chess games online

Chess Games Online

Grant and you will find!
Funny name, real interests
[ Sign up | Log in | Guest ] (beta)
gibo 54 ( +1 | -1 )
the difference in fritz's strength I know when kasparov and co play fritz or another top notch engine, the game is always played on a top notch computer e.g. 3.06ghz is that latest i think? Does the speed in your computer make that much of a difference, i know there is some difference. Also i know you can choose how much ram fritz has to play with e.g. i always play with 128mb, also an interesting thing i have found is fritz 5 seems to be stronger than fritz 7 on my computer? (I dont have fritz 8)
coperplayer 263 ( +1 | -1 )
Hi gibo,

Current chess software strength relies mainly in the positions evaluated per second.
For example, it is known that making an algorithm calculate faster is *much* more important that making it cleverer in evaluating a final position. For that reason, processing speed has a huge impact on playing strength.
It is true that a lot of positions are evaluated the same after looking 20.000.000 of nodes than after looking only a few, say 1.000.000. But there are critical positions when playing a worse move can ruin the game.

Take into acount that chess game analisys is a complex exponential problem, that is: increasing by one the search depht implies multiplying by a factor of 30 (aprox) the number of positions to be evaluated. This means that increasing computer power really does not make a lineal increasing of the search depth achieved, but it can help on deciding what to move in a critical position. In a typical game, there are some of that critical positions, and the chess software really just cannot recognize them; more computer power increases the probabilities of finding the best move. Perhaps all other avaliable moves lead to ruin.

With regard to the RAM memory used by fritz, I think you are refering to Hash Tables. These are basicaly memory structures kept by the software while it is analysing thousands of positions. Some evaluated tree nodes are stored there, and if fritz find them again, it just skips that branch evaluation, because ii already knows its value. This is a incredible optimization, because positions repeats very frecuently while analizyng different game branches. Obviously, more memory implies more storage for potentially repeatable positions and then, more nodes per second.

And with regard to your question on Fritz 5 vs. Fritz 7, it just cannot be answered. Doing fair chess software strength comparision is not an easy task, and it never should be carried out with the two applications running in the same computer, because you cannot be sure about the resources used by the two players. For example I could write a very bad chess program that uses a lot of system resources doing stupid calculous. Challenging fritz with it (on the same computer) could lead to unpredictable results; imagine my program constantly using 99% CPU proccess, then fritz could not calculate enough positions to play well, although it is a much better software.

Regards,

Daniel.

PS: Sorry for my english.
More: Chess
bogg 85 ( +1 | -1 )
gibo If I remember correctly the rule of thumb is that for roughly every seven fold increase in processor speed a chess playing engine improves about 100 Elo points. The reason the multiplier is seven is that the effective branching factor of a full width search using alpha-beta pruning methods is about seven. The MPP machines used in the matches were about three times faster than the fastest single engine PCs. That would correspond to about 56 Elo rating points (7^.56 = 3). If my math is accurate then a top chess engine playing on a high-end single CPU PC is about comparable to a 2750 player when playing at a 40/2 time control. The style of said hypothetical 2750 player is that he makes a lot of small positional errors but virtually never misses a tactical opportunity.

CT (Bogg)