The AI is yet sooo dumb

[stiv]

First I want to tell you that it is impossible to make ai script that would still be interesting

Impossible is such a strong word. What if we wrote an AI script that could write AI scripts? Or even better, an AI script that could write AI scripts that could write AI scripts?

To quote famous chess guy Boris Spassky, the advantage people have over machines is balls and cunning (I'm paraphrasing here!). This was just before a chess-playing computer kicked his ass.

But Boris is right. He was just playing the wrong game. Chess is a game of perfect information - at any given time, you can see the complete state of the game. Games like this are 'easy' - if you are a supercomputer - because you can look at all the possible possibilities.

Warzone is not like chess. You have some information (hey, that guy is shooting at me!), but you do not necessarily know other interesting things like where and how strong the rest of your opponents forces are. This makes it hard to plan and predict the future. It also makes the game interesting and surprising. Simply put, humans are much better at dealing with surprises than machines.

Bottom line: AI is hard. Good AI is even harder. And that is why AI designers cheat.

(note that we are simply talking about the theoretical problem of AI and have not mentioned the challenges of WZ's scripting langauge)

[Germanium]

if you get bored by 1 AI, why not play against 7 allied max AI's - on some maps its possible... 2v6 is more interesting cause your chances are slighly increased. I suggest 2v6 allied max ais on sandcastles with advanced bases (the other is too easy). If you have done it, make sure to tell me.

[lav_coyote25]

First I want to tell you that it is impossible to make ai script that would still be interesting

Impossible is such a strong word. What if we wrote an AI script that could write AI scripts? Or even better, an AI script that could write AI scripts that could write AI scripts?

To quote famous chess guy Boris Spassky, the advantage people have over machines is balls and cunning (I'm paraphrasing here!). This was just before a chess-playing computer kicked his ass.

But Boris is right. He was just playing the wrong game. Chess is a game of perfect information - at any given time, you can see the complete state of the game. Games like this are 'easy' - if you are a supercomputer - because you can look at all the possible possibilities.

Warzone is not like chess. You have some information (hey, that guy is shooting at me!), but you do not necessarily know other interesting things like where and how strong the rest of your opponents forces are. This makes it hard to plan and predict the future. It also makes the game interesting and surprising. Simply put, humans are much better at dealing with surprises than machines.

Bottom line: AI is hard. Good AI is even harder. And that is why AI designers cheat.

(note that we are simply talking about the theoretical problem of AI and have not mentioned the challenges of WZ's scripting langauge)

the best ai cheats like a bastard!!!! and as far as i remember the only ai i can think of - was said to never cheat... strange that, right??

[Asgorath]

I played a match on Sk-Sand-Castles-T1 against 7 Semperfi-AIs no teams on hard.

The beginning of the AI is good, it tries to grab oil and build some MG-Turrets (for attack and/or defense).
But then AI isn´t very effective in attacking at all. First 30-40 minutes the AI only attacks with 2-3 tanks always waiting 4-6 minutes until it runs the next attack.
I don´t know if the AI is busy at fighting at other places but I would like the AI to try some kind of tank-rush.
Just building 10-20 MG-tanks (or other) and give it a go

Perfectly would be if the AI wouldn´t try the same tactic in every match but vary so it gets (more) unpredictable.

[DarkCheetah]

First I want to tell you that it is impossible to make ai script that would
still be interesting. every ai plays always the same way ..
for example//if he tear down a factory it will consistently build back in
the same place no matter what happens in the database.
you certainly would not do that, right?
If you play several times using the same ai you will find his weak point,
that is the way that you win.
in order ai to have hundreds of ways it takes years for something like that..
Asked whether these documents are very large work without errors
(Not just talking about the warzone ai, its similar in most strategies).
without dependence on only the game but also on the pc, no errors would be inevitable.
and wonder if computers are stupid or the people who created them!? ..
Yet to make a long because I hope I have explained the what the problem was! ...
have you tried super ai1.5 for wz1.10??? ..
Has a very rapid production of the units.
Do not try to win, its impossible . but these will be interesting to play sometimes.
choose a smaller 4 player maps with a lot of oil.

ofcourse i'v done super ai!! i loved it! and especially winning after 2 hours of defending why dont we have it in 2.3.x series??

anyway i wanna show you how a "Insane AI" (nexus) plays like!! http://imageshack.us/photo/my-images/69 ... 00skt.jpg/ we expected a giant army... thats why i miss the cheating AI's spawning units and power

they used all of there power to rush to he farrest oil, and then they didnt had any power to build generator, epic fail?

[Black Project]

DarkCheetah, i have a strong non-cheating AI called DyDoBecomePrey v1.01 on my HD, but if you're wishing a strong cheating AI, then Genghis Khan AI is your choice.

Long time ago (last year), i was picking some of the best AI scripts and planning to release in a pack, but other things drived me off of this. Still, i can send you the AI pack if you want, just PM me and you're good to go.

Not sure if the AI works on 2.3.8, but i can guarantee those works on pre-2.3.8 releases

Regards BP

[DarkCheetah]

DarkCheetah, i have a strong non-cheating AI called DyDoBecomePrey v1.01 on my HD, but if you're wishing a strong cheating AI, then Genghis Khan AI is your choice.

Long time ago (last year), i was picking some of the best AI scripts and planning to release in a pack, but other things drived me off of this. Still, i can send you the AI pack if you want, just PM me and you're good to go.

Not sure if the AI works on 2.3.8, but i can guarantee those works on pre-2.3.8 releases

Regards BP

omg i'd love to try ur ai and eventually contact you further for any issues it has

but i'm not sure how you use a mod again, i know it has to be in mods/global...

[Black Project]

For the skirmish/coop lovers, check this thread if you are sick of the current stock AI:

viewtopic.php?f=10&t=8075

Regards BP

[JDiTo]

The AI can sometimes be challenging under circumstances where they have the upper hand on researching items (aka. you vs. 3 teamed AI) but as long as you lock down and protect most of the oil derricks on the map, the AI cannot build up enough money to launch any solid assaults on the depots.

The reason for this is that most of the AI does not understand the idea of CONSERVATION, seeing as most of the scripting reads out with oil being top priority when it becomes low, and fails to notice that sending 1 truck and 1 unit to attack a well fortified position rarely works, and even in that case, it makes sure to build one, super expensive heavy unit that succeeds at destroying mabye one defensive position of 5.

AI should check if resources are low, and attempt to gather more, but if that fails, they should check for the types of buildings around the position and build up sufficient units accordingly while building defensive positions around their base to prevent attacks from stronger players/ai.

Just my synopsis on the topic after playing many-a-match with friends vs. unteamed and teamed AI (including: dydo, nexus, semperfi)

[Tenoh]

Multiple AI vs human shouldnt be allowed to cheat with research,sick of pointlesness of it.its enough to have 7 enemy bases against you, no need to have them better tech. 30 secs and they come with vtols.. sopush the big X in the corner,frack this.

[lav_coyote25]

Multiple AI vs human shouldnt be allowed to cheat with research,sick of pointlesness of it.its enough to have 7 enemy bases against you, no need to have them better tech. 30 secs and they come with vtols.. sopush the big X in the corner,frack this.

keep in mind this site has young minds sometimes in attendance... keep it clean please. or one of the moderators will... either the way i did it, or by banning... i dont want to see this happen... so, please adhere to the rules of the forum...thanks for reading.

[NoQ]

Tenoh: the AI will not play like human anyway, and there's no way you can beat 7 humans if they're not total noobs. So do you really care what exactly are you fighting with? Just take as many enemies as you can handle and play. Or, most importantly, play against humans and see how many humans you can beat.

[zydonk]

If the AI was anyway bright, it would probably control the world (the real one) by now. While you are moving the cursor with the idea of maybe ordering a truck to do something, the AIs have most likely surveyed the total situation hundreds of times and made many many decision. One of the weakness for the AI is that many of these decisions can be contradictory - move this droid here, no move it there etc. Even so, the overall effect is that the AIs can seem purposeful and - if they for a moment gain the upper hand - can wipe you out in a remorseless way that is sometimes chilling.

Sk differs from mp in this way: to some extent, you have to play WITH the AI. Give it time to build up its base and forces, so that it will provide you with a decent challenge. Strangely, the best way to play with the AI is to let it almost win, so you are always on edge with it. Another thing is to play with maps that have an element of unpredictability: asymmetrical layouts, bases with different levels of protection, clumped wells in one place, but scattered in another. (Most recent maps are mp orientated, so everything gets leveled out.)

One way to get a good idea of what I'm talking about here is to play Pumpkin maps (which are dedicated sk maps) with the 1.10 mod. If you hang back in your assaults (but keep up with research and defense) you'll find the AI can give you some hairy moments. With practice, you'll find that you can pace the game, which can add hugely to your enjoyment. mp, on the other hand, is a frantic struggle to survive using whatever tricks you can get away with. Very Darwinian and modern, but you should remember that WZ is primarily a game - that is, a rest from kicking balls - and therefore a touch of civilization...

[test]

Sk differs from mp in this way: to some extent, you have to play WITH the AI. Give it time to build up its base and forces, so that it will provide you with a decent challenge. Strangely, the best way to play with the AI is to let it almost win, so you are always on edge with it. Another thing is to play with maps that have an element of unpredictability: asymmetrical layouts, bases with different levels of protection, clumped wells in one place, but scattered in another. (Most recent maps are mp orientated, so everything gets leveled out.)

One way to get a good idea of what I'm talking about here is to play Pumpkin maps (which are dedicated sk maps) with the 1.10 mod. If you hang back in your assaults (but keep up with research and defense) you'll find the AI can give you some hairy moments. With practice, you'll find that you can pace the game, which can add hugely to your enjoyment. mp, on the other hand, is a frantic struggle to survive using whatever tricks you can get away with. Very Darwinian and modern, but you should remember that WZ is primarily a game - that is, a rest from kicking balls - and therefore a touch of civilization...

this post made me chuckle... you are not playing enough multiplayer

[driver4]

And yet I am sooo smart.