GameKnot: club forum - Computer program vs computer program analysis complete

« TAP TO LOG IN

Play online chess!

Computer program vs computer program analysis complete

« Back to club forum

From	Message
wrecking_ball 20-Jul-14, 17:19	Computer program vs computer program analysis complete For the last 8 years or so and more so in the last 3 years,my "team"at the Marshall Club in NYC has been compiling some very interesting information regarding all of the more irregular openings to prove a point.To just list a few.....1.The Sokolsky Opening 2.Grob Opening 3.Basman Defense 4.Dutch Defense 5.Larsen's Opening 6.Birds opening and a few others. We have set up all formulations of the best opening books available and up to date,including thousands of lines analyzed that are either out of book or positions that are unclear. We have then inputted the various irregular openings and opening defenses into one custom chess program and one Deep Houdini 4 program(3199 ELO rating).The intercooled 32 core computers were tweaked over the years,including overclocking,contempt program changes from 0 to 3 and many other factors that can alter the way a strong chess program thinks.The computer programs were set to play against one another after the opening was inputted into the computer.Tests were also done on full tournament time setting such as: 120'/40 + 60/20 +30. Infinite analysis was also completed in a great majority of selected moves,with some time settings up to 8 hours per move in cases where we felt it necessary if some lines were weaker than others.A minimum of 100 games were played in each opening and opening defense and then approx 50 more games in each case with various subvariations of those styles of play. The results? Well,I do not have the exacting finished % chart yet,that is being completed and I will post it when it is officially finished.But the bottom line is this: With PERFECT or NEAR PERFECT play on BOTH SIDES of the irregular opening and irregular opening defense it is over a 95% draw and in many cases closer to 100% draw,even with the so called "weaker"openings and weaker opening defenses.This % discovery is a bit surprising to many players,even GM's because it was always statistically thought that playing a Bird or a Grob,for example,would lead to more losses in serious competition,especially in the big leagues of senior master/GM.Not necessarily true anymore!Well,kind of anyway. Big league programs like Houdini,Komodo and Stockfish can indeed play a bit better than some of the best human GM's on the planet...if the programs and large core test/play computers are set up in optimum mode.Many of those past historical losses were "human error losses"that one must factor in,whereas,the elite chess programs of today can find ways to squeak through the tiniest of irregular disadvantages and turn it into a draw or win.My team was not aiming at the win factor,but the equalization factor with the opening moves and opening defense moves that have been historically regarded as inferior.So what we are saying in NY is that the stats are not accurate as a "truism".The stats are not"completely" true unless you factor in perfect play with the best chess programs and ultra strong computers with the best hardware.So when you go to chessok or other sources for stat info,you are looking at mainly human games,not machine games.There is a big difference.And some of the stats you read do indeed,have computer vs computer stats intermixed,but they are probably not totally updated and cutting edge complete as we have in NYC.We may publish the findings in Chess Life.That is being discussed now. So the reality?Most of these irregular opening's are only "perceived" as inferior,because humans are not as good as the machine(overall) and make human errors in judgement.In actual perfect play,many many of the lines equalize,even in that weak Basman game,Larsen,Bird and others.Yet,so many players fear and stay clear of those openings and opening defenses. I am one that never will.And that is one of the many reasons I promote irregular game play in my club.Sure,irregulars are not for the beginner.But who said I was chatting about beginners? Just some FYI. WB

From

Message

wrecking_ball
20-Jul-14, 17:19

Computer program vs computer program analysis complete

For the last 8 years or so and more so in the last 3 years,my "team"at the Marshall Club in NYC has been compiling some very interesting information regarding all of the more irregular openings to prove a point.To just list a few.....1.The Sokolsky Opening 2.Grob Opening 3.Basman Defense 4.Dutch Defense 5.Larsen's Opening 6.Birds opening and a few others.

We have set up all formulations of the best opening books available and up to date,including thousands of lines analyzed that are either out of book or positions that are unclear.

We have then inputted the various irregular openings and opening defenses into one custom chess program and one Deep Houdini 4 program(3199 ELO rating).The intercooled 32 core computers were tweaked over the years,including overclocking,contempt program changes from 0 to 3 and many other factors that can alter the way a strong chess program thinks.The computer programs were set to play against one another after the opening was inputted into the computer.Tests were also done on full tournament time setting such as:
120'/40 + 60/20 +30. Infinite analysis was also completed in a great majority of selected moves,with some time settings up to 8 hours per move in cases where we felt it necessary if some lines were weaker than others.A minimum of 100 games were played in each opening and opening defense and then approx 50 more games in each case with various subvariations of those styles of play.

The results?

Well,I do not have the exacting finished % chart yet,that is being completed and I will post it when it is officially finished.But the bottom line is this:

With PERFECT or NEAR PERFECT play on BOTH SIDES of the irregular opening and irregular opening defense it is over a 95% draw and in many cases closer to 100% draw,even with the so called "weaker"openings and weaker opening defenses.This % discovery is a bit surprising to many players,even GM's because it was always statistically thought that playing a Bird or a Grob,for example,would lead to more losses in serious competition,especially in the big leagues of senior master/GM.Not necessarily true anymore!Well,kind of anyway.

Big league programs like Houdini,Komodo and Stockfish can indeed play a bit better than some of the best human GM's on the planet...if the programs and large core test/play computers are set up in optimum mode.Many of those past historical losses were "human error losses"that one must factor in,whereas,the elite chess programs of today can find ways to squeak through the tiniest of irregular disadvantages and turn it into a draw or win.My team was not aiming at the win factor,but the equalization factor with the opening moves and opening defense moves that have been historically regarded as inferior.So what we are saying in NY is that the stats are not accurate as a "truism".The stats are not"completely" true unless you factor in perfect play with the best chess programs and ultra strong computers with the best hardware.So when you go to chessok or other sources for stat info,you are looking at mainly human games,not machine games.There is a big difference.And some of the stats you read do indeed,have computer vs computer stats intermixed,but they are probably not totally updated and cutting edge complete as we have in NYC.We may publish the findings in Chess Life.That is being discussed now.

So the reality?Most of these irregular opening's are only "perceived" as inferior,because humans are not as good as the machine(overall) and make human errors in judgement.In actual perfect play,many many of the lines equalize,even in that weak Basman game,Larsen,Bird and others.Yet,so many players fear and stay clear of those openings and opening defenses.

I am one that never will.And that is one of the many reasons I promote irregular game play in my club.Sure,irregulars are not for the beginner.But who said I was chatting about beginners?

Just some FYI.

WB

« Back to club forum

GameKnot: play chess online, monthly chess tournaments, chess teams, Internet chess league, chess clubs, online chess puzzles, free online chess games database and more.