Komodo and WCCC

[Lyudmil Tsvetkov]

check the ratings, not more than 100 elo, so basically nothing substantial.

currently, SF beats Rybka by more than 85%.

if VR was a genius, what would you say about the authors of SF?

I do not understand why a so-called 'genius' would need to copy parts of other engines' code, geniuses usually do not copy, but are the ones whose ideas are copied.

the train of thought that VR(pity that we should spoil that thread with unsubstantial names and events) was at the root of later engines' success is completely made-up.

with what did he contribute to SF?
with what did he contribute to Komodo?

and, when he went back with Fritz 15, how much stronger he was able to make it?

50 elo, that is it, far below SF, far below Komodo.

if he was such of a genius, why was not he able to compete with SF and Komodo again?

VR contributed at an early stage of engine development, when engines almost completely lacked any positional chess knowledge.

as a relatively good chess played, he added some knowledge, and that was sufficient to top the rating lists for some time.

that is all, nothing more, nothing less.

If someone added some knowledge, he was Larry Kaufman, not V.R. You have some confusion. Rybka search routines made Rybka the strongest. He copied evaluation and move generator, it seems.

You can think what you want. It seems your opinion has nothing to do with facts. You cannot remember a post by Marco Costalba, writing his doubts about what was the difference between taking ideas and taking code. You cannot remember because you weren't here.

BTW, only honest people always put themselves in doubt.

The fact is, every top engine programmer studied each line of Robbolito to improve their engines. And those lines were from Rybka. Before Ippolit "revolution", other engines were much weaker. Every programmer can confirm it.

And I think this post is going very off topic. I'd like to stop talking about these old things.

it is you who started the thread, you who mentioned VR and lead the discussion this way, and now you want to stop it...

move generators are easy to write, nothing special in there, if your engine is 15% or 50% slower, does not matter that much in the long run.

it does not make sense that he copied evaluation and improved search, it is mostly quite the opposite; VR is a relatively good chess player, so that is where he could be of use, and that is what he did.

precisely because of better eval, and not search, is why R became very strong for a while.

you can easily deduce that even by the extremely low nps of the engine, only thing that can substantially slow down an engine is significantly bigger number of calculations, i.e. calculations related to specific knowledge.

search is not able to quite achieve a slowdown of 500% or so, no matter how intricate it is.

to tell you the truth, I am very bad programming newbie, but have still read Robbolito code and the codes of some 30 engines more, I do not see what is so elaborate there about Robbolito search, apart from splitting down some major routines.

by studying code and possibly implementing some new ideas, one can add 10, 30, maybe 100 elo, but not more, as there are considerable incompatibilities between codes.

only way to go rigth to the top is by complete copy, to be fully compatible.

studied, studied, studied..., if they studied so much, why not a single engine surpassed R substantially in the next 5 years?

only Houdini, SF and Komodo seem to be competely different branches.

You have no clue as usual - this is just another ruined thread after you jumped in.

Probably you are the only person in this forum, who doesn't know about Rybkas depth obfuscation to simulate a 'slow searcher' for the 'superior knowledge' hype a decade ago.

And saying Houdini started a completely different branch is hilarious, after it started as a nearly 1:1 copy of Robbolito...

read more carefully my last 2 sentences: I do make a distinction between SF, K, H, it was quite obvious, but you somehow missed it.

so many people acknowledging VR lied about something: why so many people defend him then?

but, to tell you the truth, you, as well as all other people who have not read R code first-hand, simply can not know how things stand, it is all conjectures.

judging by R games, something I can certainly do, as the games are available, in distinction to the code, I can pretty much safely assume R does introduce quite some knowledge previous engines at the time did not have, so maybe I am more right after all, and real nps numbers are closer to outputted ones.

I do not know why everyone assumes R has more search innovations than chess knowledge ones, where is your proof?

or, most probably, as with you and a range of other members, to turn you the attack favour, you are mixing supposition/personal predilection with proof?

look at R games:

- fianchettoeing its king bishop sometimes, no other engine did that at the time, even not now
- excellent handling of connected passed pawns
- some smattering of imbalances, lacking in other engines
- more reasonable psqt for pawns and pieces alike, etc., etc.

this definitely seems to me like some knowledge base, does not it?

besides, LK(Larry Kaufman) worked with R for a while, he certainly helped with imbalances and other stuff, you can not say R had no knowledge.

[Lyudmil Tsvetkov]

check the ratings, not more than 100 elo, so basically nothing substantial.

currently, SF beats Rybka by more than 85%.

if VR was a genius, what would you say about the authors of SF?

I do not understand why a so-called 'genius' would need to copy parts of other engines' code, geniuses usually do not copy, but are the ones whose ideas are copied.

the train of thought that VR(pity that we should spoil that thread with unsubstantial names and events) was at the root of later engines' success is completely made-up.

with what did he contribute to SF?
with what did he contribute to Komodo?

and, when he went back with Fritz 15, how much stronger he was able to make it?

50 elo, that is it, far below SF, far below Komodo.

if he was such of a genius, why was not he able to compete with SF and Komodo again?

VR contributed at an early stage of engine development, when engines almost completely lacked any positional chess knowledge.

as a relatively good chess played, he added some knowledge, and that was sufficient to top the rating lists for some time.

that is all, nothing more, nothing less.

If someone added some knowledge, he was Larry Kaufman, not V.R. You have some confusion. Rybka search routines made Rybka the strongest. He copied evaluation and move generator, it seems.

You can think what you want. It seems your opinion has nothing to do with facts. You cannot remember a post by Marco Costalba, writing his doubts about what was the difference between taking ideas and taking code. You cannot remember because you weren't here.

BTW, only honest people always put themselves in doubt.

The fact is, every top engine programmer studied each line of Robbolito to improve their engines. And those lines were from Rybka. Before Ippolit "revolution", other engines were much weaker. Every programmer can confirm it.

And I think this post is going very off topic. I'd like to stop talking about these old things.

it is you who started the thread, you who mentioned VR and lead the discussion this way, and now you want to stop it...

move generators are easy to write, nothing special in there, if your engine is 15% or 50% slower, does not matter that much in the long run.

it does not make sense that he copied evaluation and improved search, it is mostly quite the opposite; VR is a relatively good chess player, so that is where he could be of use, and that is what he did.

precisely because of better eval, and not search, is why R became very strong for a while.

you can easily deduce that even by the extremely low nps of the engine, only thing that can substantially slow down an engine is significantly bigger number of calculations, i.e. calculations related to specific knowledge.

search is not able to quite achieve a slowdown of 500% or so, no matter how intricate it is.

to tell you the truth, I am very bad programming newbie, but have still read Robbolito code and the codes of some 30 engines more, I do not see what is so elaborate there about Robbolito search, apart from splitting down some major routines.

by studying code and possibly implementing some new ideas, one can add 10, 30, maybe 100 elo, but not more, as there are considerable incompatibilities between codes.

only way to go rigth to the top is by complete copy, to be fully compatible.

studied, studied, studied..., if they studied so much, why not a single engine surpassed R substantially in the next 5 years?

only Houdini, SF and Komodo seem to be competely different branches.

You have no clue as usual - this is just another ruined thread after you jumped in.
Probably you are the only person in this forum, who doesn't know about Rybkas depth obfuscation to simulate a 'slow searcher' for the 'superior knowledge' hype a decade ago.

Don't waste your time, Guenther... He doesn't even know Houdini started as a modified Robbo...

one thing I can say is that Robbo code somehow reminds me of VR personality, whatever that means, but that is just a conjecture.

in any case, Houdini contributed a lot.

[Lyudmil Tsvetkov]

you can easily deduce that even by the extremely low nps of the engine, only thing that can substantially slow down an engine is significantly bigger number of calculations, i.e. calculations related to specific knowledge.

search is not able to quite achieve a slowdown of 500% or so, no matter how intricate it is.

So Vas fooled you too, and you are slow to catch on, while the rest of the world already caught on many years ago. Rybka had extremely high nps. But it disguised that by lying about it, and reporting a much lower number.

so, he lied about something after all?

how you know that nps were very high?

what is your proof? conjectural one?

I see from the games R played there is some piece of knowledge in there.

also, why would he want to lie about knowledge vs search/speed dilemma?

is knowledge that much prestigious and more important than search?

also for you?

why would not he want to claim instead that he contributed more to search?

also, if so many members here support VR contributed mainly with new search techniques and not new chess knowledge, why would he want to advance the opposite point by artificially lowering nps?

btw., how many engines out there do devide their nps by 4 or 10, or count irrelevant nodes?

[Lyudmil Tsvetkov]

what is the sense of Shredder or Komodo winning the WC, when we know SF is the best?

The point of course is that they are not really 'best', in any normal sense of the word. 10 Elo is an almost immeasurably small difference in performance. And it is obtained in a measurement made under conditions that do not always apply, through a method that is actually known to be flawed. If the rating difference between two human GM would be that small, no one would call foul play and demand his money back when the 'weaker' would win the tourney. People know it all depends on preparation, courage in choosing openings, playing style of the other participants, etc.

who is talking about 10 elo?

the difference between Shredder, the current software WC, and SF, is much larger, more like a gap.

besides, when you count contempt out, SF is leading by significantly more, even among the very top contenders.

nevermind, would not like this thread to spark any major controversies, my only problem currently is that I can not quite find one message I wanted to respond to, so my blast here...

[Lyudmil Tsvetkov]

I always assumed that WCCC had a computer science conference attached to it, but that seems to not (no longer?) be the case.

The ICGA event consists of a scientific conference, some Chess championships, and the Computer Olympiad for other games than Chess. (This year there were Chinese Chess, Shogi, Go, Draughts, Othello, Hex, NoGo, Dark Chinese Chess, Einstein Wuerfelt Nicht, Amazons, and Breakthrough tournaments, and then I probably forgot some.)

The conference lasts 3 days, also this year, and attracted some 40-50 participants (my guestimate, I did not actually count, and not everyone might have been present in the room when I took a peek anyway). About 30 people came over from Taiwan alone, where games programming seems to be a thriving research area. Most tournaments last one or two days. Most conference participants also participate in some game tournaments.

scientific conferences killed science.

40-50 participants, of which 30 Taiwanese, very representative indeed.

somehow, I can not think of a Taiwanese chess engine.

[Lyudmil Tsvetkov]

It was more important for me to understand people perception of the event. It's clear enough, now, that ICGA has to change several things for WCCC survival. While I consider Komodo deserves the title as the team is not guilty for Stockfish absence, I'm understanding this event is commercial-driven. Out of 4 participants, 3 were commercial. Organizers tried to justify Chiron presence because of its top ranking, but some free engines are stronger than Chiron.

I think this is a wrong perception. Last year there were many more non-commercial participants. Chiron was there because he was willing to come; if one of those 'stronger free engines' would have been willing to come, I am sure they would have been more than welcome.

Note that the WCCC primarily is set up as an event for programmers. Catering to the wishes of 'the public' has never been a goal in itself. It only appears on the agenda when it proves helpful for generating an income for ICGA, who can then use it to pay for the organizing costs, and distribute the surplus to the participants as prize money or travel reimbursement. But there is no paying audience. It is pretty amazing that people who do not contribute a single cent have the audacity to demand the event should be reshaped to suit their amusement... Having a large audience isn't even helpful for generating indirect income, through sponsors that want to reach the audience as a market for their products.

The problem is that having Chess programs battle it out is not a spectator sport anymore. Nothing there you could not do at home for a few dimes of electricity, 100 times more intensely. The days that strong computers were as rare as human super-GMs are long behind us. The only justification to have a special event for this is when the participants use hardware so powerful that no one can afford it at home. This is arguably what has been happening at WCCC, but in fact it turns out a deterrent rather than an attractor of public interest: people are no longer interested in things they cannot do at home themselves. I guess part of the problem is that even for strong Chess players there is no perceptible difference between a game played by 3300-Elo programs, and one played between 3500-Elo programs; it is way over their head either way.

A wrong attitude towards public is a potential nuclear explosion. If some media operator could have such an impression that'd have been the worst advertising for computer chess.

Only if you are dependent on that public for revenues. Which is not the case here. The only people paying used to be the participants. As it is now, it seems the conference and Olympiad participants are paying the Chess participants, which seems pretty outrageous.

And I think you are too hard on the participants: The playoffs were played at progressively faster TC, where operator speed and attention start to play a significant rule. So obviously distractions are unwelcome. Have you ever tried to have a chat with Carlson, while he was playing a game, or with Quintana while he was cycling in the Giro? If not, would you also say that these disrespect their audience?

3500-elo programs output better or considerably better PV than 3300-elo ones.

such PVs could be used for analysis, self-teaching, browsing through games, correspondence play, etc., etc., what countless chess players actually are doing.

what more is required so that building a stronger or much stronger engine becomes a priority.

why would you yhink that in the past interest in chess was higher?

it is quite the opposite, it is just that interest has moved from physical events and life attendance to the internet space.

definitely, inetrest in chess has only grown over the years, in considerable extent also due to mainstream 'strength-first' projects like SF.

do not you spot what the most frequent word on your own-moderated forum is: SF.

also, visit a chess forum, or even a chess tournament live.

first thing players say is: I checked with SF.

so, strength-first projects are not only very meaningful, but the essence of chess too.

[Lyudmil Tsvetkov]

And I think you are too hard on the participants: The playoffs were played at progressively faster TC, where operator speed and attention start to play a significant rule. So obviously distractions are unwelcome. Have you ever tried to have a chat with Carlson, while he was playing a game, or with Quintana while he was cycling in the Giro? If not, would you also say that these disrespect their audience?

Talking about dinosaurs, what is the purpose of operator? By the end of second decade of 21. century when AI is flourishing you need a human to enter the moves??? How retarded is that?

but then, you need someone to take pictures of, you can not just shoot a remote connection represented on a screen...

[Guenther]

I always assumed that WCCC had a computer science conference attached to it, but that seems to not (no longer?) be the case.

The ICGA event consists of a scientific conference, some Chess championships, and the Computer Olympiad for other games than Chess. (This year there were Chinese Chess, Shogi, Go, Draughts, Othello, Hex, NoGo, Dark Chinese Chess, Einstein Wuerfelt Nicht, Amazons, and Breakthrough tournaments, and then I probably forgot some.)

The conference lasts 3 days, also this year, and attracted some 40-50 participants (my guestimate, I did not actually count, and not everyone might have been present in the room when I took a peek anyway). About 30 people came over from Taiwan alone, where games programming seems to be a thriving research area. Most tournaments last one or two days. Most conference participants also participate in some game tournaments.

scientific conferences killed science.

40-50 participants, of which 30 Taiwanese, very representative indeed.

somehow, I can not think of a Taiwanese chess engine.

Somehow, I don't believe you exist, probably a troll bot experiment...

[Evert]

judging by R games, something I can certainly do, as the games are available, in distinction to the code, I can pretty much safely assume R does introduce quite some knowledge previous engines at the time did not have, so maybe I am more right after all, and real nps numbers are closer to outputted ones.

I do not know why everyone assumes R has more search innovations than chess knowledge ones, where is your proof?

or, most probably, as with you and a range of other members, to turn you the attack favour, you are mixing supposition/personal predilection with proof?

look at R games:

- fianchettoeing its king bishop sometimes, no other engine did that at the time, even not now
- excellent handling of connected passed pawns
- some smattering of imbalances, lacking in other engines
- more reasonable psqt for pawns and pieces alike, etc., etc.

this definitely seems to me like some knowledge base, does not it?

You can't conclude that from looking at games.
A well-tuned evaluation and sufficient depth will give the illusion of knowledge being present that cannot be traced to a single line of code, it becomes emergent.

Simple example: you do not need to teach a computer how to "build a bridge" in KRPKR. It will figure it out.

[Evert]

scientific conferences killed science.

40-50 participants, of which 30 Taiwanese, very representative indeed.

If that's where the research is happening, then it very well could be.

somehow, I can not think of a Taiwanese chess engine.

It's not a computer chess conference.
Also, computer chess is not scientifically interesting anymore.