ais are made the replace the gap of having a teammate or player lost to the real world but ais are terrible their programming should be reexamined and check.why will an ai deploy 9 and attack wit that 9 a neutral when it has an enemy he could use it against i hav seen cases where ai are good dat is due to the fact they largely outincome the player.
Solutions:
perhaps somehow ais should be made to tink like ladder players randomly should pick a ladder player will choose
dat is all i can think of put i wish warlight will fix there ais if they are fixed we might not even need surrendered or booted players taken over or replaced or share territories ideas circulating

We don't want AIs to be better than the average player. If that happens, people will just turn into AI on purpose to win a game.

i didn't say they should be better but it will be cool in practice to adjust difficulty like noob average pro
so that will make pratice suitable of ur level.but in non pratice ai should be as good as average players

Creating an AI is surprisingly difficult; it might well be one of the hardest problems in computer science. Just look at IBM, it took them years to create a chess AI good enough to defeat the best human player. Then it took them even longer to beat the best human Jeopardy players.
Don't forget that the WL AI is a one-person job. And not exactly full-time at that (there's much more to WL than just its AI). If you think you can do better, the APIs should offer all relevant information about a game in a machine-readable format. Note that I have no idea whether or not you'd need Fizzer's permission to run a bot, check with him if you're really interested. You should make it use an account with a clearly marked name (and even then, probably only in unranked games).

One last thing to keep in mind: the AI is "static", it's one piece of code which very rarely changes. So the slightest weakness will sooner or later be found and ruthlessly exploited by human players... *without* the AI learning from it, so it'll keep repeating the same mistakes over and over again.
There's such a weakness in the current AI for instance. If you have a territory with X army on it, bordering on an AI territory with more than 2X armies on it, which only has one connection to an enemy territory (namely, yours) it will always attack, without additional reinforcements. So, just deploy enough reinforcements to withstand its attack and (as delayed as possible) issue a counter-attack. Success is near-guaranteed.

ps. I have absolutely no idea what your proposed solution is... Ehm, I can't read it... :(

interesting rvw tnx for the info

Wow

whatever happened to the google AI challange? there was some talk like 6 months ago about warlight being a candidate for it

|> whatever happened to the google AI challange? there was some talk like 6 months ago about warlight being a candidate for it

The 2011 GAIC had already been chosen at that point. It's still a candidate for the 2012 one, but I'm not sure how the decision process works. I imagine it will be decided soon. There's a thread about it here: http://forums.aichallenge.org/viewtopic.php?f=21&t=2229

any way to beef up the AI just a little? for example, when the AI is bordering an enemy and a bunch of neutrals, just force the AI to pile its troops next to the enemy? and then, once it has about double the number of troops ("Tip: It is usually best to attack the enemy with at least twice as many troops as the enemy has."), have it attack? i think the AI's decision to spread when the enemy is near, rather than defend or attack, is the biggest weakness of the AI.

or, make 2 types of AIs:

- if a human is booted or surrenders, use the better AI
- if the game begins with the 'player' as an AI, use the current one

the AI expansion problem, is only a problem early game, and slightly mid-game... lack of defense is another problem that they have.. problem in prioritizing bonuses..

the main problem, IMHO, is the cause of the secondary problem.. the AI has no foresight of enemy movements..
if it has a bonus threatened by a 2 a 4 and a 6, then that means 1X.6+3X.6+5X.6=level of defense needed..
same problem with attack.. since you can generally predict when it will attack, since it tends to use the tip GUI mentioned.. that means you can dump a huge army in there, and then ruin their offense because they didn't foresee the possibility of you dropping it there, and *because of that didn't increase armies*.

http://warlight.net/MultiPlayer.aspx?GameID=1923998 how could let dat ai get this strong
ps i became ai after i took canada

is english your first language?

The one thing that I have seen the AI doing that I really don't understand, is only moving one direction.

I've seen it on bigger maps mostly, especially Quad Earth, but it seems the AI will pick a particular direction/front and continually use troops on that front, but none others. If the AI's backside is to an ally who can keep them safe, it works pretty well. If the AI's backside is being attacked, it will frequently not even use a single troop in defense. I took the entirety of North America from an AI in such a game without seeing a single defensive troop used.

That seems like it's a bug, but maybe the AI's behavior in intentionally coded in such a way that this happens on certain maps?

english is my second and i leave in england

@The Duke of Ben:
- Play on a map with at least 4 or 5 neutrals per territory
- Move up "behind" the AI
- As long as neutrals exist, the AI will ignore you and keep attacking neutrals instead
- Let the AI kill all the neutrals (replacing 4 or 5 neutral armies with just one of its own armies)
- You can mop up the entire map, only ever attacking territories with one army on them

I think the AI is pretty good at expanding and not too bad at fighting once all the neutrals are gone. But in between, when you've already made contact, but there's also neutrals left, you definitely have an unfair advantage. (Note, the only time I ever tried anything close to AI... well, let's just say it didn't win any prizes. :( )

I've played a game with my roommate where the AI would concentrate completely on me \*mid-late game* and let him eat up his insides.. when I am clearly not a neutral

The AI has many predictable issues, and they're quite bad. However, it's not that much of a priority. Plus, the single player levels are balanced around their current skill level (especially the last few). You already have a lower chance of winning most levels without restarting at bad starting spots (except 6). Level 3 is "merely" a 1v1v1v1, but if every AI player was equally skilled you generally have a 25% chance to win each time you play. Level 5 versus skilled players should generally mean you won't win, except if you start very luckily and both teams don't focus on you. Level 6 should clearly be impossible, 1v5.

I guess what I'm saying is that I agree with difficulty levels.

ai are very strong when there are no neutrals around and they have superior number. that s why i like to pass/airlift to them in some cases.

the main problem is that they dont defend territory with 1 army if u have 5+. so they are really weak in certain cases.

another problem is that they spread even on your path and sometime they attack u to take bonus.....

but to be onest they are indeed very good, being balanced on the single player