People understand zilch about computerchess games.

I will try to keep this succinct.
I used to enjoy pitting computer programs against each other at long time controls and watching the games.
At one time my German ELO was close to 2100 ( this one used to be slightly lower than other scales). I didn't lose in my level of understanding of chess IMHO, though I would probably be way easier to beat now. For all I know this makes me one of the reasonably strong CC fans, most aficionados are weaker players than me.
I also used to post about my observations at times at CCC or in emails to programmers. I recently went back to many of them for my private enjoyment, and they still make some sense to me.
In 2003 I did the Junior-Crafty NPS challenge on CCC e.g. Only I know how little time I spent on the comments . I can check them with god-like programs now - what I said mostly made some sense according to them, even though the programs were like 500 or 600 ELO stronger than me.
Then I took a +major+ break from computerchess.
I came back recently, at at time when Stockfish and LC0 were the real deal. I did what I used to do - I did slow games, watched them live - and thought about them.
To no avail. I actually have +no idea+ what they are doing. Stockfish's pawn sacrifices make no sense to me e.g. - for like 1.5 tempi ( I learned different things). I can see, that they work - but I really don't understand anything about current computerchess games.
People still post about these games. But am I too arrogant to think that most players won't understand them any better than I do?
Magnus Carlsen is likely to understand them. He is also likely to spot weak moves, though he'd lose a chess match against these entities.
But ordinary earthlings? I seriously doubt they understood anything better than me, and I don't understand +anything+.
Peter

I am already lost after 10-11 plies, so cheer up

Of course the ELO of Stockfish, even on a mobile phone, is far too high for us. But can you see that LC0 is outplaying most chess engines ?!

Peter Berger wrote:

I will try to keep this succinct.
I used to enjoy pitting computer programs against each other at long time controls and watching the games.
At one time my German ELO was close to 2100 ( this one used to be slightly lower than other scales). I didn't lose in my level of understanding of chess IMHO, though I would probably be way easier to beat now. For all I know this makes me one of the reasonably strong CC fans, most aficionados are weaker players than me.
I also used to post about my observations at times at CCC or in emails to programmers. I recently went back to many of them for my private enjoyment, and they still make some sense to me.
In 2003 I did the Junior-Crafty NPS challenge on CCC e.g.  Only I know how little time I spent on the comments . I can check them with god-like programs now - what I said mostly made some sense according to them, even though the programs were like 500 or 600 ELO stronger than me.
Then I took a +major+ break from computerchess.
I came back recently, at at time when Stockfish and LC0 were the real deal. I did what I used to do - I did slow games, watched them live - and thought about them.
To no avail. I actually have +no idea+ what they are doing. Stockfish's pawn sacrifices make no sense to me e.g. - for like 1.5 tempi ( I learned different things). I can see, that they work - but I really don't understand anything about current computerchess games.
People still post about these games. But am I too arrogant to think that most players won't understand them any better than I do?
Magnus Carlsen is likely to understand them. He is also likely to spot weak moves, though he'd lose a chess match against these entities.
But ordinary earthlings? I seriously doubt they understood anything better than me, and I don't understand +anything+.
Peter

Hey Peter, nice to see you join this forum! Many very strong chess players feel the same way when watching the top engines now. Perhaps, we need to just accept there is a level that only the top humans minds in chess will semi understand. However, when these engines are beating some of the best in the world, two pawns down! There is still lots for us to learn!

Peter Berger wrote:

I will try to keep this succinct.
I used to enjoy pitting computer programs against each other at long time controls and watching the games.
At one time my German ELO was close to 2100 ( this one used to be slightly lower than other scales). I didn't lose in my level of understanding of chess IMHO, though I would probably be way easier to beat now. For all I know this makes me one of the reasonably strong CC fans, most aficionados are weaker players than me.
I also used to post about my observations at times at CCC or in emails to programmers. I recently went back to many of them for my private enjoyment, and they still make some sense to me.
In 2003 I did the Junior-Crafty NPS challenge on CCC e.g.  Only I know how little time I spent on the comments . I can check them with god-like programs now - what I said mostly made some sense according to them, even though the programs were like 500 or 600 ELO stronger than me.
Then I took a +major+ break from computerchess.
I came back recently, at at time when Stockfish and LC0 were the real deal. I did what I used to do - I did slow games, watched them live - and thought about them.
To no avail. I actually have +no idea+ what they are doing. Stockfish's pawn sacrifices make no sense to me e.g. - for like 1.5 tempi ( I learned different things). I can see, that they work - but I really don't understand anything about current computerchess games.
People still post about these games. But am I too arrogant to think that most players won't understand them any better than I do?
Magnus Carlsen is likely to understand them. He is also likely to spot weak moves, though he'd lose a chess match against these entities.
But ordinary earthlings? I seriously doubt they understood anything better than me, and I don't understand +anything+.
Peter

I will try to match your brevity, and necessarily oversimplify in the process.

I know it's wrong, but imagine that the improvement was entirely down to depth of search: at a shallow depth (old, slow computers with not much memory), then to understand what's going on, you could play out alternative lines and see where they fail. However, at a deep depth (modern, quick computers with a ton of memory), when you try to play out the alternative lines to the point where they fail, the branching factor makes it prohibitively difficult to find the all the failings in the alternative moves.

The last engine I fully understood was Zarkov 6.55. It played reasonably well positionally, and it did not calculate too deep, so the way it treated most of the positions I was interested in "made sense" for me. And I could also check some of the lines.

In the cases when Zarkov was lost in closed positions of the French, or Slav, I could use Gandalf, or ProDeo, and recently I found Winter. I feel I can understand the variations I see on the screen because the evals were written differently, by humans, without using "texel tunning", "sigmoid coherence functions" (kidding) etc.

An example: I liked Zarkov and Wasp 1.02, but with each newer version Wasp was pruning more and more, being faster and faster, and John Stanback adopted the new approaches, and it just lost any sense for me.

But the baseline is that I see engines and computerchess as just a mean to play and enjoy chess. I am not interested in engine tournaments, in TCEC, rating lists, etc.

To finish, something I thought about lately: I agree more and more with Alexander Naumov who believed that open source had its drawbacks. That's of course not the he formulated it, but... Every week we hear about a clone or suspected clone, but the truth seems to be, from my perspective, that the SF team monopolized the rights to originality. I'll try to explain: while they can incorporate every single idea (they're good enough not to use code, and they have a code base) they find in other open source engines without problems, any new programmer has to open his code, and share his ideas, simply to avoid being accused of cloning. It just has killed originality, on one hand, and usability, on the other. But enough about it.

matejst wrote:

The last engine I fully understood was Zarkov 6.55. It played reasonably well positionally, and it did not calculate too deep, so the way it treated most of the positions I was interested in "made sense" for me. And I could also check some of the lines.

In the cases when Zarkov was lost in closed positions of the French, or Slav, I could use Gandalf, or ProDeo, and recently I found Winter. I feel I can understand the variations I see on the screen because the evals were written differently, by humans, without using "texel tunning", "sigmoid coherence functions" (kidding) etc.

An example: I liked Zarkov and Wasp 1.02, but with each newer version Wasp was pruning more and more, being faster and faster, and John Stanback adopted the new approaches, and it just lost any sense for me.

But the baseline is that I see engines and computerchess as just a mean to play and enjoy chess. I am not interested in engine tournaments, in TCEC, rating lists, etc.

To finish, something I thought about lately: I agree more and more with Alexander Naumov who believed that open source had its drawbacks. That's of course not the he formulated it, but... Every week we hear about a clone or suspected clone, but the truth seems to be, from my perspective, that the SF team monopolized the rights to originality. I'll try to explain: while they can incorporate every single idea (they're good enough not to use code, and they have a code base) they find in other open source engines without problems, any new programmer has to open his code, and share his ideas, simply to avoid being accused of cloning. It just has killed originality, on one hand, and usability, on the other. But enough about it.

I feel like Boban made some very good points here.

I always liked the analysis of Prodeo 1.2-1.7 and thought the moves made sense.

Same with Wasp 1.02, Frenzee 3.5.19 and other old engines...even WChess 1.06 despite being only 2330 or so.

I seem to remember years ago Boban tweaked Zarkov 6.55 a little to make it more Karpov-like which was cool.

For "sample mining" (my term, which means to have engines play tournaments *only* with the lines/tabiyas from my repertoire and thus, learn how to play it well, new plans, motifs etc) I also liked Thinker, various Rodent personalities and other "non-mainstream" engines.

With regard to open-source, yeah, I think this phenomenon is kinda weird.

How many other software industries is it the *norm* to open up your source code for all to view and plunder from?

Especially for competitive programmers.

How can a single programmer hope to compete with the TEAM of talented guys working on Stockfish, if this TEAM can *also* plunder ideas from his code?

For example, obviously, the Ethereal programmer is enormously talented, but personally, I think he should never have opened his code in the first place.

He is good enough to make something completely unique and keep the source a secret.

The same as Chris W, who recently wrote CoronaVirusChess - a clearly very strong engine.

40 years of experience wasn't enough for some idiots who still started accusing Chris, simply because he keeps the source closed (as he should).

Lance Perkins, author of Thinker was a guy who hid the PV of his (closed source) engine and drew the attention of the witchhunt crowd.

Even though he explained why he does already - to make Thinker more of a human-like opponent:

The question was: Hi folks. I'm sorry if this has been discussed already - I'm new here. The question is, are there engines which aim to play in a human-like way? If so, what ideas exist for achieving it?

Lance answered:

Quote :

1. Most humans don't play at the level of GMs. The GMs are not a representation of the majority of humans. In that sense, GMs are like computers - they don't play like humans.

2. Humans don't give out PVs while they are thinking.

3. Humans make blunders.

4. Humans are inconsistent. They get affected by all sorts of distractions, like sickness, fatigue, excitement, etc., and they end-up playing worse or better.

5. Humans converse when they play.

6. Humans are not capable of 10 moves in 1 second.

7. Humans understand closed/drawn positions. But computers play on forever.

8. Humans don't play from a copy of some endgame database. They won't give out a Queen just to get to a position that is in the endgame database.

9. Humans don't play from an opening book. They have memorized opening sequences. But is does not accomodate for a huge number of 100-ply sequences. It also cannot be swapped out at will (e.g., small book vs large book).

10. Humans don't play with fixed amount of time for each move. In contrast, when you tell an engine that the game is 40 moves in 40 minutes, it will make moves at around 1 minute each. Some move may be either too quick (easy moves) or a fixed multiple (e.g., always 3 minutes).

Having said all those, I think that ChessMaster is the closest to how humans play. It can be configured to play a whole range of styles and preferences.

In the Thinker/ThinkerBoard that plays at FICS, I also have made attempts to address some items I have listed.

(1) The engine runs on a pocket PC.
(2) Just no pv.
(3) There is a mode in Thinker, where, occasionally, the second best move is used.
(4) Different settings are used depending on the hour of the day. Best settings are used in the morning. Worse settings are used at night.
(5) Players can converse with ThinkerTalk.
(6) The engine's minimum move time is 1 second. If it is done searching for less than 1 second, it will sleep for a randmon amount of time (between 1 to 2 seconds).
(7) {no working solution yet}
( Just no eg bb/tb.
(9) Just no opening book.
(10) There is a mode in Thinker, where, the time allocation is randomized.

http://www.talkchess.com/forum3/viewtopic.php?t=20957


The "how could someone create something so strong without cloning?" accusations would come regardless of whether he was guilty or not.

This is how the CC community is.

Very true.

Also note that if it had been someone else entirely, other than Pawel, to release OpenTal, there would have been constant talk of someone cloning Rodent, open source or not.

Never mind the highly original playing style, unlike virtually all other engines, just blackwash it as being a clone. End of story. In the meantime, these people would never bother to look at a single game to see how the engine plays.

I think that for computer chess
I think that it may be better if people concentrate in other problems that are not engine strength for example how to teach humans to play better and to reply specific questions that is not about winning the game.

Some questions that I know no engine that can help me.

1)Can the side to move force at least a draw by repetition or by stalemate or by the fifty move rule or mate in n plies?

There are cases when humans play with the target to make a draw for example if they know that a draw is enough for them to win some prize so for analysis they may be interested to know if they missed a simple way to force at least a draw.

repetition during the search that is not a repetition of previous game position can be considered as a draw in this case because if you can force the root position in 8 plies then it is trivial that you can force it again later and it is going to be a draw by repetition in 16 plies.

2)Can the side to move win material when you search n plies forward when winning material can be without qsearch,with qsearch that include only captures or with qsearch that include captures and checking moves in the first ply of the qsearch(also with all replies to the check).

Examples:
After 1.d4 d5 2.c4 black can win a pawn when you search 1 ply forward with no qsearch(dxc4) even with qsearch that include captures black can win a pawn because white has no captures but if you allow checks in the first ply of the qsearch then black does not win a pawn.
dxc4 Qa4+ black move Qxc4

The same can be done with more than 1 ply and I think that it can be training for humans.

Of course 1 ply is too easy for strong humans but you can do it also for 9 plies or 11 plies and with some verification deeper search to see that you do not lose the material you won when you search 2 plies or 4 plies deeper.

The idea is that specific problems that are not winning the game can be defined by the engine to train humans about tactics and
I think that it may be also interesting to learn about cases when winning material is bad because of positional reasons when the engine analyze game(comment of some engine that analyze some game that white can force winning material by some line but it is not good for positional reasons may be interesting).

Note that there should be difference between positional reason and very deep tactical reasons even if from human point of view they are the same.

There are positional reasons that winning material is not good that the engine simply cannot see if it searches for only material evaluation because it does not search deep enough but it still understand with normal search that winning material is not good and there are tactical reasons that humans cannot but only material evaluation can see after a very deep search.

practically if the computer find that winning material is not verified with searching 21 plies forward then it can be something that is practically impossible for humans to see but it can be also something that is practically possible for humans to see if there is a single forced long line.

Note that for question 2 I remember that many years ago programmer of yace sent me an engine that can search for only material evaluation(I had no option to reply all the questions that I asked but it was a good progress for computer chess in case that many authors follow it and have at least an option to search for only material because I am sure engines today should be significantly stronger than yace in this task).

3)I would like engines to be able to tell me tactical motives of a winning material or mate problem and when they get a position they tell me if the side to move can win by back rank mate or maybe win material by a fork or by a pin and how much material can the side to move win based on some values of pieces that you choose(can be 1,3,3,5,9 but also different values).

Brendan wrote:

matejst wrote:

The last engine I fully understood was Zarkov 6.55. It played reasonably well positionally, and it did not calculate too deep, so the way it treated most of the positions I was interested in "made sense" for me. And I could also check some of the lines.

In the cases when Zarkov was lost in closed positions of the French, or Slav, I could use Gandalf, or ProDeo, and recently I found Winter. I feel I can understand the variations I see on the screen because the evals were written differently, by humans, without using "texel tunning", "sigmoid coherence functions" (kidding) etc.



An example: I liked Zarkov and Wasp 1.02, but with each newer version Wasp was pruning more and more, being faster and faster, and John Stanback adopted the new approaches, and it just lost any sense for me.

But the baseline is that I see engines and computerchess as just a mean to play and enjoy chess. I am not interested in engine tournaments, in TCEC, rating lists, etc.

To finish, something I thought about lately: I agree more and more with Alexander Naumov who believed that open source had its drawbacks. That's of course not the he formulated it, but... Every week we hear about a clone or suspected clone, but the truth seems to be, from my perspective, that the SF team monopolized the rights to originality. I'll try to explain: while they can incorporate every single idea (they're good enough not to use code, and they have a code base) they find in other open source engines without problems, any new programmer has to open his code, and share his ideas, simply to avoid being accused of cloning. It just has killed originality, on one hand, and usability, on the other. But enough about it.

I feel like Boban made some very good points here.

I always liked the analysis of Prodeo 1.2-1.7 and thought the moves made sense.

Same with Wasp 1.02, Frenzee 3.5.19 and other old engines...even WChess 1.06 despite being only 2330 or so.

I seem to remember years ago Boban tweaked Zarkov 6.55 a little to make it more Karpov-like which was cool.

For "sample mining" (my term, which means to have engines play tournaments *only* with the lines/tabiyas from my repertoire and thus, learn how to play it well, new plans, motifs etc) I also liked Thinker, various Rodent personalities and other "non-mainstream" engines.

With regard to open-source, yeah, I think this phenomenon is kinda weird.

How many other software industries is it the *norm* to open up your source code for all to view and plunder from?

Especially for competitive programmers.

How can a single programmer hope to compete with the TEAM of talented guys working on Stockfish, if this TEAM can *also* plunder ideas from his code?

For example, obviously, the Ethereal programmer is enormously talented, but personally, I think he should never have opened his code in the first place.

He is good enough to make something completely unique and keep the source a secret.

The same as Chris W, who recently wrote CoronaVirusChess - a clearly very strong engine.

40 years of experience wasn't enough for some idiots who still started accusing Chris, simply because he keeps the source closed (as he should).

Lance Perkins, author of Thinker was a guy who hid the PV of his (closed source) engine and drew the attention of the witchhunt crowd.

Even though he explained why he does already - to make Thinker more of a human-like opponent:

The question was: Hi folks. I'm sorry if this has been discussed already - I'm new here. The question is, are there engines which aim to play in a human-like way? If so, what ideas exist for achieving it?

Lance answered:

Quote :

1. Most humans don't play at the level of GMs. The GMs are not a representation of the majority of humans. In that sense, GMs are like computers - they don't play like humans.

2. Humans don't give out PVs while they are thinking.

3. Humans make blunders.

4. Humans are inconsistent. They get affected by all sorts of distractions, like sickness, fatigue, excitement, etc., and they end-up playing worse or better.

5. Humans converse when they play.

6. Humans are not capable of 10 moves in 1 second.

7. Humans understand closed/drawn positions. But computers play on forever.

8. Humans don't play from a copy of some endgame database. They won't give out a Queen just to get to a position that is in the endgame database.

9. Humans don't play from an opening book. They have memorized opening sequences. But is does not accomodate for a huge number of 100-ply sequences. It also cannot be swapped out at will (e.g., small book vs large book).

10. Humans don't play with fixed amount of time for each move. In contrast, when you tell an engine that the game is 40 moves in 40 minutes, it will make moves at around 1 minute each. Some move may be either too quick (easy moves) or a fixed multiple (e.g., always 3 minutes).

Having said all those, I think that ChessMaster is the closest to how humans play. It can be configured to play a whole range of styles and preferences.

In the Thinker/ThinkerBoard that plays at FICS, I also have made attempts to address some items I have listed.

(1) The engine runs on a pocket PC.
(2) Just no pv.
(3) There is a mode in Thinker, where, occasionally, the second best move is used.
(4) Different settings are used depending on the hour of the day. Best settings are used in the morning. Worse settings are used at night.
(5) Players can converse with ThinkerTalk.
(6) The engine's minimum move time is 1 second. If it is done searching for less than 1 second, it will sleep for a randmon amount of time (between 1 to 2 seconds).
(7) {no working solution yet}
( Just no eg bb/tb.
(9) Just no opening book.
(10) There is a mode in Thinker, where, the time allocation is randomized.

http://www.talkchess.com/forum3/viewtopic.php?t=20957


The "how could someone create something so strong without cloning?" accusations would come regardless of whether he was guilty or not.

This is how the CC community is.

Hi Brendon,

Yes, I decided not just to keep the source closed but also the executable. No part of Bella-Ciao (CoronaVirusChess) has ever left my office, and never will. I refuse to give the multiple number of malignant talkchess trolls even the tiniest bit for de-compiling, reverse engineering, to hang their accusations and witch hunts on. Only thing that gets out are games it plays online (which is quite enough testing for me) and any games I find worthy of publishing. And yes, it is a Tal engine, it’s good at manoeuvring itself into king attacks as several of the last few games in the Rapid Championship showed.

A few trivia, ideas without much importance, from a personal angle.

One of the projects that does not make too much noise on CCC, despite being one of the most interesting in the computerchess world, is LucasChess. I presume everybody here knows the software and has tried it, and it gives more to a chess player than almost anything else. [Recently, Nguyen (I hope I did not misspelled his name) made a beautiful GUI, but it has been geared by some members of CCC, in its development, toward engine matches and engine testing, instead of playing, analyzing, etc. Better database functions would have been more important for me, but it is just me I guess.] There are paths for chess software to help players play better chess, or to simply enjoy chess, and most of them remain unexplored.

CCC remains a very important place -- but the focus should be diversified, and part of the focus should be on usability. Living in a country that was destroyed by embargos and bombing and where the avg. wages are about 300 euros, free software remains very important to most chess players. Ed's efforts to provide free databases, free opening trees and other utilities, Lucas work on his software, SCID, ChessX, merit more attention.

A few words about bias and credits. When I wrote that SF monopolized originality and ideas, I meant a few concrete examples. Recently, SF started using NNs for its eval, but Jonathan Rosenthal and probably Youri Matiounine incorporated NNs in their engines years before SF. Jonathan's engine is open source, but nobody cared about the idea until the advent of NNUE. In a sense, SF monolithized the whole development paradigm. Other ways and other directions disappeared in the darkness, and we still use procedures developed by Vaclav Rajlich.

Something else, finally. As we know all -- and unfortunately -- speed gives more than evaluation. For years, the engine with the best evaluation seemed to be HIARCS, but it remains far from the top of the rating lists. The success of the NNs is, imho, due to the fact that not much was done in the development of the evaluation function for years. There was the unfortunate ERL project that could have bear fruits, but nobody but Ed was interested. It reveals a profound change in developers: most of young engine authors are engineers, most of old authors are chess players of chess lovers. The development of computerchess today is a competition of engineers and mathematicians, and chess itself lost somehow its importance in it.

I am still using Arena for engine testing. It’s a reliable software.
It runs 7/24 on all pcs, often for years. It runs uci and winboard engines and once you started a tournament you can easily add new versions and entries without losing the old data.
It generates a table where the results of all engines are counted.
Once the tournament is finished and you add 2 new engines you only see games of these 2 versus all others.

But with the difference to a match that after the match is over the 2 show their ranking in the
Tournament.
Also this way an engine is tested against several opponents and not 1.

My pc is connected with my TV set. I am sitting in my living room and can watch the engines fight.
You can bring the output so big you can easily watch it on a big tv.

Instead of watching stupid talk shows I watch engines play.

I am mostly using arena 2.01.

You have mouse and keyboard without cable so you can edit and arrange icons in any kind of window on the TV set.

The only problem is to get interesting engines.
I don’t want 5-19 Stockfish alike engines on top of the list.
This makes no sense watching.

matejst wrote:

The development of computerchess today is a competition of engineers and mathematicians, and chess itself lost somehow its importance in it.

Hey, nothing against mathematicians! The intersecting set with chess players is non-empty btw.

On a more serious note, this:

matejst wrote:

I agree more and more with Alexander Naumov who believed that open source had its drawbacks. That's of course not the he formulated it, but... Every week we hear about a clone or suspected clone, but the truth seems to be, from my perspective, that the SF team monopolized the rights to originality. I'll try to explain: while they can incorporate every single idea (they're good enough not to use code, and they have a code base) they find in other open source engines without problems, any new programmer has to open his code, and share his ideas, simply to avoid being accused of cloning. It just has killed originality, on one hand, and usability, on the other.

and that:

Brendan wrote:

How can a single programmer hope to compete with the TEAM of talented guys working on Stockfish, if this TEAM can *also* plunder ideas from his code?

For example, obviously, the Ethereal programmer is enormously talented, but personally, I think he should never have opened his code in the first place.

... are extremely important thoughts, imho. Thanks for pointing it out. Something is rotten in the state of CC.

matejst wrote:

...It reveals a profound change in developers: most of young engine authors are engineers, most of old authors are chess players of chess lovers. The development of computerchess today is a competition of engineers and mathematicians, and chess itself lost somehow its importance in it.

This is the way it should be: if you want to build a good car, you don't go to a driver!

Ras,

While I took your joke for what it is, a joke, on a more serious note I cannot say that something important is rotten in the computerchess world. The input of engineers and mathematicians -- to solve engineering and maths problems -- has been extremely important, probably crucial. I just appeal for some tolerance and some diversity: tolerance among programmers, and diversity in development.

I am fine with exchange of ideas -- but that existed already when the sources were closed. And I do think that testing procedures made the big difference and allowed for a big Elo jump, not only sources, probably more than sources. Concepts, ideas, are enough for a good developer. Andrew Grant papers, Thomas Petzke articles, the whole CPW -- and I omit the 99% of the contributions -- are a trove of knowledge.

But I also understand authors like Andrew Grant: he made his code available, he wrote explanatory papers, he developed openbench to help others, and, imho, he gave more than he is aware of. A few years ago, he could have made a living out of his work, or, at least, earn enough to cover costs. I also would probably expect, in his place, others to follow the universally acclaimed rules, although he has been heavily hampered by these same rules. I feel that the problem is about delimiting open sources GPL requests / and the concept of originality. Since SF, "clone" ceased to mean what it meant before. I am inclined to more tolerance, and a rethinking of the potential problems.

Otoh, I would like to see more GUIs, more options, more tweakabily, more tools to play, learn and improve. Not the same thing over and over, like we have now for phones, but something like the use of WDL to guide you in analysis, use of colors to reveal plans, etc. There are a lot of ideas to discover.

TheSelfImprover wrote:

This is the way it should be: if you want to build a good car, you don't go to a driver!

In a way, I feel you need both. And you need both formula 1 and family cars.

You´re absolutely right. I think (hope) I understand your point: collaboration is by no means a bad thing. More eyes and brains can produce stunning results over time, no doubt about it. And is extremely friendly, an asset on its own.

But. The main point which struck me like a lightning with your posts is that opening codes and sharing everything removes competition out of the equation at least to some, perhaps significant, degree. Maybe even too much. And, slowly and stead, this could lead to both, a loss of individuality and quality. Removing those two, competition and individuality, seems to be the dark side of the collaboration card. Most ingenious inventions and achievements in human history have been found where some form of competition took place.

Now, if you have Stockfish assimilating everything, then, in the long run, resistance is futile. Maybe they should have already renamed the program to BorgChess, or so (or am I too ugly now...).

Chris Whittington wrote:

Brendan wrote:

matejst wrote:

The last engine I fully understood was Zarkov 6.55. It played reasonably well positionally, and it did not calculate too deep, so the way it treated most of the positions I was interested in "made sense" for me. And I could also check some of the lines.

In the cases when Zarkov was lost in closed positions of the French, or Slav, I could use Gandalf, or ProDeo, and recently I found Winter. I feel I can understand the variations I see on the screen because the evals were written differently, by humans, without using "texel tunning", "sigmoid coherence functions" (kidding) etc.



An example: I liked Zarkov and Wasp 1.02, but with each newer version Wasp was pruning more and more, being faster and faster, and John Stanback adopted the new approaches, and it just lost any sense for me.

But the baseline is that I see engines and computerchess as just a mean to play and enjoy chess. I am not interested in engine tournaments, in TCEC, rating lists, etc.

To finish, something I thought about lately: I agree more and more with Alexander Naumov who believed that open source had its drawbacks. That's of course not the he formulated it, but... Every week we hear about a clone or suspected clone, but the truth seems to be, from my perspective, that the SF team monopolized the rights to originality. I'll try to explain: while they can incorporate every single idea (they're good enough not to use code, and they have a code base) they find in other open source engines without problems, any new programmer has to open his code, and share his ideas, simply to avoid being accused of cloning. It just has killed originality, on one hand, and usability, on the other. But enough about it.

I feel like Boban made some very good points here.

I always liked the analysis of Prodeo 1.2-1.7 and thought the moves made sense.

Same with Wasp 1.02, Frenzee 3.5.19 and other old engines...even WChess 1.06 despite being only 2330 or so.

I seem to remember years ago Boban tweaked Zarkov 6.55 a little to make it more Karpov-like which was cool.

For "sample mining" (my term, which means to have engines play tournaments *only* with the lines/tabiyas from my repertoire and thus, learn how to play it well, new plans, motifs etc) I also liked Thinker, various Rodent personalities and other "non-mainstream" engines.

With regard to open-source, yeah, I think this phenomenon is kinda weird.

How many other software industries is it the *norm* to open up your source code for all to view and plunder from?

Especially for competitive programmers.

How can a single programmer hope to compete with the TEAM of talented guys working on Stockfish, if this TEAM can *also* plunder ideas from his code?

For example, obviously, the Ethereal programmer is enormously talented, but personally, I think he should never have opened his code in the first place.

He is good enough to make something completely unique and keep the source a secret.

The same as Chris W, who recently wrote CoronaVirusChess - a clearly very strong engine.

40 years of experience wasn't enough for some idiots who still started accusing Chris, simply because he keeps the source closed (as he should).

Lance Perkins, author of Thinker was a guy who hid the PV of his (closed source) engine and drew the attention of the witchhunt crowd.

Even though he explained why he does already - to make Thinker more of a human-like opponent:

The question was: Hi folks. I'm sorry if this has been discussed already - I'm new here. The question is, are there engines which aim to play in a human-like way? If so, what ideas exist for achieving it?

Lance answered:

Quote :

1. Most humans don't play at the level of GMs. The GMs are not a representation of the majority of humans. In that sense, GMs are like computers - they don't play like humans.

2. Humans don't give out PVs while they are thinking.

3. Humans make blunders.

4. Humans are inconsistent. They get affected by all sorts of distractions, like sickness, fatigue, excitement, etc., and they end-up playing worse or better.

5. Humans converse when they play.

6. Humans are not capable of 10 moves in 1 second.

7. Humans understand closed/drawn positions. But computers play on forever.

8. Humans don't play from a copy of some endgame database. They won't give out a Queen just to get to a position that is in the endgame database.

9. Humans don't play from an opening book. They have memorized opening sequences. But is does not accomodate for a huge number of 100-ply sequences. It also cannot be swapped out at will (e.g., small book vs large book).

10. Humans don't play with fixed amount of time for each move. In contrast, when you tell an engine that the game is 40 moves in 40 minutes, it will make moves at around 1 minute each. Some move may be either too quick (easy moves) or a fixed multiple (e.g., always 3 minutes).

Having said all those, I think that ChessMaster is the closest to how humans play. It can be configured to play a whole range of styles and preferences.

In the Thinker/ThinkerBoard that plays at FICS, I also have made attempts to address some items I have listed.

(1) The engine runs on a pocket PC.
(2) Just no pv.
(3) There is a mode in Thinker, where, occasionally, the second best move is used.
(4) Different settings are used depending on the hour of the day. Best settings are used in the morning. Worse settings are used at night.
(5) Players can converse with ThinkerTalk.
(6) The engine's minimum move time is 1 second. If it is done searching for less than 1 second, it will sleep for a randmon amount of time (between 1 to 2 seconds).
(7) {no working solution yet}
( Just no eg bb/tb.
(9) Just no opening book.
(10) There is a mode in Thinker, where, the time allocation is randomized.

http://www.talkchess.com/forum3/viewtopic.php?t=20957


The "how could someone create something so strong without cloning?" accusations would come regardless of whether he was guilty or not.

This is how the CC community is.

Hi Brendon,

Yes, I decided not just to keep the source closed but also the executable. No part of Bella-Ciao (CoronaVirusChess) has ever left my office, and never will. I refuse to give the multiple number of malignant talkchess trolls even the tiniest bit for de-compiling, reverse engineering, to hang their accusations and witch hunts on. Only thing that gets out are games it plays online (which is quite enough testing for me) and any games I find worthy of publishing. And yes, it is a Tal engine, it’s good at manoeuvring itself into king attacks as several of the last few games in the Rapid Championship showed.

Fair enough mate.

Just be sure to share some of the sparkling games with us.

Oh. And perhaps setup a Coronaviruschess bot on lichess?

Its fun to see humans being hammered by not-so-correct engines, as I've been experimenting with recently for an upcoming project with Pawel Koziol.

\"TheSelfImprover wrote:

I will try to match your brevity, and necessarily oversimplify in the process.
I know it's wrong, but imagine that the improvement was entirely down to depth of search: at a shallow depth (old, slow computers with not much memory), then to understand what's going on, you could play out alternative lines and see where they fail. However, at a deep depth (modern, quick computers with a ton of memory), when you try to play out the alternative lines to the point where they fail, the branching factor makes it prohibitively difficult to find the all the failings in the alternative moves.

I don't think it is only about search.
To also try to answer Thorsten: LC0 has a very appealing style that looks way easier to understand for an ordinary earthling. But logic suggests that this mostly means that I miss the points where it does things I don't. Let's call this the "Capablanca effect" .
Stockfish knows what it is doing ( I have tried to use it at very short time controls, and it still does things I don't get at all and can only notice to work).
Actually I don't have much more to say about this topic.
But there is a related question, that I keep coming back to, that maybe one of you can answer. Older programs had a feature that made them easily recognizable as computer players: once they reached a nice position, they made things like Kg8-h8-g8 ( Rebel and CSTal certainly did this e.g.). Neither LC0 nor Stockfish do this at all. How come?
Peter

Mclane wrote:

My pc is connected with my TV set. I am sitting in my living room and can watch the engines fight.
You can bring the output so big you can easily watch it on a big tv.

Instead of watching stupid talk shows I watch engines play.

lol Me too! Except I connect to a projector which projects the PC screen to the wall like a cinema.

We should argue about politics less and talk more about chess and archaeology.

Hi there,

no, no ... not only pruning for higher Wasp versions.
Much more complicated!!

I like Zarkov 6.44 too.
In Zarkov Download files is the own Windows GUI.

Also nice for self playing!

Best
Frank

Note:
Have a look on the site from Fanz Huber.
Franz offer all the older chess computer for self playing and different from the older DOS programs, runs under Windows.

In my humble opinion the best work in computer chess ever!

Franz Huber:
https://fhub.jimdofree.com/

Here CBEmu, TascEmu and Chess-Dos-VirtualMachine

Best
Frank

Hi there,

again Wasp and to the logic for the development.

1. Wasp 4.00 with 3000 Elo = 150 Elo missed.
On DGT-Pi Wasp is running with 1.2Ghz with 2700 Elo.
Max. human level = 2850 Elo = 150 Elo missed ... my logic!

So, all players in the World can adjust the level with nodes per second. For Super Conny level Wasp 4.00 need around 20-30 nodes per second. In such games Super Conny are stronger in tactics but Wasp naturally with 20-30 nodes stronger in endgames. The problem in the past, the chess computer times, are solved. Because most of the chess computer are not balanced in strength (playing phases) and lost a lot strength in endgames. Max. TheKing and Genius ... because both have a better endgame as the other older programs inside the chess computers. Maybe from Ed the Risc 1Mb.

2. Wasp 4.00 have 3000 Elo = 250 Elo missed?
On King Performance with 300 Elo Wasp need 250 Elo more for the max. level from 2850. Only an example because Wasp is not available for King Performance. For feature developments of chess computers is very interesting that on 300Mhz CPU a program, simulate the style of humans and the max. humans level from 2850 Elo.

Example 1:
If Wasp 4.x have 2800 Elo on DGT Pi and a strong grandmaster have interest to play a game, the grandmaster should have a chance in endgames. Wasp have problems in endgames and can lost vs. strongest players in endgames. Not possible with Stockfish, not possible that programs like Stockfish plays with 20-30 nodes per second chess games and if ... the strength isn't good enough (to many blunders produced).

Wasp simulate 1500 - at the moment (v4.00) 2.700 Elo on DGT Pi with 1.2Ghz.

So what is important for Wasp development?
To hold exactly the same strength and weaknesses but try to make Wasp stronger.
Thats what for myself important.

The aggressive pawns in the mid-game and not to solved the endgames problems in detail. The balance between the playing phase, a strong king safty ... strong in openings, strong in mid-games  and around light lesser level in transposition into endgame and lesser level in endgame is important to hold the human style.

Example 2:
I am playing vs. Wasp 3.75 on DGT many games with 2.250 Elo Wasp Level. I made around 40%. At the moment around 25-30%. For one year I am better in form.

Most important is to search the level for self playing.
Computer should make 60-70% I think and should have a very aggressive style (what humans like).

So, with Wasp and DGT-Pi is for myself the perfect chess combination born.
Furthermore, the FEOBOS files are inside.

The admin from chesscomputer.info do that for me and I put the DGT-Pi file in my donwload area.
Many persons like that and I am very happy with this combination.

But often I am playing with Zarkov 6.55 or other chess programs on my DGT Board without DGT Pi.
I like here AnMon 5.75 most.

Best
Frank

To Peter: when the engines  have a good position, they fail to execute the opponent.
Also when tal with the high speculative score
Had a winning position the scores were high and the extensions deep. That misses accuracy. We joked and to use genius then for winning positions or tried to knock out tal function in a won position.

I think lc0 wins because it sees something stockfish or komodo cannot handle.

Hi Thorsten,

have a look in the games Hiarcs - Lc0 or Crafty vs. Lc0 ... the both lost games in my tourney. Hiarcs and Crafty comes with a good positional play and won, all the others lost in the first round at the moment.

Lc0 is for sure great but have to many holes.
On grafic cards the holes in tactic with many pieces on board, but positional strong.
And with min. performance on CPU the positional programs seems to be have an advantage, endgames are also not very strong.

So you know excactly what Lc0 can do.
That's not the best style of chess for humans but of course interesting because it's another style and often unknown for humans.

It's interesting to have programs like Junior, Hiarcs, ProDeo or Quazar in the tourney. Old programs like the good and old programmings vs. the fast searcher. I am sure that Quazar (very positional engine) will get better results vs. all the fast searcher as in the year 2012 version 0.4 is available.

I think a good idea to added now Quazar for looking a bit ...
A good mix is the secret!

Best
Frank

Chris Whittington wrote:

Brendan wrote:

matejst wrote:

The last engine I fully understood was Zarkov 6.55. It played reasonably well positionally, and it did not calculate too deep, so the way it treated most of the positions I was interested in "made sense" for me. And I could also check some of the lines.

In the cases when Zarkov was lost in closed positions of the French, or Slav, I could use Gandalf, or ProDeo, and recently I found Winter. I feel I can understand the variations I see on the screen because the evals were written differently, by humans, without using "texel tunning", "sigmoid coherence functions" (kidding) etc.



An example: I liked Zarkov and Wasp 1.02, but with each newer version Wasp was pruning more and more, being faster and faster, and John Stanback adopted the new approaches, and it just lost any sense for me.

But the baseline is that I see engines and computerchess as just a mean to play and enjoy chess. I am not interested in engine tournaments, in TCEC, rating lists, etc.

To finish, something I thought about lately: I agree more and more with Alexander Naumov who believed that open source had its drawbacks. That's of course not the he formulated it, but... Every week we hear about a clone or suspected clone, but the truth seems to be, from my perspective, that the SF team monopolized the rights to originality. I'll try to explain: while they can incorporate every single idea (they're good enough not to use code, and they have a code base) they find in other open source engines without problems, any new programmer has to open his code, and share his ideas, simply to avoid being accused of cloning. It just has killed originality, on one hand, and usability, on the other. But enough about it.

I feel like Boban made some very good points here.

I always liked the analysis of Prodeo 1.2-1.7 and thought the moves made sense.

Same with Wasp 1.02, Frenzee 3.5.19 and other old engines...even WChess 1.06 despite being only 2330 or so.

I seem to remember years ago Boban tweaked Zarkov 6.55 a little to make it more Karpov-like which was cool.

For "sample mining" (my term, which means to have engines play tournaments *only* with the lines/tabiyas from my repertoire and thus, learn how to play it well, new plans, motifs etc) I also liked Thinker, various Rodent personalities and other "non-mainstream" engines.

With regard to open-source, yeah, I think this phenomenon is kinda weird.

How many other software industries is it the *norm* to open up your source code for all to view and plunder from?

Especially for competitive programmers.

How can a single programmer hope to compete with the TEAM of talented guys working on Stockfish, if this TEAM can *also* plunder ideas from his code?

For example, obviously, the Ethereal programmer is enormously talented, but personally, I think he should never have opened his code in the first place.

He is good enough to make something completely unique and keep the source a secret.

The same as Chris W, who recently wrote CoronaVirusChess - a clearly very strong engine.

40 years of experience wasn't enough for some idiots who still started accusing Chris, simply because he keeps the source closed (as he should).

Lance Perkins, author of Thinker was a guy who hid the PV of his (closed source) engine and drew the attention of the witchhunt crowd.

Even though he explained why he does already - to make Thinker more of a human-like opponent:

The question was: Hi folks. I'm sorry if this has been discussed already - I'm new here. The question is, are there engines which aim to play in a human-like way? If so, what ideas exist for achieving it?

Lance answered:

Quote :

1. Most humans don't play at the level of GMs. The GMs are not a representation of the majority of humans. In that sense, GMs are like computers - they don't play like humans.

2. Humans don't give out PVs while they are thinking.

3. Humans make blunders.

4. Humans are inconsistent. They get affected by all sorts of distractions, like sickness, fatigue, excitement, etc., and they end-up playing worse or better.

5. Humans converse when they play.

6. Humans are not capable of 10 moves in 1 second.

7. Humans understand closed/drawn positions. But computers play on forever.

8. Humans don't play from a copy of some endgame database. They won't give out a Queen just to get to a position that is in the endgame database.

9. Humans don't play from an opening book. They have memorized opening sequences. But is does not accomodate for a huge number of 100-ply sequences. It also cannot be swapped out at will (e.g., small book vs large book).

10. Humans don't play with fixed amount of time for each move. In contrast, when you tell an engine that the game is 40 moves in 40 minutes, it will make moves at around 1 minute each. Some move may be either too quick (easy moves) or a fixed multiple (e.g., always 3 minutes).

Having said all those, I think that ChessMaster is the closest to how humans play. It can be configured to play a whole range of styles and preferences.

In the Thinker/ThinkerBoard that plays at FICS, I also have made attempts to address some items I have listed.

(1) The engine runs on a pocket PC.
(2) Just no pv.
(3) There is a mode in Thinker, where, occasionally, the second best move is used.
(4) Different settings are used depending on the hour of the day. Best settings are used in the morning. Worse settings are used at night.
(5) Players can converse with ThinkerTalk.
(6) The engine's minimum move time is 1 second. If it is done searching for less than 1 second, it will sleep for a randmon amount of time (between 1 to 2 seconds).
(7) {no working solution yet}
( Just no eg bb/tb.
(9) Just no opening book.
(10) There is a mode in Thinker, where, the time allocation is randomized.

http://www.talkchess.com/forum3/viewtopic.php?t=20957


The "how could someone create something so strong without cloning?" accusations would come regardless of whether he was guilty or not.

This is how the CC community is.

Hi Brendon,

Yes, I decided not just to keep the source closed but also the executable. No part of Bella-Ciao (CoronaVirusChess) has ever left my office, and never will. I refuse to give the multiple number of malignant talkchess trolls even the tiniest bit for de-compiling, reverse engineering, to hang their accusations and witch hunts on. Only thing that gets out are games it plays online (which is quite enough testing for me) and any games I find worthy of publishing. And yes, it is a Tal engine, it’s good at manoeuvring itself into king attacks as several of the last few games in the Rapid Championship showed.

Hey Chris, a long time ago, when I created the Zappa Dissident Aggressor settings, you warned me " don't release them!" Giving a great explanation as to why. Before, I saw your warning, I did so. Luckily, some people really liked them. However, now many years later, it seems time has proven you wise. People have forgotten CHESS, in a rage for the top elo engine race, and all the other truly amazing, unique engines are slapped aside. However, for us who actually play the game, it has spawned a generation of idiots! Watching games they cannot hope to understand. I enjoy crushing them online game, after , game.