Hi all,



Over the past months, we have read some criticism on the AI of Endless Legend. There has been an ongoing effort since release to improve the AI, but we would like to gather more feedback from you on its current state.



What we would like is that you describe situations where the AI behaves improperly, and what you think it should have done. For example, I am at war with an AI Empire (saved game provided woudl be nice), but its leader sends me a friendly message, while it should have taunted it threatened me.



We have several leads to improve the AI, but we want to make sure we prioritize tasks based on what is the important for you, and also with a clear situation to understand the issue.



Thanks for your help!

Hi, I will place other things as they come by, but one thing the AI doesn't do is to consider battle formations in combat... Each unit move order is placed without considering the other unit's position and move order and this simply ruin the day for the AI : a concrete case is a battle with the silics I had yesterday. I've got 3 marines and 1 rider, against 6 silics... And they didn't make a dent on me. Why? Because their move sequence was such that I could storm my ranged attacks agains a single unit at a time, and they came to range only one unit at a time. I've came into situations like this:

The blue stars are marines holding position and waiting for the silics to come by (this because I'm human, I've never seen the AI do that).

The red stars are silics trying to reach the marines for ruining their day.

The expected behavior is that the silics will try to enter a formation where they will enter all at the same time within striking range of the units: this way, they can cause damage, even while some of them are slain in the process... Like this.





However, what they have done, due to movement order , was this:



The first one try to reach the target, the second one do the same, and blocks completely advance of the third unit... The third unit is forced to move sideways, and the marines can fire at will to decimate the only two units that are targetable.

This is aggravated , of course, by the silics ridiculous 1 tile per turn movement... Was this made to be fun on purpose?



Another concrete example: following the "poor battlefield choice" that is explained later in this thread. Look at what this caecator did: instead of cleaning the reinforcement tile for receiving a reinforcement, it decided to attack from where it stand! I agree that in this specific situation it wouldn't matter, because the reinforcement and themselves have low initiative...

Another impression I have is that the AI seems to have a hard time upgrading units. This makes a human player steamroll the AI players, because we are facing tier 1 units with palladian tier 2 weapons... But maybe this comes from the fact that AI sucks in combat, making me have 1 army with high level units facing low level units...

One more: the AI sucks at choosing which armies will participate in combat, and when they engage in combat. They always place it all into it, including the militia garrisons... I know I've won a region when I have an opposing army at 3 tiles of a city: I simply stop the siege, attack that army directly and I'm always sure that the pesky city militia will join the fray, be slain, and get me the opportunity to occupy the city with another army.



The correct behavior, in this case, would be to let the city militia out of battle and wait for reinforcements... Unless I'm absolutely sure that with the city garrison I will have a crushing victory.



Another concrete example: Consider this combat sequence:



1) A caecator army, just formed, decide to attack my army...FIRST MISTAKE: the issue of battlefield choice appear here, heavily. Another point that appear is that the caecator's village doesn't even appear as a reinforcement...

But even if the caecator has no reinforcements from the village, at least it should be kept without attack. This, because when they attack, my army don't lose the action point



If they haven't attacked, I would:

1) Attack the caecator's army, that would consume my action point and keep the village alive for one more turn.

2) Attack the village, that would be a mistake from my part, because in that case the caecator's army would appear as reinforcement.

Since they didn't that, I have an action point to actually attack the village: and here is the outcome.

Oh , that sucks big time... Stupid reinforcement tiles placement... There are situations where the defending army only have one reiforcement tile and one single tile where to place their units... This makes me take an entire army of no matter how many opponents they have, I can simply take them out, one unit per time... A complete nonsense. One such situation is when I'm attacking a coastal city from some specific direction..

This is a concrete sample, found today during my Vaulters game... By the way, this can be considered an AI bug, because they caecator's army decided to attack me under these conditions...



And the save file Vaulters - Turn 24.zip



OPTION 1: Implement some sort of amphibious warfare, so units can be placed at water... I've given one possible solution here. /#/endless-legend/forum/6-game-design/thread/3563-naval-warfare-should-it-be-implemented-and-how&p=266931#post266931

OPTION 2: Shift the combat board in the combat direction that will give the defender /offender more tiles to play on. There should be at least enough room to place the first 4 units and 2 tile placements for each army...

Co-design at its best. Great thread.

One of the complaints I've seen a lot is that ranged units are stupidly overpowered. However, I think that when they refer to that, they refer to the dekari rangers: they have long range, and can reach everything in the map even if they don't have a line-of -sight to the target: is the defending units behind the himalayas? No problem, dekari rangers are such badasses that they can use parabolic fire and hit a target that they couldn't really see...

Hello abmpicoli and thank you for your comments!

In the future however, could you please post your feedback in a single post, please? It will be easier for us to read everything. ^^

I *know* this can be something that will probably be painful to do... But would it be possible to expose an interface for a third party AI? I'm sure that there are lots of geeks and nerds over there (hey, maybe I'm even one of them ) that would love the oportunity to develop their own AI for the many aspects of the game, such as combat , strategic decisions, etc... Then we could even place matches between the different AIs and see who is the winner..

Hello,



first 20 rounds of the game is very slow. Boring. Usualy one army. Just click and waiting end of turns.

Is it possible to speed up growt at the begining?



1. Beef up The Founder Memorial

Cost: 50% cheaper to build

Bonus: packeted bonus, 5x current packet

Bonus fade out : every 5 rounds one packet of the memorial wears off. So in 25 rounds memorial has the same bonus as it has now.

Lets say: First round memorial gives 600% of current state. 6th round it gives 500%. 11th rounds it gives 400% and so on. Finally it gives 100% for the rest of the game.





2. At 10th round player has 2 cityes.

a] wipe out an AI oponent in the game beginig is quite easy.

buy a far see amulet then wait till he leaves the city. Snipe a city and hold it.

b] An excellent/poor first region could make a lot, but two regions for everyone early in the game makes nice average -> more balanced game.

- There should be a free settler to help this. After the Founder memorial being build, his glory attracts more people so a brand new settler appears in the city. ( A submissive offer from minor close faction for cultist)





-----------------------------------------------------------------

Both changes are focused for faster and more balanced and gameplay in the first 20 rounds.





I am aware that those are not the classic AI bugs rather game play change for which AI must react properly.

I guess, AI should be prepared for such advanced game play changes.

So if you are still in the mood for improving AI of this game, also consider please changes for dynamic game play.



Thx

The AI tends to make unintelligent choices about which techs to research, which leads to unintelligent decisions about which improvements to build. I have seen Vaulters without t1 strategic extractor tech. The number of factions that have failed to research Empire Mint is alarming and not anything you would see in any competent human player. Failure to research strategic weapons and accessory tech ends up contributing to the general weakness of AI.



There are a lot of problems with the AI, but this one, to my mind, is the largest, simplest failing of AI in Endless Legend.

Thank you for your suggestions, however I think that some of you misunderstood the subject of the thread: confusing balancing & AI improvements. Let's go back on the AI!

Saylawl wrote:
Hello abmpicoli and thank you for your comments!

In the future however, could you please post your feedback in a single post, please? It will be easier for us to read everything. ^^

Well, if you say so... However, I think it is important that we keep one idea per post, because my view of a smarter AI may not be the one other people have, and this may raise a discussion, that may reach a totally different conclusion than mine. One idea = one requirement validated or invalidated by the other members...

One of the things I'm seeing here is that it seems very difficult to prove an assertion regarding the AI being good or bad if we don't see a concrete evaluation of the AI decisions... While we are playing, we obviously shouldn't see what the AI is doing, since this will ruin gameplay... However, the saves should have full event logs on what is happening to each faction, so we can eventually replay combat situations, and see the AI's decisions.



For example, with chess we can replay a chess game by using the chess movement notation... There is even a notation for a really bad mistake...



We, as players, see the incomplete effects of a possibly bad AI... But we can't really point the finger to tell what the AI is doing wrong, and why...



So, how can I log the full decisions the AI makes, and all their moves? Does debugMode provides something like that? Is there a way to filter the noise of generic debug from the ones that are really actual AI decisions?



I offer myself to play a game in debug mode, if there will be a log that provides enough info to evaluate that.

Is the thread supposed to include combat? I think there are numerous tweaks that could be made, most of which revolve around giving it a better ability to recognize when strategical logitics, such as deployment and combat movement, can be a superior advantage in comparison to a higher Morale. That and allowing the AI to change stances(offensive, defensive, hold) during a battle.



In terms of worldmap gameplay, I think it´s worth noting how often the AI exposes garrisoned armies by using them as reinforcement on battles it really shouldn´t. Most screamingly, the Cult. I don´t think I´ve ever conquered them on Endless without using this.

I´m sure you can sit there with five badass armies and eventually take those 800 fortifications down. But what´s happened to me so far is that I only need to keep a couple of reasonable armies within reinforcing range of a Hero-led one, which is positioned within sight of their city. They always send an army to engage your hero - which is understandable if they lack the vision to spot the reinforcement. But once the battle is triggered, it should be able to realize it can´t achieve victory and just sacrifice the engaging army in order to keep their garrison intact.

In Endless Space factions controlled by the AI behave according to their advantages and their lore when it comes to diplomacy. They always take into account the victory type they aim for. For example, the Amoeba is a peaceful faction, which receives huge income from trade routes and aims for the diplomatic or the economic victory. So this faction tends to sign both peace and open porders treaties with other empires asap, it will gladly accept your cooperation agreement offer and prioritizes the technologies that unlock diplomatic bonuses.



Do the Drakken really care about earning diplomatic points? Do they prioritize researching "Diplomat's Manse" and the tech that unlocks alliances? Are they peaceful? Do they use their affinity? The answer for all these questions is "no".



Diplomacy-wise I don't see a big difference between the Drakken and the Vaulters, or between the Roving Clans (this faction never offers you to sign the agreements boosting income from trade routes, it doesn't seem to care enough about researching "Imperial highways" etc.) and the Wild Walkers.

Hi,

Thank you for all your feedback. Keep in mind this thread is devoted to the AI, namely how the empires controlled by your computer use the gameplay and react to your actions.

To answer some of you:

abmpicoli wrote:
A caecator army, just formed, decide to attack my army...FIRST MISTAKE: the issue of battlefield choice appear here, heavily. Another point that appear is that the caecator's village doesn't even appear as a reinforcement...

But even if the caecator has no reinforcements from the village, at least it should be kept without attack. This, because when they attack, my army don't lose the action point

Minor factions are designed to play in that manner. They are the kind of mobs that are made to challenge you at the early game; they are not supposed to be the main threat of the game. This is why their behavior is supposed to be predictable. If they see you, they’ll attack you, whatever the situation.

Vicarious wrote:
In Endless Space factions controlled by the AI behave according to their advantages and their lore when it comes to diplomacy. They always take into account the victory type they aim for. For example, the Amoeba is a peaceful faction, which receives huge income from trade routes and aims for the diplomatic or the economic victory. So this faction tends to sign both peace and open porders treaties with other empires asap, it will gladly accept your cooperation agreement offer and prioritizes the technologies that unlock diplomatic bonuses.

ES factions don’t aim a specific victory and don’t know their own strengths. They have more evident “role-play behaviors” which, by side effects, make them better for a specific victory. But yeah, maybe we could improve the way Endless Legend’s AI role-plays to have the same side effects as in ES.

abmpicoli wrote:
I *know* this can be something that will probably be painful to do... But would it be possible to expose an interface for a third party AI? I'm sure that there are lots of geeks and nerds over there (hey, maybe I'm even one of them ) that would love the oportunity to develop their own AI for the many aspects of the game, such as combat , strategic decisions, etc... Then we could even place matches between the different AIs and see who is the winner..

You can enable AI tools, and even switch between empires to see what the AI does, by following instructions from this thread: /#/endless-legend/forum/15-modding/thread/5778-accessing-debug-modes



So, after reading your suggestions, I can conclude that the IA should:

• Plan movement of the whole army instead of each unit
• Move out of a reinforcement tile when reinforcements are still waiting
• Retrofit its units faster
• Take into consideration the best angles to attack (reinforcements, deployment zone, etc.)
• Focus more on strategic resources (on this point, the last update improved a lot how AI prioritizes strategic resources)
• Being able to not include reinforcements if the battle seems to be doomed
  
  

Thank you a lot and keep the feedback coming!

I have a question, if these things get sorted out, like AI know how to refit its units, knows what to tech, know how to properly fight battles will the bonuses on higher difficulty's stay or will they be reduced?



Also, another strange thing the ai does is they will always attack my armies in neutral areas, I understand if the cravers do this, but why would the drakken do this? They are all about diplomacy.



Another thing that's not only about the ai behaviour is that conquest and military gameplay is too powerful, and the only thing the AI can compete in, they will build massive armies over city improvements that would make them stronger in the late game. Currently the AI is very dangerous early game and very weak late game.

Manu wrote:
So, after reading your suggestions, I can conclude that the IA should:

Please, read this thread.

The main thing: AI should place independently units before the beginning of fight. Otherwise the army significantly loses in efficiency. And is worse than that: the part of army can remain locked, without a possibility to approach to the opponent - that is visible on my sreen.

Let AI places units, proceeding from following principles:

1. All units should have passed (see my sreen) a way up to spawns the opponent.

2. Units with the least health and protection should be behind.

3. Units of near fight and with small quantity of glasses of movement - as it is possible more close to the opponent.

4. Own points spawns after arrangement should remain free.

As to a choice of objectives, - let units with special capacity of the murderer of the certain type of units, - plan to themselves such objective.

Another thing: the AI very rarely uses the market. Especially rarely buys something, even when he has a lot of dust, and a strategic resource or luxury very useful.