|
Chess discussion continues here |
On November 17 2013 19:29 urboss wrote:Show nested quote +On November 17 2013 19:23 hypercube wrote:On November 17 2013 19:05 urboss wrote:
As mentioned before, the information you have during a StarCraft game is incomplete because of the fog of war. That makes StarCraft theoretically not solvable in the same way that chess would be solvable using a brute-force method. There are ways to deal with incomplete information. It's called Game Theory. Rock-Paper-Scissors has been 'solved', even though it has incomplete information. The best RSP player could not beat the computer in the long run. Imagine that the computer would have to base its decisions solely on the path a stalker has taken over the last 60 seconds. A stalker can move into (at least) 9 different directions every millisecond. The engine would have to calculate its decision based on all the paths the stalker could have taken. Yes, and a good program would likely deal with it the same way a human does: assume that most of these paths are fundamentally the same. So it would only need to look maybe 5-10 different paths (one from each class of 'essentially same' paths). Then come up with with a number of different responses and find the Nash Equilibrium between its responses and those 5-10 different paths. Yes, all of what you mention would enable computers to beat humans. But none of what you mention makes StarCraft solvable in the same way that chess is solvable.
You need to define solvability differently for games with hidden information. Rock-Paper-Scissors is solved as far as I'm concerned. There's nothing new to say about the game at all.
BTW, Chess is not really solvable with current technology. Chess engines make huge simplifications and assume some positions are clearly bad without any proof.
|
On November 17 2013 19:13 siri wrote:Show nested quote +On November 17 2013 19:02 Yurie wrote:On November 17 2013 18:52 siri wrote:On November 17 2013 17:53 kusto wrote:
bla bla bla
I already told you to google automaton 2000 to see how easy AI would beat humans. I googled it, didn't find anywhere where it played a game against a human. Just a few good micro tricks. So you are blind to the potential of this micro "tricks"? Cant you brain go any futher? Put this micro "tricks" in a protoss work rush and it would beat any human. Just like that
I honestly think I could beat a protoss worker rush if I knew it was coming. Regardless of micro.
|
On November 17 2013 19:23 hypercube wrote:Show nested quote +On November 17 2013 19:05 urboss wrote:
As mentioned before, the information you have during a StarCraft game is incomplete because of the fog of war. That makes StarCraft theoretically not solvable in the same way that chess would be solvable using a brute-force method. There are ways to deal with incomplete information. It's called Game Theory. Rock-Paper-Scissors has been 'solved', even though it has incomplete information. The best RSP player could not beat the computer in the long run. Show nested quote +Imagine that the computer would have to base its decisions solely on the path a stalker has taken over the last 60 seconds. A stalker can move into (at least) 9 different directions every millisecond. The engine would have to calculate its decision based on all the paths the stalker could have taken. Yes, and a good program would likely deal with it the same way a human does: assume that most of these paths are fundamentally the same. So it would only need to look maybe 5-10 different paths (one from each class of 'essentially same' paths). Then come up with with a number of different responses and find the Nash Equilibrium between its responses and those 5-10 different paths.
That argument about RPS isn't valid; the computer can't beat the human either. This game by definition is drawn...
|
On November 17 2013 20:05 gobbledydook wrote:Show nested quote +On November 17 2013 19:23 hypercube wrote:On November 17 2013 19:05 urboss wrote:
As mentioned before, the information you have during a StarCraft game is incomplete because of the fog of war. That makes StarCraft theoretically not solvable in the same way that chess would be solvable using a brute-force method. There are ways to deal with incomplete information. It's called Game Theory. Rock-Paper-Scissors has been 'solved', even though it has incomplete information. The best RSP player could not beat the computer in the long run. Imagine that the computer would have to base its decisions solely on the path a stalker has taken over the last 60 seconds. A stalker can move into (at least) 9 different directions every millisecond. The engine would have to calculate its decision based on all the paths the stalker could have taken. Yes, and a good program would likely deal with it the same way a human does: assume that most of these paths are fundamentally the same. So it would only need to look maybe 5-10 different paths (one from each class of 'essentially same' paths). Then come up with with a number of different responses and find the Nash Equilibrium between its responses and those 5-10 different paths. That argument about RPS isn't valid; the computer can't beat the human either. This game by definition is drawn...
Sure but that's just the way the game is. In some games if you play in the 'theoretically' best way you have less chance to win against an inferior opponent.
This is true for some chess positions too. There are positions where you can play for a draw or play a deep trap that loses against perfect play. E.g. imagine a position where the only choices are simplifying to a drawn ending or playing an unsound but dangerous sacrifice. Best play draws but the inferior move wins against much weaker opponents.
The starting position is not like this. Good play (especially with white) will increase your chances of winning and decrease your chances of losing. But this is not a general rule.
|
Solvability relates to Starcraft about as much as it relates to baseball. Theoretically there is an optimal way to play baseball based on simulating the universe. Except that this is completely nonsensical and irrelevant because many implicit assumptions you have made fail to be met: for instance that optimal play is the best way to play if you want to win. Use this approach for Starcraft and it might discover that one race is "stronger" than another race because it is theoretically always winning, except that this has nothing to do with real strength because humans (and computers too) have limitations and won't play Starcraft as if it's a turn-based game where they can use their database that has infinity more values than all the atoms in the universe. It's also completely irrelevant to building an AI (unlike chess where it is useful) These concepts just can't be used, it's honestly a scientific crime because in science you have to be very rigorous in defining not only concepts but also all the assumptions that need to be met before the concept can be applied to a problem. If you use them haphazardly it leads to ridiculous results like bringing up solvability of Starcraft.
Also, try beating this by playing about a 100 rounds without any outside help:
http://www.nytimes.com/interactive/science/rock-paper-scissors.html?_r=0
|
Why does it matter if Chess is "solved" by a computer? These are not 2 computers playing each other, they are human beings and as such they make mistakes. Thats the beauty off it.
Outside of preparation chess computers have 0 influence on this game. Once the players sit down its up to them to beat the other.
|
I'm sure if some company put $100 million into building a perfect BW AI, no one in the world would be able to touch that in a BO7 either.
Hell, we have computers that crush people in Jeopardy. Doesn't get much more incomplete information than that.
|
On November 17 2013 20:39 Grumbels wrote:Solvability relates to Starcraft about as much as it relates to baseball. Theoretically there is an optimal way to play baseball based on simulating the universe. Except that this is completely nonsensical and irrelevant because many implicit assumptions you have made fail to be met: for instance that optimal play is the best way to play if you want to win. Use this approach for Starcraft and it might discover that one race is "stronger" than another race because it is theoretically always winning, except that this has nothing to do with real strength because humans (and computers too) have limitations and won't play Starcraft as if it's a turn-based game where they can use their database that has infinity more values than all the atoms in the universe. It's also completely irrelevant to building an AI (unlike chess where it is useful) These concepts just can't be used, it's honestly a scientific crime because in science you have to be very rigorous in defining not only concepts but also all the assumptions that need to be met before the concept can be applied to a problem. If you use them haphazardly it leads to ridiculous results like bringing up solvability of Starcraft. Also, try beating this by playing about a 100 rounds without any outside help: http://www.nytimes.com/interactive/science/rock-paper-scissors.html?_r=0
Played 31 rounds, and am up 9-15-7. Wow.
|
On November 17 2013 21:05 GolemMadness wrote:Show nested quote +On November 17 2013 20:39 Grumbels wrote:Solvability relates to Starcraft about as much as it relates to baseball. Theoretically there is an optimal way to play baseball based on simulating the universe. Except that this is completely nonsensical and irrelevant because many implicit assumptions you have made fail to be met: for instance that optimal play is the best way to play if you want to win. Use this approach for Starcraft and it might discover that one race is "stronger" than another race because it is theoretically always winning, except that this has nothing to do with real strength because humans (and computers too) have limitations and won't play Starcraft as if it's a turn-based game where they can use their database that has infinity more values than all the atoms in the universe. It's also completely irrelevant to building an AI (unlike chess where it is useful) These concepts just can't be used, it's honestly a scientific crime because in science you have to be very rigorous in defining not only concepts but also all the assumptions that need to be met before the concept can be applied to a problem. If you use them haphazardly it leads to ridiculous results like bringing up solvability of Starcraft. Also, try beating this by playing about a 100 rounds without any outside help: http://www.nytimes.com/interactive/science/rock-paper-scissors.html?_r=0 Played 31 rounds, and am up 9-15-7. Wow.
You should try to outsmart it, like with "I played x1..xn last n rounds so now the computer will expect me to play a but I will play b". I don't think you can win. :o
|
Although engines can be of great help to understand the play of GMs, one needs to be careful not to watch the games purely based on engine evaluation.
In game 6, this is how people with engines may have judged the game:
"The game was drawn until the move 60. Ra4.
If Anand hadn't blundered move 60, the game would have been drawn, what an idiot!"
![[image loading]](http://i.imgur.com/zjOdbOT.png)
People perceive Anand's move as a blunder and think that this was the crucial moment of the game.
In fact, at that point both players thought that Black won the game already.
It's much harder to appreciate Carlsen's move 57.... Kf4 by just looking at the engine evaluation.
Of course, from an engine point of view 57. ... Kf4 draws just as 57. ... Kh5 or 57. ... Kf6 does.
![[image loading]](http://i.imgur.com/2Rl5zpO.png)
If you were to play at this position, you see that 57.... Kf4 gives up the c pawn and gives White 2 connected passed pawns.
At this point, you have probably discarded this option already, as did the GMs that were commentating.
Jerry was even shouting "Did Carlsen just blunder the pawn?".
It takes the genius of Carlsen to set up a trap like that in an endgame where everyone expected a draw.
|
13-7-9 for me after 30 rounds. (of the rock paper scissors thing)
Chess games have been great, shame I missed last night's, but watched a few analyses, and just love watching the Carlsen end game. Such a treat
|
The rock-paper-scissors program is really weak, at least early on. 15-10-5 after 30 rounds.
I've been enjoying the match; such a shame that it gets all the attention, while the European Team Chess Championship is also going on. And it has quite a lot of spectcular games.
|
On November 17 2013 21:31 Grumbels wrote:Show nested quote +On November 17 2013 21:05 GolemMadness wrote:On November 17 2013 20:39 Grumbels wrote:Solvability relates to Starcraft about as much as it relates to baseball. Theoretically there is an optimal way to play baseball based on simulating the universe. Except that this is completely nonsensical and irrelevant because many implicit assumptions you have made fail to be met: for instance that optimal play is the best way to play if you want to win. Use this approach for Starcraft and it might discover that one race is "stronger" than another race because it is theoretically always winning, except that this has nothing to do with real strength because humans (and computers too) have limitations and won't play Starcraft as if it's a turn-based game where they can use their database that has infinity more values than all the atoms in the universe. It's also completely irrelevant to building an AI (unlike chess where it is useful) These concepts just can't be used, it's honestly a scientific crime because in science you have to be very rigorous in defining not only concepts but also all the assumptions that need to be met before the concept can be applied to a problem. If you use them haphazardly it leads to ridiculous results like bringing up solvability of Starcraft. Also, try beating this by playing about a 100 rounds without any outside help: http://www.nytimes.com/interactive/science/rock-paper-scissors.html?_r=0 Played 31 rounds, and am up 9-15-7. Wow. You should try to outsmart it, like with "I played x1..xn last n rounds so now the computer will expect me to play a but I will play b". I don't think you can win. :o
Yes you can, you just have to change your patterns every x rounds (how much rounds you try to consider)
i played 100 rounds and had 45-28-27
|
On November 17 2013 21:54 urboss wrote:Although engines can be of great help to understand the play of GMs, one needs to be careful not to watch the games purely based on engine evaluation. In game 6, this is how people with engines may have judged the game: "The game was drawn until the move 60. Ra4. If Anand hadn't blundered move 60, the game would have been drawn, what an idiot!"
Can someone explain to me how b4 would become a draw? Me and a friend tried to figure it out but we suck at chess and failed to find it.
|
On November 17 2013 21:31 Grumbels wrote:Show nested quote +On November 17 2013 21:05 GolemMadness wrote:On November 17 2013 20:39 Grumbels wrote:Solvability relates to Starcraft about as much as it relates to baseball. Theoretically there is an optimal way to play baseball based on simulating the universe. Except that this is completely nonsensical and irrelevant because many implicit assumptions you have made fail to be met: for instance that optimal play is the best way to play if you want to win. Use this approach for Starcraft and it might discover that one race is "stronger" than another race because it is theoretically always winning, except that this has nothing to do with real strength because humans (and computers too) have limitations and won't play Starcraft as if it's a turn-based game where they can use their database that has infinity more values than all the atoms in the universe. It's also completely irrelevant to building an AI (unlike chess where it is useful) These concepts just can't be used, it's honestly a scientific crime because in science you have to be very rigorous in defining not only concepts but also all the assumptions that need to be met before the concept can be applied to a problem. If you use them haphazardly it leads to ridiculous results like bringing up solvability of Starcraft. Also, try beating this by playing about a 100 rounds without any outside help: http://www.nytimes.com/interactive/science/rock-paper-scissors.html?_r=0 Played 31 rounds, and am up 9-15-7. Wow. You should try to outsmart it, like with "I played x1..xn last n rounds so now the computer will expect me to play a but I will play b". I don't think you can win. :o
I was playing like that, but it takes such a long time that I didnt bother anymore after 6-8-2
|
rock paper scissors has nothing logical behind it, what is the use of that "computer"? lol
|
On November 17 2013 20:39 Grumbels wrote:Solvability relates to Starcraft about as much as it relates to baseball. Theoretically there is an optimal way to play baseball based on simulating the universe. Except that this is completely nonsensical and irrelevant because many implicit assumptions you have made fail to be met: for instance that optimal play is the best way to play if you want to win. Use this approach for Starcraft and it might discover that one race is "stronger" than another race because it is theoretically always winning, except that this has nothing to do with real strength because humans (and computers too) have limitations and won't play Starcraft as if it's a turn-based game where they can use their database that has infinity more values than all the atoms in the universe. It's also completely irrelevant to building an AI (unlike chess where it is useful) These concepts just can't be used, it's honestly a scientific crime because in science you have to be very rigorous in defining not only concepts but also all the assumptions that need to be met before the concept can be applied to a problem. If you use them haphazardly it leads to ridiculous results like bringing up solvability of Starcraft. Also, try beating this by playing about a 100 rounds without any outside help: http://www.nytimes.com/interactive/science/rock-paper-scissors.html?_r=0
i went 35-30-35, perfect tie. is that good or bad?
|
your Country52798 Posts
On November 17 2013 20:39 Grumbels wrote:Solvability relates to Starcraft about as much as it relates to baseball. Theoretically there is an optimal way to play baseball based on simulating the universe. Except that this is completely nonsensical and irrelevant because many implicit assumptions you have made fail to be met: for instance that optimal play is the best way to play if you want to win. Use this approach for Starcraft and it might discover that one race is "stronger" than another race because it is theoretically always winning, except that this has nothing to do with real strength because humans (and computers too) have limitations and won't play Starcraft as if it's a turn-based game where they can use their database that has infinity more values than all the atoms in the universe. It's also completely irrelevant to building an AI (unlike chess where it is useful) These concepts just can't be used, it's honestly a scientific crime because in science you have to be very rigorous in defining not only concepts but also all the assumptions that need to be met before the concept can be applied to a problem. If you use them haphazardly it leads to ridiculous results like bringing up solvability of Starcraft. Also, try beating this by playing about a 100 rounds without any outside help: http://www.nytimes.com/interactive/science/rock-paper-scissors.html?_r=0
picking randomly I went 3-16-19, mindgaming I went 19-0-7, didn't expect that to occur...
|
On November 17 2013 22:39 Wesso wrote:Show nested quote +On November 17 2013 21:54 urboss wrote:Although engines can be of great help to understand the play of GMs, one needs to be careful not to watch the games purely based on engine evaluation. In game 6, this is how people with engines may have judged the game: "The game was drawn until the move 60. Ra4. If Anand hadn't blundered move 60, the game would have been drawn, what an idiot!" Can someone explain to me how b4 would become a draw? Me and a friend tried to figure it out but we suck at chess and failed to find it.
![[image loading]](http://i.imgur.com/un7jbU3.png)
The best defense for White is to get the white rook to the 7. or 8. rank to give checks to the Black king from there.
Mind that if the white rook plays back immediately with 60. Ra8, then the Black king doesn't need to protect the pawn on f4 anymore and can step into f2 which eventually wins because the g2 pawn falls.
Anand's idea with 60. Ra4 is to keep the pawn on f4 under attack and to give checks to the king laterally. As was seen in the game, this fails because White's pawns will obstruct the white rook from giving checks.
In the actual game, 62. c4 is not the best defense. Instead the rook should come back to the 8th rank to give checks.
Let's see how Black can win after 60. Ra4 if White uses the best defense:
After 60. Ra4 h3
61. gxh3 Rg6
62. Ra8
![[image loading]](http://i.imgur.com/HrIecMv.png)
The rook comes back to give checks from behind.
62. ... f3
63. Re8+ Kf2
64. b4
![[image loading]](http://i.imgur.com/bjyzFtM.png)
Now the black rook will invade and place itself on e1.
64. ... Rg2+
65. Kh1 Rg1+
66. Kh2 Re1
![[image loading]](http://i.imgur.com/lsGern9.png)
Then the black king can sneak to f1 and the black pawn will promote.
67. Rf8 Ke2
68. Rd8+ Kf1
This is one way how Black can win.
However, what happens if 60. b4 is played instead?
60. b4 h3
61. gxh3 Rg6
62. Rc8
![[image loading]](http://i.imgur.com/aZbxXGM.png)
Again, the rook comes back to the last rank to give checks.
The only difference now is that the b pawn is more advanced.
62. ... f3
63. Re8+ Kf2
64. b5
![[image loading]](http://i.imgur.com/fAPOFiK.png)
Here, we try again the same operation as before, placing the rook on e1:
Rg2+
65. Kh1 Rg1+
66. Kh2 Re1
In this case, the white rook can take on e1.
67. Rxe1 Kxe1
![[image loading]](http://i.imgur.com/WUyP3Dy.png)
Both pawns will queen at the same time in the end.
This is just one out of many possible lines.
All the draws are based on the fact that White's b pawn is more advanced and that only checks from the 7th or 8th rank are working.
|
On November 17 2013 23:22 sharkie wrote:
rock paper scissors has nothing logical behind it, what is the use of that "computer"? lol
ppl think alike.
there is some certain things that repeat over and over again.
for example, its more likely one will play rock AGAIN after a rock-rock tie, and if it ties again, its even more likely the rock will be played again. theroretically, it should still be 33%, but in this example, humans think "there is no way he is balled enough to play rock 3 times in a row".
In my experience, ppl tend to play something else after the 3rd tie.
again, this is just examples, but there are just certain behaviours that repeat, and the "computer" learned what these tend to be.
|
|
|
|
|
|
EPT
