seeing krasi0 repairing its units made me think that bots could make good use of stuff like restoration and d-web that humans often neglect because they can't spend their attention on it
I'm not sure how good D web would be against other bots, since they would just split instantly, but maybe it could be worth it for static D. But restoration definitely sounds good against plague.
Shame the mutas couldn’t get anything done. The computer wasn’t too good at surgically attacking with them. Of course we’re used to watching people like Jaedong on this channel, so it’s a high bar 😅
as a casual broodwar viewer its a little annoying when all the action is going on elsewhere on the map and arty is talking about their base. I mean just wait til nothing is going on. Im an ape and want to watch ape go smash. It happens every single game too lol
@HalCourtney Exactly. It's one of the main reasons of why I prefer SaiyanKCM. He even does picture-in-picture if there are multiple things happening at the same time.
even if the quality of the games isnt super high, its interesting to see human programmed bots play, as opposed to self learning AI bots doing so. So thanks for doing this Arty, I appreciate the variety.
This game is not about APM, but multitasking & decision making. Thanks for the insights Artosis! If it was not for you this game would not be as amazing as it is.
@@rodrigomarchioli2707 low APM in Starcraft means you are very, very dead. There is a floor for APM in BroodWar. Most RTS there is an APM floor, some of them that floor is as low as 50 APM, others the floor is 100 APM. The pro players regularly exceed 300 APM. The key here is how precise each click is, since obviously spamming actions is no sign of skill at all...these bots prove it. A human player that is sub 50 APM can only beat bots, they will never get anywhere on ladder.
@@Ziegfried82 Yes, but my starting point is the common sense that if you have less than 100 APM you are basicaly dead. After that, as I was saying, it is all about multitasking and decision making, because if you have more than 350, 400 APM it does not make any difference. That is my point.
I wish someone made a new 2D RTS game inspired by brood war, with unit selection limits and no rock paper scissor design, cool sprites with unique animations and memorable sounds, but without the bad pathing, with different units and abilities just to spice it up. I guess we can always play brood war but it feels weird nobody tried to bring back the magic of it in another form.
It's mainly because RTS games are the hardest type of game to make, only near it is fighting games like Guilty Gear or Tekken. Granted, projects like the UED AI Project more or less imply that there exists the skill, talent and passion to do something like that, it's just a matter of 'will they actually take the time to design their game, learn their language, pick their engine and actually make it?' and the answer is usually no. Granted, the RTS genre isn't as barren of new games as people tend to believe, or want to believe, but it's certainly not as vibrant as it really could be if only blizzard weren't idiots and allowed Galaxy Editor to work like the Unreal of RTS game engines.
It almost seems like the ideal bot at this point would just be one that fine-tunes macro and a bit of micro (for moving shot, etc) but keeps the broader strokes of the base SC computer players
When i was a kid and did games against the computer in custom games, i always found the easiest way to cheese them was just send a worker over to their base, attack their main building and they all want to attack you. Just lure them away from home and they have no mining happening.
Artosis observing was way off in these games. We missed the fun parts... The final all in with probes and all by the AI. The Mutas vs Archon. The Mutas microing the 3rd base of the Toss, etc
SC2 A.I., Alpha Star beat one the best players in the world. However it was funded by a big company - not amateurs tinkering with A.I. like here (I'm not making fun of the latter, they've done very well with their limited resources).
It's crazy how these hobbyist AIs are wildly impressive in what they can do, and yet (almost) totally useless when it comes to actually playing the game. Huge respect to people who can make these bots work even at this level. It just goes to show how immensely complex a game like Starcraft actually is.
give us a game that ai actually wins)) Also, if there is such a thing I would like to see a hybrid AI-human play. You know there is a mode where two players control the same economy (Team Melee), so for example AI can macro perfectly and human will micro in parallel with available units. Or, for that matter a Team Melee human-only game.
I do wonder what a tool assisted StarCraft match would look like. It would turn the game into a weird turn based strategy game. Games would take hours but if you played the match back at normal speed you could see what “perfect” play would look like.
Having recently just played some games against the base BW AI, these bots are miles better, its not even close but maybe that just shows how bad the original AI is.
When will they make actual "intelligent" ai (not just pre programmed BO and brute force APM micro) that makes correct human like decision, like going up the ramp with 12 dragoons to kill 3 cannons
Almost 10k APM - so how much in terms of coolant measured in BTUs were needed to allow an A.I. to play that game? A few more Terran Turrets I'd say, with less tank siege-unsiege, then sprinkle in something FUNKY like Medics (with Cleanse) later on!
Hopefully we can watch the first matches Jaeyun played against these AI. It’s unfortunate we just see the subsequent matches where he’s learned how to defeat them.
I don't think there was anything exploitative with that first game. The AI is simply not coded to handle concepts like "goons going up ramps" and "carriers exist". Which would in my mind make it pretty bad at fighting anything except a strict dragoon build AI. Humans are simply too good to beat.
Looks like zerg bot didn't make changes to strat even though it scouted protoss. Like that scouting was a fiction to make the bot seem to have strategic decision making about the army conposition
I love the idea of this tournament, but I think the bots need a bit more time cooking. 9 times out of 10 they just do nothing while getting slowly killed. It's just not very fun to watch after the novelty wears off.
Well I learned that I root for terran even when it is a weird bot. I was also hoping to see 7000 apm makes ghosts suddenly OP against protoss but didn't happen. Definitely seems glitchy, almost like the AI tries to micro each individual unit too perfectly it doesn't properly understand the group dynamics. Like it misses the forest for the trees.
What do you do with 7000 apm?! Everything. You know all and see all, you are the early release of alphastar that was everywhere and doing everything, all at once.
Guys just wouldn't give the bots any opportunity to show off their macro late game. Sad. How do protosses doubt that carriers are op against terran? Just look at these games)
It’s funny watching the AI consistently have over 1000 APM, reaching over 9000 at some points and accomplishing almost nothing with it. I wonder if they capped it at a lower APM, would it more effectively units instead of attempting to micro every single unit at once. Specifically the first match where all the goons just stood there shimmying. Do note that I’m a moron and know nothing about AI so idk if that has anything to do with it or not.
The problem in the first game wasn't the 'micro', it was the decision not to move up the ramp. The 'micro' isn't the main issue (although I am surprised how bad it is; I would have at least thought the mutalisks would kill the archon without taking hits); the main issue is a lack of reasonable strategic decision-making. These AIs are just all-around bad. The fact that the 1-base carrier strategy actually worked means that the AI would lose to some extremely bad players.
they are programmed, those AI didn't learn how to do things rather there are clear commands how to do things. Thus if they get capped the commands have to be rewritten to make do with the cap.
Some points about the high APM: Bots can't do group commands, they have to order each individual unit. Bots often issue the same order every few frames. It's much easier than to check if the order is the same. Limiting the APM would essentially just be more work to prevent some of that. But without any additional benefit.
@@yevgeniygrechka6431 These are hobbyist bots that AFAIK incorporate really minimal machine learning, if any. So for example the Protoss AI is basically hardcoded to individually micro goons when they're shot at. In a vacuum this means the AI would win any dragoon vs dragoon fight on equal numbers. In practice you see stuff like this disaster on the ramp.
@@Xaan-lm4hb I'm not convinced that the AI would even win a dragoon vs dragoon fight after the ZvZ mutalisk battle yesterday. I feel like the only thing the AI can do better than humans is make non-stop workers and move units into an arc, and that by itself just doesn't win games.
Pretty underwhelming games... I'd expect from an AI with 7k apm to micro mutas one by one and completely destroy archons, for example. Or split them perfectly and attack multiple parts of the enemy base simultaniously. But they do..nothing. They can't even get up a ramp in the first game, in PvP...it's hard to get dragoons up to a ramp, but even a 200 apm guy can do that.
The Al in these matches seem like they’re on the same level as built in ai in the game. I would be curious to see how more modern alphastar would play sc1.
I think what we learned here is that AI is incapable of decision making of any kind. It is good for teaching, because it teachers you how to optimizing, and how to not make stupid mistakes. But once you have a reasonable understanding of the game, you can game the AI, because their response is predictable, because it has to be for it to be scripted. A true, generally AI playing starcraft couldn't be beaten.
Looking forward to getting back to real games with humans vs. humans... If the AIs had been better this may have been fun for a single day, but still not several days in a row.
Haven't watched any StarCraft stuff in ages, Had an Artosis video pop up in my recommended and have been binge watching for hours. Love it.
Welcome to your new daily ritual of waiting for uploads every morning
Welcome to your new found love!
Welcome
Welcome
Welcome to the daily ritual
Look at the minimap at 10:30, all of purplewave's probes choose violence
V I O L E N C E~ ✨
Yeah, surprised he didnt see that and said something about it lol
LMAO 😅😅
This is by far the best version of you in starcraft. Awesome.
I see Schnail and I like
seeing krasi0 repairing its units made me think that bots could make good use of stuff like restoration and d-web that humans often neglect because they can't spend their attention on it
I'm not sure how good D web would be against other bots, since they would just split instantly, but maybe it could be worth it for static D. But restoration definitely sounds good against plague.
Don't forget Lockdown
Shame the mutas couldn’t get anything done. The computer wasn’t too good at surgically attacking with them. Of course we’re used to watching people like Jaedong on this channel, so it’s a high bar 😅
10:42 i wanned to see the all in the bot was doing there....
as a casual broodwar viewer its a little annoying when all the action is going on elsewhere on the map and arty is talking about their base. I mean just wait til nothing is going on. Im an ape and want to watch ape go smash. It happens every single game too lol
@HalCourtney Exactly. It's one of the main reasons of why I prefer SaiyanKCM. He even does picture-in-picture if there are multiple things happening at the same time.
@@toomuchinvigilation2419yeah that picture in picture is choice. arty should try some of that.
yes schnail, snail but said like sean connery
that ish korrekt
Why does an AI with thousands of APM need to stack its mutas? Shouldn't they spread to dodge storms and archon blasts?
Stacking is partially to make it harder to target down individuals
which bots can do very well and also dodge AoE attacks like valks, archons and corsairs and the obvious storm@@illadiel6049
even if the quality of the games isnt super high, its interesting to see human programmed bots play, as opposed to self learning AI bots doing so. So thanks for doing this Arty, I appreciate the variety.
This was a fun experiment if nothing else. Hopefully the bot programmers can learn a lot and come back with some 2.0 upgrades.
Those super complex AIs end up looking as if they fared better with just amove sometimes lol
This game is not about APM, but multitasking & decision making. Thanks for the insights Artosis! If it was not for you this game would not be as amazing as it is.
The more apm the more multitasking you are able to do
@@wildcat8332 multitasking AND decision making. It is about both...
@@rodrigomarchioli2707 low APM in Starcraft means you are very, very dead. There is a floor for APM in BroodWar. Most RTS there is an APM floor, some of them that floor is as low as 50 APM, others the floor is 100 APM. The pro players regularly exceed 300 APM. The key here is how precise each click is, since obviously spamming actions is no sign of skill at all...these bots prove it. A human player that is sub 50 APM can only beat bots, they will never get anywhere on ladder.
@@Ziegfried82 Yes, but my starting point is the common sense that if you have less than 100 APM you are basicaly dead. After that, as I was saying, it is all about multitasking and decision making, because if you have more than 350, 400 APM it does not make any difference. That is my point.
I wish someone made a new 2D RTS game inspired by brood war, with unit selection limits and no rock paper scissor design, cool sprites with unique animations and memorable sounds, but without the bad pathing, with different units and abilities just to spice it up.
I guess we can always play brood war but it feels weird nobody tried to bring back the magic of it in another form.
It's mainly because RTS games are the hardest type of game to make, only near it is fighting games like Guilty Gear or Tekken. Granted, projects like the UED AI Project more or less imply that there exists the skill, talent and passion to do something like that, it's just a matter of 'will they actually take the time to design their game, learn their language, pick their engine and actually make it?' and the answer is usually no.
Granted, the RTS genre isn't as barren of new games as people tend to believe, or want to believe, but it's certainly not as vibrant as it really could be if only blizzard weren't idiots and allowed Galaxy Editor to work like the Unreal of RTS game engines.
Rusted Warfare?
it'd be nice, but it takes a lot of resources to make such a game, and the return on investment potential is limited at worst and quite risky at best.
try Mass Recall, it's BW redone in SC2
try beyond all reason, its free and rly fun
10:34 Dude, you missed the ultimate probe sacrifice right there. They are probably doing it for Aiur and nobody cared lol.
I could see burrow being pretty sick with infinite apm when your opponent doesn't have detection
It almost seems like the ideal bot at this point would just be one that fine-tunes macro and a bit of micro (for moving shot, etc) but keeps the broader strokes of the base SC computer players
Not much to see as of yet, but I'm enjoying these thus far. Keep it going. Maybe we'll get a good game and find a worthy bot.
When i was a kid and did games against the computer in custom games, i always found the easiest way to cheese them was just send a worker over to their base, attack their main building and they all want to attack you. Just lure them away from home and they have no mining happening.
Terran AI units look like there's a cat rolling around on the keyboard and mouse
It's neat to see how far AI has come
A.I: "I won that last round, btw"
Artosis giving away my life long secret of depos in front of my bunker on BGH or laddering on lost temple hits me in '99. Artosis is my spirit animal
Instead of top humans, it might be fun to grab mid-level players and see how they fare against the bots
Artosis observing was way off in these games. We missed the fun parts... The final all in with probes and all by the AI. The Mutas vs Archon. The Mutas microing the 3rd base of the Toss, etc
19:42 its over 9000!
love the vids!
APM GO BRRRRRRRRR
Shortly after it reached 10k
The fuckin goons just stuttering down at the bottom was hilarious to watch. Great call on the build by JY
These are interesting to see
I kinda was expecting the AIs to be significantly better.
SC2 A.I., Alpha Star beat one the best players in the world. However it was funded by a big company - not amateurs tinkering with A.I. like here (I'm not making fun of the latter, they've done very well with their limited resources).
It would be interesting to understand how these are trained. Maybe interview one of the AI creators?
Thanks for the consistent games Arty!
W should do this more
How does this AI starcraft work?
It's crazy how these hobbyist AIs are wildly impressive in what they can do, and yet (almost) totally useless when it comes to actually playing the game. Huge respect to people who can make these bots work even at this level. It just goes to show how immensely complex a game like Starcraft actually is.
Its really hard to go from watching snow vs light to watching this AI dancing around with goons for 2 minutes at the bottom of a ramp
Alright we're STILL safe from Skynet.
how do you hook up SCHNAIL to Remastered?
10:30 Why didn't you show the probe all in?!
give us a game that ai actually wins)) Also, if there is such a thing I would like to see a hybrid AI-human play. You know there is a mode where two players control the same economy (Team Melee), so for example AI can macro perfectly and human will micro in parallel with available units. Or, for that matter a Team Melee human-only game.
Would be interesting to see good players trying to win top SCHNAIL bots using the lowest APM as possible (without cheese)
I do wonder what a tool assisted StarCraft match would look like. It would turn the game into a weird turn based strategy game. Games would take hours but if you played the match back at normal speed you could see what “perfect” play would look like.
I keep forgetting that goons need an upgrade for 6 range or that zealots don't start with speed
pvp bot skipped detection. IMO thats more cheezy than mudman
Answer: with 7000+ APM, you WALK across the map! 6:05
man, those dragoons derping with the ramp, I feel like the original Brood War is better than PurpleWave on that -_-
Purplewave peaked over 9700 APM. What you do with that is jiggle dragoons at the bottom of a ramp without figuring out how to go up and fight.
how come this is Remastered? I thought Schnail does not support Remastered. Was it old news?
Reach 2005 vs JY modern. Who is winning the game if that happened?
Having recently just played some games against the base BW AI, these bots are miles better, its not even close but maybe that just shows how bad the original AI is.
Protoss is so hard, not even AI can figure it out 😢
When are you going to cast and upload ASL 17 round 24?
When will they make actual "intelligent" ai (not just pre programmed BO and brute force APM micro) that makes correct human like decision, like going up the ramp with 12 dragoons to kill 3 cannons
I'm only at 3:24 but WE GOT A MUDMAN???
I wonder if PurpleWave creator used to be in my clan!!
I enjoy
I was kinda hoping these AIs would do a bit better, but still very entertaining nonetheless.
Almost 10k APM - so how much in terms of coolant measured in BTUs were needed to allow an A.I. to play that game?
A few more Terran Turrets I'd say, with less tank siege-unsiege, then sprinkle in something FUNKY like Medics (with Cleanse) later on!
"A little bit of harassment here. I like that."
- Artosis, 2024 (or was it the Joker in Batman?) ;)
Some interesting micro moments here and there but overall it seems like these ai fall flat when it comes to builds and timings.
I like watching the bots
Thanks Arty
Hopefully we can watch the first matches Jaeyun played against these AI. It’s unfortunate we just see the subsequent matches where he’s learned how to defeat them.
Watching these videos makes me feel relieved that AI isn't ready to take over us just yet.
This just reminds me of playing against/abusing the crap out of the computer opponent in old BW. The AIs seem pretty dumb!
I don't think there was anything exploitative with that first game. The AI is simply not coded to handle concepts like "goons going up ramps" and "carriers exist". Which would in my mind make it pretty bad at fighting anything except a strict dragoon build AI. Humans are simply too good to beat.
Looks like zerg bot didn't make changes to strat even though it scouted protoss. Like that scouting was a fiction to make the bot seem to have strategic decision making about the army conposition
ITS OVER 9000?!!!!!!
i have 300 games at least in f rank and i can confidently say it would wipe my ass
It doesn't matter how high the apm is if the person who programmed the ai doesn't know what to do with it
I love the idea of this tournament, but I think the bots need a bit more time cooking. 9 times out of 10 they just do nothing while getting slowly killed. It's just not very fun to watch after the novelty wears off.
AI needs to play a lot more aggressive
Damn, these be some real bad bots
If anybody is interested, the A.I. bots from the SC2 hobbyists seem more fully developed (such as Eris) and they have some tournaments on youtube.
aww dude I thought I already subbed.... fk it, doing it now.
ITS OVER 9000!
Cant you give each Marine 1000 APM and they are simply immortal and kill everything ?
Well I learned that I root for terran even when it is a weird bot. I was also hoping to see 7000 apm makes ghosts suddenly OP against protoss but didn't happen. Definitely seems glitchy, almost like the AI tries to micro each individual unit too perfectly it doesn't properly understand the group dynamics. Like it misses the forest for the trees.
Feels like JY may have chosen the map for this match... it was definitely 100% an abusive style, not really " showing " how the A.I works.
human power!
What do you do with 7000 apm?! Everything.
You know all and see all, you are the early release of alphastar that was everywhere and doing everything, all at once.
Guys just wouldn't give the bots any opportunity to show off their macro late game. Sad.
How do protosses doubt that carriers are op against terran? Just look at these games)
It’s funny watching the AI consistently have over 1000 APM, reaching over 9000 at some points and accomplishing almost nothing with it. I wonder if they capped it at a lower APM, would it more effectively units instead of attempting to micro every single unit at once. Specifically the first match where all the goons just stood there shimmying.
Do note that I’m a moron and know nothing about AI so idk if that has anything to do with it or not.
The problem in the first game wasn't the 'micro', it was the decision not to move up the ramp. The 'micro' isn't the main issue (although I am surprised how bad it is; I would have at least thought the mutalisks would kill the archon without taking hits); the main issue is a lack of reasonable strategic decision-making.
These AIs are just all-around bad. The fact that the 1-base carrier strategy actually worked means that the AI would lose to some extremely bad players.
they are programmed, those AI didn't learn how to do things rather there are clear commands how to do things. Thus if they get capped the commands have to be rewritten to make do with the cap.
Some points about the high APM: Bots can't do group commands, they have to order each individual unit. Bots often issue the same order every few frames. It's much easier than to check if the order is the same.
Limiting the APM would essentially just be more work to prevent some of that. But without any additional benefit.
@@yevgeniygrechka6431 These are hobbyist bots that AFAIK incorporate really minimal machine learning, if any. So for example the Protoss AI is basically hardcoded to individually micro goons when they're shot at. In a vacuum this means the AI would win any dragoon vs dragoon fight on equal numbers. In practice you see stuff like this disaster on the ramp.
@@Xaan-lm4hb I'm not convinced that the AI would even win a dragoon vs dragoon fight after the ZvZ mutalisk battle yesterday. I feel like the only thing the AI can do better than humans is make non-stop workers and move units into an arc, and that by itself just doesn't win games.
8:55 AI at least could manage to oneshot interceptors with dragoons. Meh, dissapointment
Its not stupid it's PROGRAMED that way 😂😂😂
Programmed to be stupid. 🤣
Looks like this ai is either too small or haven't seen enough games
it's not machine learning ai
these bots are worse then the AI in game ...
Pretty underwhelming games...
I'd expect from an AI with 7k apm to micro mutas one by one and completely destroy archons, for example. Or split them perfectly and attack multiple parts of the enemy base simultaniously. But they do..nothing.
They can't even get up a ramp in the first game, in PvP...it's hard to get dragoons up to a ramp, but even a 200 apm guy can do that.
The Al in these matches seem like they’re on the same level as built in ai in the game. I would be curious to see how more modern alphastar would play sc1.
I feel cheated out of the daily human/human match :/
This series confirms that AI can't replace human
Thats what AI would want you to think)
*yet
i know you need about 5000+ APM in order to play terran properly.
The bots certainly won't be higher than B rank, my guess is that they are at a C level, but I wouldn't be surprised if they are D.
This one was even more one-sided than the first one. I feel like I could beat these bots, and I suck 😅
int
these bots don't seem much better than the base game AI
I think what we learned here is that AI is incapable of decision making of any kind. It is good for teaching, because it teachers you how to optimizing, and how to not make stupid mistakes. But once you have a reasonable understanding of the game, you can game the AI, because their response is predictable, because it has to be for it to be scripted.
A true, generally AI playing starcraft couldn't be beaten.
Probably not, but with the current tech it would not run in real-time on any consumer hardware. Not to speak of the training cost...
Looking forward to getting back to real games with humans vs. humans... If the AIs had been better this may have been fun for a single day, but still not several days in a row.
Kinda interesting, kinda not.
Artosis. Big fan 10+ years. With all due respect, these videos are boring against AI.
: )
I feel like these ai's should be able to beat the original game ai before going against pros.
Almost all of them do.
@@Bytekeeperyea, almost.