Deck testing simulator
Deck Help forum
Posted on May 11, 2016, 9:47 p.m. by firebird2k1
Anybody know of a massive deck testing simulator that could pit 2 decks together for a bunch of games (say around 100) and generate results such as win/loss and moves?
It's probably possible, I'm pretty sure that although chess has fewer rules, it has a similar decision tree and has even more possible paths to take. Chess AI are basically "solved", so I think it might be programmable. But, no nothing like what OP is asking for exists AFAIK.
May 12, 2016 12:49 a.m.
Gidgetimer says... #4
Chess AI uses brute force simulations to determine optimal lines. Such a thing would never work in MTG where the number of possible moves per turn is much higher. MTG is closer to Go in the amount of possibilities/turn. Only recently has a computer been able to play Go at the highest levels. AlphaGo runs on 1,202 CPUs and 176 GPUs though so unless you have a mainframe you couldn't simulate Go at anything approaching being considered "solved".
The major problem actually isn't even the possibilities/turn though. The real problem with simulating MTG games is that it isn't a game of perfect information. You have to make decisions based on intuition and guesswork and computers are horrible at such things.
May 12, 2016 5:23 a.m.
Cool did not know that, I'm a materials engineer by trade and only a hobbyist in the field of computing. I think I need to do more research.
May 12, 2016 8:26 p.m.
Gidgetimer says... #6
I only take a passing interest in computer simulation of games. I knew about the problems they were having with getting a computer to play Go. AlphaGo kinda came out of nowhere though. I only learned about it about 6 months after it beat the European Champ 5-0 (the first time any computer had beat any pro without a handicap) and then went on to beat the world champ 4-1.There was no real build up or competition like there was with the chess computers. Google premiered AlphaGo and it won convincingly, end of story.
The problems with simulating magic at a high level comes like I said from it not being a game of perfect information, and also with the rotating piece sets. Chess and Go have had the same pieces and rules for hundreds or even thousands of years. Magic has had its exact rules and piece set for 35 days now.
I'm sure simple simulators could exist, but they would have an accuracy and margin of error of getting an experienced player looking at both decks and evaluating them against each other. As such they are not going to be very popular and most likely anyone with enough knowledge to program such a thing wouldn't see the value of it.
Gidgetimer says... #2
I'm unaware of any such program. The logistics needed to program each deck's optimal moves would be frankly staggering.
May 11, 2016 10:49 p.m.