What human chessplayer can learn from program algorithm ?

[Hood]

Hi,

that is interesting issue.

I read E. Lasker 'Manual of chess' and found there interesting remarks
about field values, pawn values depending on the place on the chessboard.

E. Lasker was WCC and the mathematician may be because of that he mentioned thouse facts.

He started with evaluating the pawns on the initial position.

The parts of the chess engine which are coming to my mind to be usable for human are:
evaluation function and pruning method.
What are your opinions about, how to extract such an information from engine to be useable by human.

rgds h

[Roman Hartmann]

In my opinion this is one of the trickiest obstacle you have to take if you're a (strong) chess player and trying to figure out how a computer plays chess. Even though a grandmaster and a strong chess program might agree on a certain move there is a huge difference in how they reach this conclusion. While the grandmaster is highly selective and only looks at a few moves the program still needs to look at millions of positions before it comes up eventually with the same move.

The human player sees a pattern and knows how to play on while the program plays through many possible (and impossible) games and will make a decision based on the outcome of these games.

An analogy I like: Let's assume we have a computer scan a picture with a lot of faces for a certain face. The computer needs to find the faces, measure them and compare the obtained data with an existing datbase to find that certain face. A human beeing on the other hand will just recognize a familiar face in a mass of other faces without any additional effort.

A human player will see a certain pattern because he/she has seen this pattern/position before and remembers that.

What I'm trying to say is that a human player cannot make use of the algorithms a chess program is using because human play and computer play are so much different.
Planting human knowledge into a chess program on the other hand does certainly add to the playing strength of the program (assuming it's done properly, of course).

best regards
Roman

[Hood]

thanks for answer.

The pattern is one matter but chess are not memory only. Memory serves the typical situations but it is not serving exceptions. Wins are results of better served exceptions or blunders imo.

Player is selecting 3-4 candidate moves on the base of experience and the situation on the desk, known rules. It is similar to pruning at early stage.
Then he calculates variations and is estimating,evaluating the position.

Evaluation of position is based on some rules and some calculations material, king safety, pieces activity.etc

That it is what i think is similar for program and human.

Programs have tuned evaluation values. Some evaluation rules I think that knowing that values can help human in improving his play.

there are intuitive or logical plyers.

rgds
h

[jwes]

One possibility is to use computer style pondering strategy. Pick the move you think best for your opponent and and spend your time thinking as if he made that move until you have decided on the best response. If he makes the expected move, if you have thought long enough, reply instantly, else keep thinking. If he makes some other move, start thinking about that move.

[Terry McCracken]

One possibility is to use computer style pondering strategy. Pick the move you think best for your opponent and and spend your time thinking as if he made that move until you have decided on the best response. If he makes the expected move, if you have thought long enough, reply instantly, else keep thinking. If he makes some other move, start thinking about that move.

As humans we do basically that all the time...something we gave the ability to do for computers about 30 years ago.

[metax]

One possibility is to use computer style pondering strategy. Pick the move you think best for your opponent and and spend your time thinking as if he made that move until you have decided on the best response. If he makes the expected move, if you have thought long enough, reply instantly, else keep thinking. If he makes some other move, start thinking about that move.

As humans we do basically that all the time...something we gave the ability to do for computers about 30 years ago.

Well, I do not do anything like that. If I've made my move, I stand up and look at the other boards to get something else to see than just my game.

[jwes]

One possibility is to use computer style pondering strategy. Pick the move you think best for your opponent and and spend your time thinking as if he made that move until you have decided on the best response. If he makes the expected move, if you have thought long enough, reply instantly, else keep thinking. If he makes some other move, start thinking about that move.

As humans we do basically that all the time...something we gave the ability to do for computers about 30 years ago.

When you are thinking on your opponent's time, do you actually spend all your time on one possible opponent's move and then move instantly if he makes the move you expect if you have thought long enough?

[Terry McCracken]

One possibility is to use computer style pondering strategy. Pick the move you think best for your opponent and and spend your time thinking as if he made that move until you have decided on the best response. If he makes the expected move, if you have thought long enough, reply instantly, else keep thinking. If he makes some other move, start thinking about that move.

As humans we do basically that all the time...something we gave the ability to do for computers about 30 years ago.

When you are thinking on your opponent's time, do you actually spend all your time on one possible opponent's move and then move instantly if he makes the move you expect if you have thought long enough?

What I'm saying is I look at board, look at certain candidate moves mine and the opponents, work out some possible and likely lines and if the move played is expected I don't waste much time and follow up with my reply. Sometimes I may re-check but often I move instantly. If it's another move and still what I thought would be played as next best or an alternate but equal reply then I move quickly for that move as well. Only if I'm taken by surprise do I start to ponder or I find something that appears better, otherwise I just move.

I don't focus on one move or line unless it's forced.

[Hood]

I hoped that it is more to use then pondering .

[Sven]

thanks for answer.

The pattern is one matter but chess are not memory only. Memory serves the typical situations but it is not serving exceptions. Wins are results of better served exceptions or blunders imo.

Player is selecting 3-4 candidate moves on the base of experience and the situation on the desk, known rules. It is similar to pruning at early stage.
Then he calculates variations and is estimating,evaluating the position.

Evaluation of position is based on some rules and some calculations material, king safety, pieces activity.etc

That it is what i think is similar for program and human.

Programs have tuned evaluation values. Some evaluation rules I think that knowing that values can help human in improving his play.

there are intuitive or logical plyers.

rgds
h

There are even more differences between human and computer calculation. Humans do quite often use elements like these in their reasoning:

"This variation is forced and improves my position."

"This variation takes more time than the other one and may look more cumbersome but it leaves my opponent with less counterplay."

"I am not sure about this variation, it is getting quite tactical, so I'll look for some improvement to avoid these tactics."

"With this move I allow for too much simplification which reduces my winning chances. I'd better keep dangerous pieces on the board while accepting some counterplay."

It will probably take a long time until computers will be able to make use of this kind of reasing.

Another important difference is that computers always use numbers for positional evaluation (in fact they can't do anything else as of today) while humans seem to use other non-numerical approaches to determine which variation "looks" better.

Sven