Hacker News Reader

Comment by ctvo | Hacker News Reader

What people miss is that StarCraft is a physical game primarily, with tactical elements, not the other way around. There are a lot of fans who know the game well and can make all the right decisions, but will never be able to execute it in real time.

Execution of strategy is what separates the professional StarCraft players, and in that way it's much closer to professional sports than Go / Chess.

View on HN

I still play the original starcraft (broodwar) every week, and I'll have to agree with you.

once the AI is given micro mechanics and perfect execution, you just can't beat it: https://www.youtube.com/watch?v=dOIgFkyfStg

However, pros can still perform "cheese" tactics where they attack extremely early on, and the AI won't be able to properly defend because it needs cash to get an army. So if we do start to see deepmind eventually play starcraft, it will need lots of hand-holding and special programming to deter early cheese tactics.

I see you linked Berkely Overmind and said that its micro can't be beaten.

Don't you remember the Human VS AI matches that took place with Overmind right after the tournament that year?

Watch an amateur human tear through its mutalisks. https://youtu.be/hmXY_pkU2Wc?list=PL9F69781768E85502&t=591

The video just shows that the AI can perfect attack mechanics - it's still a simple AI. If deepmind was thrown at the game, it could learn to deceive/cheese/micro/drop and really do some serious damage.

I hope DeepCraft could do those things, since no other AI can at this point. And I hope it can do them better than any human. And I hope that DeepCraft can pivot its strategy on a dime when needed, in response to the human doing something. And I hope that it can do well against the myriad of play-styles that professionals have developed across the the three unique races.

My point is that DeepCraft is still gonna be quite a ways off, and is going to involve a lot more than just learning how to micro, cheese, drop, etc. (Which, by the way, would be quite a feat to even develop an AI that can learn these things since StarCraft has such a huge search space)

What if the pace of gameplay were slowed down 5x? Would execution still be an issue?

That is a really good question. I would love to see tournament players playing against each other and non-pros in a 5x slower or 10x slower game. Among the pros would be most interesting. Does everyone now play about equally well? Does some random barely pro destroy the field because that player has a better grasp on the strategy that was convoluted by the rapid aspect. If you compare to chess, good blitz players aren't always good at long games and vice versa. Computers are much better at fast blitz games (beating pros well before Kasparov in 1996). I expect that even if DeepCraft performs well in a full speed game a slower more strategic game would have the humans at an advantage (at least for a while).

This is like asking what basketball would be like if you weren't required to dribble. It might be interesting to think about, but would probably result in a different game.

This is like asking what basketball would be like if you weren't required to dribble.

Not really. The "physical" aspect of Starcraft is an information input/output feat. It's still arguably the same abstract game. Arguably, it's no longer a sport however. (And maybe it wouldn't be a "game" in the sense that basketball and football are "games" for people.)

Solving THAT game (where we get as close as possible to a brain -> machine interface and it becomes completely strategic) is interesting, but it's not StarCraft. Arguably it's the same abstract game, but when you bypass constraints (applicable only to humans), you probably should have targeted something else without those constraints in the first place. In our case, a turn-based strategy game that is equally complex and with limited information (Civilization?!).

Reply on news.ycombinator.com