why cant computers beat the endgame?
Sort:
The computer only evaluates the number of positions it can in the given timeframe an evaluates positions (not possible to only evaluate from one side) by the way it is programmed. Given that there are more possible chess positions than The number of atoms in the universe (look it up) theres no way it can calculate all possible moves. The very last endgames are however solves if The computer uses endgame tables.
Aug 9, 2015
0
#3
up to 7 pieces you cannot beat a chess computer if it has the latest tablebase in the end game.
computers are not always as strong as humans in every position. in fact humans play much beter in a 'lost' position than engines. the engine will try to find the moves that is the least worse of all, while humans start to play for 'tricks' like allowing mate in 2, but if the mate in two is missed than you win a queen. in that situation the engine would rater allow mate in 3 than to play for a 'trick'.
i sometimes teach to chess to children and i ones give a simul strating in a posstion that was lost.(enigne sayed -5,34). as first i played the simul myself against 8 children who were about 800 rating points below me. (mine us around 1500-1600 i geuss). i won 3 draw 1 and lost 4. the enigne rated 3300 lost al 8 games when the children played against it on the PC's
Aug 9, 2015
0
#5
It is interesting that Carlsen guessed incorrectly about how difficult a 7 piece tablebase was to generate and was unaware that it had already been achieved in 2012. In his defense, so was I. 
This was a prodigious task: in its highly efficient representation, the tablebase is 100 Tb.
With this handy online tool, you can practice winning in 545 moves in this position:
http://www.chess.com/forum/view/fun-with-chess/longest-mate-official---mate-in-545
Aug 9, 2015
0
#6
bb_gum234 wrote:
go_and_fun_yourself wrote:
dont computers calculate every possible move playable at least for their pieces?
Not even close. In fact today's computers, while much stronger, calculate much less. They're stronger because (on average) they ignore weak moves more quickly. They calculate less overall, but more of the important stuff.
go_and_fun_yourself wrote:
why cant computers beat the endgame?
In some endgames they are very good. Like queen endgames and pawn endgames.
In the ones they are bad at, the point, or the idea, of long term strategic moves may not "happen" until very far in the future. And between now and then are far too many non forcing moves. So the efficient engine prunes away winning (or drawing) lines too soon.
As an easy example let's say white is ahead a pawn, but the pawn can't queen. The engine wont know that until it calculates a draw by 50 move rule. This isn't just 50 moves of course, but 50 moves deep in every single line. And when the moves aren't forcing, there may be billions of individual lines. Sometimes the computer will realize if you leave it analyzing over night, but certainly not in a minute or less.
It is an important fact that you don't need to look at all of your moves to play perfectly, but you do need to look at all of your opponents' moves. (eg solving a chess problem). A good heuristic gradually introduces superficially weaker moves as stronger appearing ones fail.
This is how it was possible for computers to solve draughts. The total number of possible games was too big to fully analyse, but by being highly selective, you can get the problem down towards the square root of the full size (this is where you would get if you looked at one move for yourself and all the moves for the opponent every time.
For practical play, a computer (or a human) can trust that ignoring weak looking moves for the opponent is the most efficient use of time as well. This is unreliable for solving chess problems or for positions where a brilliancy that is not easy to see exists for the opponent.
According to this wikipedia article, all endgame positions up to 7 pieces have been solved by computers (except 6 against 1). Shredder offers a database that has all positions up to six pieces (except 5 against 1).
ive seen loads of pawn and bishop endgame puzzle where the engine says its losing but its not. dont computers calculate every possible move playable at least for their pieces?