What is the ELO of AoE II DE AI?

Age of Empires II: DE II - Discussion
1

I’ve found two sources of information on this elsewhere. The first seems to give a consensus view that Extreme is in the 1000-1100 ELO range.

This source gives a very different view:

Easiest 1100
Standard 1275
Moderate 1400
Hard 1525
Hardest 1650
Extreme 1725

I certainly can’t believe the second source is correct, as I’ve watched games with 600 ELO players, and it seems to me they would easily beat the AI on Easiest, and probably also on Standard. However, it’s rating the DE AI in terms of HD ELO not DE ELO, do they differ significantly?

The first source, though, rating DE Extreme at 1000-1100 ELO - as the median player’s DE ELO was around 1050 last time I looked, that is saying that the median ranked player can beat Extreme AI around half of the time. I’m still at the stage where I don’t even beat Moderate DE AI 100% of the time, so this would suggest I’m extremely bad at the game, as I’m playing 3 difficulty levels below the standard of the median ranked player. My feeling is the DE Extreme AI is better than 1050, because when Viper tries to beat 4 of them, I don’t think he is able to take out the first one as easily as he would take out a 1050 ELO human. That reddit thread is from 4 months ago, so maybe the Extreme AI has changed since then? I know there were issues where it could be massively easier to beat if you selected a population limit other than 200, but I think those have been fixed now, so maybe people were beating an essentially bugged AI at the time by using a different population limit?

1 Like
2

I guess it is very map dependent.

Extreme AI is much better than is HD counterpart. It can comfortably execute a drush, perfectly avoiding TC fire by micro as a pro would do. Deer pushing, boar luring and some attempt to quick walk. Now it also chooses the right counter to your army composition. The old AI was much weaker to scouts since it did not train a lot of spearmans. Now it always fights your scouts with spearman, archers with skirms. It can also comfortably defend from a smush on arena by switching to light cav. Nothing that hd AI handled.

Macro is also very good, minimizing idle TC.

Still pretty weak to towers, and vulnerable (despite less weak) to early aggression.

What is really missing is decision making. Much worse than a noob imo, but I can imagine that this is very difficult to code. So it presses the wrong spot, places castles in horrible locations, completely incapable of handling pure water maps.

Imo, several people playing online are very far from beating the extreme AI in 1v1 Arabia without tower rushing it.

In the last patches, it has improved a lot, so a post old 4 months is a bit outdated.

2 Likes
3

The second guy is talking about HD elo, not DE elo. The latter wasn’t really a thing when he wrote it.

1 Like
4

Thanks, so I just did a search, and it seems to be 600 different, so it would be
Easiest 500
Standard 675
Moderate 800
Hard 925
Hardest 1050
Extreme 1125

That seems more like it, so it’s just estimating Extreme as around 75 better than the Reddit thread. Looks like I definitely need to improve some more to even be at an average level for online play.

4 Likes
5

Both have different ratings and are not really comparable. Two main issues:

  1. HD player base is different to DE player base. The overall level at HD is lower than the overal level at DE. This means the average DE player beats the average HD player.
  2. Default rating at HD was 1600, default rating at DE is 1000.

This means you cant really subtract 600 for HD to get your DE level.

I am not really into playing against AIs. I did played some games against the AI to get the monthly event stuff. Based on that i think moderate is around 600-700 elo. I have used higher diffuculties. Based on this list i need to play against hardest or extreme AI to have them play against an AI of my own level. Maybe i will give this a try today.

I still dont know why they didnt add a AI as strong as the Barbarian AI of TheMax into the game. I think that AI will be around 1700-1800 Elo at highest difficulty, i think. It strategy choiche is also much more like a human. So it will prepare players much more for multiplayer.

I also like the idea of the guy at reddit of having just some secret bot in the ranked games, who is just the AI. Just to see how strong the AI really is. You can also measure the changes in Elo after adjusting the AI every month. Did the AI really improve? I would really love a very strong AI for practising purposes.

Just another related post of my:

I like such idea.

You dont play online? You know you can play ranked games with a match making system? It will just match you against equally skilled opponents. You first go through around 10 place matches and after that 10 games, you almost always get great matches. I would suggest just joining the MM queue and play some nice games!

If you need tips: Best tip is to practise some Build orders. I think if you know the basics of the build orders, you are already around 900-1000, which is almost the average player online players.

1 Like
6

In my opinion, it doesn’t really make sense to give an ELO rating to the AI.

The reason is that it only takes a few games against the AI to find out exploits and strategies that will do fine against it most of the time. The AI always does the same things in the same way. It only has a few build orders to pick from and you can guess what build order it will use depending on the CIV it has.

And even a low rated player can just practice and follow a build order that works 99% (or 100%) of the time vs the Extreme AI.

In other words : the success of the Extreme AI depends on whether its opponent is familiar with the AI or not. If he is, then the AI will be easy to beat/lower ELO. If he isn’t, the AI will be harder to beat/higher ELO.

It would be interesting if they would have the Extreme AI tag into 1v1 rated matchmaking queue, but undercover with a fake name. So that people who are matched against it aren’t aware it is an AI. And we can see what ELO it stabilizes at, if any.

But for someone who’s practicing against the Extreme AI for many games in a row, the level of the AI will decrease each game because you learn how to play against it. It doesn’t really adapt.

Practicing against the Extreme AI is still very useful for all kinds of ELO though. It’s a nice opponent to practice new build orders, to practice macro/micro mechanics, speed, multi-tasking, etc…
But in terms of strategy and adaptability, you don’t learn anything useful against the AI because it’s too predictble and doesn’t adapt or do unexpected things like a human would do.

EDIT : you can actually get mechanically good by purely playing against the AI. Improve your speed, macro mechanics, multi-tasking. Perfect your build orders.
But then, you’d need to transfer those purely mechanical skills against real humans, and learn to adapt, react, and develop strategically in ways that will work against humans.

1 Like
7

I’m trying to avoid getting dragged into that. I’ve been well into the top 1% in ranked gameplay in driving games, but it becomes a bit of an endless treadmill of having to work harder and harder to keep winning. But then I watch a T90 Low ELO Legends video and can’t help but wonder if I’d beat them. As you say, it would be good if the AI played online without people knowing they’re playing against AI so its ELO were known.

I guess AoE is more about knowledge and practice than driving games, where the best people become very good very quickly. So the average ranked player has probably racked up considerable hours of learning at this stage. I am gradually improving, I was able to get gold on the early economy and fast castle Art of War challenges recently, which I couldn’t do at first.

8

This is what i meant: put the AI into the ranked 1v1 queue under a fake name. You dont know you play against the AI, so you cant really use the normal exploits against the AI. If it is undercover, than its Elo can stabilize and thus is a useful to compare different AI.

1 Like
9

It probably isn’t. After watching Hera doing some coaching videos, and I’d say it’s closer to 900-1000 elo (low end of it probably). Probably less than 1000, but then again that’s where new multiplayer players come in at which makes it hard to know what ‘1000’ means.

Hard to Hardest is probably 1000-1100 range, and maybe extreme is at the top end of this, like 11xx. I imagine people that are 1200+ are comfortable beating the extreme ai on their favourite map.

There’s been lots of updates to the ai too recently, which make it much better at dealing with cheese strategies (vs ai), towering or loosely surrounding with archers.

This seems reasonable to me to be honest. I do feel like if moderate gets going or gets a good rush in it could be higher than 800.

10

I think that would be an interesting experiment, but the result wouldn’t really be useful to anyone, in my opinion. So it would just be out of curiosity.
Because they players who wonder “which ELO/level is the AI” are those who want to play against it or practice against it. Therefore they will know they are playing against the AI and the ELO we found doesn’t apply, since it’s only an ELO obtained against people who are not aware.
I guess it would only apply for the first game, until you learn some exploits. But usually people who want to practice vs the AI will play many games against it.

11

The current Extreme AI is unpredictable and dangerous, it counters almost everything. Update me if you find the AI easy in a 1v1 Arabia standard match.

12

One exploit I found is that the Extreme AI will just attack your buildings when you put them between its base and yours. It gives you so much time to just boom behind this. Then you just overcome it with units.

For example, if you go FC and you build your Barracks/Forge or Market, between its base and yours, and he does his typical Archer/skirmishers rush, his archers/skirmishers will focus your market and barracks for 10 min.

So you just have to add 2 TCs at castle age, boom up, and mass whatever strong castle age unit behind this (upgraded knights for example) and you should be able to kill him with that most of the time.

2 Likes
13

The biggest benefits will be for the devs. They get lots of data from players who play against the AI without them knowing it is the AI and thus wanna exploit the weaknesses of the AI. This means they have better access to data to see if there changes really made sense. Did their changes really improved the AI? So it is not only about ‘what rating has the AI’. To me it seems to be a fun statistics to share with the community. I would love it if we had a very strong AI, who is beating most of the players. They can monitor the progress by looking at the Elo of the undercover AI in the random ladder.

Like i said: I havent really play against the AI, so i am not really sure.

I just play against Hard (not extreme) on Arabia. It was Ethiopians (me) vs Teutons (AI). I made many mistakes, but it felt pretty easy. I never thought i would loose the game at any moment in the game. It was mostly the question: When will the AI resign? Based on this experience the hard AI will be around 850-925 i would say. I watched the replay, but the BO of the AI is pretty bad. Just a standard 23 pop archer rush will kill the hard AI with ease.

Next time i will try hardest before i go to extreme.

1 Like
14

The ease comes down to speed of execution as well as what you do, though. So I can benchmark my speed of execution against you, what time in the game are your archers reaching the AI? How many archers? With the two blacksmith upgrades?

15

First 2 archers hit the enemy around 14.30 minutes into the game. No blacksmith upgrades, i went for 2 ranges first and added blacksmith at a later moment. I kept adding archers and was around 24.00 in castle age. Around that time the AI resigned. His largest army was 2 during the game. I think in the 25 minutes he create 3 times a skirms. He didnt really create anything else.

I also played a game at hardest: Goths (me) vs Spanish (AI). I have to admit: It didnt feel much different than just the hard AI. The AI was a bit quicker to the feudal age, his BO seems just a little bit better. So that might be the biggest difference. The AI was still pretty late with army, so i just over run the AI. This time i went full M@A. The first 5 M@A hit the enemy around 12.00 (as benchmark for you). After that i kept spamming M@A. At no moment i felt i would lose this game. Around 30.00 minutes into the game (both still in feudal age) i just take down his TC with M@A and shortly he resigned. It was no real challenge. This game was still really easy. My guess for hardest AI would be around 925-1000 Elo. For sure he isnt above 1000 Elo. 1050 elo is really to high for hardest AI.

Next time i will try Extreme AI.

16

What Civs and map? It sometimes feels like other people are playing a different game to me! After seeing Yo beat Viper with Inca villagers, I tried an Inca villager rush against Moderate AI. His TC went down around the 15 minute mark. He had about 5 skirms, and also had enough resources to sneak a replacement TC up where I couldn’t see it, so I ended up losing, as I lost more vills taking down the TC than his lost production. Maybe I could have won if I’d immediately put a couple of towers up where his TC had been standing. That was against Aztecs on Serengeti.

17

Goths vs Spanish on Arabia.

18

Inca villager rush is Villagers + Towers, isnt it?

I give it a try against the extreme AI. Again pretty easy win. This time the AI didnt had any military at all. Again an easy win. This time the AI didnt made any military at all. If i look at the current AI: It paralyzed as soon as you attack. This was even more clear during building towers. Villagers just got 1 mm back and forth at my screen. The AI clearly had no idea how to counter towers. The AI was also pretty late to feudal age. It hit feudal age at after 12 minutes. I think the AI want to FC on Arabia, because he went up with 26 pop. That is a really bad strategy. No one around 1125 goes for this strategy without some walls. This game was the quickest win.

19

Thanks, I just gave that a try with your archer frush, and I can see that my execution was extremely poor compared to yours, as I didn’t make my first archer till minute 16, but nonetheless I beat the Moderate AI without too many problems. He hit Feudal before me, and was up to Castle at 19 mins, whereas it took me till 24 mins due to all the gold I was spending on archers, but even so, my constant spamming of archers and picking off the vills and skirms, trying to stay away from TC fire, caused him to resign when the score was around 2100 vs 2600.

I’ll see if I can speed up my execution, and I’ll try the harder AI levels. I guess the difference will be quite small as the early rushes mean that the higher villager count isn’t coming into play. What ELO are you at, so I know what your timings correspond to?

20

I would say on average my rating is around 1075-1100.