i cant find out how we can test our bots in privat.

I think most people are just uploading their bots to the server and testing them there. Also, you can copy/paste a game log into your local console to test a specific game against your bot locally.

But if you really want to run complete games from scratch on your own PC, you can download their build engine code and run it locally. See this post: http://theaigames.com/discussions/warlight-ai-challenge/52f4a5d65d203c41298b456b/conquest-engine-source-code/1/show

This can be useful if you want to run different versions of your bot against each other.

Edited 2014-03-18 10:45:16

Thanks :D

There is no fee to enter and the site has no ads. Where is this prize money coming from?

If I knew how to write an AI that were on par with top human warlight players, I would be a millionaire, probably even a billionaire. It's much more complex a task than what you would think.

That aside, writing an AI that can beat other AI's is much easier.

What exactly is the template theyre using? I am trying to create one but i cant work out:

Is it multiattack?
can we attack by %
can we transfer or attack only?
how is move order determined each turn?
the fog is "normal" right?
its played on an exact copy of "small earth" right?
the distribution mode is a custom distribution with 2 warlods per continent or "superbonus" as they call them?
3 starting picks with 2 armies on each with 5 base income per turn?
the luck is 75% weighted random?

I cant find out how to run my bot privately, I trying this on my MAC. I dont know what to do when I'm supposed to run the file. I tried to download the Conquest-engine but I dont know how to run it, help!

No multi-attack, no attack by percentage, no attack/transfer only, random move order, normal fog, no luck modifier (i.e. 100%), map is an exact copy of Small Earth, 2 starting armies on every territory with 5 base income, 3 picks for each player.

The distribution is a tricky one though, since it's essentially double random warlords (two random territories in each bonus.) This can't be done in WarLight exactly. You can either make a few custom scenario for a specific set-ups and test those, or you can just use full distribution or warlords to approximate it.

Also, terminology differs: a WarLight "territory" is called a "region", a "bonus" is called a "super region", and a "turn" is called a "round"

Edited 2014-03-18 12:57:39

I tried to download the Conquest-engine but I dont know how to run it, help!

EDIT: See this post http://theaigames.com/discussions/warlight-ai-challenge/531d912a5d203c346d8b45a4/compiling-java-source/1/show

Edited 2014-03-18 11:28:26

This looks like a brilliant idea, I can't wait to give this my best shot in C#. Thanks for the shout out to the contest on the home page Fizzer. Also, if anyone needs any help in C# feel free to ask me, but if you're learning the basics it would be best to Google "c# tutorials", or some similar concept, and start from there.

Hello, as bananashake asked. We got the answer on how to run the AI local. My only question is how do we run it? I am new to java :P

cool.

no mention of deadlines though.

I believe finals begin at the end of May. I think there will be more competitions after that, too, probably with different settings.

Edited 2014-03-18 12:38:48

If anyone is curious how well WarLight's AI would fare in this competition, the answer is... really badly.

But I uploaded it anyway, just for fun, with just one modification. You can see it play here: http://theaigames.com/competitions/warlight-ai-challenge/game-log/crappybot/

The only modification I made was to make it prefer smaller bonuses. This helped, but its preference to expand over fight still causes it to lose a lot. It ends up wasting armies on neutrals when it should be defending/attacking the opponent.

This AI wasn't written to be competitive in a 1v1, so I knew it would perform badly. Thus, it's named crappybot. I may change it again in the future, so this post is only accurate as of v11.

i'm divided in several points: trying something with c# or asking for javascript. doing procedural, functional or implementing a behavior tree model. and if i have enough free time to do any of it properly. had been talking with dead piggy to do a collaboration, which would definitely help on the strategic value, but still wonder what might be the best technical approach. i did a few procedural ai's in the past, i know their limitations when it comes to higher complexity. i'm a bit rusty on functional programming, but it would be the best to translate real scenario strategy into an algorythm imho. and i been working with behavior trees a lot, but am not quite sure they are the best for strategic kind of ai's and how easy i can get my gui with a behavior tree lib for one of those languages supported working, a behavior tree might grow in complexity very fast. decisions decisions.

Edited 2014-03-18 13:17:24

"behavior conditional on the opponent's identity are prohibited"
Derp, there goes my idea :(

will the winning AI become the AI for warlight? if that is possible and it is not op then i think it should

EDIT: i would like to pair up with some one (I would like to right pseudo code as i understand coding very well but don't have the skills to actually write the code yet) if we were to win (yeah right) i would not need any of the money but i just want to have the experience as i am trying out making a AI for something else (using a game engine called gamesalad) if any one wants to work with me just shoot me some mail!

EDIT 2: if i were to win a prize above 20 bucks could i chose a membership instead?

EDIT 3: i have come up with a basic list of what a AI MUST have

Basic needs of the AI

Territory priority. Needs an algorithm to assign priorities for territories taking into account how many territories left in a bonuses how much it is worth and enemies in that bonus. (Maybe take into account the chances of a enemy being there if possible?

Territory picks. Need to figure out priorities to deploy in

Deployment. Figure out a algorithm that takes into account territory priorities and enemies.

Movement. Using priorities move to the highest priority territory then next then next etc.

Non stupidity… Just what it sounds like.

Edited 2014-03-18 15:47:14

Looks fun, but as a novice to intermediate level Python user, I am not sure I have the time to do this. I will have to take a closer look.

what language would you say is the easiest to learn for a complete noob in programming? If i wanted to focus on learning a language to have a bot ready in 2 month, what language would that be?

both java and c# (the 2 example bots provided) are decent to learn programming with.
you'll still need to read a book though, atleast to get the gist of variables/functions/cycles and basic sorting algorithms. anyone can learn programing in a month or two imho, but you will need to dedicate daily to it. and you're going up against guys in university departments dedicated to AI and computing, so don't get your hopes too high on getting a high rank. but i'm sure it'll still be fun to see your code battling it out and a great opportunity to learn coding (it's much easier when you have a concrete objective in mind), so if you're serious on learning i can only advise you to give it a go.

Edited 2014-03-18 16:20:34

ok thx. I'm not a complete noob in programming, i did some basic coding in VB.net a few years ago and i just wrote a program in MATLAB for warlight 2 weeks ago. I don't know any real programming language tho, but i thought i could use the lines i wrote in matlab to help me with the warlight bot maybe. I know i can't compete with real programmers, but i have nothing to do right know (looking for a job, if you need a mechanical engineer and your company is either in south america or south-est asia, pm me :D :D). Anyway, i've always been wanting to learn programming so it sounds like a good time to do it. I think i'll give java a try

Edited 2014-03-18 16:36:38