. Go check the PBeM section of these forums to see some examples of roleplay in action. I at least can make the claim that I have won 20 grand campaigns in M2TW by 'roleplay'.

There are plenty of games that emphasize a more aggressive play style. A major strength of Total War has been the freedom to choose differently and to perhaps roleplay a little and take in the eye candy. Hence the enduring popularity of glorious achievement campaigns, the demand for 'fluff' like titles, elaborate roleplay efforts by players on the forums, and a variety of mods that promise better immersion and more realism. Computer opponents that 'roleplay' (and thus behave in a 'realistic' fashion) diplomatically thus add to the immersion and would please a sizeable percentage of players who value this style of play.

In any case, I fail to see what the fuss is all about. The computer does NOT, to my knowledge, play specifically 'to win'. It is programmed to expand, of course, and conquer the player if he gets in the way. But it appears to me that the current fragility of alliances post-v1.2 is the result of the computer faction's 'untrustworthiness' being taken into account. I have not found any trigger in the AI file that says 'if the player is about to win, stop him'.

[crpcarrot]

i think the problem is balancing and programming the AI to recognise the role player and the player who going for the victory conditions? how do u achieve that?? my programming exp is very limited but i know how the logic should work, and it would need 2 different AI's is 2 different game modes.

i would love to see the second type of AI as i dont play TD i play to my own goals which may or may not include the victory conditions.

i too wish we had the GA mode again

does the game end when an AI faction reaches the victory conditions?

i dont think foz is trying to be fair hes just stating how things are. having a game with victory conditions that the AI doesnt attempt to meet is pretty lame. wheres the challenge in that?

[Sheogorath]

Ive had this problem in my Venice campaign, namely with the Portugese and Hungarians. It seems their AI's are really into the 'Blockade one port with one ship for one turn then run away and dont do anything for the next decade or so.'
The odd bit is that they were both my best allies. I'd been allied to Hungary for nearly a hundred turns, and the only person I had declared war on was Milan (With just cause to, the bastards were using my Venice province as an army staging ground >_>) via Crusade.
Meanwhile, my vaults are full of a million or so florins and I'm wiping the floor with the Byzantines and Sicilians.
Suddenly, the Portugese come out of NOWHERE and blockade Genoa for one turn. Then the Hungarians blockaded Constantinople for one turn.
Ten turns later, a Hungarian stack shows up at Constantinople and gets its ass kicked.
I signed peace treaties with both of them (Very good terms for me and still 'Very Generous') and two turns later, theyre back at it!

About as logical as a tree full of monkeys on LSD.

[Foz]

i dont think foz is trying to be fair hes just stating how things are. having a game with victory conditions that the AI doesnt attempt to meet is pretty lame. wheres the challenge in that?

Yes, exactly. At the very least it should try hard to block me from meeting victory, as it cannot do so if another faction (currently bound to be the player) does it first.

I totally agree. In my opinion, I don't see any noticable difference between patch and pre-patch. If someone said AI will attack more intelligently in 1.2, personally I would not expect to see a 1 region faction attacking a strongest and most advanced faction. An intelligent thing to do is to go for a weaker one.

Case in point. Milan is down to only Ajaccio left, and we're at war. I made a ceasefire offer, and they turned it down. I clicked on "Make Offer" again and offered them a ceasefire again. This time they accepted. Two turns later, they sieged my full garrisoned city with a half stack, leaving only a general and a spear unit to guard their only remaining region. I obliterated their 1/2 stack unit, and send a 3-unit army (a general, a spear unit, and a heavy infantry) to take their remaining region. Their 1/2 stack can easily take one of the Sicilian's region.

Why would you not expect to see that? What you are saying is rather like saying "in world war 2, France and Great Britain shouldn't have attacked Germany, they should have attacked each other b/c they were weaker." It's simply not a valid viewpoint. The biggest nation should always be the priority target, because it is the biggest threat to the others. Even a 1-region faction should still attack the juggernaut, realizing that only by ganging up on it do the smaller factions stand any chance. To fight amongst the weaker factions is only to doom everyone, as they clearly have no chance alone against that huge faction, and any side fighting diverts vital resources from trying to win the fight against it. All attacking a smaller faction accomplishes is assuring that both of you will lack the power to repel the larger faction when it attacks, which is not even remotely intelligent.

I don't dispute that the AI's methods of implementing this strategy are often less than stellar, and it certainly plans its attacks poorly sometimes, I'm just saying I don't think you can fault the overall strategy behind it - it's far more clever than it looks upon initial examination.

That's not very fair to those of us who do enjoy roleplay. Go check the PBeM section of these forums to see some examples of roleplay in action. I at least can make the claim that I have won 20 grand campaigns in M2TW by 'roleplay'.

I feel the need to draw an important distinction. What people are trying to enforce on the AI here is in many cases not the same roleplay that they employ themselves - it is realism. People have repeatedly asked for the AI to behave like real countries. That is where my issue lies: the lopsided requirement many players are voicing. They want realism from the AI that would have it ignoring victory conditions and good gameplay strategy entirely, when they themselves do not play that way. I am only advocating that the AI should play back at you with exactly the same style you do yourself. Most players here seem to roleplay some but in the end are still playing to win the campaign - essentially some realism, but heavily tempered by the understanding that it is a game which must be won. Given the need to do so, they will often break the realism in order to accomplish the game's objectives. They don't simply try to maintain peace with everyone around them for indefinite periods of time. I'm just suggesting that if you can break realism to suit your own needs, then the AI must be able to do the same, or you'll have ruined the AI's ability to play as well as you do (not that it can anyway, I'm just saying it's absolutely impossible for it to compete if you impose more strict limits on it than the ones you play under).

In any case, I fail to see what the fuss is all about. The computer does NOT, to my knowledge, play specifically 'to win'. It is programmed to expand, of course, and conquer the player if he gets in the way. But it appears to me that the current fragility of alliances post-v1.2 is the result of the computer faction's 'untrustworthiness' being taken into account. I have not found any trigger in the AI file that says 'if the player is about to win, stop him'.

Actually, that code is in there. There are various measures of your success, and the game is constantly using them to make the AI hate you if you are successful (see the tall poppy code, for instance). The game mechanics essentially ensure that you can't hold alliances with the AI if you are doing really well, which in turn is of course meant to thrust you into war with everyone. Generally it's effective.

So while the AI itself doesn't necessarily understand the concept of winning, the mechanics that control the AI implement the key points of the concept, at least far enough to dispose the AI against whichever faction is in the lead (player-controlled or not).

[dopp]

I feel the need to draw an important distinction. What people are trying to enforce on the AI here is in many cases not the same roleplay that they employ themselves - it is realism. People have repeatedly asked for the AI to behave like real countries. That is where my issue lies: the lopsided requirement many players are voicing. They want realism from the AI that would have it ignoring victory conditions and good gameplay strategy entirely, when they themselves do not play that way. I am only advocating that the AI should play back at you with exactly the same style you do yourself. Most players here seem to roleplay some but in the end are still playing to win the campaign - essentially some realism, but heavily tempered by the understanding that it is a game which must be won. Given the need to do so, they will often break the realism in order to accomplish the game's objectives. They don't simply try to maintain peace with everyone around them for indefinite periods of time. I'm just suggesting that if you can break realism to suit your own needs, then the AI must be able to do the same, or you'll have ruined the AI's ability to play as well as you do (not that it can anyway, I'm just saying it's absolutely impossible for it to compete if you impose more strict limits on it than the ones you play under).

Yes, and some games follow your reasoning. Rise of Nations, for example, will cancel all alliances and force the computer to attack you if you are winning the game. But the game is for the benefit of the player, not the computer, and there is no problem with wanting an AI that behaves realistically, even if the player does not on occasion.

Actually, that code is in there. There are various measures of your success, and the game is constantly using them to make the AI hate you if you are successful (see the tall poppy code, for instance). The game mechanics essentially ensure that you can't hold alliances with the AI if you are doing really well, which in turn is of course meant to thrust you into war with everyone. Generally it's effective.

So while the AI itself doesn't necessarily understand the concept of winning, the mechanics that control the AI implement the key points of the concept, at least far enough to dispose the AI against whichever faction is in the lead (player-controlled or not).

Tall poppy code simply makes the smaller factions hate the larger ones. I can be the most powerful faction early game with a mere 10 territories and the computer will hate my guts just the same as if I was just 1 territory away from victory. It does not, for example, check victory conditions and then force an attack if someone is about to win, which is what some were speculating was happening in v1.2. It's harder to keep allies, perhaps, but not impossible, just give them a little more bribe money. It is not even very problematic, in fact, because the tall poppy penalty is only half that of the basic relations drain you get every turn from simply playing on very hard difficulty.

[Julius_Nepos]

I can't say I've really ever seen a case of the AI attacking "intelligently." Oh they do attack, and they attack often but there's no reasoning behind it that I can see other than an attempt to defeat the player at any cost, including the obliteration of their faction. I can, on my end, understanding the point that the smaller factions will hate and attack the player once they get very large. You'd think however that those small factions could at least strengthen their position against other weaker factions before attempting to attack the player in a hopeless show of force.

In my 1.2 (official) French campaign I've obliterated the Scots, Spaniards and English, I control all lands from the Iberian Peninsula to a long frontier line running from Antwerp to Marseilles (Genoa has been annexed but there are enemy territories in between). For a long time I did manage to hold alliances with many factions, it seems however crossing the magic 15 province barrier caused everyone and their brother to hate me. I had a war against Portugal decades ago and they still hate me, and they still wont accept a ceasefire. They're down to one territory and their odds of winning are zero. The only reason they're not destroyed is they've only got Marrakesh and I have no interest in invading Africa and opening up a new border with Sicily and the Almohades.

I personally miss glorious achievements and I don't see anything to indicate that the AI plays to win or could ever be programmed to play to win. I like to role play my factions as much as anyone else and I have more interest in internal development than capturing 45 territories and Jerusalem. My attacks are usually dictated by ecclesiastical policy, and I find even when I'm content to leave an AI faction in possession of a single province they'll never accept vassalage and they'll do all they can to attack later on (hopelessly of course). As with RTW I think vassalage should be removed from M2TW as it is a useless (and effectively AI only) option. I've read some tutorials on how to get a vassal but it seems the hoops you have to jump through are such that the vassal will then be powerful enough to attack you anyway.

I will say this though the AI in 1.2 has thrown me a curve ball in at least one respect. For some utterly bizzare reason the Scots attacked me long before I'd even conquered England. I had no border and no diplomatic contact with them yet they were landing full stacks at Bruges. They didn't survive as I outlined above but the invasion was very surprising. Anyway in conclusion I see no difference between 1.1 and 1.2 regarding alliances, they're in general useless and serve no real purpose that I can see, except amongst AI factions. Anyway cheers!

[FactionHeir]

Yes, I noticed that certain factions have favorite landing spots and tend to want to take certain regions by sea landings regardless whether it belongs to an ally or not.
Examples are Spain/Portugal's love for Caernarvon and Caen. England's love for Antwerp sea landings, Denmark for Riga and York, Venice and Byzantium for Rhodes, Byzantium for Iraklion, Moors for Cagliari (or was it Ajaccio? The southern one of the two) Milan for the northern one of the two (Ajaccio?), Sicily for Tunis, Tripoli, Durazzo and Ragusa, and Papal states for Valencia.

[Tran]

This probably has been posted somewhere but I couldn't find it, so I'lll ask:

How do you revert the AI behaviour to act like in leaked 1.02 patch instead of the (current) official 1.02 patch? What files do I need to edit and how?

Also, I agree with FactionHeir that some factions tend to attack certain territory regardless of the owner. Is there anyway to edit this to make those factions behave "more reasonable"? (e.g: check who's the territory owner first)

[FactionHeir]

Afraid not. Even with them behaving like in 1.02L, they will end up blockading the port of their target regions eventually.

But as to how to fix it....I think in 1.02L it looked more like this:

//////////////////////
	// GLOBAL AI PARAMS //
	//////////////////////
	-->

	<trusted_ally_fs_threshold float="0.5"/>
	<trusted_ally_target_fs_threshold float="0.5"/>
	<trusted_ally_target_human_fs_threshold float="0.0"/>
	<trusted_ally_gs_threshold float="0.1"/>
	<trusted_ally_target_gs_threshold float="-1.0"/>
	<trusted_ally_enemy_auto_war bool="true"/>

In the descr_campaign_ai_db.xml
Vaguely like that anyway. Don't have the leaked copy of that files anymore, but you could redownload the leaked version and get it from there.

[Tran]

About the last line:

<trusted_ally_enemy_auto_war bool="true"/>

Does it mean the allies will automatically declare war against any faction who declared war on us? Or will they just "break alliance" with that (our enemy) faction?

About the port blockade: I supposed it's a glitch, no?