A few unconnected thoughts about AI behavior:

• AI often send out settlers unprotected, making them especially easy to pick off when they're in neighboring regions.
• AI aren't very good at keeping troops where they're most needed. I've seen AI cluster significant chunks of their armies on remote island regions when they're at war with a neighbor. Of course, it's hard to say where troops are most needed.
• It would be nice to see allied AI propose agreements more often.
• Weaker AI will continue to pick on one another even when a challenger (i.e. you) poses a greater threat to each of them.
• If I intercept AI units en route and defeat them, they often still attack me with their remaining units the next turn and lead them to a terrible defeat. It seems as though they don't change moves that they set up earlier after you've engaged them in battle.
• AI seem to destroy minor faction villages indiscriminately rather than bribing them or doing quests.
• Privateers make it really easy to mess with allies as they seem to not protect cities that are near their allies.
• I never see AI use privateers themselves. They could use them to great effect. Because they don't use privateers, the Roving Clans just wind up wandering aimlessly through the territory of players they're at cold war with.
  
  

Not that I only have complaints. I was impressed recently when, after I purchased a lot of titanium from the market, two AI proposed deals offering me a lot of titanium. It was as if they learned what resources I wanted and used my need as an opening for diplomacy.

The diplomacy in EL might be limited but one good thing about it compared to Civ is that you don't have the feeling that AIs are a single blob out to get you. In EL all AIs fight the same, be it with you or with each other, making it a more believable and less frustrating system. In Civ you feel like the AIs get out of their way to get you, and they definitely treat each other much differently that they treat you. This kills the suspension of disbelief and make Civ less immersive than EL. I am not sure if this was by design, or if the AI is just too simple to have specific instructions for dealing with the players compared to each other, but it is definitely something I would like to see stay the same.

Building towards alliances is actually fairly beneficial. The trade benefits can be substantial

For trade you don't need alliances, just peace is enough.

It should be easier to see the benefits of having trade routes to OTHER empires (vs. trade routes within your empire

Trade to other empires is ALWAYS more beneficial. The single factor (afaik) in profitability of a trade route is the distance between original and destination city, and the other empires will always be farther from you than your own. But I agree that having a system similar to Civ where you can see not only the current trade routes, but pottential trade routes would be beneficial. It would help you make decisions about which empire to make peace with first and where to build ports. It is mainly an interface problem.

The Diplomatic Victory condition is poorly implemented. You just spend lots of Influence? Without any particular diplomatic goal? This forces to play very erratically, diplomatically - alternatively initiating cold wars, negotiating for peace, forming alliances, then back-stabbing, etc.

I agree with the first part, diplomacy could be more interesting. But I disagree with your second part, you don't have to be bipolar in your dealings to win diplomacy, just keep dumping points into compliments.



The biggest flaw of diplomacy right now is that there are no minimal turns when doing peace or alliance. You can spend a fortune making peace with an AI then it will cancel the agreement the turn after, thus aiming at being a "peaceful empire" with a lot of influence generation to be able to make peace with everyone and not fight is not possible, the AIs will just cancel on you right away. Only the Drakken can do it because they can force peace without spending resources or dust, so it burns down to simply having enough influence to keep spamming force peace every turn.



Also a diplomacy victory is too easy at this moment if you understand how it works. I won my 3rd game ever on EL on endless difficulty with the draken, you can win fairly easy about the same time AIs get close to a scientific victory on the first try, and with refinement you can win much much sooner. I would have suggested to reduce the impact of spending influence points and greatly enhance the impact of keeping alliances and peace going for a long time, but before that becomes possible they need to stablilise the AI so it doesn't cancel the turn after you make an alliance or peace, or at least give that agreement a minimal length of 10 turns or so on normal.

DrakenKin wrote:
I agree with the first part, diplomacy could be more interesting. But I disagree with your second part, you don't have to be bipolar in your dealings to win diplomacy, just keep dumping points into compliments.



The biggest flaw of diplomacy right now is that there are no minimal turns when doing peace or alliance. You can spend a fortune making peace with an AI then it will cancel the agreement the turn after, thus aiming at being a "peaceful empire" with a lot of influence generation to be able to make peace with everyone and not fight is not possible, the AIs will just cancel on you right away. Only the Drakken can do it because they can force peace without spending resources or dust, so it burns down to simply having enough influence to keep spamming force peace every turn.



Also a diplomacy victory is too easy at this moment if you understand how it works. I won my 3rd game ever on EL on endless difficulty with the draken, you can win fairly easy about the same time AIs get close to a scientific victory on the first try, and with refinement you can win much much sooner. I would have suggested to reduce the impact of spending influence points and greatly enhance the impact of keeping alliances and peace going for a long time, but before that becomes possible they need to stablilise the AI so it doesn't cancel the turn after you make an alliance or peace, or at least give that agreement a minimal length of 10 turns or so on normal.

It seems to me that it's just easier (once you have enough Influence accumulated, which isn't too hard) to choose those diplomatic actions that cost more. Fewer clicks, right? Vs. repeated compliments, turn after turn.



I like the idea of minimal turns for peace treaties. I was even surprised that Truces last 1 turn? Although I guess the punishment for declaring war immediately after a truce has transitioned into cold war is that the Influence cost is high.



Yes, I agree with you about increasing the importance of maintaining peace and alliances for long periods of time. Just spending Influence to win Diplomatic is uninspired design.

Each time you use compliment, the cost for it increases (counter is unique per AI). Meaning even if it starts very low, you get get it to the 100s (same cost as war / peace / ally) fairly quickly. It is annoying to have to do so, but for immersion purposes I prefer it to declaring war / peace every turn.



Influence penalty for declaring war after truce becomes fairly minimal as time goes. I don't remember exactly how much influence per turn I was making for my diplo win on endless, but it was in the high thousands, and the AIs had the same too. So a counter would be more balanced than an increased influence cost.

The AI major/minor factions should not be able to distinguish between the Human player and other AI. I all too often watch as minor AI armies wonder around and ignore the major AI faction armies, only to gang up on me. Same goes for Major AI factions. When we're in a cold war state, if I am in their territory, they attack me, but not each other.

Thank you for all your answers. With the programmers and the game designers, we have read everything and we have worked hard – and still working! – to push the AI in the good way. For now, good improvements have already been done on the military aspect and so on. Economy is still in progress and diplomacy will follow soon. I hope you’ll get these updates soon!

I filled in your forms, but I don't think I answered them very well I would like to elaborate:



1) An AI faction needs to have a battle goal for each battle. Likely goals are:



a) Minimize losses

b) Cause attrition

c) Eliminate the enemy army



2) Enemy AI also needs to estimate the likelihood of success before each battle, and then, at the end of the battle, evaluate that estimation. Clearly, if a battle is unlikely to be a success, the AI should not attempt to engage, and might attempt retreat; right now, it does a lot of suicidal charges. There is already the beginning of an estimation, in the form of the "battle wheel" to give the player an idea of relative strengths. What about evaluation? This is important because players can play so much better than the AI can, and the AI needs to learn from its mistakes. Each time it loses a battle it should have lost, that should change its future estimation of battles. You can do this in a complicated way, but I think doing it simply would do wonders. Doing it simply would be just applying a multiplier to enemy forces based on the number of unexpected losses that AI has experienced against that player.



3) There should be a simple change to the way orders are evaluated in the absence of path to the target. Right now, it leads to exploitable situations, sometimes even accidentally. If no path exists to the target, then run the path again, this time, ignoring blocking enemies. Follow the path until you reach a blocking enemy, then attack that enemy.



4) Target priority is mostly fine (at least it would be, with the addition of suggestion 3 above). However, to improve, the AI needs to think less in terms of target priority, more in terms of position and initiative. Let me explain in more depth:



An AI should divide its forces into initiative groups. That is, a group of all units that can act first, before some units but after others, or after everyone else. Usually, there's only two groups, but sometimes there's more.



Groups should, in general, split their targets against targets with lower initiative, but combine targets against targets with higher initiative.



Lower initiative units should think in terms of position, rather than targets, because they have little way of judging the position of their targets. These units should seek to maximize the number of adjacent friendlies and position on advantageous ground (in the simplest iteration of improved AI, at least).



Even for higher initiative units, targets should be judged not just on their characteristics, but on the tile the attacking unit will end up on.





5) Let me give you a sample target priority system. The numbers can vary, but notice the things I'm taking into account:



a) All assignments are made in order of descending initiative.



b) Start with target.attack*target.currentLife.



b) Apply a multiplier (let's say 2) for each of the following factors:



i) Target has lower initiative and a better main attack than counterattack

ii) Target has lower initiative period

iii) Attacker will be on ground superior to target

iii) Position attacker will probably take is bordered by at least two friendlies (including anticipated friendlies from units already assigned)



c) Apply a divisor (again, let's just say 2) for:



i) Target will be on ground superior to attacker

ii) Attacker will be on position bordered by more than three enemies

iii) For each ranged enemy unit that will be able to fire on this position



d) If target priority is


i) Cannot be occupied by enemy unit with higher initiative during this turn

ii) Superior terrain to adjacent tiles not occupied by friendlies

iii) Proximity to 2 or fewer enemies

iv) Adjacency to 2 or more friendlies

v) Is not in range of enemy ranged units



e) Make a note of anticipated position and anticipated damage to target. This anticipated outcome will drive the assignments for lower initiative units this phase, giving them anticipated bordering friendlies and anticipated enemy life.



This should be a pretty simple flowchart to play with that would improve battle AI considerably. There are a number of ways to make it more interesting-- for instance, a goal of attrition would weigh target life more heavily; a goal of defense would weigh friendly and enemy adjacency more heavily.

This is good to see. I've just completed a few games on Serious and noted how the AI doesn't seem up to snuff and posted as such in another thread (@ /#/endless-legend/forum/5-general/thread/1986-new-player-experience-questions-about-the-ai). All forms completed.

Related to Diplomacy - Economical deals: The AI hardly ever seems to research a lot of the techs needed for trading, making the related deals somewhat pointless (and yet they tend to be against them, so ironically I usually throw in one of those techs to sweeten the proposal anyway).



This also somewhat hurts the whole Roving Clan marketplace tax bonus: I've never seen them research Imperial Coinage, they just buy heroes and mercenaries (and never sell units). Even after I gifted them Imperial Coinage they only sold single unit luxury resources here and there and not much else.

Whats your opinion on the deal below?









In my opinion this is a complete exploit. Two teir 1 crappy techs for over 8,000 dust.. ridiculous.



The game needs to put a cap on what each tech is worth - dependant on the tech value in terms of its cost in science beakers. It should not be merely dependant on the factions desire for the tech or else these sorts of problems occur, where the ai unfairly pays out exorbitant amounts.

Nasarog wrote:
They have a question that I think is a warning that they see what you're doing, but it needs to be more stern than just that. When I do that, the next turn they usually close their borders.

I think it’s too hard to implement a CIV5 model since the regions are all pre-set -- walking into a region is different than walking into a district. But by the time my army’s in the district the AI (like a human) might be too late in their response.



I wonder if there’s a way to trigger a more aggressive response (shutting borders) if human army that gets within 2 hexes of district is, say, >= garrison + available region units.



Then maybe a “disperse or we will shut borders” might cause human to declare war immediately or retreat? that way AI has a bit more time to bring troops closer to home (and maybe a reason to build cities away from borders?)

Sometimes, IA make stranges choices.



Example : I'm in war against an empire, and I'm winning.

The empire send me a message that say : this war is not good for anyone, it should end.

So I go to diplomatic and send a Truce (with no condition, I don't claim anything), and then, IA refuse it !!

sgtslick wrote:


In my opinion this is a complete exploit. Two teir 1 crappy techs for over 8,000 dust.. ridiculous.



The game needs to put a cap on what each tech is worth - dependant on the tech value in terms of its cost in science beakers. It should not be merely dependant on the factions desire for the tech or else these sorts of problems occur, where the ai unfairly pays out exorbitant amounts.

each tech's value is relative to WHEN it is researched.. a crappy tech from era 1 that would have cost 28 science on turn 1 is now costing me ~ 2600 halfway thru era IV.

its also relative to the desire for a tech... for me shipyard is always low low low on my priority list but i always try to snag it in a deal when i can... its just something i dont want to have to waste turns on... you never know when you'll NEED it... i just dont wanna waste turns on it.



on a different matter...

the cultists proposed a truce while i was sacking their cap.. i didnt notice it until after i had taken the city... it must have popped up before i had taken it.. not sure... i was getting ganked somewhere else and probably panic escaped every pop up that popped up... in any event i ended up signing the treaty and making what would be an ordinarily lopsided deal in favor of the AI... however i knew they'd be going poof and i wanted to save some research turns on a couple of neglected techs with some minor sweeteners to "balance" the trade... all it truly cost me was influence.







MY SUGGESTION is in regards to the truce proposals.. you should be forced to deal with the proposal before committing acts of violence.. otherwise its fairly easy to exploit the AI by minimizing the popup, sacking a city or two, then fleecing them for whatever they've offered or you counter offered.. the proposed deal should be nullified once you've shown further aggression... or at the very minimum the conditions for truce should be weighted differently.... and in the case of the cultists if you've sacked their cap the proposal window should just disappear.



unrelated...

the popups are a huge pain in the ... sometimes.... those turns when you're waiting for the next empire plan with your lone settler in a dangerous region... you've got your mouse ready to dodge the skulking silics... you hit end turn on edge ready to go hide behind those trees if needed.. its turn 39 just one more turn... and you see it coming in your direction so you get ready to move out of its LOS then BAM some pop up steals the whole window... hit escape.... right click.... panic... anyway my point is make the popup windows less whorish in that they can be popped up but not be so intrusive and have to be dealt with so immediately... have them pop up in the background.... i dig pop ups i dont want to have to disable every one... just make the window transparent or in the background or something so that i can still move stuff around... i suppose its a focus issue.



anyway.. great game.

cheers =)

Ok, finished another game on Hard difficulty (Fast speed) as Drakken - mostly so that I could really explore the Diplomacy system. Won around Turn 100 with a Diplomatic victory, and by that time I was bringing in over 250 Influence per turn (and starting wars/forcing peace treaties in alternating turns to spend it). Did not need to research the Era VI tech that doubles your diplomacy points. Here are my thoughts:



1. I actually think that the AI's are pretty good at assessing fair deals right now. If I'm on good terms with an empire, I get better deals. AI's are always reticent to trade high-level (newer) techs, as they should be. But I can often trade for lower-level techs, if I'm giving them techs or valuable resources in return. I can often trade Dust for luxury or strategic resources, and the deals seem relatively fair.



2. If you kick an AI around in war for a little while, they get exhausted, and you propose a Cease Fire - you have a lot of bargaining power. I think this is appropriate, but it does definitely allow for further steam-rolling of AI's. I wouldn't change the Diplomatic side of this - but this is yet another negative effect of the AI's poor ability to wage war. You will not only be conquering cities/expanding your empire (which doesn't have enough negative effects, in my opinion) but you'll also be able to use muscle diplomacy to get even more of an advantage.



3. Building towards alliances is actually fairly beneficial. The trade benefits can be substantial - but the problem here is that the trade system in Endless Legend is obtuse. It should be easier to see the benefits of having trade routes to OTHER empires (vs. trade routes within your empire). These between empire trade routes should be MUCH better (provide much more Dust and Science), in order to justify the building of Docks, for example. And then using Diplomacy to enhance the value of those trade routes would be super-valuable. IMPORTANT NOTE: AI's should do this on their own, with each other! The "peaceful" AI's in particular, should be rushing to form alliances with each other so that they can get a Dust/Research advantage on the other empires. I don't get the sense AT ALL that AI's know how to use these alliances to further their economic development.



4. The Diplomatic Victory condition is poorly implemented. You just spend lots of Influence? Without any particular diplomatic goal? This forces to play very erratically, diplomatically - alternatively initiating cold wars, negotiating for peace, forming alliances, then back-stabbing, etc., etc. It's not really a fun way to play the game, except that it forces you to become familiar with the Diplomatic system. There definitely should be an overarching Diplomatic goal that you're trying to achieve - and sure, it can cost a lot of Influence, but it should make more sense from a narrative standpoint. I hesitate to add more systems into the game but... maybe if there was a "world-wide stability" index? Something that was affected negatively by cold wars and hot wars, and positively by trade and alliances? And then if there was a way to use your Diplomatic weight to encourage other empires to enter into positive diplomatic relations with each other?



This is a tough one. The Diplomatic Victory condition in Civ 5 was also unsatisfying (building the UN, have AI's vote for you?) but at least there was a "consistent" goal: be on good terms with empires, and esp. minor civs. This condition needs re-thinking or no-one will ever pursue it.







FINAL POINTS:

1. The AI should be improved to make use of "positive" diplomacy with each other (alliances and trade agreements) more often. This would have the effect of making the game more challenging for the (human) player.

2. The Diplomacy Victory condition needs to be re-designed.

I had a battle a day or two ago where the enemy melee unit was content with just standing at the base of a cliff doing nothing while my ranged unit slowly took it down. It seemed to be targeting one of my far-away low health units that it could not reach, but it did not try to re-target and attack any of my nearby units.

I had a game where i fell behind to the AI early on but battled back against a Broken Lords AI. The computer doesnt do a good job of making new stacks of units. I beat a couple stacks and then was free to take half the regions on the map. Once our scores finally flip-flopped and i passed him, the AI offered me a truce- and a payment of 100,000 gold!! Obviously the computer should of been using that to hurry out new units. This needs to be fixed!!

Actually, i have the feeling Drakkens play necrophages roles. Usually, if i have necrophage and Drakkens at neighbours, Drakkens will attack me in first ! I have the feeling necrophages attack other players on fixed timings, and Drakkens can be opportunists if they feel more powerfull than them neigbours. Maybe they are too opportunist, and should try to create solid alliances instead of destroy everything near of them if they can.

I have visited the post launch thread.

http://forums.amplitude-studios.com/content.php?489-Endless-Legend-Post-Launch-Plan



I have tried to search the boards for AI (you cant), and instead searched for other things "Improve AI" and "Single Player".

Besides minor things like slight AI tweaks in patches:

/#/endless-legend/forum/5-general/thread/1772-v1-0-7-is-out

http://forums.amplitude-studios.com/content.php?510-EL-Visions-of-the-Unseen-Focus-Diplomatic-Answers

and player complaints... there is nothing talking about AI.



Where are we? What is the latest? Was the latest patch the major diplomacy AI improvement?

I understand the AI is a work in progress.. but I honestly do not see the changes advertised in the last patch.





•Allowed garrisons of AI empires to handle a siege on their own.

---> Not sure what this means. So hard to comment.

•Improved the battle targeting of AI units.

---> Doesn't seem to work. Units still stand around doing nothing. I have a post where the whole army didn't move.

•Increased the importance of Industry when the AI chooses a city spot for its next city.

---> Still chooses horrible spots for cities (in corner of region, surrounded by water etc where cant expand properly).

•Added the use of the retrofit feature for the AI empires.

---> Seems to work. But AI doesn't use the strongest weapons/armor... So maybe not?

•Improved the way AI empires manage their Empire plan.

---> Havnt seen enemies during major war use the battle plans (my units are much stronger). This is late game so issues is either in empire plan, poor equipment or both

•Improved the way AI empires manage their construction queues.

---> Have seen BL not use 100k dust to make units (while losing) and get knocked down to 1 city.. instead of hurrying units. Issues might be related to strat resources. AI needs ability to build weaker units if out of strat. And in war situations need to be able to rush production.

•Changed the way AI empires manage their researches: must-have technologies begin with a low heuristic value but increase a lot at each completed research.

--->Have seen BL not research or build major dust researches. I can give screenshots if need be. I bet I can get examples of other empires disregarding similar important researches to their playtype.

•Added an "after battle" analysis in order to adapt the unit designs of AI empires.

---> Doesn't seem to work. AI uses same unit mix even after losing a ton of units without doing damage.

•Refactored the unit design system used by the AI empires.

---> Doesn't seem to work. Late game even on Endless AI armies are weak compared to mine.

•Improved the management of the "Slayer" abilities by the AI empires in their unit designs.

---> Cant tell a difference...there might be more infantry slayer. Its hard to say.

•Increased the weights of resources marketplace technology and the assimilation technology for the AI empires.

---> Seems to work. Its hard to say.

•Feedback have prerequisites to avoid strange messages from AI empires (for example: "We appreciate your peaceful behaviour" during a war).

---> Doesn't seem to work. Still get weird messages...."Im getting whooped, lets end the war...just kidding! give me 50k gold first."

•Open/Close Borders and Map Exchange/Embargo are now used by AI empires.

---> Works! But now I always know when AI will declare war (after closing borders).

Hi there,



I've bougth the game a few days ago and after finishing a few games, here are my thought on the AI (on aspects that weren't covered in the surveys ^_^) :



1. The AI currently seems to not be very interested in anything happening around it, and especially, what the player is doing. On a fast game, my first interaction (out of 7) was a war declaration at turn 66. As someone specified sooner, it make the diplomatic aspects of the game less interessting than they could be (since they aren't really needed to win/get a better position) and the IA tends to not try to catch up with runaway opponent (player or IA).



2. The position of IA cities are quite strange. I've often captured cities with one ressource completely missing (most of the time, industry or science). There is a definitive need for better ressources management from the AI, having them starting with poorly positioned cities is clearly not going to make them harder to beat. Having some kind of required amount of each type of ressources accessible from a certain number of case from the center of the city could help.



4. About the AI of the minor faction: I have some difficulties understanding their priorities. They seems to be a bit too agressives. In a lot of occasion, they would rush on one of my army, even though they have no chance to win (same thing with siege). The targetting in battle is very weird if you have a settler with you, their main priority seems to be to kill it, even if they barely graze the rest of your army. I would understand that from the IA of a main faction, but from a minor one that has nothing to earn from it, it seems a bit strange.



5. While I'm thinking about it, there a little thing that also perturbed me with diplomacy (it is about the GUI though). When I offered a deal, the refusal would not be directly displayed on the screen, but put on the event list (bottom right). It seems a bit strange since displaying the result directly would be more pratical.



Well, that's about it, still enjoying the game a lot and even though there may be some polish left to do, I think it's already great ^^