Mostly for Christophe Theron

[fern]

Dear Chris:
I just read in one of your posts belonging to a long thread that "it is time to retire".
That sad comment inmediately made me remember a very old discussion that we sustainned here, years ago, when I said that chess programmers, as everybody else, has ONE set of original ideas and that, once exhausted, what remains is a tedious though somewhat useful task of improving and polishing.
This "ONE set of ideas and no moresyndrome
has been studied by psychology and tends to be probed with some rare exceptions that only confirms the rule. Simone de Beauvoir wrote an excellent book about that.
You, for your part, negated such a thing. I do not remember exactly the arguments, but you thought an intelligent guy can create without limits as much as he conserves the will to do such a thing.
Since the time of that discussion we have seen the retirement of several known programmer. Lang, Schroeder, others. All of them, If I recall well his particular histories, reached a point when no more progress was obtained and so they lost stamina and eventuually left the field, sometimes telling it was so, openly, as Ed, sometimes simply stopping the development of his engines.
I would hate that you becomes a new reason to believe those psychologist and me were in the right. I understand that when others competitors comes with very succesufull products and seems to jump very far, one tend to feel a sense of disenchantement with oneself. Something like "I made the effort along years, but then suddenly this guy comes and just crush me just like that".
Then one say: "anyway, I have other things to do in my life".
I hope it is not your case. You are still too young to consider your stock of ideas exhausted. I hope you will shut my mouth with a new Tiger, sooner or later. It does not matter if that new Tiger will be the first or the second, if it will better or not than Rybka or whatever be the new darling in one or two years. What matter is not to abandon your call and keep fighting. Tiger will be definitely surpassed only the day you really does not work anymore on it.

I hope, also, you take my words in good spirit.
Fern

[mclane]

i don't think christophe retires.

i am sure he enjoys in the moment beeing together with his young daughter.
thats what i am doing too with my son.

if you are playing much outside you forget about computerchess.

we have maybe to wait a few years.

[fern]

Hope you are right.
Fern

[tiger]

Dear Chris:
I just read in one of your posts belonging to a long thread that "it is time to retire".
That sad comment inmediately made me remember a very old discussion that we sustainned here, years ago, when I said that chess programmers, as everybody else, has ONE set of original ideas and that, once exhausted, what remains is a tedious though somewhat useful task of improving and polishing.
This "ONE set of ideas and no moresyndrome
has been studied by psychology and tends to be probed with some rare exceptions that only confirms the rule. Simone de Beauvoir wrote an excellent book about that.
You, for your part, negated such a thing. I do not remember exactly the arguments, but you thought an intelligent guy can create without limits as much as he conserves the will to do such a thing.
Since the time of that discussion we have seen the retirement of several known programmer. Lang, Schroeder, others. All of them, If I recall well his particular histories, reached a point when no more progress was obtained and so they lost stamina and eventuually left the field, sometimes telling it was so, openly, as Ed, sometimes simply stopping the development of his engines.
I would hate that you becomes a new reason to believe those psychologist and me were in the right. I understand that when others competitors comes with very succesufull products and seems to jump very far, one tend to feel a sense of disenchantement with oneself. Something like "I made the effort along years, but then suddenly this guy comes and just crush me just like that".
Then one say: "anyway, I have other things to do in my life".
I hope it is not your case. You are still too young to consider your stock of ideas exhausted. I hope you will shut my mouth with a new Tiger, sooner or later. It does not matter if that new Tiger will be the first or the second, if it will better or not than Rybka or whatever be the new darling in one or two years. What matter is not to abandon your call and keep fighting. Tiger will be definitely surpassed only the day you really does not work anymore on it.

I hope, also, you take my words in good spirit.
Fern

I'm not sure about the "one set of ideas" stuff. I started chess programming because I found it fascinating that a computer could play chess and wanted to know how it was possible.

As no information was available at that time, I had to build a chess program myself, with my own ideas, with the exception of a few ones that were already there like, for example, minimax.

In the process I discovered that I was able to do things I thought I could not do. So I can say that computer chess has allowed me to better know myself, or to reveal myself.

This started in 1979 with a Sargon II program running on TRS-80.

The first time my program has reached the level of the best ones is in 1998/1999. So almost 20 years passed between the first fascination to the partial achievement of the goal it had created.

During all that time, it would be wrong to say that I started from a set of ideas. It does not make sense. I started from nothing, and then ideas emerged one by one. Each new progress, which happened after testing a handful of bad ideas, was like entering a new territory. And led to other ideas that would not have been on my mind before.

The exhaustion comes from several causes, some "internal", others "external":

External causes:
- you keep having ideas, but less of them end up being effective. It's not about the quality of your idea generator (=your brain), it's just an expected phenomenon
- improving your engine becomes slavery. It is a infinite cycle of implementing an idea or modifying some part of the program, launch a several hours long automatic test, and start over. As I wanted to try many ideas, in general it meant that ideas were waiting in line and several weeks would pass by between the idea and the moment it is tried. It kills all enthusiasm. And in order to keep the test line busy at all times you must wake up at night to keep feeding the beast. My test computers emitted loud beeps when the tests were finished so I could hear them and wake up if I was sleeping. Now add a girlfriend to the picture...
- I do believe computer chess has changed for the worst. Strong open source programs have made the field less interesting for programmers by taking away the value of some ideas that took a long time for them to find and implement. But maybe it has made the field more interesting for users, so be it. However we have seen at the same time the emergence of bad behaviour that takes advantage of the initially generous idea of open source. We have seen obvious clones and we have also seen that open source code could be hijacked to serve personal interests, which is open source used against the spirit of open source. Finally, free (but not open source) programs have been used as a way to kill the competition. It had never happened in the past. Users think it's great, but I don't think so.
- ...

Internal causes:
- Motivation is not as strong anymore when you have achieved a goal you would not even have dreamed of. I mean, Chess Tiger has topped the SSDF list twice, so the best I could achieve after that, with all the hard work it would have required, is to do it again.
- I am too much reluctant to use ideas found and implemented first by others. The problem is that in a world of strong open source programs you either do that or your program will not improve at the same pace as those who do. But doing it is no fun. I believe it's a weakness but that's how I am.
- I have always had many different interests in life, computers being the most important, it's true. And just in the area of computing I also have many interests other than chess.
- ...

The bottom line is: time to retire.

However you never know. If only I had some time to work on it, I would maybe do exactly the opposite of what I have said.



// Christophe

[Oscar L]

From what I have read, the main reason for the elo jump in Rybka 3 from the already incredible R2.3.2a seems to be the improve in the evaluation made by L Kaufman.

He uses ultrafast games New Rybka vs Old Rybka, 1 second for game! So for example overnight he gets a lot of games where he can see very small increases in elo. Adding many new evaluation terms this way that add each a few elo points, you get a >100 elo jump



In this case it seems that evaluation was better than search. What is useful in ultra fast games is useful too for testing groups times



Regards.

[Rolf]

Internal causes:
- Motivation is not as strong anymore when you have achieved a goal you would not even have dreamed of. I mean, Chess Tiger has topped the SSDF list twice, so the best I could achieve after that, with all the hard work it would have required, is to do it again.
- I am too much reluctant to use ideas found and implemented first by others. The problem is that in a world of strong open source programs you either do that or your program will not improve at the same pace as those who do. But doing it is no fun. I believe it's a weakness but that's how I am.
- I have always had many different interests in life, computers being the most important, it's true. And just in the area of computing I also have many interests other than chess.
- ...

The bottom line is: time to retire.

However you never know. If only I had some time to work on it, I would maybe do exactly the opposite of what I have said.
// Christophe

Well said. Also, why always being busy with one's own programs and achievements? These days it seems as if the World needed programmers who observate the activities of others and to so making the World a bit safer for our kids who will probably continue to walk in our own footsteps. More psychology. Less computerchess.

[Uri Blass]

From what I have read, the main reason for the elo jump in Rybka 3 from the already incredible R2.3.2a seems to be the improve in the evaluation made by L Kaufman.

He uses ultrafast games New Rybka vs Old Rybka, 1 second for game! So for example overnight he gets a lot of games where he can see very small increases in elo. Adding many new evaluation terms this way that add each a few elo points, you get a >100 elo jump



In this case it seems that evaluation was better than search. What is useful in ultra fast games is useful too for testing groups times



Regards.

I disagree.

Based on Larry kaufman's results at least half of the improvement in rybka3 is thanks to better search.

Evaluation probably helped to get 50-60 elo but not more than it.


Uri

[Ovyron]

I disagree.

Based on Larry kaufman's results at least half of the improvement in rybka3 is thanks to better search.

Evaluation probably helped to get 50-60 elo but not more than it.


Uri

I disagree with Larry Kaufman and I think that he's underestimating the help from evaluation. The improvement was already about 50-60 elo without the improvements from search. Then search came and the elo made another big jump, but it's unclear if this jump from search would have been as high without the evaluation.

They didn't make versions without evaluation or search but since eval guides search in the right direction it could be making Rybka much stronger than what is seen with the results without search alone.

[Oscar L]

I am always afraid of posting because of the possibility of you disagreeing

A pair of post by L Kaufman about this topic


I cannot tell you what makes Rybka so strong in general, but I can tell you where the estimated +140 Elo points for Rybka 3 over 2.32a came from. My best guess: fifty Elo from better eval (roughly a hundred elo of eval improvement minus fifty from slowdown), seventy from better search, fifteen from better time management, five from "contempt"


It's a bit hard to tell. The slowdown from the terms really hurts in a direct matchup against 2.3.2a; it's based on such a matchup that I estimate the gains at only fifty Elo if the search, time management, and contempt were all unchanged from 2.3.2a. However I do believe that the slowdown is less costly against unrelated engines, while the eval gains should hold up, so if this theory is correct then maybe eval contributed a net of sixty to seventy points after allowing for the slowdown. The search benefit is huge on a direct matchup basis, maybe nearly a hundred Elo points, but this too should be of less value (though still a great deal) against unrelated engines. Maybe sixty each for eval and search, plus twenty from time and contempt, might be a better estimate for the composition of the 140 Elo gain (if that figure holds upl) against non-Rybka engines.

[Uri Blass]

<snipped>

I do believe computer chess has changed for the worst. Strong open source programs have made the field less interesting for programmers by taking away the value of some ideas that took a long time for them to find and implement. But maybe it has made the field more interesting for users, so be it. However we have seen at the same time the emergence of bad behaviour that takes advantage of the initially generous idea of open source. We have seen obvious clones and we have also seen that open source code could be hijacked to serve personal interests, which is open source used against the spirit of open source. Finally, free (but not open source) programs have been used as a way to kill the competition. It had never happened in the past. Users think it's great, but I don't think so.
- ...
// Christophe

I think that strong open source programs have more than one effect on programmers.

There are programmers who are less interested in programming because
of that but there are programmers with the opposite opinion.

I also dislike the fact that free non open source like rybka2.2n2 is used to kill the competition.

I see no reason for the release of Rybka2.2n2 as free source except trying to convince programmers of Hiarcs,Shredder,Fritz,Junior to quit for the simple reason that they get the feeling that they have no chance even to compete with free programs(even if they work hard and make something better than rybka2.2n2 in one year then Vas may release new free program that is stronger at that time).

Uri