How does auto sheep scouting works? Is the auto everything that smart and sends the sheeps back to the TC so you can kill the sheeps in time?
I just saw this. Spirit of the law made some good points about the AI, but i dont really agree with all. He thinks extreme AI is just below 1000 elo. If i have a look at the BO of the AI, it is clearly not great. This is something i also noticed and why i personally disagree with 1125 elo for extreme AI. Players around that level have much better BOs.
One of the things i like was the idea of the some undercover AI into the ranked queue.
He said that early on, but I think he was just referring to how they lured the boar, which was a bit clumsy, ending up with it outside the TC. I got the impression he thought the castle and imperial times were pretty decent, and I didn’t notice him comment on what ELO it might be later on.
He also commented about the AI spotting what you’re doing and altering the units it’s making, and I was pretty sure I saw it doing that against me last night, more on Hard than Moderate. I changed to Britons for the archer frush, thinking I might as well use a civ that will give my archers a benefit in Castle age. Against Moderate, this worked fine, but against Hard, there were about 10 skirms waiting for me when my first group of 4 archers got there. I figured the AI might be assuming that Britons will make archers, so I changed to Lithuanians, but still made archers, and on Hard, what happened was that his Scout was nearby when my first archer came out, and again, there were skirms waiting for me. So I thought it seemed likely he had produced skirms as a result of seeing that I was making archers.
I’m still finding Hard pretty tough. I’ve just about hung on till Castle age in some games, i.e. had a similar score at that stage, but then his score surges ahead of mine, even if I have 3 TCs and try to keep them all making vills. To me, it seems harder than these players here, who were around 1050:
The first one clicks up to Feudal at 08:54, with 20 vills, no military made, and no loom. That can be done at 7:05 with a perfect build, so that seems pretty weak to me. I think I’ve been clicking up earlier than that with 22 vills and loom, and when I get to the Hard AI’s base, he has a lot more military than me. Having said that, the AI is probably weak in some other ways, so it’s hard for me to say if those players would beat it or not, but in terms of what they have made before clicking up to Feudal, and the time they do that, they seem worse than even Hard AI.
I think that while the ai has a lot it could improve on, there’s more to someone’s elo rating than just how well they execute a build order to castle age (or maybe imp for fast imp).
I think there’s lots of gameplay where it could top a 1000-1100 player. (I haven’t been above 1100 myself yet, I think I can if I get time to practice more, I’m fairly close.)
For instance it has godlike micro, microing every individual unit making mangonels extremely ineffective vs archers skirms etc, even if it attacks pointless buildings at times.
It’s also amazing at keeping unit production up especially in castle to imperial age. It can flood units and lots of strange unit combos that can be hard to deal with at times.
However it can be very silly when it comes to choosing where to take wood/gold/stone etc, which can leave it open to raids, although again once it’s eco is rolling it can replace vils very fast. At a lower skill level it’s easy to forget to keep making vils when fighting with military or keep up military production when focusing on eco etc.
1 Like
It’s true that the AI is strong for certain things. However, it’s very weak for others.
IF you’re lower ELO, you have to strategically play on its weaknesses rather than its strengths, and you can still succesfully beat it.
For example, if it’s good at microing against mangonels (i haven’t tried mangonels vs the AI), then just skip mangonels altogether and go for easier units. I find the AI doesn’t know how to deal / micro against a large mass of knights. Knights don’t require any micro for the human player, only patrol move, so you don’t have to be very high ELO to beat the Extreme AI with knights.
The AI is decent at reacting to small harass, so unless you have good micro it’s hard to actually harass the AI’s villagers. If you’re lower ELO and don’t have much multi-tasking or micro yet, I’d suggest skipping trying to harass the AI’s eco and going for a more eco/macro approach.
It’s not really good at attacking either, as said before, the AI will only attack buildings if you put them in front of your base.
Boom on 3 TCs. Learn to make villagers non-stop on your 3 TCs (using the “select all Town Centers” hotkey) and have them go directly on woodlines. Pick villagers from woodlines to make new farms. Once you reach 80+ villagers, add 4 to 8 stables and mass knights with +2/+2 upgrade. Then bring them all to the AI base and patrol click in the middle. Rally point all your stables to the AI’s base and produce knights non-stop. At the same time, produce villagers non-stop on your 3 TC’s.
Learn to do all of this quicker and quicker. Once you’re fast enough in your macro to be able to do that successfully everytime against he Extreme AI, practice multi-tasking and micro by adding eco harass on top of it (multi-tasking). On top of it and NOT replacing it, macro should have priority, imo, when you’re learning the game. The harass you can practice : rushing with scouts early, rushing with men-at-arms early, harassing with a few knights in early castle age, etc… while maintaining Perfect macro behind this (non-stop villagers with no idle TC, keep making new farms, houses, go for extra mines). Imo that’s the kind of basic practice a new player can do against the Extreme AI to improve mechanical skills, and probably one of the most efficient to improve fast as a beginner.
I disagree with his statement about the castle times. i dont really think 16.xx is impressive. I always hit 15.xx and have hit even some 14.xx (but 14.xx is pretty rare for me). SOTL also said that less vills when hitting castle isnt always great. I think it is only bad if you dont have a good follow up. First step in almost every Arena game is a fight for relics. Both AIs were pretty late to the fight. If he played against a human with elo 1050-1100, he would had no relics at all. Also i think his aggression was pretty late as well. So i wasnt impressed at all.
I dont know, but yesterday i played against the AI:
First game: I went archers, he went skirms.
Second game: I went M@A, he went skirms.
Third game: I went vill rushing + towers, he didnt make any military.
To me it seems like the AI just love skirms. I have to play more against the AI to really make statements about his unit choice. It could be he can react to your units, but it also can be he is just programmed to make skirms in most games.
These players are not really 1050. They have more like a peak rating aroudn 1050. I have had a look at aoe2.net for there ratings. I would consider these players as 950-1050 players, with a average rating of 1000. I feel the BO of the players is worse compared to AI, but they made military much better and just react better to harrassment than the AI. I cant really tell what would be harder. I think i beat both players pretty easily. From the two players, Detaa is probably the best and he is probably more close to 1050. I think he can beat the Extreme AI. The other player will have a much more difficult time.
Note that there is a pretty big skill gap between 1000 and 1150. If you are 1000 rated, you are just an average players. 1150 is already top 30%. So around 20% of the players is around 1000 and 1150. The skill gap between those players can be pretty big.
It really depends on your strategy. There are some strats for certain civs where you go up with 16 population (15 vills) for a scout rush. Those build orders are pretty fragile. So you need to do damage early on. 20 vills and no loom and clicking up at 8:54 is pretty late to me. So at that part they are worse than the AI.
To me the AI seems like a late game power house, but is early on pretty weak. The difference between 1000 and 1100 is for me mostly applying early pressure. That is just the weak spot of players around 1000. I can understand players around 1000 have issues with the AI, because they cant do enough damage early on. Players around 1100 can do this, so they dont really have to face the late game AI. They already beat the AI before he hits late game.
I got my first win against the AI on Hard. It was mostly sheer good luck. I played 4 games. 2 of them, he made skirms and wiped out my archers. 1 of them he didn’t make skirms, and instead built a tower, and it was near a lumber camp, so I couldn’t stop him taking that wood. But this one, as you can see, he again made a tower instead of skirms, but it was terribly positioned and didn’t protect any resources. So I was able to dot my archers about, set to stand ground, and I gradually built an advantage until he resigned. I was a bit worried when he managed to get the siege workshop up, but he never managed to produce anything with it.
One thing that would be a nice extra feature is the ability to set the stance of units coming out of each production building. It took up most of my attention just grabbing each archer before it ran at the tower, to set it to stand ground and get it to an appropriate position. If I could have set the building to make all units coming out of it start off as stand ground it would have been helpful. I had enough resources to produce from 2 ranges instead of 1, and I can see now I should have had more on gold and fewer on wood, but I’m not good enough yet at switching and performing quick actions in different places.
Anyway, my point is that you can’t judge the AI’s ELO by how it performs in one game, as it can randomly perform much better or worse. Thinking about it, I think the tower’s location in both games was around that distance from his town centre in the direction of my town centre, so if that is how the placement is chosen, it’s pure luck whether it protects resources or not.
Okay, I jumped straight from Hard to Extreme and won first time. It took a bit longer, it dealt with the rush a bit better than on Hard, notably it had multiple villagers rush my first archers, so you can see them all die in the timeline. But for some reason, they seemed to stop doing it, maybe it was when both blacksmith upgrades came in. It was slow to respond with skirms, and later in the game they were losing against my upgraded crossbowmen in a straight one on one fight anyway. You can see on the timeline when I killed its scout, I don’t know how important that was, but it couldn’t make a replacement as it didn’t have any stables. Again you can see some poor positioning. It got a castle up, but it wasn’t really protecting much that the tower wasn’t already protecting, and both were protecting stone, which obviously is less important than protecting the other resources. And the 2nd TC that it got up was also not really protecting anything new, and I think one archer stood near the lumber camp was all it took to kill every villager as they came out of there. I played better than the previous game, using what I’d learned, and building 2 forward archery ranges, forward siege, and a forward TC when it was clear that area of the map wasn’t under threat.
My feeling is that these early pressure strategies, while they’re valid in that they would also cause problems for a human if you got there early enough, are kind of cheese strategies against the AI in that you’re getting into a narrow time window before its strengths start to come into play. I think it wouldn’t take much for it to perform a lot better against these kind of strategies:
-
I don’t know how it chooses when to have a few villagers rush at an isolated military, vs when not to, but it would do better to do more of it IMO.
-
It would do much better if towers, castles and additional TCs were placed to protect the most important resources.
-
It seems to prioritise eliminating forward threats. If it had sent some military around me to counter-attack my base, it could have caused me huge problems, as I did absolutely nothing to defend my base at any point.
-
For the open maps, it should really have a “early feudal rush into moderately fast castle” type strategy, as that is what humans seem to do on these maps, and it’s how we’re beating the AI. E.g. it’s what Hera demonstrates here:
https://www.youtube.com/watch?v=8tEnjsB1y34
The AI is going up to Feudal just a bit too late, creating the window that we’re exploiting to get an early attack in before it’s ready.
Thanks for the help in this thread in getting me to my first win against the Extreme AI!
1 Like
I dont see the tower?! Is the dot on the mini map outside your screen the tower? I have seen making such useless towers as well.
Of course 1 games dont say it all. Even humans play at the same level are not that consistent. I played some games against the AI, but non of them were really difficult.
In all of them i use early aggression. This is just equal to how humans play a map like Arabia. Players around 1050 just also go for rushes. They also have to benefit to wall and thus defend their base with walls too. The AI just dont wall (because he is bad at this). The AI seems to have pretty bad build orders to me (this is also show in the vid of Spirit of the Law, i was not impressed by the build order) and if you have a look at for example the lumbercamp placements, they are also pretty bad. This are repeating points of the AI.
On both maps, the tower is to the left, and a little below, its TC.
I watched the replay of the Extreme game a couple of times, and need to change some of what I said before, though some remains the same.
I looked at the stats of my archers, and they were already fully upgraded Feudal when some of them were attacked by multiple villagers early on. My early aggression actually achieved very little, and I thought the same was true of Hera’s game - okay, he killed a few villagers, but nothing that should have caused the AI to refrain from producing any military. In Hera’s game, the AI also produced a tower, and it seems to refrain from making military when it makes a tower. Anyway, there is no obvious reason why it stopped killing my archers by rushing at them with multiple villagers. If it had continued to do that, I think it would have beaten me, and there was no difference in the stats of my archers between when it did it vs when it stopped doing it.
The tower placement wasn’t as bad as I thought, it was protecting both wood and stone. The TC protects many farms, so that just left his gold. I won largely by virtue of one archer stood at the gold mine set to stand ground. That one archer caused the AI to not obtain any more gold from that point to the end of the game. It didn’t have enough gold to even make one monk to collect relics, as it only had 84 gold. So that is why it never got to Imperial age. I wasn’t aware during the game of how critical that one element was.
I’ve played far harder games on easier levels, because it has made military instead of a tower. I think it just needs some small adjustments to become much harder to beat. It needs to stop regarding a single tower as replacing the need for military. It went up to Feudal at 24 villagers, so that is too late for quick Feudal aggression, but too early for Fast Castle. It clearly had a plan of getting to Castle age fairly quickly, then making a castle, as it was collecting the stone for that well in advance. It appeared to have no intention of making any military till that Castle was up. So it would have been better off going for a true Fast Castle, but better off still going for a faster Feudal then making some military.
My own play was quite poor. There was a period of TC idle time before reaching Feudal that I wasn’t aware of at the time, and it’s not obvious watching the replay why it happened. I think what happened is I hurriedly built a house at 13/15, and was looking at the food to get ready to make the villagers drop the boar food off if the food didn’t hit 50 before the house went up, i.e. I behaved as if I was housed when I wasn’t, I could have got the villagers to drop the boar food off and made another villager while the house was going up. Doh! I clicked up to Feudal around 8:20 with 20 vills and loom, so quite a bit better than the low ELO video I posted above, around 50 secs of lost time by that point.
If any devs read this and want the replay, I’m happy to provide it. I feel the ease with which the Extreme AI can be beaten like this is mostly bugs, I know it can play to a much higher standard than this, as I’ve seen it play much better on lower difficulty levels. It mainly needs to:
- Protect all resources.
- Don’t randomly stop killing lone military by rushing at them with multiple villagers.
- Don’t see a tower as replacing the need for military.
- On open maps, create early military before reaching Castle age.
- Counter-attack rather than focusing 100% on forward elements (this might be much harder to do, I’m not sure if it ever does this on any difficulty levels at the moment).
- Don’t spam single military towards a group of enemy military. Allow the numbers to mass until you have enough to win the fight, then bring them all forward together. Related to this - don’t build military production buildings close to enemy military where they can kill the villagers and stop the building going up, or just stand by the building and kill every unit as it’s produced. Put the buildings up out of sight of the enemy, so they go up, and then you can mass units at the building and send them out as a group.
1 Like
Now i see. both Towers are better placed then the tower is saw once. If you show the AI base this way, it would be out of your window.
Is someone into building your own AI? I know it is possible. Maybe i will give it try, just for fun.
I saw i didnt play extreme AI, but hardest for a second time (i didnt get the achievement). Now i played a game against the extreme AI. It felt just a bit harder, but still not a real struggle.
For some reason the AI went for a 26 pop (+loom) feudal age. That is just too late compared to a human. I also notice the AI love skirms. It seems to always go for skirms as defense, even if it makes no sense at all. Or better to say: It goes for 1 skirms as all it military. I dont know how you counter Incan vills and towers with just 1 skirms. There is no logic at all. Also the AI tried to send all his vills at once to rush down a tower. Than if you try to fight with your forward vills, it send all back to work and so it will go back and forth. When you have to towers within each range, you just won the game. The AI had no clue at all.
Because i was stupid and lost my scout to their TC, i was tower rushing without really knowing the place of his important resources. In the end i just made a barrack and just 1 eagles was enough to make sure all vills went back and forth between lumbercamp and TC. 1 eagles just idled around 10-20 villagers of the Extreme AI. No human around 1050-1100 will react this stupid.
What I don’t understand is, I’ve played some games on Hard where I’ve made 3 archers as fast as I can, but when they reach the AI’s base, they meet 10+ skirms and get wiped out. It’s basically impossible for me to win when the game starts like that, as it’s just too good at booming when not under any threat. I think it’s basically a bug that it sometimes chooses to build a tower, and doesn’t then mass the skirms like it does when it doesn’t build a tower. Those games are the only ones I’ve won against Hard upwards.
I think i played now 4-5 games against the AI, but i think they just build a tower in just one game. In the other games there were no towers at all. In the game with tower, they had army. In other games they had much less army.
To me the AI really loves skirms. He seems to go for skirms in every game. Every strategy which beats skirms easily is probably an easy way to win against the AI on Arabia.
I would even try some 1v2 against the extreme AI, but i do think i will loose that match up. I think i can only punish one AI in time, the other will get a free boom and crush me.
Try Extreme on Arena where you can’t get in there early and cause the same problems for the AI as you can on Arabia. I’m not really getting close to beating the Extreme AI there.
I will give it a try i think. I feel like they can boom pretty good, so their late game will be solid. Not able to early on makes the Extreme AI harder to beat, is my hope.
So i played on Arena against the Extreme AI. It was … again pretty easy.
This time i went for a krepost rush as Bulgarians. Just to do something different. My execution was bad, like i failed to do a normal FC build, but still managed to get Castle 1 minute before the AI.
Based on this Arena game i think the Extreme AI is rated around 1050.
I was, as usual, too slow with what I did on Arena. I got to Imperial well before the AI, but was trying to use the Knights line, and the resources needed to research all the blacksmith upgrades make everything too slow. I see now why people use Turks so they can go straight to hand cannoneers and bombard cannons. I found a video where TheViper did a Turks Fast Imp on Arena, so was able to compare the timings against the AI in SotL’s video:
AI
Feudal 26 vills 10:00 12:10 (2:10)
Castle 29 vills 13:28 16:08 (2:40)
Imperial 37 vills 20:09 23:19 (3:10)
TheViper
Feudal 26 vills 09:36 11:46
Loom immediately after reaching Feudal
Castle 27 vills 12:36 15:16
Imperial 30 vills 17:10 20:20
So the AI isn’t too far behind reaching Castle, but then loses 2 more minutes between Castle and Imperial. I’ll try the Turks Fast Imp myself.
I managed to beat the Extreme AI on Arena using a similar strategy to Arabia. I used Goths to get free Loom, and once I had about 10 villagers, I sent one of them up with my scout to start breaking through the AI’s wall. The AI didn’t attack me while doing that. Once inside, I built a barracks and archery range, and used the same strategy as before of using archers set to stand ground as towers, essentially. Again, I was able to get archers into positions where they prevented all gold mining, so the AI never got to Imperial Age. He kept making the odd skirm, but my archers had enough upgrades to deal with them. I was able to completely confine his activities to within his walls, and built a 2nd TC, more archery ranges, and a siege workshop just outside his walls, and eventually swiftly destroyed his entire base with something like 20 fully upgraded crossbows and 6 bombard cannons. Ultimately it all came down to the same weakness in the AI as before, that a small number of well placed stand ground archers can almost completely shut down all the AI’s activities. I didn’t feel that my play was any sort of high standard, it would be possible to refine what I did to do it much better. A human would probably defeat the strategy very easily by e.g. waiting till my villager and scout were 80% through the wall, then just building a house behind that bit.
1 Like
Did that really work against the AI? No way you get in with just 1 vill and 1 scout and that you could build a barrack and range inside his walls. That will not happen against a human at all, i hope. That seems really like an winning by exploits to me.
I hope @Chesqin will read this thread too, so he can fix the AI. The AI dont just build no walls, but also dont seems to care about his current walls if he start with walls. This seems pretty much easily exploitable for a human.