Winboard Forum

How much time do programmers need to emulate fruit's evaluation in their chess program?

Note that speed of the evaluation is not important in this discussion but it is important not to use the data structure of fruit and it is important to be correct and give the same evaluation as fruit(it may be productive to add to fruit a function that read games and calculate all the component of fruit's evaluation for every position in the games and print the sum of the evaluation of every component to a text file).

What about other strong free programs with source code?

Did programmers try this task as an exercise in order to improve themselves in avoiding bugs in the evaluation?

Uri

Uri Blass wrote:How much time do programmers need to emulate fruit's evaluation in their chess program?

Why would anybody want to do that? The evaluation function, more than everything else, is what makes a chess engine unique and defines the engine's style and personality. To emulate the evaluation function of another engine seems pointless and boring to me. If I want a program with Fruit's evaluation, I can just use Fruit.

Tord

The reasons to do it are:

1)Training in writing evaluation without bugs
I believe that a lot of evaluations have bugs(note that fear of new bugs is one of the reasons that I changed very little in the evaluation of Movei in the last year)

If I try to do well defined task when it is possible to check for no bugs by comparing with fruit's evaluation then I may get more confidence in writing good evaluation without bugs.

2)I believe that fruit's evaluation is simply better than evaluation of most engines and implementing parts of fruit's evaluation and checking for no bugs force you to understand it and understanding it can help you to have better evaluation and ideas for better evaluation.

Uri

Tord Romstad wrote:

Uri Blass wrote:How much time do programmers need to emulate fruit's evaluation in their chess program?

Why would anybody want to do that? The evaluation function, more than everything else, is what makes a chess engine unique and defines the engine's style and personality. To emulate the evaluation function of another engine seems pointless and boring to me. If I want a program with Fruit's evaluation, I can just use Fruit.

Tord

Agree Completely!

Its better to think up of your own eval and if there are bugs in them fix it . Compare the program's eval to your personal evaluvation of the position instead of another engines. Humans, even low rated players, can do evals of quiet positions much better than any engine can, including Fruit.

Pradu wrote:

Tord Romstad wrote:

Uri Blass wrote:How much time do programmers need to emulate fruit's evaluation in their chess program?

Why would anybody want to do that? The evaluation function, more than everything else, is what makes a chess engine unique and defines the engine's style and personality. To emulate the evaluation function of another engine seems pointless and boring to me. If I want a program with Fruit's evaluation, I can just use Fruit.

Tord

Agree Completely!

Its better to think up of your own eval and if there are bugs in them fix it . Compare the program's eval to your personal evaluvation of the position instead of another engines. Humans, even low rated players, can do evals of quiet positions much better than any engine can, including Fruit.

I do not believe in that theory

I am a candidate master but I think that Fruit is clearly superior relative to me in evaluation of many quiet positions.

I already learned from fruit king attack evaluation(before reading about Fruit king's attack evaluation I did not think about evaluating king attack in that way) and I am sure that I can learn from other evaluation parts of fruit.

I do not claim that fruit's evaluation is optimal or close to optimal and it can be improved but it clearly considers ideas that I did not consider in the past.

Note also that the problem in evaluation is not only what to evaluate but how much importance to give to details that you evaluate.

I do not think that humans are better than fruit in not overestimating or underestimating factors and I am sure that humans including masters may not understand things that Fruit understand.

After all chess is 99% tactics so humans do not need to know much about evaluation and if they see more than their opponents it is enough for them to win.

Uri