Excellent job

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Excellent job

Post by Lyudmil Tsvetkov »

Adam Hair wrote:Only Nelson (Cato), IM Erik Kislik and possibly now Martin know which openings will be used in the Finals.
Is not it the same openings for all stages?

Btw., I do not think it would be fair to change the book from stage to stage, unless there are special reasons for it.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Excellent job

Post by Lyudmil Tsvetkov »

Lyudmil Tsvetkov wrote:
Adam Hair wrote:Only Nelson (Cato), IM Erik Kislik and possibly now Martin know which openings will be used in the Finals.
Is not it the same openings for all stages?

Btw., I do not think it would be fair to change the book from stage to stage, unless there are special reasons for it.
OK, I just read the opening book section of TCEC (should have done it long before, but replaying the games is more important for me).

And from what I read, my conclusion is that the same selection criteria would apply for the final, i.e. the 32 positions picked up by Erik will come from the same main database of 1800+ games with evals within the 30cps margin.

Strange that I did not notice any position discrepancies in earlier stages, at least in the evals of the main engines and the overall database outlook.

Obviously, the database is made up of positions with similar outlook, as complex tactical, emphatically asymmetric positions were left out in the selection process, as Komodo's and SF's evals on such positions quite likely would have diverged significantly, even if some of those positions had been objectively acceptable.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Excellent job

Post by Lyudmil Tsvetkov »

Btw., Adam, do you know what the incidence rate of A,B,C,D and E opening is?

My impression is that A+E openings (both prone to closed play developments) dominate in numbers the cumulative weight of B,C and D openings, or maybe this was just due to the small sample?

Does your main database follow the human databases, where A and E openings are significantly less frequent?
Nelson Hernandez
Posts: 101
Joined: Sun Nov 14, 2010 9:36 pm
Location: U.S.

Re: Excellent job

Post by Nelson Hernandez »

Lyudmil,

Thank you for your kind words and analysis at the start of this thread. I did have to laugh when I read further on that, after all the analysis and verbiage, you had not yet read my opening notes! You're probably in the majority so don't be embarrassed.

To answer your questions...

Combining stages 1, 3, 4 and the SuperFinal 'A' ECOs are seen 24%, B 24%, C 14%, D 17% and E 22%. Note that in rare instances the ECO I thought I was delivering subsequently transposes to another line, so my numbers may not correspond exactly to what you might see by downloading the games and counting.

Your observation that A+E dominate is not quite accurate but you would be correct if you said they were overrepresented compared to human and especially computer databases. When selecting a position (methodology described in my notes) I'm looking for one that meets a set of criteria. Some openings meet the requirements easier than others. For example, you will rarely (if ever) see a Berlin or a Petroff played in TCEC because I actively steer away from draw-prone openings.

As for your third question, my main database is a consolidation of human/correspondence/engine games, where engine games represent a significant majority. ECO B openings are clearly the leaders numerically, so relative to the database B is quite underrepresented. Reason: if I reflected what was in the database we'd be seeing many Sicilian Najdorf openings, which is not a good idea for competition purposes.

Thanks again for your questions and comments.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Excellent job

Post by Lyudmil Tsvetkov »

Nelson Hernandez wrote:Lyudmil,

Thank you for your kind words and analysis at the start of this thread. I did have to laugh when I read further on that, after all the analysis and verbiage, you had not yet read my opening notes! You're probably in the majority so don't be embarrassed.

To answer your questions...

Combining stages 1, 3, 4 and the SuperFinal 'A' ECOs are seen 24%, B 24%, C 14%, D 17% and E 22%. Note that in rare instances the ECO I thought I was delivering subsequently transposes to another line, so my numbers may not correspond exactly to what you might see by downloading the games and counting.

Your observation that A+E dominate is not quite accurate but you would be correct if you said they were overrepresented compared to human and especially computer databases. When selecting a position (methodology described in my notes) I'm looking for one that meets a set of criteria. Some openings meet the requirements easier than others. For example, you will rarely (if ever) see a Berlin or a Petroff played in TCEC because I actively steer away from draw-prone openings.

As for your third question, my main database is a consolidation of human/correspondence/engine games, where engine games represent a significant majority. ECO B openings are clearly the leaders numerically, so relative to the database B is quite underrepresented. Reason: if I reflected what was in the database we'd be seeing many Sicilian Najdorf openings, which is not a good idea for competition purposes.

Thanks again for your questions and comments.
Hi Nelson.

Thanks for the explanations.

You have really done a great job!
I must admit that I follow all the games with interest, that was not quite the same last season. One of the reasons is the high level of artistic satisfaction from the game contents. When everything develops according to strict rules, you have that feeling. Any big eval deviations certainly would spoil that. I am certain the engines are also happy they are threated in a fair manner. :)

I did not read the book rules until yesterday, but I have replayed all major TCEC games twice...

By now there is no doubt whatsoever you possess the highest-quality and most even book in the world with such length. Only problem I see with it is the uneven spread of A,B,C,D,E openings, but maybe it is time for engines to learn to play closed positions better. If you could also include some Berlin/Petrov positions in the future, as well as some sharper ones, and a bit bigger portion of QGD/QGA, just for a higher sample representativeness, that would be gorgeous.

Kind regards to Reston.
carldaman
Posts: 2288
Joined: Sat Jun 02, 2012 2:13 am

Re: Excellent job

Post by carldaman »

Lyudmil Tsvetkov wrote:
Nelson Hernandez wrote:Lyudmil,

Thank you for your kind words and analysis at the start of this thread. I did have to laugh when I read further on that, after all the analysis and verbiage, you had not yet read my opening notes! You're probably in the majority so don't be embarrassed.

To answer your questions...

Combining stages 1, 3, 4 and the SuperFinal 'A' ECOs are seen 24%, B 24%, C 14%, D 17% and E 22%. Note that in rare instances the ECO I thought I was delivering subsequently transposes to another line, so my numbers may not correspond exactly to what you might see by downloading the games and counting.

Your observation that A+E dominate is not quite accurate but you would be correct if you said they were overrepresented compared to human and especially computer databases. When selecting a position (methodology described in my notes) I'm looking for one that meets a set of criteria. Some openings meet the requirements easier than others. For example, you will rarely (if ever) see a Berlin or a Petroff played in TCEC because I actively steer away from draw-prone openings.

As for your third question, my main database is a consolidation of human/correspondence/engine games, where engine games represent a significant majority. ECO B openings are clearly the leaders numerically, so relative to the database B is quite underrepresented. Reason: if I reflected what was in the database we'd be seeing many Sicilian Najdorf openings, which is not a good idea for competition purposes.

Thanks again for your questions and comments.
Hi Nelson.

Thanks for the explanations.

You have really done a great job!
I must admit that I follow all the games with interest, that was not quite the same last season. One of the reasons is the high level of artistic satisfaction from the game contents. When everything develops according to strict rules, you have that feeling. Any big eval deviations certainly would spoil that. I am certain the engines are also happy they are threated in a fair manner. :)

I did not read the book rules until yesterday, but I have replayed all major TCEC games twice...

By now there is no doubt whatsoever you possess the highest-quality and most even book in the world with such length. Only problem I see with it is the uneven spread of A,B,C,D,E openings, but maybe it is time for engines to learn to play closed positions better. If you could also include some Berlin/Petrov positions in the future, as well as some sharper ones, and a bit bigger portion of QGD/QGA, just for a higher sample representativeness, that would be gorgeous.

Kind regards to Reston.
+1

I concur re: more closed positions (more Classical KIDs, Stonewalls and so on)! :) Such positions can occur a lot more frequently than many people think, especially when engines are playing with a short book vs a skilled human player who can steer the game into favorable pawn structures, as Lyudmil has demonstrated many times in this forum.

Great job by Nelson & Co., otherwise. :D

Regards,
CL
yanquis1972
Posts: 1766
Joined: Wed Jun 03, 2009 12:14 am

Re: Excellent job

Post by yanquis1972 »

carldaman wrote:
Lyudmil Tsvetkov wrote:
Nelson Hernandez wrote:Lyudmil,

Thank you for your kind words and analysis at the start of this thread. I did have to laugh when I read further on that, after all the analysis and verbiage, you had not yet read my opening notes! You're probably in the majority so don't be embarrassed.

To answer your questions...

Combining stages 1, 3, 4 and the SuperFinal 'A' ECOs are seen 24%, B 24%, C 14%, D 17% and E 22%. Note that in rare instances the ECO I thought I was delivering subsequently transposes to another line, so my numbers may not correspond exactly to what you might see by downloading the games and counting.

Your observation that A+E dominate is not quite accurate but you would be correct if you said they were overrepresented compared to human and especially computer databases. When selecting a position (methodology described in my notes) I'm looking for one that meets a set of criteria. Some openings meet the requirements easier than others. For example, you will rarely (if ever) see a Berlin or a Petroff played in TCEC because I actively steer away from draw-prone openings.

As for your third question, my main database is a consolidation of human/correspondence/engine games, where engine games represent a significant majority. ECO B openings are clearly the leaders numerically, so relative to the database B is quite underrepresented. Reason: if I reflected what was in the database we'd be seeing many Sicilian Najdorf openings, which is not a good idea for competition purposes.

Thanks again for your questions and comments.
Hi Nelson.

Thanks for the explanations.

You have really done a great job!
I must admit that I follow all the games with interest, that was not quite the same last season. One of the reasons is the high level of artistic satisfaction from the game contents. When everything develops according to strict rules, you have that feeling. Any big eval deviations certainly would spoil that. I am certain the engines are also happy they are threated in a fair manner. :)

I did not read the book rules until yesterday, but I have replayed all major TCEC games twice...

By now there is no doubt whatsoever you possess the highest-quality and most even book in the world with such length. Only problem I see with it is the uneven spread of A,B,C,D,E openings, but maybe it is time for engines to learn to play closed positions better. If you could also include some Berlin/Petrov positions in the future, as well as some sharper ones, and a bit bigger portion of QGD/QGA, just for a higher sample representativeness, that would be gorgeous.

Kind regards to Reston.
+1

I concur re: more closed positions (more Classical KIDs, Stonewalls and so on)! :) Such positions can occur a lot more frequently than many people think, especially when engines are playing with a short book vs a skilled human player who can steer the game into favorable pawn structures, as Lyudmil has demonstrated many times in this forum.

Great job by Nelson & Co., otherwise. :D

Regards,
CL
i think one reason you dont see much of the two you named is that they dont pass the evaluation test. most engines will give +.50 to white in any typical KID or stonewall position.
shrapnel
Posts: 1339
Joined: Fri Nov 02, 2012 9:43 am
Location: New Delhi, India

Re: Excellent job

Post by shrapnel »

Lyudmil Tsvetkov wrote:This is really a wonderful book
:shock: :shock: :shock: You've GOT to be KIDDING !
A weaker, sillier Book I have yet to see.
Your chess knowledge may be good, Mr Tsvetkov, but you have NO idea how strong chess opening Books have become !
Why, using one of the latest Sicilian.ctg commercial books its possible to DESTROY even a strong opponent, WITHOUT even ONCE coming out of Book ! :shock:
Strong Books like the Sicilian series and the l33t books have practically made Engines obsolete....I kid you NOT !
I remember going over one of the present TCEC games where Komodo lost to Stockfish ; I almost fell of my chair LAUGHING at the ridiculous Opening moves that the TCEC book was forcing the hapless Komodo to play !
:lol: :lol: :lol:
Try on of the latest commercial .ctg Books for yourself, Mr Tsvetkov, and you will see how they WIPE the floor with the TCEC book !
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
carldaman
Posts: 2288
Joined: Sat Jun 02, 2012 2:13 am

Re: Excellent job

Post by carldaman »

yanquis1972 wrote:
carldaman wrote:
Lyudmil Tsvetkov wrote:
Nelson Hernandez wrote:Lyudmil,

Thank you for your kind words and analysis at the start of this thread. I did have to laugh when I read further on that, after all the analysis and verbiage, you had not yet read my opening notes! You're probably in the majority so don't be embarrassed.

To answer your questions...

Combining stages 1, 3, 4 and the SuperFinal 'A' ECOs are seen 24%, B 24%, C 14%, D 17% and E 22%. Note that in rare instances the ECO I thought I was delivering subsequently transposes to another line, so my numbers may not correspond exactly to what you might see by downloading the games and counting.

Your observation that A+E dominate is not quite accurate but you would be correct if you said they were overrepresented compared to human and especially computer databases. When selecting a position (methodology described in my notes) I'm looking for one that meets a set of criteria. Some openings meet the requirements easier than others. For example, you will rarely (if ever) see a Berlin or a Petroff played in TCEC because I actively steer away from draw-prone openings.

As for your third question, my main database is a consolidation of human/correspondence/engine games, where engine games represent a significant majority. ECO B openings are clearly the leaders numerically, so relative to the database B is quite underrepresented. Reason: if I reflected what was in the database we'd be seeing many Sicilian Najdorf openings, which is not a good idea for competition purposes.

Thanks again for your questions and comments.
Hi Nelson.

Thanks for the explanations.

You have really done a great job!
I must admit that I follow all the games with interest, that was not quite the same last season. One of the reasons is the high level of artistic satisfaction from the game contents. When everything develops according to strict rules, you have that feeling. Any big eval deviations certainly would spoil that. I am certain the engines are also happy they are threated in a fair manner. :)

I did not read the book rules until yesterday, but I have replayed all major TCEC games twice...

By now there is no doubt whatsoever you possess the highest-quality and most even book in the world with such length. Only problem I see with it is the uneven spread of A,B,C,D,E openings, but maybe it is time for engines to learn to play closed positions better. If you could also include some Berlin/Petrov positions in the future, as well as some sharper ones, and a bit bigger portion of QGD/QGA, just for a higher sample representativeness, that would be gorgeous.

Kind regards to Reston.
+1

I concur re: more closed positions (more Classical KIDs, Stonewalls and so on)! :) Such positions can occur a lot more frequently than many people think, especially when engines are playing with a short book vs a skilled human player who can steer the game into favorable pawn structures, as Lyudmil has demonstrated many times in this forum.

Great job by Nelson & Co., otherwise. :D

Regards,
CL
i think one reason you dont see much of the two you named is that they dont pass the evaluation test. most engines will give +.50 to white in any typical KID or stonewall position.
:lol:

Yes, of course. It isn't always reliable to turn to the engines for accurate eval scores, especially in openings they're known to have trouble with. Screening these openings out for reasons of lopsided eval can create a vicious loop where the engines never get tested in those openings they struggle to evaluate correctly, and thus they hardly ever improve their play in those types of positions.

They do have a human IM on their team to help with the filtering, though, so hopefully he'll see the light. :)
Adam Hair
Posts: 3226
Joined: Wed May 06, 2009 10:31 pm
Location: Fuquay-Varina, North Carolina

Re: Excellent job

Post by Adam Hair »

shrapnel wrote:
Lyudmil Tsvetkov wrote:This is really a wonderful book
:shock: :shock: :shock: You've GOT to be KIDDING !
A weaker, sillier Book I have yet to see.
Your chess knowledge may be good, Mr Tsvetkov, but you have NO idea how strong chess opening Books have become !
Why, using one of the latest Sicilian.ctg commercial books its possible to DESTROY even a strong opponent, WITHOUT even ONCE coming out of Book ! :shock:
Strong Books like the Sicilian series and the l33t books have practically made Engines obsolete....I kid you NOT !
I remember going over one of the present TCEC games where Komodo lost to Stockfish ; I almost fell of my chair LAUGHING at the ridiculous Opening moves that the TCEC book was forcing the hapless Komodo to play !
:lol: :lol: :lol:
Try on of the latest commercial .ctg Books for yourself, Mr Tsvetkov, and you will see how they WIPE the floor with the TCEC book !
Books can be judged in more than one way, Anil. The openings used in TCEC serve a different purpose than those found in Sicilian Avenger.