5 ms·

>With such a cheat, roaches punch way above their weight, completely breaking the rock-paper-scissors balance of the game. Against marines, sure. But against

>With such a cheat, roaches punch way above their weight, completely breaking the rock-paper-scissors balance of the game. Against marines, sure. But against siege tanks/disruptors? Sirlin's spiel about there being more counters than people think (http://www.sirlin.net/ptw-book/introducingthe-scrub http://www.sirlin.net/ptw-book/introducingthe-scrub) seems applicable here.

vkou 8y ago
Yes, they start punching way above their weight against tanks, too. Even moreso if you throw in an auto-scatter script. https://www.youtube.com/watch?v=IKVFZ28ybQs https://www.youtube.com/watch?v=IKVFZ28ybQs
- invalidOrTaken 8y ago
  What about tanks on high ground and out of reach? What if the roaches have to funnel through a walled choke guarded by some MMLib as well? That might seem contrived, but there is a player whose job is to contrive it! No, I don't think the next patch should include "improved" roach AI. But deep strategy games like Starcraft tend to have multiple levels on which to do battle, and you can often nullify an opponent's insurmountable advantage on one level by doing battle on another level. I honestly think that if super-roaches like we've been discussing were patched into the game, with no further changes made, winrates would stabilize around 50% in a few years (assuming everyone hadn't quit in disgust). Now, that might not be the game you'd want to play. I'll freely admit I wouldn't want to play it. But that's you and me---perhaps whoever's into chess would love it. And "This would no longer be fun for me," while perfectly legitimate, is a very different claim than "the game balance breaks."
  - lodi 8y ago
    > ...deep strategy games like Starcraft tend to have multiple levels on which to do battle... That's exactly the point; the "other levels" are macro, micro, and multi-tasking--the "realtime" components of an RTS game. (I consider positioning and scouting a factor of micro and multi-tasking respectively.) Those are the facets of the game that let you take two equally matched strategies, execute slightly better than the opponent, and thereby eke out a gradually compounding advantage. If you remove those, the "deep strategy" of starcraft is basically just doing the one or two counterplays that you obviously need to do to survive. "He built too many early game marines so now I build banelings or I die." The strategic "if he does this, I'll do this, but then he'll do that" decision tree is very shallow in a game like brood war or sc2. > What about tanks on high ground and out of reach? What if the roaches have to funnel through a walled choke guarded by some MMLib as well? Then I'll be forced into one of a small handful of tictactoe-like responses: brood lords, vipers, doom drop on top of your army, or pull you apart with muta/nydus multi-taski... no wait. Just the first three options I guess.
ABCLAW 8y ago
Is it, though? Sirlin's scrub is a strawperson player that will not adapt to any changes in the game, instead asking that the game itself is changed to suit them. Running cheats is the ultimate scrub move - you change your play experience asymmetrically to benefit yourself rather than practicing at the game as offered to get better. There are a ton of interactions in games that are degenerate if performed at TAS level. These aren't areas looking for disruption; they're just impossible to perform with human inputs and they confer game advantages that are not surmountable by non-assisted players. The speed running community segregates TAS content from played content specifically because of this. You will not beat the robot that can perform a 60 input 1 frame trick that gives you a .5 second time save.
- invalidOrTaken 8y ago
  >Is it, though? Sirlin's scrub is a strawperson player that will not adapt to any changes in the game, instead asking that the game itself is changed to suit them. Running cheats is the ultimate scrub move - you change your play experience asymmetrically to benefit yourself rather than practicing at the game as offered to get better. That's a fair point; I agree. But the ancestors are debating a scenario with sanctioned mechanical aids. vkou classifies them as "cheats" above, but I humbly suggest that he used the term improperly, given the context. Also, remember that we're talking about RTS games, which have an S component as much as they have an RT component. OpenAI can beat Dendi in the early phases of solo mid, but can a team of AIs beat Navi in a full game of DotA? Perhaps they will eventually, but if they do, it will take a whole lot more than just reaction time.
  - de_watcher 8y ago
    > but can a team of AIs beat Navi in a full game of DotA? Yes it can, especially Na`Vi (Dota-reddit jokes that AI won't play with Na`Vi because developers want to test AI with a pro team). There were matches with pro players last month. Players say that mechanics in 5v5 fight are perfect and the global strategy is there. I won't say that the strategy has many states in Dota: it's the items and position on the map (by choosing one of the objectives). It's not ready for the real Dota 2 tournaments though: AI was trained for the specific 5 heroes.

vkou 8y ago

Yes, they start punching way above their weight against tanks, too. Even moreso if you throw in an auto-scatter script. https://www.youtube.com/watch?v=IKVFZ28ybQs https://www.youtube.com/watch?v=IKVFZ28ybQs

invalidOrTaken 8y ago
What about tanks on high ground and out of reach? What if the roaches have to funnel through a walled choke guarded by some MMLib as well? That might seem contrived, but there is a player whose job is to contrive it! No, I don't think the next patch should include "improved" roach AI. But deep strategy games like Starcraft tend to have multiple levels on which to do battle, and you can often nullify an opponent's insurmountable advantage on one level by doing battle on another level. I honestly think that if super-roaches like we've been discussing were patched into the game, with no further changes made, winrates would stabilize around 50% in a few years (assuming everyone hadn't quit in disgust). Now, that might not be the game you'd want to play. I'll freely admit I wouldn't want to play it. But that's you and me---perhaps whoever's into chess would love it. And "This would no longer be fun for me," while perfectly legitimate, is a very different claim than "the game balance breaks."
- lodi 8y ago
  > ...deep strategy games like Starcraft tend to have multiple levels on which to do battle... That's exactly the point; the "other levels" are macro, micro, and multi-tasking--the "realtime" components of an RTS game. (I consider positioning and scouting a factor of micro and multi-tasking respectively.) Those are the facets of the game that let you take two equally matched strategies, execute slightly better than the opponent, and thereby eke out a gradually compounding advantage. If you remove those, the "deep strategy" of starcraft is basically just doing the one or two counterplays that you obviously need to do to survive. "He built too many early game marines so now I build banelings or I die." The strategic "if he does this, I'll do this, but then he'll do that" decision tree is very shallow in a game like brood war or sc2. > What about tanks on high ground and out of reach? What if the roaches have to funnel through a walled choke guarded by some MMLib as well? Then I'll be forced into one of a small handful of tictactoe-like responses: brood lords, vipers, doom drop on top of your army, or pull you apart with muta/nydus multi-taski... no wait. Just the first three options I guess.

ABCLAW 8y ago

Is it, though? Sirlin's scrub is a strawperson player that will not adapt to any changes in the game, instead asking that the game itself is changed to suit them. Running cheats is the ultimate scrub move - you change your play experience asymmetrically to benefit yourself rather than practicing at the game as offered to get better. There are a ton of interactions in games that are degenerate if performed at TAS level. These aren't areas looking for disruption; they're just impossible to perform with human inputs and they confer game advantages that are not surmountable by non-assisted players. The speed running community segregates TAS content from played content specifically because of this. You will not beat the robot that can perform a 60 input 1 frame trick that gives you a .5 second time save.

invalidOrTaken 8y ago
>Is it, though? Sirlin's scrub is a strawperson player that will not adapt to any changes in the game, instead asking that the game itself is changed to suit them. Running cheats is the ultimate scrub move - you change your play experience asymmetrically to benefit yourself rather than practicing at the game as offered to get better. That's a fair point; I agree. But the ancestors are debating a scenario with sanctioned mechanical aids. vkou classifies them as "cheats" above, but I humbly suggest that he used the term improperly, given the context. Also, remember that we're talking about RTS games, which have an S component as much as they have an RT component. OpenAI can beat Dendi in the early phases of solo mid, but can a team of AIs beat Navi in a full game of DotA? Perhaps they will eventually, but if they do, it will take a whole lot more than just reaction time.
- de_watcher 8y ago
  > but can a team of AIs beat Navi in a full game of DotA? Yes it can, especially Na`Vi (Dota-reddit jokes that AI won't play with Na`Vi because developers want to test AI with a pro team). There were matches with pro players last month. Players say that mechanics in 5v5 fight are perfect and the global strategy is there. I won't say that the strategy has many states in Dota: it's the items and position on the map (by choosing one of the objectives). It's not ready for the real Dota 2 tournaments though: AI was trained for the specific 5 heroes.