Even crazier
Moderators: hgm, Rebel, chrisw
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Even crazier
I might post here a couple of even crazier notions than the ones exposed in an earlier 'Crazy theories' thread. Maybe there will be only a single message, maybe more. I wanted to keep it distinct from the earlier thread not to water it down additionally with the amount of unanswered subthreads.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Semi, but real
Many people do not like semi things, things half true, half real, half wanted. It is the same in chess.
For example, I think most chess gurus, even authorities, would not suppose the existence and importance of semi-backward pawns, but they are both very real, and very meaningful in terms of evaluation scores.
The definition for a semi-backward pawn would actually be quite simple: a pawn whose advance is stopped by one enemy pawn, while at the same time supported by another friendly one, could be called semi-backward. I think with semi-backward pawns it makes sense to consider them only on the 7th and 6th ranks, in distinction to both fully backward and backward-fated pawns, as in most cases the validity of a semi-backward pawn depends on the piece control of the square in front of the pawn, where usually the enemy might have better such only for the 2 above-mentioned ranks.
[d]6k1/1pp5/p7/P7/8/8/8/6K1 w - - 0 1
Here b7 is a semi-backward pawn: a5 stops its advance, while c7 supports it.
[d]6k1/1pp5/p7/P7/2N5/8/8/6K1 w - - 0 1
You add another piece (usually a minor) controlling the square in front of the semi-backward pawn (b6), and the b7 pawn already becomes fully backward - it can not advance without being lost. It is very simple, how can you assert that semi-backward pawns do not exist? They do, and are indeed very meaningful. Sometimes, with other terms being equal, they can make the distinction.
2 other examples for semi-backward pawns.
[d]6k1/1pp5/8/P7/8/8/8/6K1 w - - 0 1
[d]6k1/1pp5/8/2P5/8/8/8/6K1 w - - 0 1
If you can claim that semi-backward pawns are important, they gain even more in importance when they are part of the king shelter, but I think most or even all engines, judging by the games I have seen, ignore this factor, much to the detriment of their positional understanding.
[d]6k1/5pp1/4p2p/4P3/8/8/8/6K1 w - - 0 1
f7 is a typical semi-backward pawn, part of the king shelter. If an usual semi-backward pawn on the 7th rank might be worth a penalty of some 10-15cps, when it is part of the king shelter its penalty might rise by some 1/3 to 2/3, that would make some 20-25 cps. f7 is a real and vital weak pawn to be penalised, but how many engines would actually do this? My experience is that on many occasions engines fail precisely because of a lack of similar knowledge. It is bad to have semi-backward pawns, part of the king shelter, because they make the shelter considerably less flexible.
[d]6k1/5pp1/4p2p/4P3/4N3/8/8/6K1 w - - 0 1
You add a minor piece controlling the square in front of the semi-backward pawn (f6), and f7 already becomes fully backward, as it can not advance without being lost. How can you assert in such a case that the f7 weakness does not exist? It is there, and it is clearly felt positionally.
Someone would say that a semi-backward-fated pawn part of the king shelter is a pretentious term and an illusion at the same time, but this is not so.
[d]6k1/5ppp/4p3/4P1P1/8/8/8/6K1 w - - 0 1
f7 is such a pawn. A semi-backward-fated pawn would be a pawn whose advance is stopped by 2 enemy pawns (e5 and g5 in the above example), while only one friendly pawn supports it (g7). It is similar to a fully backward pawn (the standard backward pawn considered by most engines), but even weaker, as the possible captures by 2 alternating enemy pawns make it even more exposed to advancing safely. When it is part of the king shelter, its meaning rises sharply, and actually the f7 pawn above could reasonably be penalised by some 35cps, a value that is not to be easily dismissed and that could completely alter the game situation. It is not at all difficult to implement as a term, in any case not more difficult than implementing a potential passer, it is frequent, and brings added value.
My question to all thinking chess friends upon leaving this discussion would be: why neglect the existence of a term that quite frequently is as important as a potential passer, a normal double pawn part of a group or a rook on a semi-open file? One could as well neglect the above-mentioned terms, that are however much more familiar among developers.
All comments very much appreciated, although I know the thread will remain silent.
Anyone currently implementing semi-backward pawns, or something similar?
Best, Lyudmil
For example, I think most chess gurus, even authorities, would not suppose the existence and importance of semi-backward pawns, but they are both very real, and very meaningful in terms of evaluation scores.
The definition for a semi-backward pawn would actually be quite simple: a pawn whose advance is stopped by one enemy pawn, while at the same time supported by another friendly one, could be called semi-backward. I think with semi-backward pawns it makes sense to consider them only on the 7th and 6th ranks, in distinction to both fully backward and backward-fated pawns, as in most cases the validity of a semi-backward pawn depends on the piece control of the square in front of the pawn, where usually the enemy might have better such only for the 2 above-mentioned ranks.
[d]6k1/1pp5/p7/P7/8/8/8/6K1 w - - 0 1
Here b7 is a semi-backward pawn: a5 stops its advance, while c7 supports it.
[d]6k1/1pp5/p7/P7/2N5/8/8/6K1 w - - 0 1
You add another piece (usually a minor) controlling the square in front of the semi-backward pawn (b6), and the b7 pawn already becomes fully backward - it can not advance without being lost. It is very simple, how can you assert that semi-backward pawns do not exist? They do, and are indeed very meaningful. Sometimes, with other terms being equal, they can make the distinction.
2 other examples for semi-backward pawns.
[d]6k1/1pp5/8/P7/8/8/8/6K1 w - - 0 1
[d]6k1/1pp5/8/2P5/8/8/8/6K1 w - - 0 1
If you can claim that semi-backward pawns are important, they gain even more in importance when they are part of the king shelter, but I think most or even all engines, judging by the games I have seen, ignore this factor, much to the detriment of their positional understanding.
[d]6k1/5pp1/4p2p/4P3/8/8/8/6K1 w - - 0 1
f7 is a typical semi-backward pawn, part of the king shelter. If an usual semi-backward pawn on the 7th rank might be worth a penalty of some 10-15cps, when it is part of the king shelter its penalty might rise by some 1/3 to 2/3, that would make some 20-25 cps. f7 is a real and vital weak pawn to be penalised, but how many engines would actually do this? My experience is that on many occasions engines fail precisely because of a lack of similar knowledge. It is bad to have semi-backward pawns, part of the king shelter, because they make the shelter considerably less flexible.
[d]6k1/5pp1/4p2p/4P3/4N3/8/8/6K1 w - - 0 1
You add a minor piece controlling the square in front of the semi-backward pawn (f6), and f7 already becomes fully backward, as it can not advance without being lost. How can you assert in such a case that the f7 weakness does not exist? It is there, and it is clearly felt positionally.
Someone would say that a semi-backward-fated pawn part of the king shelter is a pretentious term and an illusion at the same time, but this is not so.
[d]6k1/5ppp/4p3/4P1P1/8/8/8/6K1 w - - 0 1
f7 is such a pawn. A semi-backward-fated pawn would be a pawn whose advance is stopped by 2 enemy pawns (e5 and g5 in the above example), while only one friendly pawn supports it (g7). It is similar to a fully backward pawn (the standard backward pawn considered by most engines), but even weaker, as the possible captures by 2 alternating enemy pawns make it even more exposed to advancing safely. When it is part of the king shelter, its meaning rises sharply, and actually the f7 pawn above could reasonably be penalised by some 35cps, a value that is not to be easily dismissed and that could completely alter the game situation. It is not at all difficult to implement as a term, in any case not more difficult than implementing a potential passer, it is frequent, and brings added value.
My question to all thinking chess friends upon leaving this discussion would be: why neglect the existence of a term that quite frequently is as important as a potential passer, a normal double pawn part of a group or a rook on a semi-open file? One could as well neglect the above-mentioned terms, that are however much more familiar among developers.
All comments very much appreciated, although I know the thread will remain silent.
Anyone currently implementing semi-backward pawns, or something similar?
Best, Lyudmil
-
jdart
- Posts: 4375
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: Semi, but real
In my eval, both White pawns would be considered as attacking the Black king, because they control squares that are near it. But this is a small factor in the overall eval.
In general I have found it difficult to correctly evaluate opponent pawn advances on the king side (aka "pawn storms"). Many times the pawns are blockable and advancing them does not help: in fact if the attacker has castled on the same side then his own kingside is being weakened. But for every one of those cases you can find one where the pawn advance is good.
In general I have found it difficult to correctly evaluate opponent pawn advances on the king side (aka "pawn storms"). Many times the pawns are blockable and advancing them does not help: in fact if the attacker has castled on the same side then his own kingside is being weakened. But for every one of those cases you can find one where the pawn advance is good.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: Semi, but real
Hi Jon.jdart wrote:In my eval, both White pawns would be considered as attacking the Black king, because they control squares that are near it. But this is a small factor in the overall eval.
In general I have found it difficult to correctly evaluate opponent pawn advances on the king side (aka "pawn storms"). Many times the pawns are blockable and advancing them does not help: in fact if the attacker has castled on the same side then his own kingside is being weakened. But for every one of those cases you can find one where the pawn advance is good.
Thanks for posting.
That is the trick, doing more meaningful terms that would account more precisely for the complete picture. The complete picture would mean possible ramifications when going deeper into the tree, where in different ramifications only certain terms would be important. When you have more terms, it is more difficult to go wrong somewhere.
For example, you have in the eval only a storming pawn (or pawn close to the enemy king), you have 100 positions, in half of which the ability to attack the enemy king with the pawn might be meaningful, while in the other half the lower flexibility of the shelter because of the semi-backward might be of importance. When you do only one term for the 2, with different value, you necessarily miss part of the true logic behind the position and will go wrong in half of the cases (or whatever the relevant number of positions). When you have both terms, it is more difficult to go wrong.
In the position you are referring to, I would say that g5 is a storming pawn (because being able to advance), e5 is gaining space (because not able to advance, consequently not helpful in opening files; but would be a storming pawn, if there was a black pawn on d6 or f6, supposing renewed mobility in capturing and ability to open files; thus, a storming pawn blocked would only gain space, lower value than for storming, but when in a lever with an enemy pawn, even if bloacked, it again becomes a stormer), while f7 is a separate story on top of this, making the shelter inflexible.
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Blocking of pawns galore
I know some engines know how to efficiently block (here I will refer to blocking of pawns with pieces, in distinction to blocking/fixing enemy pawns with own pawns) enemy passers, but my general idea is that the popular understanding of blocking with pieces other, different types of pawns is somewhat limited (but, of course, I would be happy to be corrected).
For me, blocking is extremely important in chess, it creates a physical obstruction in front of the pawn that is blocked, meaning that its value and potential in advancing forward are temporarily or even permanently challenged. And pawns are basically there on the board to do something, to advance and attack, attempt promotion, etc., without moving they become pretty much useless over prolonged periods of time. Thus, blocking an enemy pawn with an indicated piece would mean that you deprive the opponent of some potential that would remain only latent until the pawn is blocked. Some might not believe it, but blocking of different types of pawns, not only passers, provides real added value, and is sometimes the only plausible solution of a range of game situations.
I would basically distinguish between 5 main types of blocking different pawns. In almost all cases the only blockers to be considered are the minor pieces, as other pieces are easily removed by enemy pieces of lower power from their blocking positions. Usually, the blocking bonus for the knight would be some 1/3 higher than that of the bishop, in some cases maybe more. It is very important at that that the blocking minors could not be attacked not only by enemy pawns, but even by enemy pieces of equal power, which would decrease their blocking capacity. I think blocking of whatever type of pawn should not even be considered, if an enemy pawn is able to attack it, and some additional bonus might be due if no enemy pieces of equal power can attack the blocker.
The 5 main types, for me, would be:
- blocking of separate passers, including connected passers that are straddled
- blocking of horizontally isolated pawns
- blocking of backward pawns
- blocking of pawns that are neither of the above types (normal pawns), but with the blocker unable to be attacked by enemy pawns
- blocking when part of bigger fixed structures
Below some diagrams to illustrate the use of the main types of blocking.
[d]6k1/3p4/3Bp3/2P3p1/p5N1/N5P1/5P1P/6K1 w - - 0 1
The knight on a3 blocks a separate passer. Here blocking is efficient namely because by definition the blocker can not be attacked by an enemy pawn, as there is not such. I think this should be considered as the main type of blocking, the most important at that. Bonus for blocking could be in the range of 25-30cps, roughly equal the value of the passer blocked.
The bishop on d6 blocks a backward pawn. The meaning of such blocking is that the already backward pawn (difficult to advance) becomes already altogether impossible to advance, mechanically obstructed, which frequently might have some further implications. In a sense blocking means the demise of the backward pawn as a moving entity. Hard to say what the exact bonus should be, maybe 1/5 to 1/3 the standard one for blocking a separate passer, some 5-10cps.
The knight on g4 blocks an isolated pawn. Here, too, blocking is efficient, because by definition no enemy pawn could attack the blocker. The meaning of this type of blocking is that the already vulnerable isolated pawn becomes even more vulnerable structurally, as it can not hope in taking shelter to another square by advancing, as this is occupied by the blocker. Sometimes this might have implications. Concerning the bonus value, close to blocking a backward pawn, maybe some 5-10cps. (but here I think of mentioning that it makes sense to block only fully backward pawns, and not backward-fated, where this would result in a redundancy)
[d]6k1/8/p7/R4p2/4pB2/4N3/1P6/6K1 w - - 0 1
The rook on a5 blocks an isolated pawn, maybe the only blocking function for the rook that should be considered, but bonus should be lower than that for a minor doing the same job.
The bishop on f4 and the knight on e3 block connected passers that are straddled (meaning that none of those passers can attack one of the blocking minors). It is important at that that there are no other enemy pawns on adjacent files that could attack the blockers. Bonus values could be similar to blocking a separate passer.
[d]6k1/8/5p2/p3p3/Np2B1P1/5P2/P7/6K1 w - - 0 1
The knight on a4 blocks a normal pawn with the blocker unable to be attacked by an enemy pawn. The meaning of this type of blocking is simply that the pawn can not advance, but, as it is not a pawn of special importance, the blocking bonus could be very modest, some 3-5cps.
The bishop on e4 blocks a normal pawn with the only pawn being able to attack it itself backward (suggesting very low probability of an efficient attack). Same 3-5cps. (here again it springs to my mind that in no way bonus points for blocking should be mixed up with bonus points for outposts, those are simply separate things)
[d]6k1/8/1p3p2/pP2pP2/N3B3/5P2/P7/6K1 w - - 0 1
Same position, but with only enemy pawns able to attack the blockers being fixed themselves by enemy pawns (meaning low probability of attack, currently inexistent).
[d]6k1/8/2p2p2/1pP1pBp1/pN2P1P1/P7/8/6K1 w - - 0 1
The knight on b4 and the bishop on f5 fulfill the function of blocking when part of bigger fixed structures of pawns. In this case, by definition, the blockers are unattackable by enemy pawns, as on both adjacent files to the blocker the enemy pawns would be fixed/blocked themselves. The meaning of this type of blocking is that the blocker artificially maintains the wholeness of the own pawn structure/chain, in the sense that, if captured, the chain would reproduce itself to the benefit of the side with the blocker (with the respective bonus points for chains, lead pawns, etc.), while, if not captured, it is still useful in connecting the ranks of the own forces (making enemy penetration difficult). That could have a particular effect in positions with many files closed (and, actually, the file where the blocker is should be considered as closed itself). The bonus for this type should be quite substantial, in the range of the standard one for blocking a separate passer, some 20-25cps.
How many types of pawns does your engine block?
Best, Lyudmil
For me, blocking is extremely important in chess, it creates a physical obstruction in front of the pawn that is blocked, meaning that its value and potential in advancing forward are temporarily or even permanently challenged. And pawns are basically there on the board to do something, to advance and attack, attempt promotion, etc., without moving they become pretty much useless over prolonged periods of time. Thus, blocking an enemy pawn with an indicated piece would mean that you deprive the opponent of some potential that would remain only latent until the pawn is blocked. Some might not believe it, but blocking of different types of pawns, not only passers, provides real added value, and is sometimes the only plausible solution of a range of game situations.
I would basically distinguish between 5 main types of blocking different pawns. In almost all cases the only blockers to be considered are the minor pieces, as other pieces are easily removed by enemy pieces of lower power from their blocking positions. Usually, the blocking bonus for the knight would be some 1/3 higher than that of the bishop, in some cases maybe more. It is very important at that that the blocking minors could not be attacked not only by enemy pawns, but even by enemy pieces of equal power, which would decrease their blocking capacity. I think blocking of whatever type of pawn should not even be considered, if an enemy pawn is able to attack it, and some additional bonus might be due if no enemy pieces of equal power can attack the blocker.
The 5 main types, for me, would be:
- blocking of separate passers, including connected passers that are straddled
- blocking of horizontally isolated pawns
- blocking of backward pawns
- blocking of pawns that are neither of the above types (normal pawns), but with the blocker unable to be attacked by enemy pawns
- blocking when part of bigger fixed structures
Below some diagrams to illustrate the use of the main types of blocking.
[d]6k1/3p4/3Bp3/2P3p1/p5N1/N5P1/5P1P/6K1 w - - 0 1
The knight on a3 blocks a separate passer. Here blocking is efficient namely because by definition the blocker can not be attacked by an enemy pawn, as there is not such. I think this should be considered as the main type of blocking, the most important at that. Bonus for blocking could be in the range of 25-30cps, roughly equal the value of the passer blocked.
The bishop on d6 blocks a backward pawn. The meaning of such blocking is that the already backward pawn (difficult to advance) becomes already altogether impossible to advance, mechanically obstructed, which frequently might have some further implications. In a sense blocking means the demise of the backward pawn as a moving entity. Hard to say what the exact bonus should be, maybe 1/5 to 1/3 the standard one for blocking a separate passer, some 5-10cps.
The knight on g4 blocks an isolated pawn. Here, too, blocking is efficient, because by definition no enemy pawn could attack the blocker. The meaning of this type of blocking is that the already vulnerable isolated pawn becomes even more vulnerable structurally, as it can not hope in taking shelter to another square by advancing, as this is occupied by the blocker. Sometimes this might have implications. Concerning the bonus value, close to blocking a backward pawn, maybe some 5-10cps. (but here I think of mentioning that it makes sense to block only fully backward pawns, and not backward-fated, where this would result in a redundancy)
[d]6k1/8/p7/R4p2/4pB2/4N3/1P6/6K1 w - - 0 1
The rook on a5 blocks an isolated pawn, maybe the only blocking function for the rook that should be considered, but bonus should be lower than that for a minor doing the same job.
The bishop on f4 and the knight on e3 block connected passers that are straddled (meaning that none of those passers can attack one of the blocking minors). It is important at that that there are no other enemy pawns on adjacent files that could attack the blockers. Bonus values could be similar to blocking a separate passer.
[d]6k1/8/5p2/p3p3/Np2B1P1/5P2/P7/6K1 w - - 0 1
The knight on a4 blocks a normal pawn with the blocker unable to be attacked by an enemy pawn. The meaning of this type of blocking is simply that the pawn can not advance, but, as it is not a pawn of special importance, the blocking bonus could be very modest, some 3-5cps.
The bishop on e4 blocks a normal pawn with the only pawn being able to attack it itself backward (suggesting very low probability of an efficient attack). Same 3-5cps. (here again it springs to my mind that in no way bonus points for blocking should be mixed up with bonus points for outposts, those are simply separate things)
[d]6k1/8/1p3p2/pP2pP2/N3B3/5P2/P7/6K1 w - - 0 1
Same position, but with only enemy pawns able to attack the blockers being fixed themselves by enemy pawns (meaning low probability of attack, currently inexistent).
[d]6k1/8/2p2p2/1pP1pBp1/pN2P1P1/P7/8/6K1 w - - 0 1
The knight on b4 and the bishop on f5 fulfill the function of blocking when part of bigger fixed structures of pawns. In this case, by definition, the blockers are unattackable by enemy pawns, as on both adjacent files to the blocker the enemy pawns would be fixed/blocked themselves. The meaning of this type of blocking is that the blocker artificially maintains the wholeness of the own pawn structure/chain, in the sense that, if captured, the chain would reproduce itself to the benefit of the side with the blocker (with the respective bonus points for chains, lead pawns, etc.), while, if not captured, it is still useful in connecting the ranks of the own forces (making enemy penetration difficult). That could have a particular effect in positions with many files closed (and, actually, the file where the blocker is should be considered as closed itself). The bonus for this type should be quite substantial, in the range of the standard one for blocking a separate passer, some 20-25cps.
How many types of pawns does your engine block?
Best, Lyudmil
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Why unopposed?
I am convinced a separate category of pawns, called unopposed, have their full right to existence on the chess board, I do not know about engine evals.
For me, an unopposed pawn would be any pawn with no enemy counterparts (pawns) on the same file. I think that has some meaning, but first we have to ask the question what is a normal pawn?
[d]6k1/1pp5/p7/8/8/1P6/1PP5/6K1 w - - 0 1
I think on the above diagram c2,c7 and b7 could be called normal pawns, b2 and b3 are double, while a6 is obviously something distinct. A normal pawn could be any pawn in the initial position, or a pawn that has one or more enemy pawns on the same file where it is. a6 in the above example does not comply to such a rule, so what is it basically in terms of general chess knowledge?
I think that if the category of unopposed pawns is introduced into use, that might solve some pending problems. It would make sense to consider unopposed pawns separately from double pawns, as the 2 do different things on the board. Double pawns are due some penalty for structural deficiency, making them easier to attack, as well as for their clumsiness in advancing. Unopposed pawns, on the other hand, would be due some bonus for quite different reasons: the ability to advance easier by definition (as the main obstacle to advance would be the enemy pawn on the same file), and the potential, under given circumstances, to help in creating a potential passer and a passer. That is quite a definite function, it is real, and meaningful. Maybe it would be opportune to consider both double and unopposed pawns, with assigning lower penalty for doubles and a bit of bonus for unopposed, instead of a single bigger penalty for doubles. I think this would be more correct in terms of general chess rules.
[d]6k1/1pp5/p7/8/8/1P6/2P5/6K1 w - - 0 1
Of course, a6 is a potential passer (candidate) on the above diagram. It still complies to the definition of an unopposed pawn, with the specific clarification that the number of pawns on adjacent files able to support its advance forward is at least equal to the number of enemy pawns able to stop this advance. So, in a sense, a potential passer is a supspecies of unopposed pawns, the general category. A full passer is obviously also a subspecies of this, but of course, with much bigger importance.
Thus, an unopposed pawn would be one where the number of enemy pawns able to stop its advance forward is bigger than the number of friendly pawns able to support its advance. Points might be dispensed in function of the prevalence of pawns stopping its advance.
I think the standard one would be the case when 1 more enemy pawn stops the advanvce than the pawns supporting. The bonus for this could be somewhere half the bonus for a potential passer, some 8-10cps (if a potential passer is worth half a passer, or some 15cps). When 2 more enemy pawns stop the advance than the pawns supporting, the unopposed pawn could be worth some 6-8cps (still not to be neglected). In the case of 3 more pawns stopping its advance, the bonus could fall to 4cps. In any of the cases, the pawn would enjoy bigger than normal ability to advance.
[d]6k1/1pp3p1/3p3p/8/4P3/2P5/1PP4P/6K1 w - - 0 1
On the above diagram, e4 is a standard unopposed pawn with one more enemy pawn stopping its advance (no friendly pawns supporting it, while 1 enemy pawn,d6, stopping it), some 8cps bonus due. d6, on the other hand, is a less important unopposed pawn, with 2 more enemy pawns stopping its advance (c2,c3 and e4 all stop it, while only c7 supports it), bonus would be some 6cps. g7 is, of course, a potential passer.
[d]6k1/1pp3p1/3p3p/8/4P3/2P1P3/1PP4P/6K1 w - - 0 1
On the above diagram, with very small changes, a white e3 pawn added, d6 is already stopped by 3 more enemy pawns (including e3), so that its bonus would fall to some 4cps. e4 would still be due the standard bonus for unopposed, but not e3, as the less advanced double pawn, even if unopposed, does not enjoy the capacity of the general unopposed pawn to move forward actively and freely.
And one last example, to illustrate the validity of unopposed pawns.
[d]6k1/ppp2ppp/3pp3/8/8/2P2P2/PPP2PPP/6K1 w - - 0 1
On the above diagram, neither d6, nor e6 are potential passers, as a significant number of enemy pawns stop their advance. Are not they due some additional bonus, they scream for it, they do not quite look like normal pawns. I think they are. If you dispense some 8cps for each pawn as unopposed, I still think that those pawns are due an additional further bonus, because of their work in tandem. My experience tells me that in some 80% of cases such 2 unopposed pawns on adjacent files will quite probably give rise to the creation of a friendly potential passer, and eventually a full passer. They are quite a feature of the position. So that I think 2 unopposed pawns on adjacent files are due a further bonus of some 10-15cps for working in tandem, apart from the separate bonus points for opportunity to actively advance.
I will enjoy the tiny 1% opportunity of someone commenting here. Is someone doing such pawns in eval, or maybe something similar?
Best, Lyudmil
For me, an unopposed pawn would be any pawn with no enemy counterparts (pawns) on the same file. I think that has some meaning, but first we have to ask the question what is a normal pawn?
[d]6k1/1pp5/p7/8/8/1P6/1PP5/6K1 w - - 0 1
I think on the above diagram c2,c7 and b7 could be called normal pawns, b2 and b3 are double, while a6 is obviously something distinct. A normal pawn could be any pawn in the initial position, or a pawn that has one or more enemy pawns on the same file where it is. a6 in the above example does not comply to such a rule, so what is it basically in terms of general chess knowledge?
I think that if the category of unopposed pawns is introduced into use, that might solve some pending problems. It would make sense to consider unopposed pawns separately from double pawns, as the 2 do different things on the board. Double pawns are due some penalty for structural deficiency, making them easier to attack, as well as for their clumsiness in advancing. Unopposed pawns, on the other hand, would be due some bonus for quite different reasons: the ability to advance easier by definition (as the main obstacle to advance would be the enemy pawn on the same file), and the potential, under given circumstances, to help in creating a potential passer and a passer. That is quite a definite function, it is real, and meaningful. Maybe it would be opportune to consider both double and unopposed pawns, with assigning lower penalty for doubles and a bit of bonus for unopposed, instead of a single bigger penalty for doubles. I think this would be more correct in terms of general chess rules.
[d]6k1/1pp5/p7/8/8/1P6/2P5/6K1 w - - 0 1
Of course, a6 is a potential passer (candidate) on the above diagram. It still complies to the definition of an unopposed pawn, with the specific clarification that the number of pawns on adjacent files able to support its advance forward is at least equal to the number of enemy pawns able to stop this advance. So, in a sense, a potential passer is a supspecies of unopposed pawns, the general category. A full passer is obviously also a subspecies of this, but of course, with much bigger importance.
Thus, an unopposed pawn would be one where the number of enemy pawns able to stop its advance forward is bigger than the number of friendly pawns able to support its advance. Points might be dispensed in function of the prevalence of pawns stopping its advance.
I think the standard one would be the case when 1 more enemy pawn stops the advanvce than the pawns supporting. The bonus for this could be somewhere half the bonus for a potential passer, some 8-10cps (if a potential passer is worth half a passer, or some 15cps). When 2 more enemy pawns stop the advance than the pawns supporting, the unopposed pawn could be worth some 6-8cps (still not to be neglected). In the case of 3 more pawns stopping its advance, the bonus could fall to 4cps. In any of the cases, the pawn would enjoy bigger than normal ability to advance.
[d]6k1/1pp3p1/3p3p/8/4P3/2P5/1PP4P/6K1 w - - 0 1
On the above diagram, e4 is a standard unopposed pawn with one more enemy pawn stopping its advance (no friendly pawns supporting it, while 1 enemy pawn,d6, stopping it), some 8cps bonus due. d6, on the other hand, is a less important unopposed pawn, with 2 more enemy pawns stopping its advance (c2,c3 and e4 all stop it, while only c7 supports it), bonus would be some 6cps. g7 is, of course, a potential passer.
[d]6k1/1pp3p1/3p3p/8/4P3/2P1P3/1PP4P/6K1 w - - 0 1
On the above diagram, with very small changes, a white e3 pawn added, d6 is already stopped by 3 more enemy pawns (including e3), so that its bonus would fall to some 4cps. e4 would still be due the standard bonus for unopposed, but not e3, as the less advanced double pawn, even if unopposed, does not enjoy the capacity of the general unopposed pawn to move forward actively and freely.
And one last example, to illustrate the validity of unopposed pawns.
[d]6k1/ppp2ppp/3pp3/8/8/2P2P2/PPP2PPP/6K1 w - - 0 1
On the above diagram, neither d6, nor e6 are potential passers, as a significant number of enemy pawns stop their advance. Are not they due some additional bonus, they scream for it, they do not quite look like normal pawns. I think they are. If you dispense some 8cps for each pawn as unopposed, I still think that those pawns are due an additional further bonus, because of their work in tandem. My experience tells me that in some 80% of cases such 2 unopposed pawns on adjacent files will quite probably give rise to the creation of a friendly potential passer, and eventually a full passer. They are quite a feature of the position. So that I think 2 unopposed pawns on adjacent files are due a further bonus of some 10-15cps for working in tandem, apart from the separate bonus points for opportunity to actively advance.
I will enjoy the tiny 1% opportunity of someone commenting here. Is someone doing such pawns in eval, or maybe something similar?
Best, Lyudmil
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Leading the chain
This will be more of a question to you: how do you actually implement chains of pawns in your engines? I tried to read something on the matter (but not in engine codes), and all I managed to find was that a chain would be important and it has base and peak pawns (supposedly the least and the most advanced pawns of the chain respectively). Is that all, how do you actually score chains?
For me, chains of pawns are extremely important, and definitely one of the elements that modern engines still misunderstand. I harassed my mind for quite some time to think how could chains actually be scored, but the only idea I came up with was assigning bonus points exclusively for the peak pawns of the chain (I will call those here lead pawns, you will excuse my personal jargon, it is just easier for me that way to imagine things).
Of course, chains are due some quite nice bonus for the main reason that, while being solid, they are capable of efficiently restricting the activity of the enemy pieces, while helping the activity of the friendly ones. Sometimes, that is quite meaningful, and has a longer-term effect, allowing friendly pieces to mobilise gradually to the detriment of enemy pieces. That is not always successfully registered by engines, sometimes they fail badly here. (And, of course, the definition for a chain would be a number of friendly pawns along the same diagonal)
[d]6k1/8/2P5/3P4/4P3/1P4P1/P4P1P/6K1 w - - 0 1
On the above diagram, there are 4 chains of white pawns: a2-b3, f2-g3, h2-g3 (but you might also consider the apex g3 pawn), and e4-d5-c6. I think it really does not make sense to consider lead pawns (peaks) when they are only on the 3rd rank, for the plain reason that there they do not fulfill at all their main function of restricting the enemy pieces' activity, there they are simply too unimportant. I think it makes sense to start counting lead pawns from the 4th rank onward.
My system for dispensing bonus points to lead pawns (the most advanced pawns of the chain, the most importyant one, where all of the strength of the chain is concentrated; you eliminate the enemy lead pawn, and you are certainly better off) is quite simple:
- you start with the 4rh rank and assign maybe some 10cps for a lead pawn there
- you increase the bonus with some 10-20% for the 5th rank, another 20% for the 6th rank, etc.
- you increase further the bonus for the number of pawns the most advanced pawn of the chain is leading (and that is why I like the word lead), if it leads just one friendly pawn, the bonus will be standard, when it leads 2 friendly pawns, increase by 10%, another 10% increase when it leads 3 friendly pawns, etc. I think this is quite meaningful, because the more pawns a lead pawn leads, the stronger it becomes, limiting enemy interference and increasing the chances for regeneration of the chain in case some of the members of the chain is captured.
- very importantly, you increase still further the score, when the lead pawn is close to the enemy king (say just one or 2 squares away); please note, that a normal pawn, not part of a chain, and a lead pawn close to the enemy king, are not one and the same thing. The lead pawn is superior, not lastly because of its regeneration ability, when it is captured by an enemy pawn, it regenerates instantly and efficiently.
- you increase the score, when all or most of the pawns of the chain are fixed/blocked by enemy pawns, by maybe 1/3, as in this case the restricting capacity of the lead pawn and the chain as a whole is even bigger, because pieces of the two sides come into contact in a more difficult way, which prolongs the life of the chain
- lastly, but not least inportantly, you do all of the above things
The conclusion could be, that a good lead pawn of a bigger chain on the 6th rank, when close to the enemy king and with most pawns of the chain fixed, could well get an astoundingly big bonus, maybe some full pawn (not at all exaggerated), with other pawns in between also getting some very nice bonus points. Once you miss to score this, you miss really an awful lot.
[d]6k1/8/7P/6P1/3P1P2/2P1P3/1P6/6K1 w - - 0 1
On the above diagram, d4 is a lead pawn on the 4th rank of the b2-c3-d4 chain, as well as the e3-d4 chain (but you might also consider the d4 apex pawn), while h6 is a lead pawn on the 6th rank of the nice and large e3-f4-g5-h6 chain, leading 4 friendly pawns in all (or another 3), close to the enemy king, with all chains not fixed by enemy pawns. Of course, h6 would be due much higher bonus than d4, that is obvious and essential.
[d]6k1/8/4p3/3pP3/2pP4/2P5/8/6K1 w - - 0 1
On the above diagram, e5 leads 3 pawns in all on the 5th rank, while c4 does the same for black, with both chains fixed. What is the difference then? The difference is, of course, the direction of the white chain, which is close and looks toward the enemy king. That would make for a substantial rise in score for the e5 pawn. Engines very often disregard this rule.
[d]6k1/5p2/4pP2/3pP3/2pP4/2P5/8/6K1 w - - 0 1
Similar diagram, but with longer chains. Please note, that f6 is extremely powerful, I think it might be worth a full pawn (so you could sometimes sacrifice the exchange for it, or a pawn), as it is not only close to the enemy king, creating nasty threats, but is there too remain for quite some time, as any capture of the pawn will immediately regenerate the dangerous lead pawn with some effect. Engines tend not to recognize such important patterns.
[d]6k1/5p2/4pP2/3pP3/3P4/2P5/8/6K1 w - - 0 1
Just to mention again, when most, but not all of the pawns of the chain are fixed, you could still assign higher bonus for fixed chains.
Any comments very much appreciated. Please, shed some light upon me on how you implement chains. I really did not found a more useful approach than doing primarily lead pawns.
I think here at least Carl (if not already migrated from the forum) could say a few words.
Best, Lyudmil
For me, chains of pawns are extremely important, and definitely one of the elements that modern engines still misunderstand. I harassed my mind for quite some time to think how could chains actually be scored, but the only idea I came up with was assigning bonus points exclusively for the peak pawns of the chain (I will call those here lead pawns, you will excuse my personal jargon, it is just easier for me that way to imagine things).
Of course, chains are due some quite nice bonus for the main reason that, while being solid, they are capable of efficiently restricting the activity of the enemy pieces, while helping the activity of the friendly ones. Sometimes, that is quite meaningful, and has a longer-term effect, allowing friendly pieces to mobilise gradually to the detriment of enemy pieces. That is not always successfully registered by engines, sometimes they fail badly here. (And, of course, the definition for a chain would be a number of friendly pawns along the same diagonal)
[d]6k1/8/2P5/3P4/4P3/1P4P1/P4P1P/6K1 w - - 0 1
On the above diagram, there are 4 chains of white pawns: a2-b3, f2-g3, h2-g3 (but you might also consider the apex g3 pawn), and e4-d5-c6. I think it really does not make sense to consider lead pawns (peaks) when they are only on the 3rd rank, for the plain reason that there they do not fulfill at all their main function of restricting the enemy pieces' activity, there they are simply too unimportant. I think it makes sense to start counting lead pawns from the 4th rank onward.
My system for dispensing bonus points to lead pawns (the most advanced pawns of the chain, the most importyant one, where all of the strength of the chain is concentrated; you eliminate the enemy lead pawn, and you are certainly better off) is quite simple:
- you start with the 4rh rank and assign maybe some 10cps for a lead pawn there
- you increase the bonus with some 10-20% for the 5th rank, another 20% for the 6th rank, etc.
- you increase further the bonus for the number of pawns the most advanced pawn of the chain is leading (and that is why I like the word lead), if it leads just one friendly pawn, the bonus will be standard, when it leads 2 friendly pawns, increase by 10%, another 10% increase when it leads 3 friendly pawns, etc. I think this is quite meaningful, because the more pawns a lead pawn leads, the stronger it becomes, limiting enemy interference and increasing the chances for regeneration of the chain in case some of the members of the chain is captured.
- very importantly, you increase still further the score, when the lead pawn is close to the enemy king (say just one or 2 squares away); please note, that a normal pawn, not part of a chain, and a lead pawn close to the enemy king, are not one and the same thing. The lead pawn is superior, not lastly because of its regeneration ability, when it is captured by an enemy pawn, it regenerates instantly and efficiently.
- you increase the score, when all or most of the pawns of the chain are fixed/blocked by enemy pawns, by maybe 1/3, as in this case the restricting capacity of the lead pawn and the chain as a whole is even bigger, because pieces of the two sides come into contact in a more difficult way, which prolongs the life of the chain
- lastly, but not least inportantly, you do all of the above things
The conclusion could be, that a good lead pawn of a bigger chain on the 6th rank, when close to the enemy king and with most pawns of the chain fixed, could well get an astoundingly big bonus, maybe some full pawn (not at all exaggerated), with other pawns in between also getting some very nice bonus points. Once you miss to score this, you miss really an awful lot.
[d]6k1/8/7P/6P1/3P1P2/2P1P3/1P6/6K1 w - - 0 1
On the above diagram, d4 is a lead pawn on the 4th rank of the b2-c3-d4 chain, as well as the e3-d4 chain (but you might also consider the d4 apex pawn), while h6 is a lead pawn on the 6th rank of the nice and large e3-f4-g5-h6 chain, leading 4 friendly pawns in all (or another 3), close to the enemy king, with all chains not fixed by enemy pawns. Of course, h6 would be due much higher bonus than d4, that is obvious and essential.
[d]6k1/8/4p3/3pP3/2pP4/2P5/8/6K1 w - - 0 1
On the above diagram, e5 leads 3 pawns in all on the 5th rank, while c4 does the same for black, with both chains fixed. What is the difference then? The difference is, of course, the direction of the white chain, which is close and looks toward the enemy king. That would make for a substantial rise in score for the e5 pawn. Engines very often disregard this rule.
[d]6k1/5p2/4pP2/3pP3/2pP4/2P5/8/6K1 w - - 0 1
Similar diagram, but with longer chains. Please note, that f6 is extremely powerful, I think it might be worth a full pawn (so you could sometimes sacrifice the exchange for it, or a pawn), as it is not only close to the enemy king, creating nasty threats, but is there too remain for quite some time, as any capture of the pawn will immediately regenerate the dangerous lead pawn with some effect. Engines tend not to recognize such important patterns.
[d]6k1/5p2/4pP2/3pP3/3P4/2P5/8/6K1 w - - 0 1
Just to mention again, when most, but not all of the pawns of the chain are fixed, you could still assign higher bonus for fixed chains.
Any comments very much appreciated. Please, shed some light upon me on how you implement chains. I really did not found a more useful approach than doing primarily lead pawns.
I think here at least Carl (if not already migrated from the forum) could say a few words.
Best, Lyudmil
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: Leading the chain
Here just a small case in point.
[d]r1b2rk1/p1q1bp1p/1pn1pRp1/2ppP1P1/3P4/2PBP3/PP1N3P/R1BQ2K1 b - - 0 14
This is from my game with Stockfish (I apologise again to the Stockfish team, this is just a good example, although not perfect, for Stock could have captured probably on e5 after capturing on f6), here the engine captures on f6 without much thought, which is not fully reasonable.
Stockfish seems not to recognise that after gf6 white builds a tremendous chain of 5 pawns led by f6, which is due some full pawn additional bonus. Thus, instead of the g5 pawn, modestly gaining space, and the similarly modest e5 lead pawn, leading just 3 pawns on the 5th rank and more distant from the enemy king, suddenly a much more powerful f6 lead pawn arises, a pawn already on the 6th rank, leading 4 other friendly pawns and very close to the enemy king. This adds some almost full pawn to the white score, which Stock seems not to fully grasp. When you add some open or semi-open files after gf6, playing Bf6 for black seems dubious. If Rf6, sacrificing the exchange, could be categorised as a positional move, in order to make it, you do not need anything special other than have the precise eval terms. Considering high bonus for the f6 pawn is such a term.
Actually, I do not know what a positional term is. It seems that people refer to some subtler terms, not generally accepted in the eval, as positional terms, and positional play, but those terms are absolutely like the rest of the more popular terms, just not that much popular.
Below the game again (I already posted it in the tournament section), just for reference.
[pgn][PlyCount "79"]
[Event "Blitz 2m+2s"]
[Site "Sofia"]
[Date "2013.09.22"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish 4 64 SSE4.2"]
[Result "1-0"]
[ECO "D00"]
[TimeControl "120+2"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "39"]
[MLFlags "000100"]
{1024MB, Dell XPS 4Cores} 1. e3 {[%emt 0:00:00]} 1... e6 {[%emt 0:00:08]} 2. d4
{[%emt 0:00:04]} 2... c5 {[%emt 0:00:00]} 3. f4 {[%emt 0:00:07]} 3... d5
{[%emt 0:00:06]} 4. Nf3 {[%emt 0:00:01]} 4... Nf6 {[%emt 0:00:05]} 5. c3
{[%emt 0:00:02]} 5... Nc6 {[%emt 0:00:11]} 6. Bd3 {[%emt 0:00:01]} 6... Bd6
{[%emt 0:00:04]} 7. O-O {[%emt 0:00:02]} 7... O-O {[%emt 0:00:04]} 8. Ne5
{[%emt 0:00:04]} 8... Qc7 {[%emt 0:00:03]} 9. Nd2 {[%emt 0:00:04]} 9... b6
{[%emt 0:00:06]} 10. g4 {[%emt 0:00:14]} 10... g6 {[%emt 0:00:00]} 11. g5
{[%emt 0:00:02]} 11... Nd7 {[%emt 0:00:05]} 12. Rf3 {[%emt 0:00:09]} 12... Ndxe5
{[%emt 0:00:05]} 13. fxe5 {[%emt 0:00:15]} 13... Be7 {[%emt 0:00:06]} 14. Rf6
{[%emt 0:00:10]} 14... Bxf6 {[%emt 0:00:08]} 15. gxf6 {[%emt 0:00:02]} 15... Bb7
{[%emt 0:00:05]} 16. Nf3 {[%emt 0:00:03]} 16... Qd7 {[%emt 0:00:12]} 17. Qe1
{[%emt 0:00:10]} 17... Rfc8 {[%emt 0:00:07]} 18. Qh4 {[%emt 0:00:12]} 18... Qd8
{[%emt 0:00:04]} 19. Ng5 {[%emt 0: 00:51]} 19... h5 {[%emt 0:00:00]} 20. Bd2
{[%emt 0:05:09]} 20... Rc7 {[%emt 0:00:07]} 21. Kf2 {[%emt 0:00:24]} 21... cxd4
{[%emt 0:00:10]} 22. exd4 {[%emt 0:00:02]} 22... Nxe5 {[%emt 0:00:03]} 23. dxe5
{[%emt 0:00:02]} 23... Rc4 {[%emt 0:00:19]} 24. Bxc4 {[%emt 0:00:11]} 24... dxc4
{[%emt 0:00:02]} 25. Qd4 {[%emt 0:00:00]} 25... Qxd4+ {[%emt 0:00:04]} 26. cxd4
{[%emt 0:00:02]} 26... Bd5 {[%emt 0:00:04]} 27. Bb4 {[%emt 0:00:12]} 27... a5
{[%emt 0:00:03]} 28. Bd6 {[%emt 0:00:02]} 28... Rc8 {[%emt 0:00:03]} 29. Rc1
{[%emt 0: 00:09]} 29... b5 {[%emt 0:00:07]} 30. Ke3 {[%emt 0:00:14]} 30... b4
{[%emt 0:00:00]} 31. Ne4 {[%emt 0:00:05]} 31... c3 {[%emt 0:00:05]} 32. bxc3
{[%emt 0:00:07]} 32... Bxa2 {[%emt 0:00:00]} 33. Bc5 {[%emt 0:00:18]} 33... b3
{[%emt 0:00:07]} 34. Ba3 {[%emt 0:00:02]} 34... a4 {[%emt 0:00:10]} 35. Nd6
{[%emt 0:00:21]} 35... Rc7 {[%emt 0:00:03]} 36. c4 {[%emt 0:00:03]} 36... Rc6
{[%emt 0:00:02]} 37. d5 {[%emt 0:00:08]} 37... Rc7 {[%emt 0:00: 05]} 38. dxe6
{[%emt 0:00:09]} 38... fxe6 {[%emt 0:00:02]} 39. f7+ {[%emt 0:00:02]} 39... Rxf7
{[%emt 0:00:03]} 40. Nxf7 {[%emt 0:00:01]} 1-0
[/pgn]
[d]r1b2rk1/p1q1bp1p/1pn1pRp1/2ppP1P1/3P4/2PBP3/PP1N3P/R1BQ2K1 b - - 0 14
This is from my game with Stockfish (I apologise again to the Stockfish team, this is just a good example, although not perfect, for Stock could have captured probably on e5 after capturing on f6), here the engine captures on f6 without much thought, which is not fully reasonable.
Stockfish seems not to recognise that after gf6 white builds a tremendous chain of 5 pawns led by f6, which is due some full pawn additional bonus. Thus, instead of the g5 pawn, modestly gaining space, and the similarly modest e5 lead pawn, leading just 3 pawns on the 5th rank and more distant from the enemy king, suddenly a much more powerful f6 lead pawn arises, a pawn already on the 6th rank, leading 4 other friendly pawns and very close to the enemy king. This adds some almost full pawn to the white score, which Stock seems not to fully grasp. When you add some open or semi-open files after gf6, playing Bf6 for black seems dubious. If Rf6, sacrificing the exchange, could be categorised as a positional move, in order to make it, you do not need anything special other than have the precise eval terms. Considering high bonus for the f6 pawn is such a term.
Actually, I do not know what a positional term is. It seems that people refer to some subtler terms, not generally accepted in the eval, as positional terms, and positional play, but those terms are absolutely like the rest of the more popular terms, just not that much popular.
Below the game again (I already posted it in the tournament section), just for reference.
[pgn][PlyCount "79"]
[Event "Blitz 2m+2s"]
[Site "Sofia"]
[Date "2013.09.22"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish 4 64 SSE4.2"]
[Result "1-0"]
[ECO "D00"]
[TimeControl "120+2"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "39"]
[MLFlags "000100"]
{1024MB, Dell XPS 4Cores} 1. e3 {[%emt 0:00:00]} 1... e6 {[%emt 0:00:08]} 2. d4
{[%emt 0:00:04]} 2... c5 {[%emt 0:00:00]} 3. f4 {[%emt 0:00:07]} 3... d5
{[%emt 0:00:06]} 4. Nf3 {[%emt 0:00:01]} 4... Nf6 {[%emt 0:00:05]} 5. c3
{[%emt 0:00:02]} 5... Nc6 {[%emt 0:00:11]} 6. Bd3 {[%emt 0:00:01]} 6... Bd6
{[%emt 0:00:04]} 7. O-O {[%emt 0:00:02]} 7... O-O {[%emt 0:00:04]} 8. Ne5
{[%emt 0:00:04]} 8... Qc7 {[%emt 0:00:03]} 9. Nd2 {[%emt 0:00:04]} 9... b6
{[%emt 0:00:06]} 10. g4 {[%emt 0:00:14]} 10... g6 {[%emt 0:00:00]} 11. g5
{[%emt 0:00:02]} 11... Nd7 {[%emt 0:00:05]} 12. Rf3 {[%emt 0:00:09]} 12... Ndxe5
{[%emt 0:00:05]} 13. fxe5 {[%emt 0:00:15]} 13... Be7 {[%emt 0:00:06]} 14. Rf6
{[%emt 0:00:10]} 14... Bxf6 {[%emt 0:00:08]} 15. gxf6 {[%emt 0:00:02]} 15... Bb7
{[%emt 0:00:05]} 16. Nf3 {[%emt 0:00:03]} 16... Qd7 {[%emt 0:00:12]} 17. Qe1
{[%emt 0:00:10]} 17... Rfc8 {[%emt 0:00:07]} 18. Qh4 {[%emt 0:00:12]} 18... Qd8
{[%emt 0:00:04]} 19. Ng5 {[%emt 0: 00:51]} 19... h5 {[%emt 0:00:00]} 20. Bd2
{[%emt 0:05:09]} 20... Rc7 {[%emt 0:00:07]} 21. Kf2 {[%emt 0:00:24]} 21... cxd4
{[%emt 0:00:10]} 22. exd4 {[%emt 0:00:02]} 22... Nxe5 {[%emt 0:00:03]} 23. dxe5
{[%emt 0:00:02]} 23... Rc4 {[%emt 0:00:19]} 24. Bxc4 {[%emt 0:00:11]} 24... dxc4
{[%emt 0:00:02]} 25. Qd4 {[%emt 0:00:00]} 25... Qxd4+ {[%emt 0:00:04]} 26. cxd4
{[%emt 0:00:02]} 26... Bd5 {[%emt 0:00:04]} 27. Bb4 {[%emt 0:00:12]} 27... a5
{[%emt 0:00:03]} 28. Bd6 {[%emt 0:00:02]} 28... Rc8 {[%emt 0:00:03]} 29. Rc1
{[%emt 0: 00:09]} 29... b5 {[%emt 0:00:07]} 30. Ke3 {[%emt 0:00:14]} 30... b4
{[%emt 0:00:00]} 31. Ne4 {[%emt 0:00:05]} 31... c3 {[%emt 0:00:05]} 32. bxc3
{[%emt 0:00:07]} 32... Bxa2 {[%emt 0:00:00]} 33. Bc5 {[%emt 0:00:18]} 33... b3
{[%emt 0:00:07]} 34. Ba3 {[%emt 0:00:02]} 34... a4 {[%emt 0:00:10]} 35. Nd6
{[%emt 0:00:21]} 35... Rc7 {[%emt 0:00:03]} 36. c4 {[%emt 0:00:03]} 36... Rc6
{[%emt 0:00:02]} 37. d5 {[%emt 0:00:08]} 37... Rc7 {[%emt 0:00: 05]} 38. dxe6
{[%emt 0:00:09]} 38... fxe6 {[%emt 0:00:02]} 39. f7+ {[%emt 0:00:02]} 39... Rxf7
{[%emt 0:00:03]} 40. Nxf7 {[%emt 0:00:01]} 1-0
[/pgn]
-
jdart
- Posts: 4375
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: Leading the chain
As I have discussed before, I have a small "space" bonus that encourages advanced pawns. It is a very small eval term though.
I think most of your discussion needs to be qualified depending on what pieces are present. For example it is well known that a "bad bishop" blocked by its own pawns deserves a penalty. In general too most programs give significant weight to piece mobility, and the mobility of sliding pieces, especially Bishops, is affected by pawn structure.
Locked positions with many pawns that cannot advance further tend to be drawish (but not always). Some programs discourage playing into this kind of position. I believe I used to have such a penalty but don't anymore.
Re king safety: a pawn near the King may or may not be a significant threat. If one or more pieces can be added to the attack, especially a Queen, then it is very important. Otherwise it may not be.
--Jon
I think most of your discussion needs to be qualified depending on what pieces are present. For example it is well known that a "bad bishop" blocked by its own pawns deserves a penalty. In general too most programs give significant weight to piece mobility, and the mobility of sliding pieces, especially Bishops, is affected by pawn structure.
Locked positions with many pawns that cannot advance further tend to be drawish (but not always). Some programs discourage playing into this kind of position. I believe I used to have such a penalty but don't anymore.
Re king safety: a pawn near the King may or may not be a significant threat. If one or more pieces can be added to the attack, especially a Queen, then it is very important. Otherwise it may not be.
--Jon
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: Leading the chain
Hi Jon.jdart wrote:As I have discussed before, I have a small "space" bonus that encourages advanced pawns. It is a very small eval term though.
I think most of your discussion needs to be qualified depending on what pieces are present. For example it is well known that a "bad bishop" blocked by its own pawns deserves a penalty. In general too most programs give significant weight to piece mobility, and the mobility of sliding pieces, especially Bishops, is affected by pawn structure.
Locked positions with many pawns that cannot advance further tend to be drawish (but not always). Some programs discourage playing into this kind of position. I believe I used to have such a penalty but don't anymore.
Re king safety: a pawn near the King may or may not be a significant threat. If one or more pieces can be added to the attack, especially a Queen, then it is very important. Otherwise it may not be.
--Jon
Thanks for keeping the discussion alive.
Of course, pieces present are important. I was referring to a rich middlegame, which is most difficult for both humans and engines.
I think pawns close to the enemy king would be relevant whenever the side with such a pawn has at least a single heavy piece (rook). This would comprise most constellations. On the other hand, my experience is that chains in general would be relevant even when heavy pieces are absent, with only some minors. Even in simple pawn endgames, chains have their, although more limited, relevance.
Of course, a good idea would be to scale up and down the score for chains in terms of the material present, more material would be beneficial to chains, as their restricting role increases.
I do not know, I like chains, at least they are good to look upon...