Dear Devs,

After playing the OpenDev for 30+h and also watching many hours of live streams I have some suggestions on how to improve the AI so that the harder difficulties also pose a challenge for experienced 4X players:

1) War Trade-offs

I've had it a few times that even on the highest difficulty setting the AI is declaring an unjustified war without having the military to actually fight the war. So without any fighting at all, I could just declare them a vassal after a few turns. The suggestion here is to let the AI only declare an unjustified war if they have such a strong military that they could actually fight and win battles against the player. I do like the game design behind the grievance and war support a lot and I think that should probably stay roughly like it is. It's just that the AI should be taught to make smarter use of it by making better trade-offs when a war could result in their benefit. Seeing AI-vs-AI battles also very often result in vassals, I would guess that this does not only happen when an AI is playing against a human player but that they just generally declare more wars than they can actually sustain.

In general I think the AI should prioritize unit production and researching upgrades before declaring a war against a strong opponent.

2) Specialization of Territories for Districts

As a human player it is relatively easy to see that in this game one really good strategy is having cities with 2-4 territories and specializing the territories so that one is generating most of the food, one is generating most of the production, and one is generating most of the science / money. Due to the adjacency bonuses of districts and the fact that districts remove other yields, this strategy scales the best with current balancing. If balancing is supposed to stay this way (which I think makes a lot of sense from a logical perspective), I think the AI should make use of this knowledge by also specializing their territories for certain purpose and ideally also plan the outpost locations so that expanding from it would already help them with the specialized goal of that territory. Since the AI can do more of those calculations "in their head" than humans, this game mechanic would result in a very good infrastructure for them since they can plan very consistently. So this strategy would give the AI and advantage that would compensate for other more advanced planning and smart synergies that humans can see easier than an AI.

This advanced strategy would probably only make sense to use on the higher difficulties since casual players might not keep up with this specialization, but for the more experienced players it would certainly encourage us to play the game for more time and maybe also buy more DLCs later if we feel like after a few games we are already in auto-win mode.

I think this change would address the issue that in the late game the gap between the player and the AI (even on humankind difficulty) grows larger and larger. Humans are just able to plan better. The AI hardly can catch up once the game reaches a certain amount of complexity in its decisions.

3) Assimilating Free Cities

In most of my own games and the live streams I have seen the AI very rarely was assimilating other free cities even though they certainly have the resources to do so and it would benefit them a lot. I think if there would be a little bit more competition for those, the AI would probably end up in a better position that makes them more able to compete with the players. I think that is another factor in their late-game decline.

4) Wonder Synergies  

Currently there are super strong synergies between Angkor Wat, Notre-Dame, and Machu Picchu. I think if the AI would be able to value wonders not just from their own value but also assign pair-wise or group-wise weights to wonder combinations, the AI would be so much stronger. In the Victor OpenDev we only saw about half of the game, so those synergies did not have much implications on the following game, but having all those three wonders (which is certainly possible in a single game) brings one in the primary position to have an uncontested lead for the second half of the game since every city would grow about every single turn. So the suggestion is basically to let AIs compete stronger for wonders and also teach them how to use them effectively. 

5) Counter-Strategies

This is a somewhat more advanced strategy and I am not sure it's super easy to implement since it assumes the AI has enough knowledge of the player's actions, but maybe its makes sense to have a difficulty above Humankind (like Deity or so :D) where the AI would get full knowledge, making them able to implement this strategy so that they can compete even with stronger players who are able to make use of most game mechanics. The strategy is to analyze the weak-point of a player and effectively counter it. For example a player who has a lot of influence and is very rapidly expanding will definitely outgrow the opponent in the long term. But since they don't have many cities but mostly unattached outposts, they have very weak unit production. So they are unable to defend themselves against military aggression. So if the AI sees that the human player is playing on rapid expansion they should at least consider to fight them militarily to steal away that large un-managed territory. In my games the AI very rarely did this, and if they tried, they had not prioritized military enough to be effective with this (see point 2).

The-Cat-o-Nine-Tales wrote:

• How did the AI evolve throughout the game? Did it feel equally challenging at the start and end of your game? Does increasing the difficulty increase the challenge as much as you expect?
• Does the AI commit any grave strategic missteps? What is the AI doing and what do you think it should do? (Please provide saves if you can.)

For my feedbacks here, I mostly play on Metropolis Difficulty or on normal game and tend to play in more peaceful style. So this would be sharing my experience as more of a casual gamer who generally likes building pretty cities or a city builder who focus more on economic aspect than military.

1. AIs not prioritise building up their cities or economy much

Based on my own observation out of many normal gameplays, I noticed that AIs of any regular empires do not prioritise building up their cities even in early part of the game. I allied with a couple of other AIs to gain vision sharing with them. It was wierd to see their territories to be mostly empty or void of any improvements in their lands.

Out of my curisiousity, I asked some people in Discord about this particular issue. It appears that this does not apply to just Metropolis game difficulty but also all level of difficulties.

Since AIs do not build up their economy or their cities in this case, they are most likely to not pose much of a threat. This is especially true when playing peacefully. Not to mention if AIs focus too much on building up their military rather than their economy counter part when preparing for war, these moves can potentially cripple them more in the long run.

I do not expect AIs to be very smart like human players when it comes to exploiting or taking full advantage of certain features like:

- Attach/Detach mechanic
- Outpost relocation
- City merging
- Dedicated region for city specialisation
- Chaining up Wonders/Holysites for maximum effects

However, what I would like to see from AIs is they at least should prioritise building up their cities first in order to gain more resources from their lands. As many 4x players know, having strong economy base is often a good idea.

If anything, AIs of Independent People seem to take more initiative to build up despite their limited lifespan.


This is what I would like to see or expect from AIs in regalar empires.

2. AIs pregress to next Era too fast and instantly when obtain 7 Stars

This is also partly linked to fast game pace issue that a fair number of players complained about. Whenever AI opponents get their 7 stars enough to progress to next Eras, they ALWAYS do it instantly. This kind of moves lose the AIs potential or opportunities to earn more fame points than human players. Not to mention that when all AIs progress to next Era whenever they can, this can give players of any difficulty levels a sense of game pace being very fast.

Personally, I think it is fine to see a few AI opponents advance to next Era instantly which is to simulate some of them being hasty or taking initiative, but NOT WHEN ALL OF THEM ALWAYS DO IT REGARDLESS OF DIFFICULTY LEVELS OR THEIR PERSONALITIES.

Anyway, I would like to ask from Devs about AI improvement is to:

- Change AI behaviours concerning Era Progression
- AI planning how they play their game in order to earn Fame Stars and Fame pts according to difficulty levels
- AI behaving more based on their personilities

Additionally, I would like to add that you can also increase the number of required Fame Stars in order to advance to next Era (Not just 7 stars), which should have direct effects on how AIs behave. I will cover more about this issue of game pace in the other Feedbacks thread.


Thank you for reading.

Well.. I think that the AI is the main issue with this game. Right now, it is pretty bad. I understand that since game mecanics are not yet permanent, it is not worth it to really make the AI use them very well, but if this stays like this, I have serious concerns about replayability. On my first game, I played on metropolis difficulty and I won. It is no necessarily a problem, but I knew I had won at early medieval because I already almost had one era of advance. This is way too early for a normal AI (I will assume here that since metropolis is the default difficulty, it is the normal difficulty, with AI getting exactly the same boni as the player) againt a player which had never played this game and made a lot of mistakes. For my second game, I played on Humankind difficulty and the AI was way better. I finished 3rd but 1) I was getting out of control in production and was currently crushing the AI that bullied me from the beginning of the game, so if the game continued further, chances are I would eventually overtake them and 2) this was my second game ever, I am not supposed to be this close to the strongest AI the game can offer. On my third game, I went down to metropolis again, because I wanted to know how easy it would be to make a world conquest in normal difficulty. I had not the time to finish this game, but I think I won it as early as ancient era, because:

Finally, I saw on this forum that a lot of people won consistently against Humankind difficulty AI. The problem is that if the AI is not good enough and I get into auto-win mode, I will probably play a few games to try different strategies and then I will get bored of winning and only play in multiplayer (ie not very often) despite the game qualities. So this is something that must be

adressed, at the very least by adding more difficulties, where AI gets more buffs. But even then, improving the unbuffed AI would also be very cool. I haven't play enough to know exactly what it does wrong, but I noticed some things:

First, it begins the game badly in the neolithic era. I don't really know what it does, but even doing all that, I was not the last to enter ancient era:

Also, as other people mentionned, it does not stay in this era very long (an by that I don't mean that it is fast to reach its first star, but rather that once it got its first star, it quickly moves one). I don't know if it really moves on as soon as it can, but if it does it should take its time to profit of the exponential tribe growth, unless pressured because other players already entered the next era (this is also true for the following era, usually staying longer in the era allow you to boost significantly the fame you gain: what I usually do is that I move on to the next era when a) Another player is close (6 stars) to entering the next era and I absolutely want one culture or b) I feel like other era stars will be too difficult to complete (expansionnist and militaristic when I do not plan to go to war)). But anyway, this bad starts makes a few things difficult to the AI in the ancient era. First, it does not have enough scouts to convert them into population and boost their starting city. It also has less scouts than the player, so it will leave it open for an attack or unable to defend if the player really wants to take them down. It will also consider itself weaker, making it easier for the player to refuse AI's demand and overall impacting negatively AI agressivity and diplomacy.

 Another thing I spotted is that the AI is very bad at religion and influence. Religion is not that impacting: they only loose the possibility to have tenets and construct holy sites (since in normal difficulty I never saw an AI construct the holy site before I do, this is not true on Humakind difficulty however)(Not having holy sites can be very penalizing when holy sites give a hundred science and another player have them all). I think influence could be one of the main factors that make them lose the game: first, it is quickly under the player's influence (I do not really know them, I was never influenced, but it is probably not that impactful). More impactful is the fact that it shows their influence production is lacking, and the fact that influence is crucial in the early game to obtain territories, construct cities and attach outpost: in two word, to develop the basis of your empire. And guess what: AI seems unable to extend:

These screenshots are from my first game: look at the number of territories these 3 AI have, compared to me. Special mention to Green to still not having attached the beautiful territory it could have for itself, while I am about to enter the Early Modern era. The same scenario repeated on my third game, I do not really know about the second, I didn't explored at all. If the AI does not have the influence to acquire territories or cities, it is not that surprising that they fall behind in other stats too.

Beside these two points, I don't really know what could be the problem with economic AI.

Now, about the war AI. There is one big mistake that comes to my mind:

Here, you can see an AI that decided to attack my scout with another scout alone. The problem here is I can win the battle without doing aything: I stay on my flag and just wait for the scout to attack. It does not have the manpower to finish me (it even loses more health than me because of the bonus to defend). So he just retreats. There are actually two mistakes here: first, if you plan to attack with your scouts, you should group them: one scout alone will not accomplish anything.

The other, bigger mistake is to attack when you can not win, or even worse when you will be crushed (once or twice brown sended one unit against three of my own). The first reason not to do this is because you won't do anything: you will just lose more health than your enemy. If you want to weaen an enemy to kill him later, just group the two armies and kill him in one battle, it decreases the chances to fail (because the ennemy have healed, because he retreats when he is attacked for the second time....). So yeah, if it does not give you any advantage, why do this? Even worse, it has a big impact on war support. First, the AI loses 8 war support an I gain 8 because I won the battle. Secondly, I can make a demand to have 100 gold in compensation. If the AI accepts, I won 100 gold. If it refuses (and it is our first clash so neither of us had demands on the other), I just have to wait a little, and I will have 100 war support. It both makes the AI very susceptible to an attack and makes it feel weaker than it actually is, which can have effects on the way it behaves diplomatically. So the AI should definitely avoid engage in battles it cannot win. That does not mean it must only engage in battles it is sure to win (it would probably be frustrating for the player), but at least it needs to check if it has enough of an advantage (in troops or terrain) to be able to win if the player does not act at all (that way it ensures that the player must act strategically in order to win).

In term of strategy on the battlefield, the AI is way better than I would have though. However, there are still some things it is not very good at.

First, something very minor and probably difficult to reproduce because it only happened once: in a big battle, I took the AI flag but I was loosing to its raw elephantine power (this was on Humankind difficulty). It was later on the battle from this photo:

However, I won the battle simply because the AI killed the unit on the flag using its ranged units, and forgot to retake the flag! That was good for me but I shouldn't have won this. However, most of the time it retakes the flag properly so I don't know what happened.

What is worse and much more reproducable is that the AI tend to send its units outside of fortifications to attack ranged units, as in the following screenshot:

Needless to say, this unit got destroyed much more easily on this terrain (river, instead of fortifications. I understand the idea behind this: the AI does not want a player to shoot down its units one by one without reacting, but right now it is more detrimental to the AI than standing still.

This one is a bit of a tricky one though, I would probably adapt my strategy to the exact situation, but I will try to break down what I would do:

- If the ennemy has melee units and is actively using them, keep my units inside the wall to make it melee units suffer more than mine.

- If the ennemy does not use its melee units an just shoot, here it becomes complicated: I would try to estimate how many units I would lose if I went out of the city to try and defeat the ennemy outside, and I will try to estimate what losses I could do to the ennemy. I will compare this to what kind of losses I would expect from staying inside the city for the remaining rounds. If I think that going outside is more favourable, I would send all my units to create a real threat to all its army. If I think it is better to stay in the city, I will defend the walls accessible to enemy units, retreat elsewhere and cycle my units to avoid them to die. If neither of this option is acceptable (because of difficult terrain for example), I would consider letting the ennemy enter the city to fight it on a better terrain.

* Cases where I would typically stay inside the city: enemy forces are overwhelming, I also have range units, the terrain outside is not favorable (rivers, low ground,chokepoints...), there is only a few rounds reamaining

* Cases where I would typically go outside of the city: I greatly outnumber the enemy, the ennemy has few melee units and its ranged units are on flat terrain so easily accessible.

I know this feedback is negative, but I hope it can help you to improve the game

I think a unique feature would be the ability to create custom AI with archetypes, & also maybe invest say 50 points into seven possible sub-features of 0/14 that will define the AI "Character".

1. Diplomacy (Chance too seek peace, or to accept or counter deals of fruition to self and if loyalty & trade check is high then increase fruition to allies.)
2. War (Aggressiveness, if low loyalty increase commitment of military and strategy/ tactics dependant on specialized military archetypes.)
3. Expansion & Research (More emphasis on seeking fertile food, industry and science tiles to maximize on building economy and researching.)
4. Adaption (Adapt to changes on revealed map, such as new player encounter, borders near own or where near planned expansion or event's. Plus ability for specialized adaptability archetype too effect need for more aggression or diplomacy due to allies destroyed or weakened economy etc.)
5. Loyalty (To allies and treaties, repeat trade or if enemies likelyhood of skirmishes or attacks/sieges etc.)
6. Espionage (Surprise war, saboteur, stolen research or fermented uprising, line of sight in enemy cities due to use of spies etc.)
7. Trade/Mercantilism (Likelihood of new trades, treaties regarding trade or seeking resources to increase state power/affluence & sharing with allies.)