More than anything else, I think that the AIs have to recognize the threat of a runaway faction (be it the player or the AI) before it becomes too late. In my current game, The Necrophages (played by the AI) have three times as many cities as anyone else and about 5 times the next strongest army. They won't be stopped unless the other factions unite against them; but they're not doing so.



Thanks in advance.

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.

Maybe I'm just "lucky" but they just don't close borders on me when they obviously should.

I don't have a ton of experience with the game as of yet (1 complete game, several partially finished), so please take with grain of salt. But I also don't consider myself a "hard-core" player, so I think feedback from my demographic is somewhat important/useful if you guys want to reach a broad audience.



Thoughts on AI:



1. I agree with one of the posters above that I like the "quirky" race-specific AI's you guys have put into the game. I MUCH prefer this to a overarching AI system that all races uses to "win the game" just like the player (see Civ V). The unique value of Endless Legend is that it has great immersion and lore, and race-specific AI (in terms of pursuing different kinds of victory conditions, minimally) is a contributor to that.



2. I would recommend not making cities more intrinsically defensible. I love how quick and fluid warfare is and I HATE long slogs. That is for other 4x games (Warlock, Civ, etc.). I think you can create a stronger challenge for players without just giving an AI access to tougher cities that require longer sieges.



3. Success in warfare seems dependent on a few primary issues:



A. Putting a priority on techs that enhance army strength - esp. the ones that increase army size and weapon strength. AI's planning to pursue war should research those ASAP.

B. Putting together 1-2 powerful stacks. Maximizing army size, and upgrading units to have the best weapons. From my limited experience, the AI does not seem proficient at combined arms. In my current game, for example, I have only seen stacks of ranged units. This makes it relatively easy to counter AI armies. I think if AI's were trained with a simple rule to have 50% ranged and 50% infantry/cavalry/flier, that would work pretty well.

C. Choosing bonuses that enhance your ability to engage in warfare. Civ plan, use of appropriate luxury resources, etc.



There's obviously more behind successful warfare (including combat tactics, which I'm avoiding entirely here because I actually think it's a less relevant issue), but these 3 seem super-important to me.



4. With regards to defense, I think that AI's pursuing a more turtle strategy should protect their cities better (esp. regions on the border) with heroes/governors, units, and fortification. That's really it. From my standpoint, I just don't want to be able to wander into an AI region and find a key city defended by 2 ranged units + militia (vs. my stack of hero + 4-8 uber-units). If I invade, their cities should probably switch to building units immediately and they should use dust (if they have it) to speed up unit generation.







That's all I can think of right now, I'm certainly curious to hear what other people have to say.

I've just finished another game (post 1.0.8 patch) on Hard difficulty, Fast speed. It was a very enjoyable game, and it does seem like the AI is improving a bit. Here are the issues I continue to notice regarding AI.



1. I'm not sure about the algorithm the AI uses to determine where to settle cities. Has this been discussed elsewhere? I've run into a a number of AI cities that seem poorly placed.



2. Are AI civs effectively using strategic and luxury resource boosts? Do they have a simple rule to just use them when available? That would probably work well enough - but it might be useful to reserve the warfare-specific ones (Blood Crystal, Grassilk) for when they are at War. Just wondering here.



3. AI seems fairly proficient at upgrading their units (research techs and retrofitting their units). However, they really need to focus on researching Meritocratic Promotion (and the other tech) that increases max Army size. That is super-super-important for effective warfare.



4. I'm still not seeing great evidence of mixed arms. For example, I was at war with the Broken Lords for the latter third of my most recent game. They had strong, high-defense upgrade Stalwarts - but nothing else! And max army size of 4. This made it pretty easy for me to build some anti-infantry armies (Infantry Slayer, etc.) when invading them subsequently.



5. I did see AI's defending their cities more effectively - often with full garrisons of tough units. It made conquering them harder and a more satisfying challenge.

Combat targeting :



1) Near dead targets / damaged target / weaker or easily killed targets... focus fire those to reduce overall damage.

2) Easy access targets, start by attacking what is nearby. (AI often ignores distance in its target acquisition it seems. There is one bug, if you surround a ranged healer with all your other units, AIs won't move some units because they keep targeting that inaccessible choice target and there is no path to reach it.)

3) Focus generally on what is easier to kill. If a hero is near, I will focus his units and leave him last, unless he's weaker or very dangerous.

4) I will attack depending on my tiles positioning... if I can land in a forest, i might attack something just because i can reach it from a position of strength. (Take into account morale bonuses, staying near allies or landing near allies can influence who you attack.)



There you go, attack what is weak, dangerous, but mostly what is near, and under good conditions.



---



QUESTION : Are the AI script files accessible? If they are I would very much like to make an "AI enhancement" mod.

The poll questions were not really easy for me to answer, so here is some additional clarification.





Warfare and Defense



1. I try not to have more than 1 war going on at the same time. If I'm at war with one enemy and another attacks me, for example, I will try to negotiate a peace treaty with the first enemy as quickly as possible. Especially in the early game, I don't usually have enough armies to wage war on 2 fronts and adequately defend my territory.



2. In the early game (up to turn 50 or so), I try to keep 1-2 units (+ militia) in every city as defense. I also hopefully have an army of 3-4 units + hero at this time that can wander and deal with minor factions.



3. By mid-game, assuming my empire is 5+ regions, I will primarily defend my capital and border regions that are adjacent to potential threats. Defense will be 3-4 units (+ militia). At this point, I usually have 1 large roving army (6 units + general). Maybe a second if I'm pursuing a heavy conquest strategy.



4. Talking about late-game seems irrelevant, but at that point I often have 3 roving armies of 8 units + general each.





Tactical Combat



1. Targeting: My primary suggestion here is simply to have the AI focus fire on the closest enemy unit. If there's a tie in which unit is closest, then focus fire on the one with the lowest HP. The biggest mistake the AI makes is not finishing off units that are close to dying. If often tried to get around my infantry and attack ranged or support units, but its efforts are usually wasted. Even if the AI isn't going to win its battles, if it whittled down your army from the first engagement, it would make a big war more challenging.





Diplomacy



1. Maybe I'm odd, but I rarely use Diplomacy in this game. The market makes negotiating with AI for strategic/luxury resources largely moot. Trading techs sounds great, but is a difficult thing to balance. I love the idea of having to spend Influence to do anything, but there does come a point (perhaps too soon) where you have so much Influence that it's a meaningless resource in the mid/late game. There should be stronger incentives to form peace treaties (besides the bump in trade route income) - and harsher consequences to breaking them.

Following up the great answers in this post and the threat level concept.



threat level = dmg the unit is going to do for the rest of the battle / the number of turns the current unit will take to kill the enemy unit



That should account for almost dead units, units that have move etc. Include bonus for the rest of the army in the dividend.



Now check for distance. If it cant be reach go for the second. Consider using another attack to another unit as a proxy to get closer.



In any case I've developed AI and it is really difficult business. We would be happy if the Ai is just not too dumb. Just domething as simple as attacking the closest target could be acceptable.



I am not sure which is your testing framework for the AI but I would suggest creating a test harness that simulates battles under most conditions. It is not that difficult to implrment. You create battle templates changing unit numbers composition factions heroes terrain. Then use different versions of the Ui to battle each other using permutations. Check which Ai does higer dmg kills. Finally have a human test for exceptional tactics like healer fully protected and add exceptions for this



Hope it helps

Can someone answer me about my modding question? Can we mod the AI at this time?

You can. I already modded some files. Take a look in the modding forums.