I just recently startet playing after the Draken were released. Overall I really enjoyed EL so far and still have high hopes for it.

Yet my biggest concern is the AI:



Right now playing on endless usually is less "you shall not pass" and more "you already won". I know that developing an AI that is challenging for exerienced players and wont resort to insane cheating is sadly a herculean task. So far only very few games got near that goal and even recent games usually fail at it (Sometimes I get the feeling that AI-opponents in strategy games made no progress during the last 10 years - but I am getting off topic).

I know Endless Legend (and therefore the AI) is still in development. But given that even games like Civ5 offer an incompetent AI (esp. during war) I really hope the developers devote enough time to the AI-opponents.



What I really hope for Endless Legend (on higher difficulties) is to offer a constant challenge from opponents who:

- know how to expand and develop their cities

- have no problems with neutral creatures rampaging through their lands

- own adequate defensive armies and are able to defend their cities

- are able to plan and execute conquests, especially against opponents who neglect their military

- at least try to work towards a victory



If the AI has to cheat to achieve these goals thats fine. Just make it reasonable and solution oriented:

For example Civ5 on deity simply gives the AI extra units, workers and techs. While this solution makes the AI stronger it also makes it almost impossible for human players to focus on early wonders - the AI simply has too much of a head start. This is an example of a bad solution to the problem, since it takes away a certain aspect of the game (in this case early wonder focus).





So far the gravest AI-Errors (on endless difficulty) I witnessed were:



- The AI often founds its cities in terrible locations - in corners / nowhere near anomalies or rivers - even though the area offers much better locations.

-> This might have something to do with the AI using settlers as soon as they are threatened. If this is the case then it would be better to simply make AI-settlers immune to neutral creatures instead of cippling the AI with terrible cities.



- The AI is eager to accept peace a few turns after they atacked you, even if they took you by total surprise and could easily conquer your cities (same in Endless Space).



- The AI often can't handle the neutral barbarians (esp. on hard). I would see the AI cities constantly sieged by barbarians and I even won one game by simply waiting until all AI opponents were defeated by barbarians.



- Most of the time the AI seems to have no army or almost no army. This makes their cities practically fair game for human players.

-> If you are worried about the AI wasting too much dust on unit maintenance then give the AI a big discount for unit maintenance.

-> If you are worried about the AI wasting too much production for units then you could give the AI occasional free units. Lowering production costs for AI-units even more would work too, but might create a problem of AI-players spamming too many units at some point.



- During trades the AI occasionally makes no sense.

-> While buying a city from the AI for a large sum of dust is certainly valid (the roving clans main quest kinda requires this), the AI selling all its cities for enough dust and techs is certainly not. I didn't test it, but I assume buying all the AI capitals should be much cheaper than reaching the dust for the economic victory.

-> Also the AI sometimes values a certain luxury ressource (in my case spices) too highly and will give almost anything for it. It seems to be either a bug or a really really obscure refference to Frank Herbert's Dune: "He who controlls the spice ..."



- On a (kinda) positive note: In one game the AI actually won - at turn 50 (normal speed) one draken AI got a diplomacy victory. While I am glad to have lost against an AI at least once I can't shake the feeling that this is not working as intended.





The reason why I am writing all this is simple: At least for me a weak opponent-AI quickly errodes my motivation to start a game in singleplayer. I have the exact same problem with Endless Space already and I really want to enjoy Endless Legend games as long as possible.



So please devote some time to the AI

From what I have heard, AI is still mostly a placeholder while they were working on features. Now that we are more or less feature complete, they are working on AI.

mdrxl wrote:
Sometimes I get the feeling that AI-opponents in strategy games made no progress during the last 10 years

It's not just a feeling. It's an observation I also made. If I compare competence of the Civ 3 and the Civ 5 AI, I'd actually say the Civ 3 AI was better.



Making one game and improve it more and more is not economically beneficial. So instead of doing that there's always new games released. New games have new rulesets and thus need new AIs, so they have to be written from scratch most of the time. Also you can't really have good results with the AI before all mechanics are in.

Then look at how much time they spend on it. If the current AI is still just a placeholder with the release coming really close, you know that even if they switch all their manpower on it, the result can't be that great.



An actually good AI for a 4x game could be made if there's years of observation of player-strategies and adopting the knowledge of the very best players into the AI.



I think that competition instead of cooperation is especially noticeable in the field of game-AI's.



If there was a way for other programmers to access the original source-code and constantly improve it, great things could be done.



Sometimes I wish they would make one game and then just improve it over and over. They might even exchange the engine and the art-assets but they should keep the core intact so they can constantly improve the AI to make it really good. Like chess AI's. It's no surprise that in Chess-AIs a cooperative-non-profit-project has surpassed all commercial AIs.