Some of you are interested in how much APM the bot has. For 1 Marine Ursadak needs 300 APM. For example in the last big battle there were 76 Marines, that makes a total amount of 300*76 = 22800 APM
Lol this is nothing. I can do that too: Just slow the game down to 1:1000 ratio. YEAAAAA Oh wait this is real life. WOWOWOWOW I bow down to our imminent robot AI overlords
On July 25 2011 12:28 Rain.cz wrote: wow this reminds me IMhappy playing
U mean Happy right? The one with the youtube micro video? IMHappy is korean. Happy is eu WC3 turned sc.
On July 25 2011 12:30 Jhohok wrote: Lol this is nothing. I can do that too: Just slow the game down to 1:1000 ratio. YEAAAAA Oh wait this is real life. WOWOWOWOW I bow down to our imminent robot AI overlords
Next you should do zerg micro with all of the diffferent units or protosss blink micro and zealot micro. I assume FF's are impossible to do in a ssytem like this.
Why did you split marines against zerglings? That seems like it would increase melee attackable surface area instead of reduce it? You may have been able to save some marines against the pure zergling armies.
1 marine is the counter to an ultralisk. Standard.
Those were some sick automation replays of pretty reasonable in-game situations. Just shows the amazing skill ceiling that SC2 has to offer, even if this is probably not humanly attainable.
I am very interested to see custom starcraft AI develop into an AI that can play a 1v1 match without cheating vs decent human players (and also against each other). I hope that this topic draws a lot of interest from programmers and also perhaps receives some help from Blizzard via custom tools. I am not sure how cumbersome it is to program via the galaxy editor, but I assume it's pretty bad.
I remember a tournament type thing for BW where programmed AI's fight. Would really like to see that sort of thing for SC2, with epic macro/micro bots battling it out. ^^
now if only you could somehow outdo this by having one bot stutter micro zerglings forward while the other bot stutters marines back i.e. the bot is good enough to have every zergling move past the marines individually and abuse their faster speed to its fullest potential
On July 25 2011 13:58 Wolf wrote: Skill ceiling? Nonexistent.
You're comparing a human to a machine. It's like saying there's no cap on how fast humans can run because a car can go hundreds of miles per hour. With that said, while it's a little amusing to watch how ridiculous the marines look with all that APM, it would be very interesting to see how optimized a 400 apm capped AI could make their micro.
Hopefully we will see pros concentrate on micro more at some point (perhaps in a few years after they get more experience) because they can definitely do better than they do right now.
What if the marines with higher hitpoints stood in front of marines with lower hitpoints in the marine vs zergling battles?
Yeah that would make things more interesting, having them smart enough to interact with eachother so they can also realize one group being targetted can simply run and let other marines attack rather than every single marine simply stutter stepping and focus firing the whole time.
On July 25 2011 12:19 godemperor wrote: lol @ the ending with marine vs ultra, great vid!
OH lol! i was like, huh? where?
then i realized that there was stuff going on behind the credits xD
poor ultralisk
Also OP I agree with the guy, can you also make a program that uses only 250 or so APM (maybe even 300) to see how players COULD "realistically" do if they optimize their micro?
Edit: Actually, probably 150-200 APM would be more accurate, as the bot would be able to do everything effortlessly (and instantly) while a player wouldn't be able to, so something less than 250 I think would be good?
On July 25 2011 14:35 bkrow wrote: Ursadak would be pretty mad if the zerg built infestors :p nice video haha really awesomr to watch
Or if the zerg tried to surround. I would like to see Ursadak fight against micro'd zerg units and see where that goes. I'm sure it would still wreck but I don't think it would be nearly as powerful if the lings and banes were on anything but attack move.
On July 25 2011 13:58 Wolf wrote: Skill ceiling? Nonexistent.
You're comparing a human to a machine. It's like saying there's no cap on how fast humans can run because a car can go hundreds of miles per hour. With that said, while it's a little amusing to watch how ridiculous the marines look with all that APM, it would be very interesting to see how optimized a 400 apm capped AI could make their micro.
Interested to see other units that could be used like this. Or is the marine the sole exception that a higher apm makes the unit infinitely more powerful?
My guess is perhaps stalkers, hellions, and reapers. Wanna see ursadak use these units to the fullest.
On July 25 2011 12:19 godemperor wrote: lol @ the ending with marine vs ultra, great vid!
OH lol! i was like, huh? where?
then i realized that there was stuff going on behind the credits xD
poor ultralisk
Also OP I agree with the guy, can you also make a program that uses only 250 or so APM (maybe even 300) to see how players COULD "realistically" do if they optimize their micro?
Edit: Actually, probably 150-200 APM would be more accurate, as the bot would be able to do everything effortlessly (and instantly) while a player wouldn't be able to, so something less than 250 I think would be good?
I think ur mistaken if you think 300 is the apm in the video. It's 300 apm PER marine. There's about a ton of marines on those maps..
On July 25 2011 13:58 Wolf wrote: Skill ceiling? Nonexistent.
You're comparing a human to a machine. It's like saying there's no cap on how fast humans can run because a car can go hundreds of miles per hour. With that said, while it's a little amusing to watch how ridiculous the marines look with all that APM, it would be very interesting to see how optimized a 400 apm capped AI could make their micro.
I wanna see this!
It doesn't make the skill ceiling exist more. It is always possible to play better.
On July 25 2011 14:47 Crissaegrim wrote: Interested to see other units that could be used like this. Or is the marine the sole exception that a higher apm makes the unit infinitely more powerful?
My guess is perhaps stalkers, hellions, and reapers. Wanna see ursadak use these units to the fullest.
Marines move fast when stimmed, have a high fire rate, and don't have a projectile with their attack/don't take time to turn around which basically makes them the greatest for stutter-step micro. Stalkers attack too slow but an AI could certainly be several times more effective with them than a player because on attack-move they overkill so much (all the stalkers hit the same one unit on the attack-move command even though half of them would've killed it). Hellions stop for too long before attacking although I suppose a decent number of them controlled by an infinite APM AI would also be able to take out an infinite number of lings, given the space. Reapers can be used the same way as marines but simply can't attack air, so.
I'm surprised no one asked yet- You used this AI in only melee games, which leads to the question: What sort of AI does it use for build orders, and other strategy, if any? Was it just a user controlling it and then relinquishing control to auto-micro in fights? It didn't seem at all like Blizzard AI. It sorta seemed like a human controller considering the unit placement (particularly sieged tanks).
Also, the whole title and premise at the beginning was Ursadak vs Automaton 2000. Why Didn't you compare their differences in micro in the same situations (put ursadak in the very same situation(s) Automaton 2000 was in)?
Does this micro AI do anything other than micro marines (or marauders) right now?
Lastly, with regards to the video, it would have been good to include the army size tab, so people can see the losses for each side, and things like how even the fight was.
On July 25 2011 13:58 Wolf wrote: Skill ceiling? Nonexistent.
You're comparing a human to a machine. It's like saying there's no cap on how fast humans can run because a car can go hundreds of miles per hour. With that said, while it's a little amusing to watch how ridiculous the marines look with all that APM, it would be very interesting to see how optimized a 400 apm capped AI could make their micro.
? He jsut means it's still significantly higher than it currently is in terms of mechanics ( not necessarily strategies )
There's always plenty of room for improvement. Some people hit a wall while others may transcend higher...
That was sexy... The fact that all marines are microed individually makes for a weird looking video, like if all the banelings were in slow motion. But really nice job, I would actually love to see a real battle between those two AI, don't know if it's possible to implement that in a game though.
This is one example of how Humans will never ever ever ever ever ever reach the skillcap of pretty much any game that does is not extremely simple (e.g tic tac toe.) , SC2 included.
On July 25 2011 13:59 Herculix wrote: now if only you could somehow outdo this by having one bot stutter micro zerglings forward while the other bot stutters marines back i.e. the bot is good enough to have every zergling move past the marines individually and abuse their faster speed to its fullest potential
eh, i don't ever think melee units will outmicro ranged units >_>
On July 25 2011 18:04 Knutzi wrote: it would be funny if they showed that to DRG filming his reaction while telling him its a foreign player going to mlg ^_^
On July 25 2011 18:04 Knutzi wrote: it would be funny if they showed that to DRG filming his reaction while telling him its a foreign player going to mlg ^_^
Do this plz.
This would be a epic troll move. Hot_Bid go for it !
Holy shit, I can't wrap my head around how a human could to that even to some small engagement, but we'll see in 10years I guess. Next rines v sieged tanks? lol
On July 25 2011 12:19 godemperor wrote: lol @ the ending with marine vs ultra, great vid!
OH lol! i was like, huh? where?
then i realized that there was stuff going on behind the credits xD
poor ultralisk
Also OP I agree with the guy, can you also make a program that uses only 250 or so APM (maybe even 300) to see how players COULD "realistically" do if they optimize their micro?
Edit: Actually, probably 150-200 APM would be more accurate, as the bot would be able to do everything effortlessly (and instantly) while a player wouldn't be able to, so something less than 250 I think would be good?
I think ur mistaken if you think 300 is the apm in the video. It's 300 apm PER marine. There's about a ton of marines on those maps..
That's not what he said at all, he wants to see the program have a limited APM that is closer to a pro's APM and not a bot with unlimited APM
On July 25 2011 20:40 Beakerbite wrote: I kinda want to see an evenly matched set of rines and medivacs being controlled by each bot against each other.
Yeah that's what I was expecting, because of the word, "duel".
Regardless, these AI bots are amazing, so excited to see them vs eventually.
Blizzard was talking about a progamer level AI for the campaign IIRC, maybe they can implement something like this, since the next campaign is zerg based. Imagine the ultimate level that only the best pros can beat
Nice concept, I thought you were going to be letting two bots duel against each other, which would be pretty awesome niche to spectate. AI designer vs AI designer
Very enjoyable to watch. I'd like to see a more direct duel though if possible, although I don't know what battle would be most enjoyable. Maybe blink stalker duels? Who knows? ^_^
Seriously..what happens if they get flanked and attacked by both sides? It just looks like they kite backwards forever. I want to know what happens when they are slightly obstructed and can't just walk backwards!
the very first marine vs bane was the best, because microing back and forth like that is good, but doing back and forth micro while rotating the entire arc, now thats impressive.
Not very respectful to talk down on Automaton 2000 like that, since it was the first one to do marine splitting well and is obviously the inspiration behind your bot(which Isn't insanely better).
Hopefully we will see pros concentrate on micro more at some point (perhaps in a few years after they get more experience) because they can definitely do better than they do right now.
*stares at boxer* :D
yes, we know that people can concetrate more on marine splitting and blink micro, now on what zerg players need to concentrate more ? muta micro is already explored and almost non existant besides stacking vs marines and magic boxing vs thors/archons, banes can be moved or a-moved depending on situation, lings are not microable in large numbers i wish to know what can be done for zerg that is humanly possible, hint hydras are to slow, ultras suck, corruptors needs target-firing only, broodlords -_-, infestors ? well infestors are only about 1 fungal and die/dont die(thx to speed nerf), only thing that has some potential is muta vs phoenix micro but its so rare and depends mostly on how good protoss is because he has advantage, and above all the prerequisite to have potential micro is being on creep... which even with efforts can be denied
On July 25 2011 23:57 shizna wrote: i don't understand why people are so impressed...
have you never seen aimbots in FPS games?
the AI can be programmed to act perfectly, human players can not act perfectly. it's not physically possible...
you might aswell watch a 1000m sprint man versus pagani zonda.
I think it's much harder to make an micro bot than an aimbot. But I might be pretty ignorant with this statement as well.
Pretty much this. An aimbot is really easy to make. A micro bot on this level is insanely difficult to make (Even blizzard never came close with their own AI, and had to resort to the AI cheating instead)
Hopefully we will see pros concentrate on micro more at some point (perhaps in a few years after they get more experience) because they can definitely do better than they do right now.
*stares at boxer* :D
yes, we know that people can concetrate more on marine splitting and blink micro, now on what zerg players need to concentrate more ? muta micro is already explored and almost non existant besides stacking vs marines and magic boxing vs thors/archons, banes can be moved or a-moved depending on situation, lings are not microable in large numbers i wish to know what can be done for zerg that is humanly possible, hint hydras are to slow, ultras suck, corruptors needs target-firing only, broodlords -_-, infestors ? well infestors are only about 1 fungal and die/dont die(thx to speed nerf), only thing that has some potential is muta vs phoenix micro but its so rare and depends mostly on how good protoss is because he has advantage, and above all the prerequisite to have potential micro is being on creep... which even with efforts can be denied
WOWOWOW stop right there. That's a bunch of bold statements you made. In fact I was going to write a paragraph explaining why I disagree with you but the more I read your post the more I want to punch my screen so please stop spreading uninformed and misleading bullshit.
On July 25 2011 23:57 shizna wrote: i don't understand why people are so impressed...
have you never seen aimbots in FPS games?
the AI can be programmed to act perfectly, human players can not act perfectly. it's not physically possible...
you might aswell watch a 1000m sprint man versus pagani zonda.
I think it's much harder to make an micro bot than an aimbot. But I might be pretty ignorant with this statement as well.
Pretty much this. An aimbot is really easy to make. A micro bot on this level is insanely difficult to make (Even blizzard never came close with their own AI, and had to resort to the AI cheating instead)
Yeah, I think this is quite a distinct problem from that of making an aimbot. Calculating a vector from player A's gun to player B's head is simple, HEADSHOT BOOM. Now for Marine Micro, can you even easily define the function you are trying to optimize?
I'm curious about the 300 APM per marine - it seems like there must be useless actions in there. Does the AI act only when it needs to, to perform optimal stutter step, or is it constantly updating the marines path at set intervals?
I'll be interested to see how you coded that, can we get the code ourselves ? Whereas it seems not very efficient (code wise) as 300 APM per marine is way to much, it is very effective against basic attack-moves. The number of calc isn't that high surprisingly.
On July 26 2011 02:08 Otolia wrote: I'll be interested to see how you coded that, can we get the code ourselves ? Whereas it seems not very efficient (code wise) as 300 APM per marine is way to much, it is very effective against basic attack-moves. The number of calc isn't that high surprisingly.
Anyway, a very good job. Congratulations.
300 APM per marine actually isn't all that much, which sounds totally wierd consider that actually is a lot >_>
It breaks down into 5 APS per marine, which makes perfect sense considering how each marine is essentially constantly calculating it's movements and moving back -> a move forward, and repeat. That's already 2 actions right there, and I can totally see this doing 2.5 iterations in a second.
On July 26 2011 01:50 ThePhan2m wrote: Too bad bots suck in strategy Even with that micro, a human would beat it
No you wouldn't. In TvZ, this bot would go 2rax into 4rax and kill you with marine micro before you get any tech out. Not only that, it will be constantly attacking/harassing every second of the game.
On July 25 2011 13:25 Ulfsark wrote: I wonder if this was inspired by a post i made in the automaton thread about having 2 people program bots and have them fight haha.
Having watched this....a new fear of Terran has been planted in my Overmind.
Queens have unionized and refuse to use their energy injecting larvae. Drones will no longer harvest and have burrowed themselves far from their mineral lines.
Our creep no longer conveys the message "Private property. No Trespassing." Now it is a skim to be Scanner Swept off Terran's latte.
Oh man, that very last battle was like a BW one, it spanned across multiple screens and every unit was fighting, I can't wait for the day stuff like that becomes the norm in pro games.
On July 26 2011 05:16 hybersnack wrote: Is it really possible for a bot to micro a marine vs an ultralisk if it has medivac support? insane.. Shows how useless ultralisks are
I don't think you can judge balance based on this. You can micro 1 viking and kill 200/200 BCs or 1 stalker against 200 marines without stim or a marauder with concussive shells against zealots without charge.
On July 26 2011 05:16 hybersnack wrote: Is it really possible for a bot to micro a marine vs an ultralisk if it has medivac support? insane.. Shows how useless ultralisks are
no its actually not possible. a stimmed marine is way faster than a normal medivac. for the video i had to give the medivac a small speed buff
all other videos are not stat cheated or outupgraded.
Great video. Amazing work. This doesn't really raise concern for imbalance but it makes me feel sad as a Zerg player though knowing Terran units in the hand of a player have unlimited potential.
On July 26 2011 05:16 hybersnack wrote: Is it really possible for a bot to micro a marine vs an ultralisk if it has medivac support? insane.. Shows how useless ultralisks are
I don't think you can judge balance based on this. You can micro 1 viking and kill 200/200 BCs or 1 stalker against 200 marines without stim or a marauder with concussive shells against zealots without charge.
On July 26 2011 05:16 hybersnack wrote: Is it really possible for a bot to micro a marine vs an ultralisk if it has medivac support? insane.. Shows how useless ultralisks are
no its actually not possible. a stimmed marine is way faster than a normal medivac. for the video i had to give the medivac a small speed buff
all other videos are not stat cheated or outupgraded.
- darglein
If the marine kited in a tight circle, the medivac wouldn't need a speed boost, so i would say it is possible.
On July 26 2011 05:16 hybersnack wrote: Is it really possible for a bot to micro a marine vs an ultralisk if it has medivac support? insane.. Shows how useless ultralisks are
no its actually not possible. a stimmed marine is way faster than a normal medivac. for the video i had to give the medivac a small speed buff
all other videos are not stat cheated or outupgraded.
- darglein
If the marine kited in a tight circle, the medivac wouldn't need a speed boost, so i would say it is possible.
that's right, but to pefrom this i had to use an other AI. running in circles in a big baneling fight is not very effizient.
Wow. That's insane. Those marines attacks were canceled so fast they just looked like little spinning stars. Just shows there is always room for improvement in the game. Would like to see other units, like stalkers not over killing or mutas. Would love to see a zerg AI versus that also and see if it's just as effective when the zerg is surrounding and microing its units.
I have to admit, that display was pretty badass and even though no human will ever achieve that level of control, it truly shows the potential of the marine as a unit...
On July 25 2011 12:26 zhurai wrote: t.t I think people should make a micro bot for banelings vs tank+marine =.=; (with the banelings being controlled by the bot)
T_T;
Automaton 2000 has one of Pure Ling vs Pure Tank. It's pretty crazy how well it does avoiding tank splash damage.
Maybe its still possible if you let the marine run circles around tower or destructable rock in wich case the medivac maybe could just stay in 1 place?
even without yamato i am sure 1 viking cant kill 200 bc,s if the bc,s are also controlled in the perfect way starting with 200 bc,s in 1 corner and the viking at range 7 away from them the bc,s would not tail the viking till all of them are shot down, they would first spread across the map to be able to aproach the viking from all directions guaranteeing to get at least a few shots in and eventually killing the viking (is my guess with 200 suply bc,s against 1 viking) It would be interesting to see these battles played out without 1 side amoving and the other side having the perfect bot annyway:lthx for the tread and the vids love this ai aspect of starcraft and would be awesome to have some microtournament for bots Strategic ai in sc is difficult but verry interesting and looking forward to publications about that subject also the brute force wich bought so much succes in chess is simply impossible in sc (to manny posibilities)
On July 26 2011 02:08 Otolia wrote: I'll be interested to see how you coded that, can we get the code ourselves ? Whereas it seems not very efficient (code wise) as 300 APM per marine is way to much, it is very effective against basic attack-moves. The number of calc isn't that high surprisingly.
Anyway, a very good job. Congratulations.
300 APM per marine actually isn't all that much, which sounds totally wierd consider that actually is a lot >_>
It breaks down into 5 APS per marine, which makes perfect sense considering how each marine is essentially constantly calculating it's movements and moving back -> a move forward, and repeat. That's already 2 actions right there, and I can totally see this doing 2.5 iterations in a second.
The stimmed marine attack cooldown is 0.5739s so he can perform 2 attacks per second.
0s : Attack 0.0001 : Movement 0.5739 : Attack 0.5740 : Movement -- 1.1479s : Attack
That's approx. 4. If you had the moving back (only possible off-creep) then you have your 5 APS. Everything is clear;
One interesting thing I noted is that in the 5 marines 2 rax early push, the AI stutter step regardless of attack cooldown (much longer steps but better positioning between steps)
??? is any one else not impressed? Wow, you can make a bot micro stimmed marines against blings off creep... that is the one situation where micro is really the only important thing, but how often is that realistic?
what kinds of builds does this do? can it play tvp and tvt? do you have any plans to make it ladder?
On July 26 2011 01:50 ThePhan2m wrote: Too bad bots suck in strategy Even with that micro, a human would beat it
No you wouldn't. In TvZ, this bot would go 2rax into 4rax and kill you with marine micro before you get any tech out. Not only that, it will be constantly attacking/harassing every second of the game.
Actually, some units you cannot beat with micro. The human player has to adapt (of course he cannot go banelings / zerglings against this) Though once he gets his tech up to infestors, some abilities cant be countered with micro. Not to mention units that outrange rines, like sunkens. Mass them!
Also video is kinda biased, no human would attack those rines straight on. He would have to flank from several directions, and that would make micro harder, even for that bot.
I really enjoyed all those micro battle so far but it was always tvz... can you maybe show some perfect blink stalker micro in the future? or maybe perfect kiting/splitting of mmm vs a protoss deathball?
Reminds me of watching MarineKing in GSL Open Season 2 LOL. Granted, this is way more intense, but at the time his marine micro and splitting was insane! Still is, but it was worlds ahead of anyone else at the time.
On July 26 2011 10:02 Tristan wrote: Ok so who wants a future where the esport of SC2 is just Watson-like AIs by different IT companies duking it out
Can you imagine a "perfect" game of sc2? oh my god, imagine the multitask. You would need at least 5 different observers and maybe 10+ casters for a single match
1. That was just a great video and great demonstration of some seriously impressive AI. 2. This should end all argument that the skillcap in SC2 is too low. Forever.
On July 26 2011 10:02 Tristan wrote: Ok so who wants a future where the esport of SC2 is just Watson-like AIs by different IT companies duking it out
Isn't there an annual Brood War AI tournament? You'd think it's about time someone gets an SC2 competition going.
so human can do it too if you have 1 pro controlling each marine in a team melee? nice
Not if - your humans make timing mistakes (like missing one of those actions by a tenth of a second, ever) - your humans make spatial errors (like misclicking by a few pixels, ever) - the AI used any form of shared state between marines to decide on the actions (good luck exchanging information at that rate with humans)
On July 26 2011 04:57 BlueyD wrote: Someone needs to make a bot like this which queen-transfuses instead of marine-spreading.
I made an AI for my custom Zone Control map that has some impressive transfusion micro. Combined with speedlings for lots of map control it easily owned 4 real players even though it doesn't cheat in any way. I had to nerf the queen so hard to be able to beat it 1v1(it's still quite hard though)
Where did the OP & creator Darglein go? I would like an answer to my question, an I assume other people may want to know other things as well.
On July 25 2011 15:23 Xapti wrote: I'm surprised no one asked yet- You used this AI in only melee games, which leads to the question: What sort of AI does it use for build orders, and other strategy, if any? Was it just a user controlling it and then relinquishing control to auto-micro in fights? It didn't seem at all like Blizzard AI. It sorta seemed like a human controller considering the unit placement (particularly sieged tanks).
Also, the whole title and premise at the beginning was Ursadak vs Automaton 2000. Why Didn't you compare their differences in micro in the same situations (put ursadak in the very same situation(s) Automaton 2000 was in)?
Does this micro AI do anything other than micro marines (or marauders) right now?
Lastly, with regards to the video, it would have been good to include the army size tab, so people can see the losses for each side, and things like how even the fight was.
Would reduce the speed of the game raise the skill gap? Right now we play on Faster, what if we played on Fast or Normal? I was playing that marine splitting map by Griffith(probably mispelled that), and I usually only get to level 14 or so on Faster, I played on "Fast" once without knowing and got to level 18.
Would it make blink, splitting marines and bane/ling micro in ZvZ more skilled?
On July 29 2011 05:23 rRod wrote: Would reduce the speed of the game raise the skill gap? Right now we play on Faster, what if we played on Fast or Normal? I was playing that marine splitting map by Griffith(probably mispelled that), and I usually only get to level 14 or so on Faster, I played on "Fast" once without knowing and got to level 18.
Would it make blink, splitting marines and bane/ling micro in ZvZ more skilled?
For micro, maybe, but it would also enable everybody to be a macro god. And macro games would be painfully long to watch.
On July 29 2011 05:23 rRod wrote: Would reduce the speed of the game raise the skill gap? Right now we play on Faster, what if we played on Fast or Normal? I was playing that marine splitting map by Griffith(probably mispelled that), and I usually only get to level 14 or so on Faster, I played on "Fast" once without knowing and got to level 18.
Would it make blink, splitting marines and bane/ling micro in ZvZ more skilled?
Actually you have this backwards, INCREASING the speed of the game would raise the skill gap because it would be a lot harder and take more practice to do proper micro. Slowing the game down just makes it easier for more people to do the same micro with a lot less practice = smaller skill gap.
This is the ultimate skill cap for SC2. Unreasonably high, but we are not even close to this type of stuff yet. (Can we see some Toss related stuff in the future?)
On July 29 2011 05:23 rRod wrote: Would reduce the speed of the game raise the skill gap? Right now we play on Faster, what if we played on Fast or Normal? I was playing that marine splitting map by Griffith(probably mispelled that), and I usually only get to level 14 or so on Faster, I played on "Fast" once without knowing and got to level 18.
Would it make blink, splitting marines and bane/ling micro in ZvZ more skilled?
For micro, maybe, but it would also enable everybody to be a macro god. And macro games would be painfully long to watch.
Well playing on "normal" isnt that crazy "painfull" , try it out, its not really a bad thing if players can use more macro.
An allready strong macro player will have plenty of more time for micro ; so the games get actual a lot more intresting as you can do much more multitasking if the game is simply slower.
The actual "bad" thing is just that is really is slower, games take longer, as the ingame time really matches the real time (which is ironic anyway, that SC has a faster minute than real life).
But overall i would rate "normal speed" better for the most players ; and for the real strong, it just gives more room for micro.
first of all the AI can only micro marines. the ai doesent do anything else. for the replays i created modified version of the maps where the players can enable and disable the ai with buttons. so i controlled the terran players until the big fight. in the big fight i just pressed a button and the marines were microed by themselves.
why is it a micro bot duell??
of course its not a direct micro bot duell. its more a comparison of this 2 bots. i made ursadak because a lot of people said after seeing automaton 2000 something like that " OMG this is the best micro i have ever seen. SOO SICK!!!!!!" . So my intention was to show you that a good marine micro AI needs more then running away from units in a certain range until the distance is big enough to fire again, all of course in a completly flat huge empty area.
automaton 2000 is much better than ursadak!!!!
its true that automaton dont lose any marines in the mariens vs. banelings battle. so its true that automaton is better if the mission is to kill only banelings with stim on a wide area without obstacles. but the problem is automaton would completly fail on realistic ingame szenarios like i showed in the videos. my ai was developed to do well in a wide range of szenarios on all kind of different terrain. for example watch the part with the baneling bust again and you will see some marines running to the left behind the mineral line of the command centre and finding a way back to the other marines without getting surrounded or stuck in a corner.
will there be a unbeatable non cheating AI?
thats definitly not possible, but there are plans that i will work with the creator of green tea AI together to create the most powerfull micro/macro AI out there.
sorry for my bad english i hope you understand it all
Darglein I just wanted to say thank you for all the custom micro challenges you have made. They have helped my micro intensely. Everyone can see the passion and appreciate the time you spent. Your english is just fine! Thanks Again!
Everytime I see those AI Videos, i always think "Why must those "incredible A.I.s" only defend against "A-Move Attacks".
Are that A.I.s especially trained to fight against A-Move Attacks/Blizzard Attackmove A.I. and can't handle better/individuell attacks (like defending Banelingattacks/micro from Nestea or another "incredible A.I")?
As long as this isn't happening... for me its just a (ONE) script. No A.I. (many scripts). Just a videoanimation. But a nice one.
i cried the manliest of man tears as i watched this video. R.I.P all those brave zerglings who gave their lives for the ursadak. Better the ursadak then the lyote. Really cool bot though, fun to show people and tell them that it isnt a bot.
this is why zerg sucks, you just don't have the ability to do this kind of crazy micro so as micro skill increases they just get worse. like banelings are only good if your opponent isn't really good. such a crap shoot.
EPT
