I guess I get bored to easily...
http://netfiles.uiuc.edu/tlantz/www/myth/tflmeleestrategy1.pdf
Maybe funny to someone...
6 pages to basically say that Myth is like Chess because at each timestep you choose an action to improve your position, and your opponent(s) do the same, then asking if a computer could play Myth perfectly. (Myth is actually closer to Backgammon because of the chance element in Myth. But computers can play Backgammon too.)
I realized this myself a while ago. So here are a few thoughts:
- To program our AI player, it would be much simpler to use the Myth game state directly than to recover it from camera control and visual processing (like we humans do). The game state has the position of units we cannot see and even the random seed (making it possible to predict if a bottle will dud or not). It would be tempting for the AI to use that information (but would it be fair to do so?).
- With access to the random seed, the game becomes deterministic again, like Chess instead of Backgammon.
- Because of unpredictable network lag, it is never possible to know for sure that any action you do will occur at the right timestep, reintroducing chance into the game.
- All finite-state games can be solved perfectly in theory. The problem is that the amount of computation and the memory requirement can exceed what current machines offer, or even what the universe can fit. Currently Tic-Tac-Toe can be played perfectly, but not Chess. (Computers can play damn good Chess, but not *perfect* chess). I believe that Myth is even harder to play perfectly than Chess because there are many more game states and many more timesteps per game.
- Even if it is not perfect I have no idea how good a Myth AI would be if someone smart worked on this for a while. I wish I had enough free time to try this myself. A simpler related project would be to write a bot that would micromanage my units to dodge incoming projectiles.
I realized this myself a while ago. So here are a few thoughts:
- To program our AI player, it would be much simpler to use the Myth game state directly than to recover it from camera control and visual processing (like we humans do). The game state has the position of units we cannot see and even the random seed (making it possible to predict if a bottle will dud or not). It would be tempting for the AI to use that information (but would it be fair to do so?).
- With access to the random seed, the game becomes deterministic again, like Chess instead of Backgammon.
- Because of unpredictable network lag, it is never possible to know for sure that any action you do will occur at the right timestep, reintroducing chance into the game.
- All finite-state games can be solved perfectly in theory. The problem is that the amount of computation and the memory requirement can exceed what current machines offer, or even what the universe can fit. Currently Tic-Tac-Toe can be played perfectly, but not Chess. (Computers can play damn good Chess, but not *perfect* chess). I believe that Myth is even harder to play perfectly than Chess because there are many more game states and many more timesteps per game.
- Even if it is not perfect I have no idea how good a Myth AI would be if someone smart worked on this for a while. I wish I had enough free time to try this myself. A simpler related project would be to write a bot that would micromanage my units to dodge incoming projectiles.
-
iron
- Site Admin
- Posts: 2006
- Joined: Thu Feb 26, 2004 1:21 am
- Location: diving out of the Sun at 10 o'clock high!
- Contact:
This reminds me ... the webpage of the old TFL order Foundation *F* had some seriously good strategy articles for melee and many other aspects of TFL. I remember one on carpet bombing espousing the L method that was really in depth.
Pity is that site disappeared some years ago, and without even knowing its URL I can't even look it up on web.archive.org :\
Edit: Amazing what a bit of googling will do ... Here it is
Can't find the article on cbing, but there's plenty of other good stuff there.
Edited By iron on 1139663232
Pity is that site disappeared some years ago, and without even knowing its URL I can't even look it up on web.archive.org :\
Edit: Amazing what a bit of googling will do ... Here it is
Can't find the article on cbing, but there's plenty of other good stuff there.
Edited By iron on 1139663232
Cool paper. Reads like nostaglia, and academia at the same time. Now if I could just get a license for Power Animator.
Iron you are master of the archive.org. I hope someone's backing up these forums cause it just kills me to see those forum titles and not be able to read them. If only spammers hadn't killed Usenet we could all still be posting at alt.games.myth and saved forever by Google or Henry Spencer.
Iron you are master of the archive.org. I hope someone's backing up these forums cause it just kills me to see those forum titles and not be able to read them. If only spammers hadn't killed Usenet we could all still be posting at alt.games.myth and saved forever by Google or Henry Spencer.
-
Eddaweaver
- Posts: 1026
- Joined: Mon Sep 13, 2004 6:05 am
- Location: M.E.
On the TFL nostalgia subject, this was one of the first Myth websites on the web. It sat unchanged for about eight and a half years until it went down sometime within the last few months:
http://web.archive.org/web....ttp
You can read alot of interesting things about the early history of Myth @ groups.google.com . Did you know the original Myth TFL demo won't work if your Windows 95 language settings at set to German?
" Is that what they're called? How do you kill them? I had this whole big
mob of guys and I was kicking butt,then I went after this big blue
motherf&^&**er and he turned my big mob into a pile of blood and guts!
ARRRHHGGHH! "
Hehe
Does anyone have this TFL demo?:
http://groups.google.com/group....9c2cfe7
I know there was an original TFL demo release, and a re-release with a TCP/IP BSOD bug fixed, but I haven't heard of a 3rd release with Shadow Of The Mountain before. Someone also mentioned a German-language release of the demo.
Anyone know what multiplayer map he's talking about in this?:
http://groups.google.com/group....4&fwc=1
Edited By Eddaweaver on 1139766689
http://web.archive.org/web....ttp
You can read alot of interesting things about the early history of Myth @ groups.google.com . Did you know the original Myth TFL demo won't work if your Windows 95 language settings at set to German?
" Is that what they're called? How do you kill them? I had this whole big
mob of guys and I was kicking butt,then I went after this big blue
motherf&^&**er and he turned my big mob into a pile of blood and guts!
ARRRHHGGHH! "
Hehe
Does anyone have this TFL demo?:
http://groups.google.com/group....9c2cfe7
I know there was an original TFL demo release, and a re-release with a TCP/IP BSOD bug fixed, but I haven't heard of a 3rd release with Shadow Of The Mountain before. Someone also mentioned a German-language release of the demo.
Anyone know what multiplayer map he's talking about in this?:
http://groups.google.com/group....4&fwc=1
Edited By Eddaweaver on 1139766689
-
iron
- Site Admin
- Posts: 2006
- Joined: Thu Feb 26, 2004 1:21 am
- Location: diving out of the Sun at 10 o'clock high!
- Contact:
Sounds like Ground Zero to me.Eddaweaver wrote:Anyone know what multiplayer map he's talking about in this?:
http://groups.google.com/group....4&fwc=1
iron wrote:This reminds me ... the webpage of the old TFL order Foundation *F* had some seriously good strategy articles for melee and many other aspects of TFL. I remember one on carpet bombing espousing the L method that was really in depth.
Nice find! (the links to the foundation strategy pages) This is just the kind of stuff I am trying to preserve at the myth graveyard.
Lots of Myth stuff at http://mythgraveyard.org.
Sometimes I put hard to find stuff in my my Udogs folder.
Sometimes I put hard to find stuff in my my Udogs folder.