Hate Speech: Theory Fighter University: Remedial Math
- Thread starter Dr. Hates
- Start date
Sp1d3r
Demon Beaver
Yeah. The one on tiers is probably his best so far (as far as helping me progress)
Belial
King of Hundred Swords
I didnt mean to post much, really, but that "Hate speeches" are the best thing that happened to 8wr in some time. Discussion here is something I've been pondering for YEARS now, so I'll drop my 2 cents.
1) About Hates: yes, his math is _flawed_ in a way, but given its an article he's doing tremendeously good job at delivering the idea. I think he suggest the best concept so far. And also he is totally right on nonstandart positions. it is something that's better adressed with logic, rather than math.
2) About(and to) Nirf: In fact is right. I actually described that here http://caliburforum.com/forums/showthread.php?t=34608 take a look, last part where I describe matrices equations to solve risk-reward.
It is in no way contradictional to your ideas, but I want to point out one important thing. I came to realize over the years (and I've done matrices, like, since SC2) so I think there is a problem. SC generally has too much variables. And its really easy to dig into options forever and at uneccesary times. It is hard to keep too much stuff in mind. Its very difficult to calculate proper risk-reward for things like step, moves that give mixup on hit/block, delayed atacks and many other things. Sadly, some randomness still happen.
I want to separately talk about moves postion on hit/block. It is very hard to represent that with numbers. Hit gives you mixup, KD give you wakeup, if your move gets blocked/punished, your opp has mixup/wakeup. You dont want to calc all this. Just dont.
What you really want to do is to take 2-3 BEST reactions to most difficult situations and discrard the rest. If your opponent learns to beat those 2-3 options you adjust your game naturally.
But, speaking from experience - it is very hard for opponent to find a solution to 3 diverse options, that offer enough multi-coverage. Lets speak about when you're on defence: Its fair to say 99% opponents I ever played end up _doing nothing_ when faced with reverse mixup.
So here you dont want to seek for "perfect equilibrium", which, in case of freezing opponent, would be extremely complex (or applying your own mixup, which has terrible risk-reward you see). Usually what you want to do is to note your opponent habit - does he freeze and atack if nothing happen. how much it takes. does he throw? etc. And do something to punish him. Actually I've developed a number of habits, like using TC/TJ (MST for example) moves or step-atack when opponents freeze etc. If needed it all could be put into a matrix as well, though.
Another reason why its good to only limit yourself to so much math is b/c
a) You can only handle a few options yourself at a time
b) your health is finite.
You dont really want to calc all the small things. In reality most of the time you want to apply your primary mixup, with just a little adjustments. There are many "small" mixups in the game, that are favorable, but I just usually take a risk at randomness to deal 50 than guaranteed mixup for 20 or so. For example dashing/stepping before atacking (multicoverage in a sort, since it gives better tracking and will beat GI's).
It is really important how comfortable an option feel to a player. If there is an option that's only used in one situaion, one matchup its not that good, since its a lot of distraction.
3) Bibulus: Your point is also good. You cant completely substract human factor from SC. Though it is possible to ultimately solve things with math, it is much easier in effort/result regard, when you use both math _and_ logic.
PS: There is a very good book, that helped me understand a lot about how humans learn and apply this to my view on fighting games - On Intelligence by Jeff Hawkins. Recommended to every dedicated theory fighter (at least the first part or the book is, the second is a bunch of scientific in-depth stuff)
PS2: I've been trying to put together an article about, well, lots of things in fightings. Hopefully threads like this finally will get me motivated enough to finish it.
1) About Hates: yes, his math is _flawed_ in a way, but given its an article he's doing tremendeously good job at delivering the idea. I think he suggest the best concept so far. And also he is totally right on nonstandart positions. it is something that's better adressed with logic, rather than math.
2) About(and to) Nirf: In fact is right. I actually described that here http://caliburforum.com/forums/showthread.php?t=34608 take a look, last part where I describe matrices equations to solve risk-reward.
It is in no way contradictional to your ideas, but I want to point out one important thing. I came to realize over the years (and I've done matrices, like, since SC2) so I think there is a problem. SC generally has too much variables. And its really easy to dig into options forever and at uneccesary times. It is hard to keep too much stuff in mind. Its very difficult to calculate proper risk-reward for things like step, moves that give mixup on hit/block, delayed atacks and many other things. Sadly, some randomness still happen.
I want to separately talk about moves postion on hit/block. It is very hard to represent that with numbers. Hit gives you mixup, KD give you wakeup, if your move gets blocked/punished, your opp has mixup/wakeup. You dont want to calc all this. Just dont.
What you really want to do is to take 2-3 BEST reactions to most difficult situations and discrard the rest. If your opponent learns to beat those 2-3 options you adjust your game naturally.
But, speaking from experience - it is very hard for opponent to find a solution to 3 diverse options, that offer enough multi-coverage. Lets speak about when you're on defence: Its fair to say 99% opponents I ever played end up _doing nothing_ when faced with reverse mixup.
So here you dont want to seek for "perfect equilibrium", which, in case of freezing opponent, would be extremely complex (or applying your own mixup, which has terrible risk-reward you see). Usually what you want to do is to note your opponent habit - does he freeze and atack if nothing happen. how much it takes. does he throw? etc. And do something to punish him. Actually I've developed a number of habits, like using TC/TJ (MST for example) moves or step-atack when opponents freeze etc. If needed it all could be put into a matrix as well, though.
Another reason why its good to only limit yourself to so much math is b/c
a) You can only handle a few options yourself at a time
b) your health is finite.
You dont really want to calc all the small things. In reality most of the time you want to apply your primary mixup, with just a little adjustments. There are many "small" mixups in the game, that are favorable, but I just usually take a risk at randomness to deal 50 than guaranteed mixup for 20 or so. For example dashing/stepping before atacking (multicoverage in a sort, since it gives better tracking and will beat GI's).
It is really important how comfortable an option feel to a player. If there is an option that's only used in one situaion, one matchup its not that good, since its a lot of distraction.
3) Bibulus: Your point is also good. You cant completely substract human factor from SC. Though it is possible to ultimately solve things with math, it is much easier in effort/result regard, when you use both math _and_ logic.
PS: There is a very good book, that helped me understand a lot about how humans learn and apply this to my view on fighting games - On Intelligence by Jeff Hawkins. Recommended to every dedicated theory fighter (at least the first part or the book is, the second is a bunch of scientific in-depth stuff)
PS2: I've been trying to put together an article about, well, lots of things in fightings. Hopefully threads like this finally will get me motivated enough to finish it.
Dr. Hates
Jerk.
I'd love to see your organized thoughts on a lot of this stuff. As we discovered at Evo a few years back, you and I take a very similar approach to a lot of things, theory-wise, but I've always found the crucial differences useful and interesting.
As regards my article, I at least tried to foreground the ways I "cheat" with the numbers in order to make this a useful thought experiment. The trickiest thing to deal with regarding mixups is that the more you know what your opponent will do, the more effectively you can stop it. As such, optimal strategies are incredibly fragile things. I'm actually planning to engage with this concept in the relatively near future, because you can actually turn an opponent's good decision-making against him with a bit of effort.