Even crazier

[Lyudmil Tsvetkov]

Let me post a brief reply to simulate discussion.

Of course, one caveat would be that even if it is possible to implement unopposed pawns in the eval, which I do not know if is the case, but suppose so, one has to very well tune at least 3 combined values: double pawns, unopposed pawns, and rooks on semi-open files.

Hopefully, this will not have a further impact on other evaluation terms...

[bob]

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...

[Lyudmil Tsvetkov]

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

[Lyudmil Tsvetkov]

2 more example on the importance of chains.

[d]r4r1k/ppp3p1/2nq1pPp/b3pP2/2bpP2P/P2N4/1PP1P1B1/R1BQ1RK1 w - - 0 18

Do you recognize white has a considerable, maybe even winning advantage here? The white plan will be to transfer the queen to h5 (a bit long, but possible), the Nd3 to g4, and then sacrifice a piece on h6, with huge winning chances. If the black king does not retreat hastily to the opposite, queen side, where the white chain along the b1-h7 diagonal, and especially the g6 pawn, already would not exert enormous pressure, black seems like doomed.

The white double pawns on the e file are not felt in the least as they are inaccessible for attack, while the Nd3 kind of prolongs the white chain of pawns, linking c2 to e4 (so that in case the Nd3 is captured, a longer chain will regenerate instantly). It is difficult to see Nd3 is very useful, unless you consider it being part of the chain. (But you might still think so, if you assign a bonus for a knight on d3)

White is really very close to winning here, but how would you recognize this, if you do not assign a legitimate, and very real, huge bonus for the g6 lead/peak pawn? How on Earth would you do that? The g6 pawn is not the same in its present status, supported by the entire chain, and f5 in particular, and when it would have been alone, unsupported instead. The real difference is huge, and the difference in score should also be.

[d]r3rnk1/1p1b2p1/1Pp1pp1p/p1Pp3q/P2P1P2/2Q1PN2/R5PP/1B3RK1 w - - 0 25
White has big, winning advantage. How many engines would think so?

The black a5 pawn, which is vertically isolated, is soon going to fall (when the white knight is transferred to b3), and white should win without much difficulty. But, how would you be able to know that white is winning, especially in advance, relying simply on the score, if you do not evaluate the excellent e3-b6 chain, and especially the b6 lead pawn, with a very nice bonus? (might be worth some 30cps there, apart from the space on the 6th rank b6 is gaining)

Do you think the white e3-b6 chain is not real? And what about the black d5 apex pawn? (or whatever it is called)

To do a still further regression, I think actually the main problem with most of today's engines, even the top, is their lack of proper understanding of subtler positional elements, respectively lack of appropriate eval terms. And you can not go more positional than scoring chains of pawns, backward and backward-fated pawns (especially the second ones), blockade bonus points for pieces blocking passers and different other pawns, closure of files/sides (with dispensing bonus for the side having advantage, usually space in terms of pawns, but also minors, on the side which is still not fully closed when the other side of the board is fully closed).

Those are really the positional elements that are still missing from most modern engines (you could add also compensation here, sacrificing material for positional or tactical compensation along logical lines), and that is the part of engines that is currently most underdeveloped, sometimes virtually inexistent. It is true that if you manage to raise the general strength of the program in some other way, you also somehow bridge in part of the gap in areas limked to positional play, but only a very tiny part, and painfully slow at that. I think implementing subtler positional factors could really bring huge added value to engine play, but, of course, that is my personal opinion.

In any case, most of the human wins against engines of today would come through the knowledge and use of the above-mentioned factors, i.e. the weaknesses in software are there, and that should already ring a bell.

[Lyudmil Tsvetkov]

For many engines, but also for many humans, the closure of files/sides is a phantom. Does this phantom really exist, or maybe it just occasionally and unexpectedly appears in a random (and probably statistically irrelevant) nightmarish lost game?

For me, the phantom is real, for others this might be just the next nightmarish loss.

I will just formulate here a very simple rule, and later possibly post some adequate examples. (Please note, that the below diagrams are not real games, but just an illustration)

[d]6k1/5pp1/4p2p/3pP3/p1pP4/PpP4P/1P3PP1/6K1 w - - 0 1
White is better here, probably winning (supposing a rich middlegame). The matter of fact is that the only white pawn on e5, gaining space advantage, is worth more than the combined 3 black pawns gaining space on the queen side (a4,c4,b3), for the simple reason that the entire queen side is closed (all 4 files on it), and thus the black space advantage there comes to nothing. Black can not hope to open files there, or develop any other reasonable activity. At the same time, just the single e5 pawn gaining space on the king side promises white much, as having space advantage on the side which is still not fully closed (where the other is already fully closed) supposes a serious advantage for the whole game, as this space gained will usually be beneficial to opening files, attacking, etc., while the opponent will only have to passively defend.

I do not know if this rule is applicable technically in software (but programming wizards are sure to find a way of doing so), but in any case it is true in terms of general chess knowledge, and I really can not imagine general chess knowledge failing in software for some reason. If you do not know it, you are about to lose a number (probably not negligeable) of nightmarish games, unless of course, phantoms do not have sway over you.

[d]6k1/p6p/1p2p1pP/3pPpP1/2pP1P2/2P5/PP6/6K1 w - - 0 1
Black has considerable advantage: the king side is fully closed, so that the huge white space advantage in terms of the e5,g5 and h6 pawns really does not matter, as white will not be able to open files there and attack. At the same time, the only black pawn on c4 gaining space promises black good chances to attack and open files there. Thus, black has the advantage. The logic behind this is that space gained on a particular side is generally beneficial to initiative, attacking and opening files, while thwarting the opponent, but space gained on a fully closed side will fail to do so. Therefore, one must be very attentive (including engines), when closing the last still remaining open file on a particular side. First check who has space advantage on the other side, and then close.

My rule would be to assign some nice bonus in some situations, for the player gaining space on the side which is still not fully closed: maybe some 50cps for such an arrangement. Of course, as the side where the enemy king is is usually more important, gaining space on such a side that is still not fully closed might give you a bonus of 60cps, while gaining space on the opposite side just 40cps.

Any highly critical comments very much appreciated.

Best, Lyudmil

[Matthias Hartwich]

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?

No, stick to them as long as they discribe the structures well.

If do you have a better definition AND other think it fits, than it will be used. Otherwise the older definitions are just better as many know what they mean.

The progress in chess knowledge won't change everything and special cases will get their names eventually.

[jdart]

Sorry, but I don't get it. What can Black do here?

[d] 6k1/p6p/1p2p1pP/3pPpP1/2pP1P2/2P5/PP6/6K1 w - -

Black has no majority so can't create a passer, if I am not mistaken.

--Jon

[Lyudmil Tsvetkov]

[d]6k1/p7/1p2p1p1/3pPp1p/2pP1P1P/2P3P1/PP6/6K1 w - - 0 1
On the above diagram you see another version of a fully closed side: the king side should be considered as fully closed, as both pawns on the g file - g3 and g6 - are backward-fated. In such a case it is practically impossible to open the side without unreasonable sacrifices, and sacrificing a full pawn in most cases will be unfeasible. Therefore, the king side could be judged as fully closed.

Now, how can you say that backward-fated pawns do not exist? You see they are real and useful, at least you can use them when determining if a specific side is or not entirely closed. And, in any case, the difference between normal backward pawns and backward-fated pawns is obviously very substantial. They are different pawns and there is no way you can assimilate them.

[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.

[d]6k1/8/4p1p1/1p1pPp1p/pPpP1P1P/P1N3P1/1P6/6K1 w - - 0 1
Another possibility for closing a file could be seen above: the c file could be regarded as closed (and that might help in determining if a side is fully closed or not, or even if all files on the board are closed/locked, to assess drawing chances), because the knight on c3 blocks the enemy c4 pawn, whenever that knight is impossible to attack by enemy pieces, or well protected. That could also be used when deciding the status of closure.

So that, basically, apart from own and enemy pawns blocking/fixing each other, a file could be deemed closed in the additional 2 situations of backward-fated pawns, as well as piece blocking an enemy pawn. Now, how can you assert that blocking with pieces of enemy pawns, even distinct from passers, is irrerelevant? It is relevant, but maybe not sufficiently applied and understood.

[Lyudmil Tsvetkov]

Some very small example of the validity of apex pawns.

[d]6k1/6b1/2p1p3/2Pp4/3P4/2N1P3/8/6K1 w - - 0 1
d5 is an apex pawn there, it is due a bonus not only because it is very solid (which is still something), but also because of the flexibility of the structure as a whole. For example, black can now play e5, as d5 would remain defended, so that the apex pawn supposes flexibility in moving, which might have some bearing on the game.

[d]6k1/1p4b1/4p3/3p4/3P4/1PN1P3/8/6K1 w - - 0 1
Now, d5 already is not apex, but a simple lead/peak pawn, less influential here, with very few members of the chain. Now playing e5 for black becomes impossible, as d5 is already not defended and e5 loses a pawn. This shows that the current structure is less flexible in advancing than a structure with an apex pawn.

So, seemingly, apex pawns are quite real. And they might have quite some influence on the way the game develops, the success or failure of entire variations might hinge on such a pawn. Seeing this asset could actually make to accept a variation that would otherwise have been discarded to the benefit of another one, realistically less good. Of course, the importance of apex pawns would nowhere come close to, say, double pawns, if a double pawn scores some penalty of 20cps, an apex could score a bonus of 5-10cps, but still, if it is a real asset, it does have implications on the development of the game and the choice of moves.

[Lyudmil Tsvetkov]

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?

No, stick to them as long as they discribe the structures well.

If do you have a better definition AND other think it fits, than it will be used. Otherwise the older definitions are just better as many know what they mean.

The progress in chess knowledge won't change everything and special cases will get their names eventually.

Hi Matthias.

But the problem is that in many cases those older notions do not describe in a sufficiently perfect way the available structures in chess. Progress have been made both in search and evaluation, as well as hardware, over the years, and that is why Houdini of 2012 beats Shredder of 10 years ago by a margin of 99.5/0.5. Houdini has both better evaluation, and better search. If you had to stick with the eval of Shredder from 2002, then progress would be very little.

And now, the challenge is to create an engine that would appear in 2022 that would be able to beat Houdini of 2012 with a similar margin (not quite, but still). I do not know about the search, but, definitely, you can not use the evaluation notions of 2012 in a 2022 engine.