5 ms·

It's really not possible (or practical) to plot 20 moves ahead in a chess game. Every different piece moved by your opponent would represent a new 'line' that y

by nanijoe 17y ago

It's really not possible (or practical) to plot 20 moves ahead in a chess game. Every different piece moved by your opponent would represent a new 'line' that you would have to compute 20 moves for.

pavs 17y ago
> It's really not possible (or practical) to plot 20 moves ahead in a chess game. 20 'best' moves ahead, according to the observer/player/chess-engine; assuming opponent played the best moves (according to the observer) in reply. Remembering a game played 6 years ago or being able to "think" 20 moves ahead may seem an amazing feat to regular chess players. But for professional players of GM or IM status - I don't think it is something amazing. Professional chess players memorize 100s (if not 1000s) of chess opening moves and their variations for many years. Remembering a chess game is not very impressive in that context.
- robryan 17y ago
  I'd imagine at the top level being able to see say the 20 best moves ahead would regularly occur similarly as predicted by the player and if things deviate it could be there there advantage anyway.
  - iron_ball 17y ago
    Yes, and in fact that's how openings work. Each opening is a sequence in which both players are making an optimal response to the previous move. If a player deviates from the opening, then by definition he just made a suboptimal move.
    - KC8ZKF 17y ago
      Openings aren't demonstrably optimal. (Or there would be only one.) Players make improvements on openings all the time.
      - swombat 17y ago
        There are a relatively small number of equally optimal moves (compared to the total number of possible moves) for a given opening, after white has opened in a certain way. The vast majority of moves other than the ones covered in the opening and/or its known variations are indeed sub-optimal, and quite often if someone deviates from one of the great many standard openings, this can be exploited by the opponent.
        nanijoe 17y ago
        The truth of the matter is that chess players seldom play out a single opening..Since most openings are pretty well known (along with their optimal moves), the beginning part of a chess game is usually spent trying to maneuver your opponent into an opening of your liking (or trick them into thinking you are playing one thing, then switching on them). What starts out looking like the Slav , may end up Queens Indian (as an example), and this makes the 20 moves ahead claim even more ludicrous. There are whole books on (slaying) the dragon variation of the Sicilian defense and the Sicilian is just one option that can result from the first move being e4..of course your opponent can also transpose to a different formation as the game progresses. In the end game where there are relatively few pieces on the board you may be able to plan ahead with more precision, but other than that trying to think 20 moves ahead is not just impractical, it may well be dumb.
        iron_ball 17y ago
        This is what I meant. Openings aren't traps, and openings can branch as players choose different variations, but to stray from the opening book entirely is seldom a wise move. (That's one reason I never got too far into chess; I realized that competitive play would require a lot of opening knowledge.)
- fh 17y ago
  nanijoe is still pretty much technically correct (the best kind of correct). It's just that when chess players talk about plotting "n moves ahead", they mean something different than a programmer talking about a program computing "n moves ahead". Top chess players evaluate a handful of possible lines of play very very deeply. But it's impossible for the human mind to evaluate every response and counter-response to any reasonable depth, which is why computer players dominate so much in tricky tactical positions.
nazgulnarsil 17y ago
im a casual player and i plot 10 moves. 20 isn't surprising to me.
- roundsquare 17y ago
  Well, the number of possibilities grows exponentially with the number of moves you plan ahead. Going from 10 to 20 is a huge difference. Obviously no one plots out every move, but still...
  - nazgulnarsil 17y ago
    greatly depends on how large the branches you trim are. some moves often force other moves, greatly trimming the search tree.
nandemo 17y ago
I'm not a chess player but I'm pretty sure that's not how it works. 20 is only the depth of the search, the width is variable. Good players can prune the game tree aggressively -- i.e. they select only a few candidate moves at each step -- while still getting a good approximation of the best move.
- pk2200 17y ago
  You're correct. The branching factor in a typical chess middlegame is 35-36, but top players usually only consider 1-3 moves per position. This is why the best humans were able to beat computers at chess for such a long time. (In the last 5-10 years the computers have pulled ahead, for good.)
mightybyte 17y ago
In the variant called suicide chess it's routinely possible for humans to recognize wins 30 moves in the future. I've seen strong players do this in fractions of a second during lightning games. One factor contributing to this is a comparatively low branching factor in terms of raw move legality. But I think it's similar (maybe slightly lower) to the effective branching factor a grandmaster considers in regular chess. Another factor is that it's not necessary to see the actual moves. Humans can immediately evaluate patterns that have a known outcome that many moves in the future without needing to enumerate the exact moves of the principle variation in their mind. You might argue that this isn't really what is meant by "x moves ahead". But I think this more "fuzzy" definition may very well be more accurate.
- Perceval 17y ago
  In Gödel Escher Bach, Hofstadter has three sections on "chunking." Chunking is distinguished from the type of branching calculation of future moves. Rather than evaluating linear move-countermove-move branches, humans can also evaluate in terms of previously recognized patterns or configurations of pieces [essentially what you're talking about, mightybyte, if I understand you correctly]. This kind of pattern recognition or pattern seeking is a substantial shortcut, reducing the amount of mental computation necessary in branching approaches. It's also an area in which human still potentially retain an efficiency advantage over computer opponents like Deep Blue. There's some explanation on Wikipedia: http://en.wikipedia.org/wiki/Chunking_(psychology) http://en.wikipedia.org/wiki/Chunking_(psychology)

by nanijoe 17y ago

It's really not possible (or practical) to plot 20 moves ahead in a chess game. Every different piece moved by your opponent would represent a new 'line' that you would have to compute 20 moves for.

pavs 17y ago

> It's really not possible (or practical) to plot 20 moves ahead in a chess game. 20 'best' moves ahead, according to the observer/player/chess-engine; assuming opponent played the best moves (according to the observer) in reply. Remembering a game played 6 years ago or being able to "think" 20 moves ahead may seem an amazing feat to regular chess players. But for professional players of GM or IM status - I don't think it is something amazing. Professional chess players memorize 100s (if not 1000s) of chess opening moves and their variations for many years. Remembering a chess game is not very impressive in that context.

robryan 17y ago
I'd imagine at the top level being able to see say the 20 best moves ahead would regularly occur similarly as predicted by the player and if things deviate it could be there there advantage anyway.
- iron_ball 17y ago
  Yes, and in fact that's how openings work. Each opening is a sequence in which both players are making an optimal response to the previous move. If a player deviates from the opening, then by definition he just made a suboptimal move.
  - KC8ZKF 17y ago
    Openings aren't demonstrably optimal. (Or there would be only one.) Players make improvements on openings all the time.
    - swombat 17y ago
      There are a relatively small number of equally optimal moves (compared to the total number of possible moves) for a given opening, after white has opened in a certain way. The vast majority of moves other than the ones covered in the opening and/or its known variations are indeed sub-optimal, and quite often if someone deviates from one of the great many standard openings, this can be exploited by the opponent.
      - nanijoe 17y ago
        The truth of the matter is that chess players seldom play out a single opening..Since most openings are pretty well known (along with their optimal moves), the beginning part of a chess game is usually spent trying to maneuver your opponent into an opening of your liking (or trick them into thinking you are playing one thing, then switching on them). What starts out looking like the Slav , may end up Queens Indian (as an example), and this makes the 20 moves ahead claim even more ludicrous. There are whole books on (slaying) the dragon variation of the Sicilian defense and the Sicilian is just one option that can result from the first move being e4..of course your opponent can also transpose to a different formation as the game progresses. In the end game where there are relatively few pieces on the board you may be able to plan ahead with more precision, but other than that trying to think 20 moves ahead is not just impractical, it may well be dumb.
      - iron_ball 17y ago
        This is what I meant. Openings aren't traps, and openings can branch as players choose different variations, but to stray from the opening book entirely is seldom a wise move. (That's one reason I never got too far into chess; I realized that competitive play would require a lot of opening knowledge.)
fh 17y ago
nanijoe is still pretty much technically correct (the best kind of correct). It's just that when chess players talk about plotting "n moves ahead", they mean something different than a programmer talking about a program computing "n moves ahead". Top chess players evaluate a handful of possible lines of play very very deeply. But it's impossible for the human mind to evaluate every response and counter-response to any reasonable depth, which is why computer players dominate so much in tricky tactical positions.

nazgulnarsil 17y ago

im a casual player and i plot 10 moves. 20 isn't surprising to me.

roundsquare 17y ago
Well, the number of possibilities grows exponentially with the number of moves you plan ahead. Going from 10 to 20 is a huge difference. Obviously no one plots out every move, but still...
- nazgulnarsil 17y ago
  greatly depends on how large the branches you trim are. some moves often force other moves, greatly trimming the search tree.

nandemo 17y ago

I'm not a chess player but I'm pretty sure that's not how it works. 20 is only the depth of the search, the width is variable. Good players can prune the game tree aggressively -- i.e. they select only a few candidate moves at each step -- while still getting a good approximation of the best move.

pk2200 17y ago
You're correct. The branching factor in a typical chess middlegame is 35-36, but top players usually only consider 1-3 moves per position. This is why the best humans were able to beat computers at chess for such a long time. (In the last 5-10 years the computers have pulled ahead, for good.)

mightybyte 17y ago

In the variant called suicide chess it's routinely possible for humans to recognize wins 30 moves in the future. I've seen strong players do this in fractions of a second during lightning games. One factor contributing to this is a comparatively low branching factor in terms of raw move legality. But I think it's similar (maybe slightly lower) to the effective branching factor a grandmaster considers in regular chess. Another factor is that it's not necessary to see the actual moves. Humans can immediately evaluate patterns that have a known outcome that many moves in the future without needing to enumerate the exact moves of the principle variation in their mind. You might argue that this isn't really what is meant by "x moves ahead". But I think this more "fuzzy" definition may very well be more accurate.

Perceval 17y ago
In Gödel Escher Bach, Hofstadter has three sections on "chunking." Chunking is distinguished from the type of branching calculation of future moves. Rather than evaluating linear move-countermove-move branches, humans can also evaluate in terms of previously recognized patterns or configurations of pieces [essentially what you're talking about, mightybyte, if I understand you correctly]. This kind of pattern recognition or pattern seeking is a substantial shortcut, reducing the amount of mental computation necessary in branching approaches. It's also an area in which human still potentially retain an efficiency advantage over computer opponents like Deep Blue. There's some explanation on Wikipedia: http://en.wikipedia.org/wiki/Chunking_(psychology) http://en.wikipedia.org/wiki/Chunking_(psychology)