An AI is a majorly time-consuming thing to make. The only reason I got mine done in time was because I'm on study leave! While a decent cinematic could be done in the space of a day, an AI requires a small project's worth of work!
My plan was to load up the competing AI's into Mimic, and let it learn their build orders xD. It would be beaten in terms of micro however. It's kind of a "jack of all trades, but master of none".
what things should a AI could do for this contest? If it's only about doing somewhat then i can maybe join with my absolutly crap only building lings, drones, lords and spincecrawler AI. =)
Is it aloud that the AI can only handle one race?
I really wonder that the "Overmind" AI hasn't joined in this contest.
Designed (almost) entirely in AI files as a sixth level of difficulty rather than in triggers
It's not as complete as I wanted it to be, but this is what I've got. Known caveats:
Cheats: It can see the whole map. This is because I'm not done with scouting techniques yet.
A couple AI functions are still in triggers and haven't been ported to the AI files yet.
It's not nearly as good as it will be.
Phoenix micro basically ignores all units except mutas (For kiting. Its attack AI is normal non-micro for all non-muta air units). This will be expanded upon in the future.
Sentry AI is unoptimized (and unfinished). Large battles with many sentries may result in heavy lag.
Only PvZ (Zerg player, Protoss computer) is currently supported!
I named it 'fledgling' to reiterate the fact that it's nowhere near completion or even a state I'd consider ready for an alpha release. I think it will probably be able to beat most bronze/silver players regardless. I haven't done extensive testing with mutas or anything past T2 so feel free to try, but I can't promise it will counter very well. :)
I would have joined the competition but I'm studying for exams and I just played around only a little bit with the editor until now... I wrote down some sketches for an AI system that use different strategies, tactics and even micro, but it's just some drawing on the paper i didn't had the time to do something, it would have been funny :D
I saw very good work here, rrowland micro is very nice, and just the idea of mimic ai is original, gj guys! :D
rrowland, I did some messing around with your AI today, and the computer seems to have no idea what to do if you trap it in its main and macro up a larger army. It would be awesome if you incorporated some drop-tech expansion rules, or some economy harassment to prevent the player from winning the game by keeping the computer at the choke.
I'm not sure if this helps, but I played it as pvp, went 2gate zealot push into 4gate stalker/zealot and destroyed the expansion. Then I just sat at his expo and macro'd up. The only things that left his base were the occasional probes, in addition it didn't appear that he was creating very many units at that point.
On another note, the first game I played against it, where I didn't destroy the expansion, was awesome. I went mass roaches and he did a stalker/zeal/sentry/VOID RAY timing push. I was impressed. :)
rrowland, I did some messing around with your AI today, and the computer seems to have no idea what to do if you trap it in its main and macro up a larger army. It would be awesome if you incorporated some drop-tech expansion rules, or some economy harassment to prevent the player from winning the game by keeping the computer at the choke.
I'm not sure if this helps, but I played it as pvp, went 2gate zealot push into 4gate stalker/zealot and destroyed the expansion. Then I just sat at his expo and macro'd up. The only things that left his base were the occasional probes, in addition it didn't appear that he was creating very many units at that point.
On another note, the first game I played against it, where I didn't destroy the expansion, was awesome. I went mass roaches and he did a stalker/zeal/sentry/VOID RAY timing push. I was impressed. :)
The AI is only programmed for PvZ right now, so if you were protoss it probably didn't build any units past its initial build order including 3 zealots and a couple sentries. That being said, it's also at a very early point in its alpha stage as I mentioned in my entry post, so a lot of possibilities such as drops and harassment are not yet implemented. Thank you for the feedback though, I'm glad someone has tried it and I'm always open to suggestions. :)
Edit: Updated my entry post to show PvZ only in bold :)
I just tried your entry out, and found it to be very challenging. The first game I had no idea what to expect, and it just steamrolled me with zealot/sentry lol.
The second game I went roach hydra to see how different it would be and it came walking up with 5 or so immortals and sentry/stalker/zealot. I almost stopped it but not quite. It even micro'd the immortals a bit haha.
I'm guessing it would hold up well against various rushes since it knows exactly what your doing lol, that seems to be the biggest flaw is it would always counter exactly what you do, but I suppose that makes it that much more of a challenge to win :D
I did a 14-pool into 15-gas, thinking about going 1-base roach. I changed my mind and decided to grab a lair and then a hatch at about 20 or so, transitioning into 2 base muta. It rushed me with LOTS of sentries and 3-4 zealots. I lost my expansion to it, but I pooled the 7-9 mutas from the larva at it, and destroyed the attacking force. I eliminated the pylons powering its stargate and won en masse with mutas.
The thing I thought was buggy was that it didn't rush me because I had no defending units early game at all. I built no zerglings, spinecrawlers, etc.
Additionally, the 3 phoenixes it did manage to build never attacked me. They played defensively; they only attacked when I moved to attack them. When I did this, their AI was superb, although their chances would have been a lot better if the 3 phoenixes had grouped up. The point is, I just ignored them and destroyed all the buildings to win the game and they never attacked me back. If they had been harassing my mutas from different directions with their extreme micro already implemented into their AI, I probably would have lost (I'm not very good at all).
I'd send you a replay, but yeah... Battle.net...
I think the reason I won was that I was soooooo bad that it didn't know what I was doing until it was too late.
Edit: The first game I tried mass speedling was a disaster. He completed DESTROYED me with his sentry AI. I think I managed an amazing ONE sentry kill to nearly 50 speedlings versus 6-8 sentries.
I just wanted to give a huge shoutout to the ppl that have invested time into these AI's. I got into beta very late and never really got a chance to fully grasp the intricacies of all the macro/micro in this game. These will be extremely valuable practice tools for me (build orders etc...) while the beta is down and until release.
Thanks again!
btw rrowlands AI beat me down several time before I won.
I did a 14-pool into 15-gas, thinking about going 1-base roach. I changed my mind and decided to grab a lair and then a hatch at about 20 or so, transitioning into 2 base muta. It rushed me with LOTS of sentries and 3-4 zealots. I lost my expansion to it, but I pooled the 7-9 mutas from the larva at it, and destroyed the attacking force. I eliminated the pylons powering its stargate and won en masse with mutas.
The thing I thought was buggy was that it didn't rush me because I had no defending units early game at all. I built no zerglings, spinecrawlers, etc.
Additionally, the 3 phoenixes it did manage to build never attacked me. They played defensively; they only attacked when I moved to attack them. When I did this, their AI was superb, although their chances would have been a lot better if the 3 phoenixes had grouped up. The point is, I just ignored them and destroyed all the buildings to win the game and they never attacked me back. If they had been harassing my mutas from different directions with their extreme micro already implemented into their AI, I probably would have lost (I'm not very good at all).
I'd send you a replay, but yeah... Battle.net... I think the reason I won was that I was soooooo bad that it didn't know what I was doing until it was too late.
Edit: The first game I tried mass speedling was a disaster. He completed DESTROYED me with his sentry AI. I think I managed an amazing ONE sentry kill to nearly 50 speedlings versus 6-8 sentries.
Yeah a lot of work still needs to be done on the phoenix AI. Though, I'm not sure why the phoenixes weren't defending their base against you while you were attacking, as even the default AI should be telling them to do that. I will definitely look into that.
I guess it's too late to enter the contest, but I still want to show off WOPR. How many people can say that they made an entire functional AI without any AI functions or triggers? Remember, no peeking. I want the unique tactics involved to be a surprise; I promise it isn't cheating. The terrain is essentially desert oasis, with some modifications.
NOTE: At the beginning of the game it will tell you that you won. Just hit return to game and ignore this.
If it beats you, it'll only beat you the first time - but it was really only meant to fight other AI and people that don't pay close enough attention; while I haven't tested this, I doubt that the other competitors will be able to adapt to my unorthodox strategy, though the rush-friendly terrain modifications might almost be considered cheating, especially as I was too lazy to make it 100% symmetrical. Not that it matters, the way WOPR fights, and not bad for a single day's work.
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
This contest is really going to fail... 2 submissions 3 days away from the end?
Quote from Eldrazor:
This contest is really going to fail... 2 submissions 3 days away from the end?
----
Or it would be an awesome 1v1 head to head battle.
@Eldrazor: Go
An AI is a majorly time-consuming thing to make. The only reason I got mine done in time was because I'm on study leave! While a decent cinematic could be done in the space of a day, an AI requires a small project's worth of work!
@zifoon: Go
My plan was to load up the competing AI's into Mimic, and let it learn their build orders xD. It would be beaten in terms of micro however. It's kind of a "jack of all trades, but master of none".
Hi guys,
what things should a AI could do for this contest? If it's only about doing somewhat then i can maybe join with my absolutly crap only building lings, drones, lords and spincecrawler AI. =)
Is it aloud that the AI can only handle one race?
I really wonder that the "Overmind" AI hasn't joined in this contest.
@Demonette: Go
I think the rules are relatively lax; if it can fight someone in melee, submit it!
Meh. I know a bunch of people who told me they were working on AI's, :X.
Mhmm! And i'm almost positive rrowland is submitting his...
Well, IDK if I was one of those 2 but I'm not in this one. I misunderstood and the map I made wasn't melee AI.
@Vortexx2010: Go
Don't sweat it! Even if it doesn't turn out to be eligible in the end, I had good fun playing your map. (Although I did get completely slaughtered)
Alright, I'm gonna be hanging out with friends and busy the rest of the contest period so here's my entry: http://www.sc2mapster.com/assets/behemoth-ai/files/1-behemoth-fledgling/
Features:
It's not as complete as I wanted it to be, but this is what I've got. Known caveats:
I named it 'fledgling' to reiterate the fact that it's nowhere near completion or even a state I'd consider ready for an alpha release. I think it will probably be able to beat most bronze/silver players regardless. I haven't done extensive testing with mutas or anything past T2 so feel free to try, but I can't promise it will counter very well. :)
Phoenix micro video:
Sentry micro video:
I would have joined the competition but I'm studying for exams and I just played around only a little bit with the editor until now... I wrote down some sketches for an AI system that use different strategies, tactics and even micro, but it's just some drawing on the paper i didn't had the time to do something, it would have been funny :D
I saw very good work here, rrowland micro is very nice, and just the idea of mimic ai is original, gj guys! :D
rrowland, I did some messing around with your AI today, and the computer seems to have no idea what to do if you trap it in its main and macro up a larger army. It would be awesome if you incorporated some drop-tech expansion rules, or some economy harassment to prevent the player from winning the game by keeping the computer at the choke.
I'm not sure if this helps, but I played it as pvp, went 2gate zealot push into 4gate stalker/zealot and destroyed the expansion. Then I just sat at his expo and macro'd up. The only things that left his base were the occasional probes, in addition it didn't appear that he was creating very many units at that point.
On another note, the first game I played against it, where I didn't destroy the expansion, was awesome. I went mass roaches and he did a stalker/zeal/sentry/VOID RAY timing push. I was impressed. :)
The AI is only programmed for PvZ right now, so if you were protoss it probably didn't build any units past its initial build order including 3 zealots and a couple sentries. That being said, it's also at a very early point in its alpha stage as I mentioned in my entry post, so a lot of possibilities such as drops and harassment are not yet implemented. Thank you for the feedback though, I'm glad someone has tried it and I'm always open to suggestions. :)
Edit: Updated my entry post to show PvZ only in bold :)
@rrowland: Go
I just tried your entry out, and found it to be very challenging. The first game I had no idea what to expect, and it just steamrolled me with zealot/sentry lol.
The second game I went roach hydra to see how different it would be and it came walking up with 5 or so immortals and sentry/stalker/zealot. I almost stopped it but not quite. It even micro'd the immortals a bit haha.
I'm guessing it would hold up well against various rushes since it knows exactly what your doing lol, that seems to be the biggest flaw is it would always counter exactly what you do, but I suppose that makes it that much more of a challenge to win :D
@rrowland: Go
I did a 14-pool into 15-gas, thinking about going 1-base roach. I changed my mind and decided to grab a lair and then a hatch at about 20 or so, transitioning into 2 base muta. It rushed me with LOTS of sentries and 3-4 zealots. I lost my expansion to it, but I pooled the 7-9 mutas from the larva at it, and destroyed the attacking force. I eliminated the pylons powering its stargate and won en masse with mutas.
The thing I thought was buggy was that it didn't rush me because I had no defending units early game at all. I built no zerglings, spinecrawlers, etc.
Additionally, the 3 phoenixes it did manage to build never attacked me. They played defensively; they only attacked when I moved to attack them. When I did this, their AI was superb, although their chances would have been a lot better if the 3 phoenixes had grouped up. The point is, I just ignored them and destroyed all the buildings to win the game and they never attacked me back. If they had been harassing my mutas from different directions with their extreme micro already implemented into their AI, I probably would have lost (I'm not very good at all).
I'd send you a replay, but yeah... Battle.net... I think the reason I won was that I was soooooo bad that it didn't know what I was doing until it was too late.
Edit: The first game I tried mass speedling was a disaster. He completed DESTROYED me with his sentry AI. I think I managed an amazing ONE sentry kill to nearly 50 speedlings versus 6-8 sentries.
I just wanted to give a huge shoutout to the ppl that have invested time into these AI's. I got into beta very late and never really got a chance to fully grasp the intricacies of all the macro/micro in this game. These will be extremely valuable practice tools for me (build orders etc...) while the beta is down and until release.
Thanks again!
btw rrowlands AI beat me down several time before I won.
Yeah a lot of work still needs to be done on the phoenix AI. Though, I'm not sure why the phoenixes weren't defending their base against you while you were attacking, as even the default AI should be telling them to do that. I will definitely look into that.
I guess it's too late to enter the contest, but I still want to show off WOPR. How many people can say that they made an entire functional AI without any AI functions or triggers? Remember, no peeking. I want the unique tactics involved to be a surprise; I promise it isn't cheating. The terrain is essentially desert oasis, with some modifications.
NOTE: At the beginning of the game it will tell you that you won. Just hit return to game and ignore this.
@Andernerd: Go
Haha, very interesting strategy. I still won though :)
I've also won using this strategy against a friend before. Very amusing.
Andernerd, I threw you in the final results anyway, :P.
Vote now!
If it beats you, it'll only beat you the first time - but it was really only meant to fight other AI and people that don't pay close enough attention; while I haven't tested this, I doubt that the other competitors will be able to adapt to my unorthodox strategy, though the rush-friendly terrain modifications might almost be considered cheating, especially as I was too lazy to make it 100% symmetrical. Not that it matters, the way WOPR fights, and not bad for a single day's work.