Glad to hear the Diplomacy A.I is being improved a bit in shifters but I'm worried that's not gonna be enough.



My biggest gripe however is the A.I.'s tendency to drastically under perform late game, is this being addressed? As this is probably the biggest killjoy of playing against the A.I, building a massive empire only to find the A.I is no longer any threat at all.



This also ties in with another big problem I've noticed to do with late game which is the fact A.I makes no acknowledgement of a player getting close to a victory condition. This seems odd since in player based games people will obviously become worried when another player gets close to victory so will likely change there tactics, most likely in the form of attacking said player. The A.I. however do not do this and carry on there merry way even if your a few turns from winning.



Surely the A.I should be programmed to become very hostile to the player or in fact any other faction that come close to a victory condition? I was expecting armies to come baring down upon me when I was close to a science victory yet the A.I. continued thinking me as peaceful and had there armies just casually wonder inside there own territory without me having to sweat at all.



Are these issues being addressed in shifter? If not I strongly advice they are managed as resolving these issues would make single player games far more enjoyable.



(Also worth noting I usually play in the top 2 difficulties so it's not a problem of having the difficulty set to low.)

someguy216 wrote:
Glad to hear the Diplomacy A.I is being improved a bit in shifters but I'm worried that's not gonna be enough.

Bear in mind that while work on the AI has continued it's not the main focus of the update - where the free stuff is concerned we've mostly been looking at faction balance and winter, plus fixing bugs and improving the GUI.

someguy216 wrote:
My biggest gripe however is the A.I.'s tendency to drastically under perform late game, is this being addressed? As this is probably the biggest killjoy of playing against the A.I, building a massive empire only to find the A.I is no longer any threat at all.

We'll need some more information in order to look into this: in what way do you feel the AI is under-performing? Poor technology choices? Poor production choices? Or maybe it is not handling diplomacy correctly? If, for instance, the lag is in the score screen then maybe it's a matter of how the score is calculated and not in the AI's actual performance.

someguy216 wrote:
Surely the A.I should be programmed to become very hostile to the player or in fact any other faction that come close to a victory condition? I was expecting armies to come baring down upon me when I was close to a science victory yet the A.I. continued thinking me as peaceful and had there armies just casually wonder inside there own territory without me having to sweat at all.

The concern is that it would feel cheesy if all AI opponents immediately went berserk when the player approached victory. Some early 4X games did this and it could be very frustrating. We've chosen to make the AI more role-play than competitive - that is it plays like the ruler of a fantasy kingdom and not like a player controlling the ruler of a fantasy kingdom.

wilbefast wrote:
... Some early 4X games did this and it could be very frustrating...

Some early 4x games like Civ V and Civ BE. :-D

steelstiletto wrote:
Some early 4x games like Civ V and Civ BE. :-D

neither of those two games have AI that go berserk when you approach win conditions.

Yes, they will try and stop you and your relations decrease but they don't throw everything they got at you.

wilbefast wrote:
... plus fixing bugs and improving the GUI.

Can you tell more about improving the GUI, pls?

wilbefast wrote:




The concern is that it would feel cheesy if all AI opponents immediately went berserk when the player approached victory. Some early 4X games did this and it could be very frustrating. We've chosen to make the AI more role-play than competitive - that is it plays like the ruler of a fantasy kingdom and not like a player controlling the ruler of a fantasy kingdom.

The problem is you just spend 50 turns just playing out the victory condition.It is very boring end to the game.I think it is still roleplaying that the races would try to stop the dominate power at the end.

Yeah. It is like written here. If you don't decide to play aggressive early game, the AI won't start to harm you (only minors do if set on "hard" difficulty).So regularly you can grow succesfully, until you are strong enough to kill one AI faction after the other...



BUT I have to admit AI is stronger in general then before last patch. But more aggressivity against human players would be nice.



P.S. About diplomacy AI I won't talk about. I don't use diplomacy with any faction

wilbefast wrote:
Bear in mind that while work on the AI has continued it's not the main focus of the update - where the free stuff is concerned we've mostly been looking at faction balance and winter, plus fixing bugs and improving the GUI.

I understand but that's why I was a little worried about current A.I problems not being fixed. As much as I'm excited to see a new faction and new mechanics I would hate to see big problems overlooked.

We'll need some more information in order to look into this: in what way do you feel the AI is under-performing? Poor technology choices? Poor production choices? Or maybe it is not handling diplomacy correctly? If, for instance, the lag is in the score screen then maybe it's a matter of how the score is calculated and not in the AI's actual performance.

Perhaps the word 'drastic' is a bit dramatic but there's still certainly a problem. I've noticed in part it is to do with the score screen which isn't wholly accurate as I stated in the previous post about the global score mechanic. But otherwise I think its a small bit of everything I'm afraid. A.I rarely make use of anything better then rank 3 iron. I can't really tell if they go for inappropriate or appropriate technologies as I can't see all of them and I understand that its hard to program the A.I. to make use of more complex/less generalisable techs (like the dust transmitter the one time build dust building which adds +10 dust to all tiles with dust.)



But I think diplomacy is a biggy, if they managed to wage war more often against appropriate factions (i.e. close factions that are weaker in score) they could grow more quickly near the end game by taking over an empire's cities and thus consuming them into one big empire. (Which is often what the player does to get ahead, the A.I. however are very slow in there wars so never seem to completley conquer another faction.)

The concern is that it would feel cheesy if all AI opponents immediately went berserk when the player approached victory. Some early 4X games did this and it could be very frustrating. We've chosen to make the AI more role-play than competitive - that is it plays like the ruler of a fantasy kingdom and not like a player controlling the ruler of a fantasy kingdom.

It's not cheesy at all as long as you don't make them LITERALLY go all out to get you. It wouldn't make sense if a trusting ally or very weak faction charged at your gates. But everyone else should within reason. I know it wouldn't be easy in this game to have them ally and gang up on you due to alliances being tied to a specific research, but if they all attacked you then that would be the equivalent of the alliance anyway. They'd just be enemies fighting a common enemy "An enemy of my enemy is my friend." As far as roleplay and immersion are concerned it still makes perfect sense that they'd be concerned of a neighbour becoming to powerful as this neighbour will likely come and destroy and enslave there kingdom if they don't do something about it.



It's quite a lot to deal with so I don't expect it all to be fixed within a day but it would be nice to see some progress in these areas gradually at least one by one. Also want to point out otherwise the A.I. is good for a strategy game but unfortunately still has some holes.

wilbefast wrote:
The concern is that it would feel cheesy if all AI opponents immediately went berserk when the player approached victory. Some early 4X games did this and it could be very frustrating. We've chosen to make the AI more role-play than competitive - that is it plays like the ruler of a fantasy kingdom and not like a player controlling the ruler of a fantasy kingdom.

Why would it be less frustrating that the AI lets you casually achieve a victory it should see coming a mile away?



There are two issues with that POV from my perspective.



1) Prior to Shifters (I'm not in the beta so I can only go by what I currently have), the AI was very unimpressive in the tactical side of waging war. Even with them having superior numbers and a superior army, you could regularly beat the AI in the field with some decent tactics. It makes waging war against the AI much less costly and much less of a risk if you're not losing troops and easily taking their cities.



2) It is very anti-climatic for a game to wind down in this way. In AoW 3, when you get close to charging the sacred seals which let you take control of all magic in the realm, the other empires break peace or outright declare war because why would you want an empire you're not allied or close friends with to dominate you from a role-playing perspective? Do all the AI's get a secret "submissive" trait I'm not aware of?

ss7877 wrote:
Can you tell more about improving the GUI, pls?

Hmm... off the top of my head we now have tooltips on all the status effects in the tactical battles, you can detach generals from armies, there are details on the game settings in the pause screen and army attributes are more clearly displayed in their army panel - and there have been many many tooltip/localisation fixes.

Ashbery76 wrote:
The problem is you just spend 50 turns just playing out the victory condition.It is very boring end to the game.I think it is still roleplaying that the races would try to stop the dominate power at the end.

The thing about AoW is that the end-game is referred to heavily in the lore - not so in Legend: the AI doesn't know about victory conditions or the approach of the end of the game, because it doesn't really "know" that it is in a game. It is the leader of its people. As such its goals are to ensure their survival and to prosper, not to prevent other empire from reaching a score threshold. Score is simply used as a heuristic for evaluating the economic development of an empire.

This is an issue though, gameplay-wise. To solve it though we'll need to do something a little more complex that just making the AIs hate your guts when the game is about to end. Note that the AI (by design) never knows which empire is a human and which is an AI, as an AI could be replaced by a human at any time in multiplayer (and vice-versa).

someguy216 wrote:
A.I rarely make use of anything better then rank 3 iron. I can't really tell if they go for inappropriate or appropriate technologies as I can't see all of them and I understand that its hard to program the A.I. to make use of more complex/less generalisable techs (like the dust transmitter the one time build dust building which adds +10 dust to all tiles with dust.)

Retrofitting is a tricky one. We've put a huge amount of work into it and are still not wholly satisfied. Trouble is you need dust and strategic resources to retrofit: it's very tricky to figure out exactly how much of these resources to spend on retrofits versus buyouts, heroes, buildings, market resources, pillars and mercenaries. Raising the priority of one of these tends to break the others.

Post Forges Of Creation I'm pretty satisfied with (and proud of) the technology and building choices though. And the AI is more intelligent about how it creates unit design and which units it retrofits (ie. veterans). Better doesn't mean ideal of course

someguy216 wrote:
But I think diplomacy is a biggy, if they managed to wage war more often against appropriate factions (i.e. close factions that are weaker in score) they could grow more quickly near the end game by taking over an empire's cities and thus consuming them into one big empire. (Which is often what the player does to get ahead, the A.I. however are very slow in there wars so never seem to completley conquer another faction.)

War requires a lot of things to line up - the big weakness of our AI's decentralised decision-making architecture is that it never focuses on one task to the exclusion of everything else. This is also its big strength, but the trouble is that to wage war effectively you tend to need to be pretty single-minded up it. I have some ideas for making the AI more aggressive late-game but I'm not sure what can be done about it better and coordinating war efforts - it's very much a structural issue.

someguy216 wrote:
It's quite a lot to deal with so I don't expect it all to be fixed within a day but it would be nice to see some progress in these areas gradually at least one by one. Also want to point out otherwise the A.I. is good for a strategy game but unfortunately still has some holes.

I really wanted to focus on the battle targeting for the release of Shifters, but you're right that the end-game AI could use some love. We'll discuss what can be done about this issue moving forward.

Glad to hear that battle targeting is being worked on.

Just fought some battles as WW against AM, where my rangers would actually move forward on the battlefield in order to attack a warlock way in the back, while ignoring other enemy units which were 2 hexes in front of them ...

icarus86 wrote:
Glad to hear that battle targeting is being worked on.

Just fought some battles as WW against AM, where my rangers would actually move forward on the battlefield in order to attack a warlock way in the back, while ignoring other enemy units which were 2 hexes in front of them ...

The update should fix this. We went through a bunch of test-cases with natev on the VIP forums, trying to figure out why the AI was making certain decisions. Overall it should be a lot cleverer.



Just a heads up though: we're still tracking an issue with distance calculations in tactical battles (terrain movement penalties aren't taken into account correctly), but we're waiting for a programmer to be free so they can fix it. Unfortunately it'll mean I need to do another pass on all the targeting weights afterwards (because the results from distance calculations are likely to change) so it's not going to be ready for Shifters (which comes out tomorrow!) Hopefully we can get it in a hot-fix.



Also if you ever see any weird behaviour in the tactical battles please-please-please-please send me your save so I can have a look and tweak the targeting weights

wilbefast wrote:
The update should fix this. We went through a bunch of test-cases with natev on the VIP forums, trying to figure out why the AI was making certain decisions. Overall it should be a lot cleverer.



Just a heads up though: we're still tracking an issue with distance calculations in tactical battles (terrain movement penalties aren't taken into account correctly), but we're waiting for a programmer to be free so they can fix it. Unfortunately it'll mean I need to do another pass on all the targeting weights afterwards (because the results from distance calculations are likely to change) so it's not going to be ready for Shifters (which comes out tomorrow!) Hopefully we can get it in a hot-fix.



Also if you ever see any weird behaviour in the tactical battles please-please-please-please send me your save so I can have a look and tweak the targeting weights

Sounds great, THANK YOU !

Will definitely send savegames in case I notice any weirdness after Shifters.

wilbefast wrote:
The update should fix this.

Already bought. Thanks and grats. But still waiting compact city banners ...



P.S. Yes, I'm persistent