(This is a remake of the original post. I apologise for spamming but the restrictions on new accounts forced me to redo the thread.)
With the constant inflation of player elo it’s hard to keep track of the skill level of different elo ranges. So to have some form of baseline for comparison it would be cool to know:
What’s your 1v1 DE matchmaking elo on DE and can you beat the DE extreme ai in over 50% of 1v1 arabia games with or without the use of ai exploits like tower rush and douching? Doing other rush strategies does not count as exploits. NOTE THAT THESE ARE THE ONLY SETTINGS FOR WHICH THE POLL IS VALID. IF YOU DON’T PLAY THESE CRITERIA JUST CHOSE THE “I DON’T PLAY RANKED / I DON’T PLAY ARABIA” OPTION
What’s your matchmaking elo and can you beat the DE extreme ai in over 50% of 1v1 arabia games
I’m less than 700 and lose with or without exploits.
I specify in the description of the poll that only tower rush and douching count as exploits so thats the answer to your question. The reason only these to are counting is that those to are strategies all ai’s are inharently bad against while other types of early agression is not as much a general weakness of ai’s but something that the DE ai (which is based on the Promi ai) is particularly weak against.
I’m well aware that the DE ai is much stronger late game but that is for another pole to cover.
This right here is the reason I didn’t express “exploit” as an option in the poll, instead asking whether the player needed a “specific strategy” or could otherwise beat it “generally.”
The AI is more randomized than you’d expect at face value, but it’s still hard code. If you find a strategy that works against them, it’ll always work against them. Further, the definition of Exploit, as mentioned, is extremely vague. Certainly the inca tower rush is a very unique strategy but I’d have a hard time using a loaded term like that. Against the AI you might call it that, but you can call a lot of things exploits against the AI.
You consider tower rushing as an exploit but not spreading archers into the base in groups of 1’s? Especially when beating them up to feudal by two vills and having the archers in their base before the archery range is down? Be reasonable. It’s standard play to us but when the AI isn’t programmed to expect it, you’ve exploited a hole in the AI’s capabilities.
That’s why I use the term “specific strategy.” It asks the player if they’ve found a way they can beat the AI, and use that, whatever it is. I can beat the AI generally and no unique caveats apply. That’s a better way to phrase it to reach a reasonable answer as to the strength of players in respect to the AI.
I think the DE AI would rank somewhere around 1050-1100 on the ladder, most players around or above that range should have little to no trouble taking it down in a 1v1 Arabia with standard settings.
Of course, like previously mentioned, team games are a whole different story, mostly because they tend to go later and the AIs tend to have better macro in late game, and I think they are pretty good at slinging resources to each other when needed.
You misunderstand my intentions for this pole. I always wanted to know at what elo players could beat the ai with using any strategy.
The only reason I excluded tower rush and douche was that those strategies abuses a part of ai behavior that is hard coded. I used the term “exploit” to make the poll options short and easy to read, but I see now this was a mistake as people mistake this to mean exploits in general or don’t read the full poll description. I gues I may have to remake the poll again.
To be entirely honest, I thought the devs took steps to improve the AI against tower rushes, and I haven’t tried to do it in a long time (I don’t have a build order down for it or anything) so I really don’t know how abusive it still is to this point. I just know that if you know even a semi-decent fast feudal rush build the AI is totally useless and I don’t really see how trushing is much worse than that.
You can certainly do alot in terms of anti trush scripting and most scripters has done so. The DE ai could still get much better at this. However the teoretical skill cealing of an ai is capped when it comes to anti trush in a way that is not the case to nearly the same degree against other strategies. A good example of this is the Barbarian ai. It’s basicly better than the DE ai at everything and can handle any rush strategy at a lvl similar to a 1000-1100’ish player except trush and douche. (On UP, not DE) The fact that ai’s dont automatically detect if an object (tower) is inacceable due to being walled in combined with the way villagers gets retasked when hit by a tower or tc just makes it so much harder to script around.
It’s not that good. Especially in the late-game you would expect it to be able to have a perfect eco and spam from all buildings, but it gets ruined by dumb decisions. Example: you have an army of paladins and halberdiers, the AI has imp camels + crossbows. Sounds like an appropriate composition on the AI’s part, right? But if you hide your paladins for literally just two minutes and just kill some camels the AI will decide to change its army and waste its time and res teching into elite ele archers.
It does answer tower rushes, by rushing it down with tons of villagers. But this all falls apart when you have military. Especially obvious when you donjon rush, the AI keeps its villagers dancing below the TC like in the good ol’ times because it wishes to send them all at once, but serjeants are too spoopy scary. Eventually it will gather the courage to do so after like 15 minutes, but it’s pretty much too late by then.
Combining the results so far from aoeezone and the official forums the results are:
: < 900 elo:
3/8= 37,5 % claim to win
: 900-1000 elo
7/9= 77,8 % claim to win (100 % with trush)
: > 1000 elo
125/131= 95,4 % claim to win
First of all I really wanna stress that the margin of error in these findings is so large that any conclusions drawn here are speculative at best. I think the only way to get proper data would be to individually seek out players of the relevant elo to get a proper sample size, as these players seem to be fairly rare on both forums.
With that being said, so far it seems like the elo needed to consistently beat the DE extreme ai on 1v1 arabia without trushing or douching is around the lower 900’s. It seems plausable that the number could be higher among players that doesn’t go for early aggresion.
The best experiment would be an undercover AI in ranked. So you dont have any idea you are playing against the AI. This way you make sure you cant really exploit its weaknesses. This also limited the unconsciously change in playstyle because you play against the AI. Also this way you will get easy access to a proper sample size containing data of multiple players.
I think 800-900 elo players should be able to beat the extreme AI pretty consistently. But if you dont know it the AI, then its ratings might go up to 900-1100, i could image. This claim also means that many 800-900 elo players will can have a little higher rating if they play a bit more aggressive.
I have beaten extreme AI twice. Both involved some cheese so I chose with exploits. First time I wanted the achievement so I set up an Islands battle and set the AI to Aztecs while I selected Portuguese.
The second time was just an all out Bulgarian men at arms rush to see if it would work.