Even crazier

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

Moderators: hgm, Rebel, chrisw

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

Re: Leading the chain

Post by Lyudmil Tsvetkov »

syzygy wrote:
syzygy wrote:The real question is whether awarding points to this flexibility gains Elo. Only if it does, the concept is "valid". That it does might be intuitively obvious to you, but this is not how top engines make progress nowadays.
And I should add that before trying this out, it must be determined whether the concept is not already implicit in other pawn evaluation terms. If it is, a new term would merely slow down parameter tuning by adding a redundant parameter.
Of course, everyone is welcome to check it and report the results.
Obviously, there any redundancies should be avoided for precise implementation.

All those principles I have tried to formulate have been tested in my games against engines, and whatever games I have been able to win have been due to the engine largely neglecting the existence of such terms, while myself making use of them. For me this is experience, I test in this way, do not run games against other engines.

Engines have made substantial progress in the areas and using the terms in widespread consumption (like double, open files, etc.), and not a substantial progress, or entirely lacking, in areas and using eval terms not of widespread use and with accentuated positional connotations.

PS. Btw., one of the very reasons for the inexistence until now of similar positional terms is the simple fact that engines are not tested in real games against humans, and thus the human feedback has been completely missing from engine development in recent years. You see and analyse some engine losses, but it is difficult to find there anything even closely related to the human approach of positional understanding, and the related terms.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Semi, but real

Post by Lyudmil Tsvetkov »

bob wrote:
Lyudmil Tsvetkov wrote:
bob wrote:Two things.

1. Why try to invent a NEW pawn vocabulary? There are several good books on pawn structure. Kmoch comes to mind.

2. I don't like terms like "semi-backward." Why not just do as most upper-level players do and consider weak pawns. A backward pawn is weak. An isolated pawn is weak. An artificially isolated pawn is weak. Kmoch defines all of those. Ditto for majority, crippled majority, mobile majority, immobile majority, etc...

But these new terms are not going to catch on very well since there is already a lot of pawn structure analysis in GM books...
Mr. Hyatt, I respect you very much for your enormous contribution to computer chess, but, what would you suggest instead? Stick to the notions people invented 30-50-100 years ago and not progress any more, simply because we regard those doctrines as a dogma?

Do you really suggest that a scientific mind doing research some 100 years ago would really surpass persons that would inquire into interesting aspects of knowledge that probably exists, but is still unveiled? It would be really funny if it is like that. Actually, most branches of science and even art have already attained levels where there are very few substantial things to do, but, fortunately, computer chess is not one of those.

Do you realize the enormous progress being made in computer chess every year, including in terms of elo increase to the point where many programs already surpass the best humans? How much of this progress do you think is due to implementing eval terms in general, and, if you want to make a distinction, in applying existing and tinkering with new terms, previously unused? I guess eval would be responsible for 1/3 of all progress, including search and hardware. That is not negligeable.

I myself do not have any ambitions at all, I would be glad to help a bit some engines, if possible, as engines, including Crafty (thanks for this software!), have helped me quite a lot in my development. But is not it evident that some eval terms are really outdated? You can not do a 21-st century engine with concepts from the middle of the 20th century. Could you tell me what auothors of renown have suggested pawn concepts like backward-fated, unopposed, apex pawns, proximity of peak pawns to the enemy king, etc.? Do you think they are not valid concepts? Do they have to be suggested by Hans Kmoch in a paper edition to be valid?

I would suggest to do the following: try measuring the impact of, say, apex pawns, in Crafty, very easy to do, and see what happens. I suppose, if correctly implemented, with no redundancies, you will see some elo increase in Crafty, it might be 10 elo, but will be there. Why reject the existence of concepts that are useful?

Best, Lyudmil
Sorry, but nothing in your "new terms" is new. Read some good books on pawn structure, and you will see that you have not identified anything new at all...

Most authors I have read discuss your concepts. The "apex" is a part of a common attacking theme. Which is often modified by the direction of the pawns. For example, white pawns at d3,e4,f5 are in a chain pointing toward the enemy king, helping an attack, while black pawns at f6, e5, d4 also form a chain, but it points in the wrong direction.

As I said, using new terms for existing ideas doesn't help at all...
And why in this case engines in general, even all top, do very badly in terms of understanding chains?

Where are apex, uopposed and backward-fated pawns specifically mentioned, just for the record?
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: The phantom of closure

Post by Lyudmil Tsvetkov »

Tord Romstad wrote:
Lyudmil Tsvetkov wrote: [d]6k1/p7/1p2p1p1/3pPp1p/2pP3P/2P1P1P1/PP6/6K1 w - - 0 1
Now, on the above position, g3 is still backward-fated, but g6 already is not, and you can not call the king side fully closed. There is a substantial difference. Could you possibly say that g6 in this diagram and g6 in the former one are one and the same pawn? This is ludicrous. So that, for me, backward-fated pawns have their right of existence, actually they are one of the most positional elements in chess, if chess programs are reluctant in general or unable to implement them, that is another question.
This reminds me of an evaluation term I once had, which I called "pawn structure mobility". What I did was to give every pawn on the board a bonus measuring its ability to move without getting lost or creating weaknesses. I don't remember the exact numbers I used, but the general idea, applied to the above position, would be something like this:
  • The black pawns on a7 and b6 would get a significant bonus, because they are able to move freely without creating any backward, isolated or doubled pawns or important weak squares in their own camp.
  • The black pawn on g6 would get a somewhat smaller bonus. If supported by a friendly piece, it can safely advance, but only at the cost of accepting an isolated pawn on h5.
  • No other black pawns have any safe moves, they therefore don't get any bonus
  • The white pawn on a2 only gets a small bonus, because moving it means weakening the square b3, which is attacked by a black pawn, and would render the b2 pawn locked in place.
  • The white pawn on b2 also gets a small bonus, because moving it would give white (after b3 cxb3 axb3 b5) a backward pawn on c3.
  • No other white pawns get bonuses, because they're all immobile, even with the support of friendly pieces.
This may seem expensive to compute (it even included a simple pawn-move-only search, as illustrated in the discussion of the white b2 pawn), but the nice thing about evaluation terms which *only* consider pawn structure is that you don't have to compute them very often. Even though the number of positions encountered during the search is astronomical, the number of distinct pawn structures are much smaller. The program remembers the pawn structures it has seen before, and only needs to perform the expensive pawn structure evaluation on the rare occasions when it encounters a previously unseen structure. This makes very complicated pawn structure evaluation terms like this one attractive to investigate. The programmers among you already knew all of this, of course, but I thought it would be interesting and useful for Lyudmil to learn.

Unfortunately, as is so often the case with complicated evaluation terms, I was never able to make it work well in practice. The program played some impressive-looking positional games from time to time and looked more intelligent, but the test results were still disappointing. Of course this doesn't mean that the idea is unworkable; it's quite possible that I was just too stupid and/or didn't work hard enough to make it work.
Hi Tord.
Thanks for the intervention!

I really like such profound, topical replies. This is what makes discussion meaningful, and not petty exchanges.
Very interesting approach to pawn structure you have had, a thoughtful one, which, btw., supports the general assessment that the black pawn structure is better here.

One distinction is that all of the terms I suggest (with very few exceptions, actually) are perfectly statical, you do not need to move any pawns or pieces to assign a value. Actually, as statical as a double pawn.

I think, when a discussion is based on ideas and fact, it is always useful to all parties.

I still remember sparring some 10 years ago against Gothmog_016 :D Stockfish has made a long way from that point.

Thanks again for the contribution.

Best, Lyudmil
carldaman
Posts: 2283
Joined: Sat Jun 02, 2012 2:13 am

Re: Leading the chain

Post by carldaman »

Lyudmil Tsvetkov wrote:
All those principles I have tried to formulate have been tested in my games against engines, and whatever games I have been able to win have been due to the engine largely neglecting the existence of such terms, while myself making use of them. For me this is experience, I test in this way, do not run games against other engines.

Engines have made substantial progress in the areas and using the terms in widespread consumption (like double, open files, etc.), and not a substantial progress, or entirely lacking, in areas and using eval terms not of widespread use and with accentuated positional connotations.

PS. Btw., one of the very reasons for the inexistence until now of similar positional terms is the simple fact that engines are not tested in real games against humans, and thus the human feedback has been completely missing from engine development in recent years. You see and analyse some engine losses, but it is difficult to find there anything even closely related to the human approach of positional understanding, and the related terms.
Bingo! Hitting the nail right on the head. Human quality control is largely absent from engine development. "If a change gains Elo vs other engines, --> then it is good, if a change fails to gain or loses Elo, --> then it must be bad." This seemingly logical paradigm is too simplistic, and may 'not see the forest for the trees'.

This may sound too radical and counterproductive to most, but I still hope that at least some of the developers will see that there is more to engine development than mere Elo chasing. How useful and instructive an engine is for analysis, and how effective it can be against a ~ 2000-2300 rated human should also be major factors. Many of us would gladly buy an engine that's somewhat weaker tactically (and thus not #1 on rating lists), but very smart strategically and/or in closed positions.

The current mode of thought is that anything that slows the engine down is bad. Thus a lot of potentially useful knowledge is rejected right off the bat. I believe there has to be a better balance and a happy medium somewhere, but it can only be pursued by (temporarily, at least) giving up on the 'Elo-chasing at all costs' paradigm.

Nice post, Lyudmil. Glad I'm not alone on this.

Regards,
CL
carldaman
Posts: 2283
Joined: Sat Jun 02, 2012 2:13 am

Re: The phantom of closure

Post by carldaman »

Tord Romstad wrote:
Unfortunately, as is so often the case with complicated evaluation terms, I was never able to make it work well in practice. The program played some impressive-looking positional games from time to time and looked more intelligent, but the test results were still disappointing. Of course this doesn't mean that the idea is unworkable; it's quite possible that I was just too stupid and/or didn't work hard enough to make it work.
Very interesting, Tord. You mention 'test results', but were these simply engine vs engine tests? I'm more convinced than ever that these can be misleading. Lyudmil himself has demonstrated with a substantial number of games, that even the latest Stockfish, despite its very high rating, is not quite that effective against a human player of around 2300 (master) strength.

I am a proponent of a different approach, where taking one step backwards to add (slow) knowledge may be advisable, so that two or more steps can be taken forward.
Otherwise, the apparent strength gains derived from engine-only tests can turn out to be an illusion (to a large extent). I think that Lyudmil's results tend to support my viewpoint.

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

Re: Leading the chain

Post by Lyudmil Tsvetkov »

carldaman wrote:
Lyudmil Tsvetkov wrote:
All those principles I have tried to formulate have been tested in my games against engines, and whatever games I have been able to win have been due to the engine largely neglecting the existence of such terms, while myself making use of them. For me this is experience, I test in this way, do not run games against other engines.

Engines have made substantial progress in the areas and using the terms in widespread consumption (like double, open files, etc.), and not a substantial progress, or entirely lacking, in areas and using eval terms not of widespread use and with accentuated positional connotations.

PS. Btw., one of the very reasons for the inexistence until now of similar positional terms is the simple fact that engines are not tested in real games against humans, and thus the human feedback has been completely missing from engine development in recent years. You see and analyse some engine losses, but it is difficult to find there anything even closely related to the human approach of positional understanding, and the related terms.
Bingo! Hitting the nail right on the head. Human quality control is largely absent from engine development. "If a change gains Elo vs other engines, --> then it is good, if a change fails to gain or loses Elo, --> then it must be bad." This seemingly logical paradigm is too simplistic, and may 'not see the forest for the trees'.

This may sound too radical and counterproductive to most, but I still hope that at least some of the developers will see that there is more to engine development than mere Elo chasing. How useful and instructive an engine is for analysis, and how effective it can be against a ~ 2000-2300 rated human should also be major factors. Many of us would gladly buy an engine that's somewhat weaker tactically (and thus not #1 on rating lists), but very smart strategically and/or in closed positions.

The current mode of thought is that anything that slows the engine down is bad. Thus a lot of potentially useful knowledge is rejected right off the bat. I believe there has to be a better balance and a happy medium somewhere, but it can only be pursued by (temporarily, at least) giving up on the 'Elo-chasing at all costs' paradigm.

Nice post, Lyudmil. Glad I'm not alone on this.

Regards,
CL
Hi Carl.
Thanks for the precious support!

2 against the pack. :shock: Now I am not afraid any more.

My only further comment to this would be that you can made an engine both appealing and strong, both human-like and engine-like, just like humans might learn some of the tactical tricks of engines.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: The phantom of closure

Post by Lyudmil Tsvetkov »

jdart wrote:[d] 8/6kb/p3n1p1/Pp1nPp2/1PpPp2p/2P4P/3B2P1/2R1R1K1 b - - 0 1

White pawn score:
general = 0
midgame = 70
endgame = 91
blend = 91
Black pawn score:
general = 0
midgame = 54
endgame = 72
blend = 70

White gets a bonus for the passer (diminished because it is blocked by piece). Black has a majority here (unlike the other position I commented on) so gets a bonus for a potential passer. Overall White gets a little higher pawn score. Maybe wrongly.

Overall the position is scored as better for White because White has the two Rooks vs two minors.

--Jon
But how can than black win, if white has a material advantage of some 3 pawns, plus some advantage in terms of pawns?
jdart
Posts: 4375
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: The phantom of closure

Post by jdart »

I am not sure Black can win, but I agree White will have great difficulty making progress.

--Jon
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Semi, but real

Post by bob »

Lyudmil Tsvetkov wrote:
bob wrote:
Lyudmil Tsvetkov wrote:
bob wrote:Two things.

1. Why try to invent a NEW pawn vocabulary? There are several good books on pawn structure. Kmoch comes to mind.

2. I don't like terms like "semi-backward." Why not just do as most upper-level players do and consider weak pawns. A backward pawn is weak. An isolated pawn is weak. An artificially isolated pawn is weak. Kmoch defines all of those. Ditto for majority, crippled majority, mobile majority, immobile majority, etc...

But these new terms are not going to catch on very well since there is already a lot of pawn structure analysis in GM books...
Mr. Hyatt, I respect you very much for your enormous contribution to computer chess, but, what would you suggest instead? Stick to the notions people invented 30-50-100 years ago and not progress any more, simply because we regard those doctrines as a dogma?

Do you really suggest that a scientific mind doing research some 100 years ago would really surpass persons that would inquire into interesting aspects of knowledge that probably exists, but is still unveiled? It would be really funny if it is like that. Actually, most branches of science and even art have already attained levels where there are very few substantial things to do, but, fortunately, computer chess is not one of those.

Do you realize the enormous progress being made in computer chess every year, including in terms of elo increase to the point where many programs already surpass the best humans? How much of this progress do you think is due to implementing eval terms in general, and, if you want to make a distinction, in applying existing and tinkering with new terms, previously unused? I guess eval would be responsible for 1/3 of all progress, including search and hardware. That is not negligeable.

I myself do not have any ambitions at all, I would be glad to help a bit some engines, if possible, as engines, including Crafty (thanks for this software!), have helped me quite a lot in my development. But is not it evident that some eval terms are really outdated? You can not do a 21-st century engine with concepts from the middle of the 20th century. Could you tell me what auothors of renown have suggested pawn concepts like backward-fated, unopposed, apex pawns, proximity of peak pawns to the enemy king, etc.? Do you think they are not valid concepts? Do they have to be suggested by Hans Kmoch in a paper edition to be valid?

I would suggest to do the following: try measuring the impact of, say, apex pawns, in Crafty, very easy to do, and see what happens. I suppose, if correctly implemented, with no redundancies, you will see some elo increase in Crafty, it might be 10 elo, but will be there. Why reject the existence of concepts that are useful?

Best, Lyudmil
Sorry, but nothing in your "new terms" is new. Read some good books on pawn structure, and you will see that you have not identified anything new at all...

Most authors I have read discuss your concepts. The "apex" is a part of a common attacking theme. Which is often modified by the direction of the pawns. For example, white pawns at d3,e4,f5 are in a chain pointing toward the enemy king, helping an attack, while black pawns at f6, e5, d4 also form a chain, but it points in the wrong direction.

As I said, using new terms for existing ideas doesn't help at all...
And why in this case engines in general, even all top, do very badly in terms of understanding chains?

Where are apex, uopposed and backward-fated pawns specifically mentioned, just for the record?
I don't consider your idea of "unopposed" to be relevant. You gave an example of black pawns on a7, b7 and c7, and white pawns on b2, b3 and c2. That black pawn on a7 is NOT any sort of "positional advantage" whatsoever. It does not threaten to create a passed pawn. It does not threaten to disrupt white's already disrupted pawns.

So what, exactly, is the point. For backward or "backward fated", the ideas are well known. And fall under the classification "weak pawns".

Whether chess books cover the topics has NOTHING to do with whether engines understand the concepts. There is no connection. My computer doesn't read chess books and learn how to play better.
Joerg Oster
Posts: 948
Joined: Fri Mar 10, 2006 4:29 pm
Location: Germany
Full name: Jörg Oster

Re: Leading the chain

Post by Joerg Oster »

This reminded me of a post by Chris Whittington, a most remarkable one.
See here: http://www.talkchess.com/forum/viewtopi ... 77&t=44969
Jörg Oster