New Hardest AI Difficulty

I know almost nothing about team games, I’ve never played them, but it’s not supposed that when a player logs out they are replaced by the AI, if a 4 player game the AI replaces two players which AI do you want to replace them, the new or the old?

The solutions is so simple.

Just allow us to play any AI smartness level without cheats. So that’s its possible to play the hardest with no cheat

And add another setting “gather bonus” that would be either 100% 125% 150% or 200% gather rate.

this way it scales a lot better if you want to start improving progressively. Right now its impossible to play the smartest ai without cheats.


The problem is that making a better AI requires more work, which can be occupied by other demands of the community.

The vast majority of your answers go towards the same conclusion: The AI gathering rate boost should be an additional option to be added to the Hardest AI!

This seems totally reasonable in terms of development. As many say, creating an actually more clever AI isn’t super easy, and it will slow down more important development. Maybe this gathering rate can be kept BUT it needs to be both: mentioned clearly to players and adjustable/togglable.


The gathering rate bonus should also be able to be given to human players, like AoE 2 DE’s handicap can, so someone can make a fairer game against a friend who is a different skill level, or if the friends want to play as a team against the AI, the bonus can be used to bring a weaker player up to the right standard to be able to properly participate in the game.


Yeah i like the idea of being able to add the gather bonus to anyone. (except ranked /quick play)

If you want to give yourself 200% gather bonus against easy AI, why not…

1 Like

It’s definetly very hard. French vs English, swarmed with 45 Longbowmen in 7 minutes. That was nuts.


in the mod-advanced settings it is done

Yea all of this belongs in custom lobby options IMO. How are they going to make this work in quickplay?

Suddenly you divide the co-op vs hardest AI playerbase into 100%, 125%, 150%, 175%, 200% categories?

Probably the easiest solution for them is to reduce the bonus to 150% and separately call it “insane” ai. Then it’s just 1 new option in quickplay.

Then “hardest” doesn’t make sense anymore since it’s not the hardest, so just delete hard and rename hardest “hard”.

You’re implying that this much negative feedback must be because ‘fake accounts’?

I never go on these forums. I made an account specifically to voice how much I hate this change. That’s how strongly I feel about it and I’m sure a lot of people are in the same boat


A big issue with how AI bonuses are implemented is how often they are all front loaded to give the AI a huge early game advantage. Civilisation 6 does this too by giving the AI free starting cities and extra units which gives them a massive early game advantage but if the player is able to survive and keep pace in this initial stage they basically win as the AI doesnt get anything else as the game advances.

I like @EricGonzalezM idea of having the gather rate be incremental which might make the difficulty ‘smoother’

But ultimately as has already been said this needs to be seperated from the Hardest AI so Hardest AI remains the best non-cheating AI. Its a shame its currently impossible to test the Hardest AI with the improvements minus the cheats.

1 Like

I did not say that. Although there’s been a huge issue with that here.

It’s also possible that there are people creating accounts just because they care about this one issue so much, even though they’ve never posted before.

Or maybe most likely, a combination of the two.

What other improvements? I use to play against Hard and the former Hardest a lot and with the PUP, Hard behaves exactly as it always have. Don’t get me wrong, there have been improvements, but in past releases. I see nothing new in this PUP.

I have never ever seen the AI properly taking down a keep. Ever. It always sends all of its units to torch it right under the boiling oil until it realizes the whole army is going to disappear, so it retreats very late with 5% of the units it started with. When the AI brings up siege, it never utilizes it or focuses on the wrong buildings. Want to exploit how dumb it is? Build mills and houses and other cheap stuff around your keeps and watch the AI prioritize those.

Beasty has been casting Hardest AI matches were he evidences all the shortcomings and dumb logic the AI has had forever (he gets a scout to run circles around one of his outposts and AI spearmen chasing it down and dying). So forgive me if I don’t believe that besides the 2x resource multiplier, the AI has seen no other tangible changes.

Wait… there was one change: The AI now builds outposts in the middle of nowhere, protecting absolutely nothing, but hey, that’s something new even though it insistently sends all its villagers to rebuild them and die in the process.

1 Like

I can’t link to them from mobile because the AoE page links to Steam, which then opens the native app, but uh everything under the resource rate changes? Here’s the start of it; there’s a good page and a half or so:


I saw that. Mine was mostly a rhetoric question as a lot of those changes, although welcome, are corner cases or apply to specific situations/environments like water maps.

There was a lot mentioned in those notes about AI prioritizing siege more than before, but in practice even if the AI has 6 mangonels, 3 trebs and 2 cannons, they are not utilized and mostly parade around the map waiting to poof out of existence, or as I said, they fire on houses while a keep is sitting pretty destroying everything else.

There is still a long, long way to go and adding a resource multiplier clearly didn’t please anyone. I would be surprised if devs don’t backtrack and leave this as a separate handicap option.

To me a good AI should even have the same clicking/hotkey delays as a human player has, gradually becoming lower as difficulty increases (as a human player would) and even make a few mistakes here and there. I see that even in Easy, villagers react instantaneously, landmarks are placed just exactly when the resource caps are met, etc. A beginner human player would take seconds to make decisions.

A good AI should outsmart you in equal conditions, not by cheating. That’s why chess AI is so respected, because it can’t make extra moves or have twice as many pieces. AI in a lot of driving games has become very good too, often making the same mistakes you make (crashing, taking the wrong line) but being ruthless when conditions are right. Their cars are not twice as fast and barring the shitty rubber band effect some games still use, you can either be years ahead or years behind.

1 Like

chess is arguably way more programmable than an RTS like AOE4. But there are things they can do, like making a target priority list that has keeps above houses, and monks/scouts/mehters above infantry, etc. Program them to ‘see’ towers they’re pathing into. Or optimizing their build orders would be easy since every civ has 1-3 best builds that can separate gold from platinum players in 6 minutes. The builds AI uses automatically puts them sub-gold difficulty.

1 Like


I’m not quite sure what you’re trying to ask of me. My talking of the patch notes being overshadowed was no more than that. I’m not blaming players for it. I get why it’s happening. But I think it’s a shame, and if people think there genuinely weren’t other changes, I like showing them that there are. I have nothing to say to “i know there were other changes even though i think they’re not impactful”.

Also, I’ve already (repeatedly) said my position is that they should separate resource scaling and the difficulty levels. Like, I support that.

That said, chess is a completely different game. Solving a video game is nothing like solving chess. This is a key reason in why so many strategy game AIs cheat. The probability space alone is a huge thing. But that’s a fun tangent for another thread!

Of course, the chess comparison is not 1:1 and it was only to showcase that you can have good, non-cheating AI. I recognize the exponential increase in difficulty to get even near that in something as complex as an RTS instead of starting off with fixed squares and fixed moves.

Definitely not asking anything from you. I’m taking to the wider audience while acknowledging some of the improvements that have been made to the AI that you mentioned, but even though adding the 2x resource multiplier was probably just a few clicks of effort for devs, time would’ve better spent just reprioritizing siege targeting. Just by having the risk of properly utilized siege would make harder AIs much more challenging, whereas now I know I can defend absolutely anything with a few outposts or a well-placed keep and completely ignore that area for the rest of the match.

It’s good that we all agree the handicap should be separate. I am happy that chances are high it becomes an option in the future so you can mix-and-match continuously improving AI with a bit more of a challenge if you so wish.

The previous hardest mode is not much of a looker. It definitely needed improvements. Sadly, the cheating improvement is not the way to go. This cheating option should disappear in favor of AoE II’s handicap system.

1 Like

Quickplay co-op vs AI is a tricky one. By far the largest number of players is those queuing vs the hardest AI rather than lower difficulties. There was a time when I was playing a co-op game against AI most days, and because of the difficulty of finding a game against anything below hardest, a lot of people were queuing against hardest when there was no way they could beat it 1v1. Some of these people would have 15 villagers and no military by 10 minutes then quit when the AI attacked them. After a while, though, the general standard of players improved to the point where it was actually too easy to win these games, and I haven’t played co-op for a while.

The point of all that is that people will probably queue against the hardest available difficulty no matter what, so it needs to remain an appropriate standard for that. I don’t think having 100, 125, 150, 175, 200% would work, as the hardest available difficulty would be too hard, but player numbers would be low for anything easier, due to the number of options.

I think maybe rename hardest to very hard, and a new “extreme (125%)” might work for co-op, it would need to be tried, and the win rate of players monitored. I’d expect that most people playing co-op want to be able to win most games, but have it feel like it wasn’t stupidly easy. I think it does need just one more difficulty above the current hardest, rather than any greater number of options, as we have to expect the new hardest available difficulty will become the most commonly chosen level. Alternatively an incremental boost as the game progresses, as suggested by another poster, could be a good option for the single new hardest available difficulty for co-op, but I’d suggest starting off conservatively, err on the side of making it “too easy”, and gradually increase it over time if player feedback and win rate supports the view that it remains too easy. As soon as the win rate for games where no human has quit early on starts to fall significantly below 100%, I’d say that’s probably too hard for what most people are looking for from co-op games.