New idea for a fair Human-Machine battle without handycap

[Laskos]

We can resolve our dispute if a top GM will be allowed to take back on a >0.3 blunder and 3 times at any time against Rybka on any hardware. I can bet that the top GM will win in a 6 games match.

Kai

Losing becomes impossible.
Do you know a reason why any elo 2000 player could lose against Rybka3 under these conditions ?

Matthias.

Try to play Rybka under these rules and you will see . An amendment, unforced errors >0.3 are taken back, because when Rybka will going to mate you, you have to take back every possible move .

Try, and play, if you are so skeptical. I guess that only strong GM's can stand even in these conditions. If I am not wrong, Uri was exactly of opposite opinion to yours, that taking back tactical blunders will not change anything, because engines are positionally superior to top humans .

Kai

[Laszlo Gaspar]

The best way for humans to be able to compete with computers is to have different time handicaps. This ofcourse is increasing in the computer's favor linearly as hardware increases. I can still win tons of games on a slow P4 against R3 if it has 1 minute and I have 15 minutes. On an Octa things change dramatically in the engine's favor. I expect within a few years, even the strongest GM will not be able to compete against Rybka using classical time controls and Rybka using only 5 minutes per game. That is how wide the gap between humans and engines is becoming.

I think there are some misunderstandings here. Short time controls favour the computers this is true. On the other hand humans are better at long ones (say 1 move per day). This is because they understand chess while programs don't. Humans assess only important lines and so have a better - so called - branching factor.
As I stated already in the past engines can be considered better than humans when they are better in all time controls.

My proposal is to give a little help for the human player to make the match more fair and see what he can do.

[Uri Blass]

The best way for humans to be able to compete with computers is to have different time handicaps. This ofcourse is increasing in the computer's favor linearly as hardware increases. I can still win tons of games on a slow P4 against R3 if it has 1 minute and I have 15 minutes. On an Octa things change dramatically in the engine's favor. I expect within a few years, even the strongest GM will not be able to compete against Rybka using classical time controls and Rybka using only 5 minutes per game. That is how wide the gap between humans and engines is becoming.

I think there are some misunderstandings here. Short time controls favour the computers this is true. On the other hand humans are better at long ones (say 1 move per day). This is because they understand chess while programs don't. Humans assess only important lines and so have a better - so called - branching factor.
As I stated already in the past engines can be considered better than humans when they are better in all time controls.

My proposal is to give a little help for the human player to make the match more fair and see what he can do.

Your assumptions that humans are better in 1 move per day is not proved
and I do not believe that this assumption is correct today when rybka is available.

Uri

[Laszlo Gaspar]

Your assumptions that humans are better in 1 move per day is not proved
and I do not believe that this assumption is correct today when rybka is available.

Uri

Hi Uri,

this is what we don't know for sure and if we want to know it we should provide fair conditions for both sides and my idea is a possible way.

As far as Rybka is concerned its search seems to be nothing extraordinary compared to other engines (relatively not stronger at long time controls) and its advantage might disappear with longer time control.
So I don't feel the presence of Rybka a strong argument in this debate.

[PauloSoare]

See what happened in this match between Stephen Ham and some engines.
The match was almost ten years ago! Engines runned on a Pentium III 500 MHz with 256MB RAM.

http://www.correspondencechess.com/campbell/ham/ham.htm

[Uri Blass]

Your assumptions that humans are better in 1 move per day is not proved
and I do not believe that this assumption is correct today when rybka is available.

Uri

Hi Uri,

this is what we don't know for sure and if we want to know it we should provide fair conditions for both sides and my idea is a possible way.

As far as Rybka is concerned its search seems to be nothing extraordinary compared to other engines (relatively not stronger at long time controls) and its advantage might disappear with longer time control.
So I don't feel the presence of Rybka a strong argument in this debate.

I disagree that rybka's search is not something extraordinary compared to engines before rybka.

I think that it can be proved easily by matches with unequal time control

play Rybka3(32 bits ponder off) against fruit2.1 at unequal time control and give fruit2.1 50:1 time advantage.


I expect Rybka3 to lose 100 game match
at 1 minute against 50 minutes of fruit2.1 but
I also expect Rybka to win at 20 minutes against 1000 minutes of fruit2.1

Maybe somebody can check it.

Edit:of course losing or winning is dependent on the hardware and I want to give fruit time advantage that is enough for it to score result between 60-40 and 70-30 in the fast match so I suggest first to play the faster match and based on the result to decide maybe you need to change the handicap.

After deciding about the correct handicap to cause fruit2.1 to be something like 100 elo better than rybka then repeat the match at time control that is 20 times slower.

Uri

[bob]

Dear chess fellows,

I have an idea which I think should be tried out before we can state with certanity that machines are better than humans.

As far as I know, still all Human-Machine battle was organized so far the same way as a human-human compatition. We expect the GM to calculate all variations in his mind and make the chosen move on the board.
Why is that when a chess engine as its natural operation can make a move and take it back on a board which is real for itself and let it evaluate the upcoming positions...?

My proposal is to let the human use a special computer tool (a GUI or database program or whatever) to be able to analyze his decisions before making the real move on the board.
Certainly no chess engines can be used for help only databases.

This sounds like correspondence play but with tournament time control and without engine help.

I can imagine that special software should be developed for this purpose based on GMs' requirements.

What is your opinion?

That is contorted logic. Last time I played a game, I 'moved pieces" around in my mind also. I can visualize the chess board, and then analyse variations by moving pieces around on that mental image. How is that different from what the computer is doing???

The human GMs are already better at doing this than you can imagine. A GUI is going to slow them down and make them play weaker, not stronger.

[Laszlo Gaspar]

That is contorted logic. Last time I played a game, I 'moved pieces" around in my mind also. I can visualize the chess board, and then analyse variations by moving pieces around on that mental image. How is that different from what the computer is doing???

The human GMs are already better at doing this than you can imagine. A GUI is going to slow them down and make them play weaker, not stronger.

Dear Prof. Hyatt,

I don't think that it is different I even state that calculating with a graphical aid is not different from what computers do. That is why I propose it.
You state that GMs can do it better without it but tireness should be taken into account (1 move per day) and long variations might require some assistance.
I don't say that a usual GUI helps! It has to be designed to be useful (for example freezing certain positions in a new window) by GMs need!

Uri,

In case of Fruit2.1 you might be right but what about Zappa and Naum? They are not considered any special and have better branching factors.

[Uri Blass]

That is contorted logic. Last time I played a game, I 'moved pieces" around in my mind also. I can visualize the chess board, and then analyse variations by moving pieces around on that mental image. How is that different from what the computer is doing???

The human GMs are already better at doing this than you can imagine. A GUI is going to slow them down and make them play weaker, not stronger.

Dear Prof. Hyatt,

I don't think that it is different I even state that calculating with a graphical aid is not different from what computers do. That is why I propose it.
You state that GMs can do it better without it but tireness should be taken into account (1 move per day) and long variations might require some assistance.
I don't say that a usual GUI helps! It has to be designed to be useful (for example freezing certain positions in a new window) by GMs need!

Uri,

In case of Fruit2.1 you might be right but what about Zappa and Naum? They are not considered any special and have better branching factors.

Zappa and Naum are engines from the time of rybka and I expect them also to be able to beat humans at 24 hours per move.

Uri

[bob]

That is contorted logic. Last time I played a game, I 'moved pieces" around in my mind also. I can visualize the chess board, and then analyse variations by moving pieces around on that mental image. How is that different from what the computer is doing???

The human GMs are already better at doing this than you can imagine. A GUI is going to slow them down and make them play weaker, not stronger.

Dear Prof. Hyatt,

I don't think that it is different I even state that calculating with a graphical aid is not different from what computers do. That is why I propose it.
You state that GMs can do it better without it but tireness should be taken into account (1 move per day) and long variations might require some assistance.
I don't say that a usual GUI helps! It has to be designed to be useful (for example freezing certain positions in a new window) by GMs need!

Uri,

In case of Fruit2.1 you might be right but what about Zappa and Naum? They are not considered any special and have better branching factors.

What is the point? In correspondence, your idea might have merit and is already being done by human players. But in a game/60? Such a GUI adds a significant penalty over a strong human's built in chess playing skills. Chess is not "one move per day". Correspondence chess is pretty close. But humans already use multiple programs, multiple boards, so there's nothing new there at all. But for normal chess, this is a non-starter.