Username:
B
I
U
S
"
url
img
#
code
sup
sub
font
size
color
smiley
embarassed
thumbsup
happy
Huh?
Angry
Roll Eyes
Undecided
Lips Sealed
Kiss
Cry
Grin
Wink
Tongue
Shocked
Cheesy
Smiley
Sad
page  <- 1234567891011 -> <- 1 .. 3 .. 11 ->
--
--
List results:
Search options:
Use \ before commas in usernames
Gombill, you do indeed sound like the perfect person to do this run.

I figure a couple of obvious questions about the AI are:

* How does it decide what to build? I've read stories about the AI screwing up and producing mass scout cavalry armies. Are there ways to manipulate the AI into building crap units?
* How does it decide where to attack? Obviously it'd be helpful on some of the harder missions to be able to get the AI to consider a lone house in a distant corner of the map as your 'base' and go attack that, letting you slip into their base while their army is away without worrying about your actual base being levelled.
Did some preliminary runs today, got the first Tutorial mission in 1:29 (real time) and Joan of Arc 1 in 5:10 (game time).  I believe JOA1 is nearly perfect, maybe up to 3 more seconds can be saved unless I'm entirely missing a better strategy somewhere.

The tutorial missions don't give you an endgame screen, kind of unfortunate as far as running is concerned because we don't know what the game time is.  I suppose I could just turn on the in-game timer.  Also unfortunate that it doesn't appear the game speed can be changed from Normal in the tutorial campaign.

@Patashu - the AOE2 AI system was very much meant to be played with.  I think the game CD comes with a sample AI (named SampleAI) which has lots of comments inside it about how to do certain things.  It also comes with a large, 130 (I think) page document about everything you could ever want to know about AI making, called CPSB.  I could attach my own AI here if anyone would be interested in having it, and I'll give a description of how to use it.

@ExplodingCabbage - It is my understanding that each faction in the campaign missions has its own AI file attached to it so they will all act differently (of course), but there are hard-coded things about the AI which will never change, like the attacking scheme. 

Two AI faults which are definitely exploitable:
1) The AI will choose a target when it attacks and 'parade' his army to this target, often ignoring attacks against the army until it gets there.  I still don't know exactly how the AI chooses targets for this attack, but he seems to like castles, markets, and stables most of all.  Setting an army of archers out in front of your town can often pick off 75% of the attack because it has a town-based target it's parading towards.

2) If you have a wall (which you start off with quite often in the campaigns), the AI will attack it somewhat intelligently UNLESS the wall has a hole in it.  In this case, the AI tends to ignore the wall and funnel all units through the small hole.  It's often possible to intentionally make a hole and create a 'maze' in such a fashion, lining the maze with towers, and the enemy units will traverse the maze without attempting to bring the walls down (as long as the maze has a 'finish' inside your town)

There are many other AI exploits, too many to list here.
Any chance of some WIP videos? Perhaps we could optimize a few strats.
Edit history:
Gombill: 2010-06-13 11:09:56 pm
Joan of Arc 2 in 8:20 of game time (4:10 real time). 

It's too slow to wait for the castle age to get battering rams, the trick is that orange does not get murder holes quickly enough and I can simply take the castle with knights and the two heroes.  For some reason if you attack his south-most castle he doesn't even respond with his military so there's nothing to it.  There's enough villagers walking in and out of their gate so you never have to attack their wall to get in.

This was just meant to be a practice run but turned out pretty well for me.  Maybe sub-8 is possible, but it would be extremely difficult and require tons of micro.

I will post a video of this (and my other current runs) as soon as I have some time.

Edit:  Looks like someone on youtube has JOA2 in 2:42 real time.  Uses pretty much the same strategy but better execution.
Quote from Gombill:
Did some preliminary runs today, got the first Tutorial mission in 1:29 (real time) and Joan of Arc 1 in 5:10 (game time).  I believe JOA1 is nearly perfect, maybe up to 3 more seconds can be saved unless I'm entirely missing a better strategy somewhere.


Would be interesting to see vids. Dunno if you know anything about video encoding; if you don't, then I recommend downloading Anri-Chan. Its Low Quality option should churn out a compressed video very quickly at a very small size which you can then attach to a post here. That's probably the least painful way of sharing WIP vids.

Quote:
Edit:  Looks like someone on youtube has JOA2 in 2:42 real time.  Uses pretty much the same strategy but better execution.


Hmm, sugagsr. I'm guessing that's the same suga who did the phase 1 HL2:Ep1 run, seeing as he's got HL2 vids in his channel. Btw, check his uploads, he's got runs of all the tutorial missions and Joan 1 as well, though his Joan 1 is a few seconds slower than yours, I believe.

(Edited to fix grammar.)
defying gravity
Wow, these videos by suga made for an awesome watch. I had no idea the ai was that abuseable. I noticed some things when I played the game, like he is not evading when you attack ground instead of rightclicking, or it following a single unit around, but that was just brillant. With these techniques you can possibly bring down the levels in which you really have to build up an army and fight to -rough guess- below 10, severely reducing the time investment of that project (and the run time of course :D)
Still its quite a big project of course, I really hope you got what it takes and we can prepare for some awesome videos (I'm really looking forward to some offensive town hall or castle action :D)
Edit history:
Gombill: 2010-07-29 01:59:28 pm
I improved Joan 1 a bit more, it's now 5:07 (game time) but still has a few flaws.

De Metz gets caught in the red cavalry, small detail wasting no more than 1 game second
Joan won't get on the boat right away, biggest mistake of the run, takes 3 game seconds
Joan stands under one of the gates and won't move, costs about 1 game second

The run is attached for inspection...

EDIT:  Removed the attached run because it was obsoleted.
Just had a pretty near perfect run of JOA1.  De Metz was a tiny bit slow getting through the 'turn back' trigger but otherwise perfect.

Time of 5:02 (game time)

I guess I jumped too soon on the previous post.  The new one is attached.

Attachment:
The run looks sweet but the FRAPS frame counter at the bottom right is pretty ugly. Do you reckon you could redo it without?

BTW, I was attempting to replicate suga's Wallace 5 on my unpatched AoK install but failed because of what seems to be an AI difference; the red archers came up and shot my scout as soon as it started attacking the wall. Any ideas why?
Here is why he defended with his archers against you:

Are you sure you attacked exactly the same wall piece that he did?  It makes a difference.  The way the AI responds to an attack is he summons all units within a specified constant "response distance" of the point of attack (Since I've made AIs, I actually know exactly how to change this constant) and defends with those units.  The key is to attack a wall piece that is outside the response distance of the archers.

I've beaten suga's Wallace 6 with a very nifty trick.  I build a market, go to the castle age, and build extra monasteries closer to the two far away relics so the monks don't have to walk so far.  There's some difficult details in between, of course, but I've figured them out pretty well.

His time was a 4:30 (game time) and I've achieved 4:16 so far, but I think I may even be able to get sub-4.
Edit history:
Gombill: 2010-10-08 12:27:44 am
OK, here's a fresh 3:52 (game time) for Wallace 6.  I doubt the integrity of my long-term strategy, but the relics are all temporarily in monasteries!

In principle it should be possible to do it about 10 game seconds faster, I will keep working on this.

EDIT:  Video removed because the time has been improved
Quote from Gombill:
Are you sure you attacked exactly the same wall piece that he did?  It makes a difference.


Yes, the very same corner piece.

Quote:
The way the AI responds to an attack is he summons all units within a specified constant "response distance" of the point of attack (Since I've made AIs, I actually know exactly how to change this constant) and defends with those units.  The key is to attack a wall piece that is outside the response distance of the archers.


In that case, what's especially odd, watching his reaction using the 'marco' and 'polo' cheats to remove the fog of war, is that he sends his most distant pair of archers to attack my scout first, then whenever they overtake a subsequent archer, that archer will start moving in to attack too (such that they all arrive at the same time). It must be a version difference. I'm running on a fresh, unpatched AoK install but with a NoCD crack, so I don't know what version I'm effectively on (I kept the original uncracked game exe so when I get back home and have access to the CDs again, I can be sure I'm running an unpatched game). My guess is that it's an AI change brought in in Conquerors (which I assume you have installed) that is allowing your tactics to work.

BTW I'll provide a video of the archer behaviour I'm describing if you'd like.

Quote:
I've beaten suga's Wallace 6 with a very nifty trick.  I build a market, go to the castle age, and build extra monasteries closer to the two far away relics so the monks don't have to walk so far.  There's some difficult details in between, of course, but I've figured them out pretty well.

His time was a 4:30 (game time) and I've achieved 4:16 so far, but I think I may even be able to get sub-4.


Sweet, I was wondering if anything like this was possible. In a similar vein, I was wondering whether on Wallace 5 it could be faster to delete your Town Centre and build a new one right outside his wall.
Edit history:
Gombill: 2010-06-15 02:30:38 pm
I've already tried deleting and building a new one near his town, right now using that strategy I have a 6:45 that had plenty of flaws but I don't see improving by 1 1/2 minutes to make it better than suga's.

The way I've been doing it is first attacking the south-most wall piece with my scout and getting all the archers to commit to that side of the base.  They are then outside the "response distance" of the northmost wall piece and won't respond to attacks on it.  Doing the opposite (attacking north first) hasn't worked for me because two of the archers are too far away to respond to that and stay at the tower.
Just beat Hamorad's Genghis Khan 2 and almost halved his time (from 5:00 game time to 3:02).  Here's the vid.  Maybe a few more seconds of improvement possible, but as you can see by the minimap it's already pretty well a straight line from start to finish.




Attachment:
Edit history:
Gombill: 2010-11-09 03:02:25 am
Gombill: 2010-07-29 02:00:30 pm
3 New Runs to post (not done before by people posting on this site)

I continue with the 'easy' missions, possible to beat basically with what you start with and not involving any base building.

Saladin 1 in 8:28 (game time)
Saladin 3 in 7:16
Barbarossa 3 in 4:29

Small flaws present in each, these are not finished products but will probably represent my basic strategy for each one.

I think I can get as low as 4:10 on Barbarossa 3 by converting the lumber camp, then deleting it and moving through it.  In practice, though, it has attracted too much attention from the blue army.

EDIT:  Removed the videos because they suck now.
These are fun to watch, hope you finish the run as the game really deserves one.
Edit history:
Gombill: 2011-12-06 06:39:53 pm
Joan of Arc 6 in 7:52 (game time)

This was a pretty entertaining one to make.  Who would have guessed 80% of my casualties in the mission would be self-inflicted?

EDIT:  Video obsoleted.
jesus christ you raped the AI...
Quote from Gombill:
Joan of Arc 6 in 7:52 (game time)

This was a pretty entertaining one to make.  Who would have guessed 80% of my casualties in the mission would be self-inflicted?


I have 2 questions: Why did you kill your own troops? And how did that trade cart survive with 1 hp? It looked like it clearly got hit.
I killed my own troops because I needed to get my population down below 20.  This way I only needed to build 4 houses to get the transport ship, instead of 10 houses.

And I don't know how the trade cart survived.  Sometimes arrows randomly miss, I guess I just got lucky (the Thumb Ring upgrade makes arrows never miss, but the britons can't get this upgrade).  Maybe the archers started firing at the cavalry instead.
we have lift off
Great so far but just one question. How are you chosing which levels to do? Are you going to go through them in order or just in your own order until they are all done?
ridd3r, by the looks of it he's trying to get the easier levels out of the way first, and leave the harder (and perhaps more boring?) levels that'll require proper base building and combat to the end.

By the way, Gombill, great stuff, I've enjoyed every video in this thread I've watched. Smiley

Any ideas from anyone on why the AI difference between my game and suga's/Gombill's?
@  ridd3r:  A few days ago I went through all the levels and prioritized them based on how easy I thought the runs would be to accomplish.  Now I'm going through them in that order.  Once I eliminate all the easy ones I'll probably just hit the remaining ones in order.  I'm going to try to get through all the Age of Kings campaigns and see if a submission will be accepted on those separately before running the Conquerors levels.

Some new runs (no videos at the moment):

Barbarossa 6 in 9:05 (game time)
Genghis Khan 1 in 11:30
Genghis Khan 5 in 29:55
Joan of Arc 6 in 7:21

Hamorad had already done Barbarossa 6 in 13:20 by knocking down yellow's wall with mangonels while staying allies, but I turned yellow to enemy and got it down much faster with paladins while keeping yellow's units off me.  This saved 4 minutes without really revising any of hamorad's other strategies.

Genghis Khan 1 is definitely improvable, but I'm moving on for the time being.

The above Genghis Khan 5 run was just a first practice run, I will probably have it under 20 game minutes by the time I'm done with it.  It's necessary to get to the imperial age from a start of just 4 villagers and get enough stone for 3 castles to beat this level, so it will be pretty long in any case.

I realized I could improve the Joan of Arc 6 run by not bothering with purple's wall and just setting up the trebuchet from outside his town.  I'm confident I can get sub-7 with enough tries at it; I had one promising run land on the shore with the trade cart at 6:00 precisely, but it got killed by the archers. 

@Cabbage:  I tried attacking that corner piece you talked about and the archers came after me.  My only idea at this point is that maybe the AI being on Standard difficulty had something to do with it in suga's video.  After that I have no clue (I'm updated to newest patches).
Cabbage:

I set the AI to easy and the archers on Wallace 5 didn't respond to me, no matter where on the wall I attacked.  So that's the reason.

It doesn't appear at first that you can set the AI difficulty on the Wallace campaign (there's no dropdown menu like the other campaigns) but it just defaults to the difficulty of the last game you played elsewhere in Age of Empires. 
Interesting. Are all your Wallace videos on Hard or were some of them done before you knew about this?

Edit: Checked suga's video again. The text at the top-left clearly claims the difficulty is 'standard', unless I'm misunderstanding it. Is this wrong?