Open Side Menu Go to the Top
Register
can game theory be used in fighting games can game theory be used in fighting games

02-24-2017 , 11:09 PM
I know it works best in zero sum games. For a contrived example say we are on offense and our opponents only options are to block high or low. Our only options are to attack mid or low. If our mid attack lands we get a juggle that leads to 110 damage points but if our mid is blocked they get 40 damage points. If our low hits we get 30 damage points. If it's blocked they get 80 damage points. To sum it up.


Mid hits=110 damage points mid is blocked we lose 40 damage points
Low hits =30 damage points low is blocked=we lose 80 damage points. Can we use math to figure the optimal ratio of mid attacks to low attacks?

Last edited by grumpy64; 02-24-2017 at 11:19 PM.
can game theory be used in fighting games Quote
02-24-2017 , 11:22 PM
we have to know the probability of the various attacks landing which depends on opponent and varies based on multiple factors including your previous ratio of the different moves

so no, you can't really math out that kind of fighting game decision

if you really want to apply math to fighting games, there are rough risk/reward calculations you can do
can game theory be used in fighting games Quote
02-24-2017 , 11:26 PM
Quote:
Originally Posted by Sun Tzu
we have to know the probability of the various attacks landing which depends on opponent and varies based on multiple factors including your previous ratio of the different moves

so no, you can't really math out that kind of fighting game decision

if you really want to apply math to fighting games, there are rough risk/reward calculations you can do
So we can't find a Nash for high low mixup
can game theory be used in fighting games Quote
02-24-2017 , 11:34 PM
OK in the above situation our advantage is 35 points if we do an exact 50/50 and our opponent blocks at an exact 50/50 ratio
can game theory be used in fighting games Quote
02-25-2017 , 01:11 AM
I think you could, in theory. But in practice, it's not going to help you.
can game theory be used in fighting games Quote
02-25-2017 , 02:25 AM
There's a very decent article out there about applying GTO to super smash brothers melee. Sry no link tho.
can game theory be used in fighting games Quote
02-25-2017 , 04:09 AM
If you were to break it down, you could say each frame in the game is a simultaneous move by both players and analyze that using game theory. But since there are (typically) 60 frames per second, the decision tree you are analyzing would be enormous. So it's not of much practical use even though mathematically it's possible.

In the specific scenario given, assuming that the payoff function is the net damage done, I'm getting the Nash equilibrium as you attack mid 42.31% of the time (11/26) and the opponent blocks mid 73.08% of the time (19/26). Not 100% sure if I did the algebra right but it's definitely possible to calculate with a bit of math to see that the payoff for those percentages is independent of the opponent's action (average payoff is equal no matter what the opponent does).

Last edited by ballin4life; 02-25-2017 at 04:17 AM.
can game theory be used in fighting games Quote
02-25-2017 , 04:26 AM
Quote:
Originally Posted by Chip Niche
There's a very decent article out there about applying GTO to super smash brothers melee. Sry no link tho.
http://alexspuffstuff.blogspot.com.a...tive-play.html
can game theory be used in fighting games Quote
02-25-2017 , 12:27 PM
Quote:
Originally Posted by ballin4life
If you were to break it down, you could say each frame in the game is a simultaneous move by both players and analyze that using game theory. But since there are (typically) 60 frames per second, the decision tree you are analyzing would be enormous. So it's not of much practical use even though mathematically it's possible.

In the specific scenario given, assuming that the payoff function is the net damage done, I'm getting the Nash equilibrium as you attack mid 42.31% of the time (11/26) and the opponent blocks mid 73.08% of the time (19/26). Not 100% sure if I did the algebra right but it's definitely possible to calculate with a bit of math to see that the payoff for those percentages is independent of the opponent's action (average payoff is equal no matter what the opponent does).
Just curious what formula did u use? I only know how to solve simple things using a matrix.
can game theory be used in fighting games Quote
02-25-2017 , 03:50 PM
Quote:
Originally Posted by grumpy64
Just curious what formula did u use? I only know how to solve simple things using a matrix.
I just wrote out the payoffs for each player and then set the payoffs equal to each other and solved using a bit of algebra.

If x = percentage of attacking mid and y = opponent percentage of blocking mid.

The attacking player's payoff function is
x*(110(1-y) - 40y) + (1-x)(30y - 80(1-y))

So the defending player's unexploitable strategy is the one where it doesn't matter what value the attacking player chooses for x, i.e. it's the one where
110(1-y) - 40y = 30y - 80(1-y)
110 - 150y = 110y - 80
190 = 260y
19/26 = y

so that's how I got 19/26 or 73.08% blocking mid as the unexploitable strategy. Same idea to find the unexploitable value of x as well, just using the opponent's payoff function. I think this method of setting the opponent's payoffs equal to each other works for all zero sum games.
can game theory be used in fighting games Quote
02-26-2017 , 12:21 AM
Quote:
Originally Posted by ballin4life
I just wrote out the payoffs for each player and then set the payoffs equal to each other and solved using a bit of algebra.

If x = percentage of attacking mid and y = opponent percentage of blocking mid.

The attacking player's payoff function is
x*(110(1-y) - 40y) + (1-x)(30y - 80(1-y))

So the defending player's unexploitable strategy is the one where it doesn't matter what value the attacking player chooses for x, i.e. it's the one where
110(1-y) - 40y = 30y - 80(1-y)
110 - 150y = 110y - 80
190 = 260y
19/26 = y

so that's how I got 19/26 or 73.08% blocking mid as the unexploitable strategy. Same idea to find the unexploitable value of x as well, just using the opponent's payoff function. I think this method of setting the opponent's payoffs equal to each other works for all zero sum games.
TY
can game theory be used in fighting games Quote
02-28-2017 , 11:00 AM
One thing that would immediately complicate this is even if the opponent is forced to eat a mixup they have multiple options Wether they block high or low. For example they may sacrifice damage on a blocked low for better oki opotions or wall carry
can game theory be used in fighting games Quote

      
m