LCP

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

LCP

Post by Lyudmil Tsvetkov » Mon Feb 09, 2015 10:39 am

You might think I am kidding.

This has nothing to do with the LCD, though very lot in common with KID structures.

This topic is not discussed for the first time, but maybe here an approach that is more software-friendly.

LCP is an abbreviation for longer chain pawn, i.e. it would treat positions with arising longer chains, in many cases KID structures.

One caveat is that this could concern only people who have implemented in their engines chain/defended pawn bonus.

Here some diagrams:

[d]6k1/8/8/8/3P4/2P5/8/6K1 w - - 0 1
Basic chain structure: you have one defending pawn, c3, also called the base, and one defended pawn, d4

bonus is usually given to the defended pawn, and some engines do it in terms of ranks and files

So that, basically, chain pawns fall into 2 main categories: defended pawns and defending pawns.

[d]6k1/8/8/4P3/3P4/2P5/8/6K1 w - - 0 1
However, there is one third category of chain pawns, namely the defended and defending pawn.

above, c3 is a defending pawns, e5 is a defended pawn, d4 is also a defended pawns, however it is also a defended and defending pawn at the same time, as it is defended by c3 and defends e5.
So the d4 pawn plays here more functions than one could suppose.

Bonus is usually given to the defended d4 and e5 pawns because of being defended, but as far as I know no one gives bonus for the d4 pawn for being defended and defending at the same time. And that is essential, as this is an entire new function.

Actually, defended and defending pawns appear with longer pawn chains, consisting of at least 3 pawns. I would give some 10cps bonus for the defended and defending pawn as an uniform bonus additionally to other bonus points.
And indeed, longer chains are much stronger than shorter ones, so some additional bonus is really due.

[d]6k1/8/5P2/4P3/3P4/2P5/8/6K1 w - - 0 1
However, the really useful and meaningful chains are the chains consisting of more than 3 pawns, chains of 4, 5 and 6 pawns, when one of the pawns in the latter case is a passer on the 7th rank.

Usually, the most meaningful chains are chains of 4 and 5 pawns.
Here the so called LCP (longer chain pawn) appears

c3 is a defending pawn, d4,e5 and f6 are defended pawns, d4 and e5 are also defended and defending pawns, but, please note, e5 is also a pawn defended by a defended and defending pawn, d4. So it is different. I call this the longer chain pawn and it is very powerful.

I would give such a pawn some 20 cps additional bonus.

So a longer chain pawn is basically a pawn defended by a defended and defending pawn.

[d]6k1/8/5P2/4P3/3P4/2P5/1P6/6K1 w - - 0 1
a longer chain of 5 pawns

Here b2 is defending pawn, c3, d4, e5 and f6 are defended pawns, d4 and e5 are also pawns defended by defended and defending pawns, or so called LCPs.

So I would give here 2 times 20cps bonus for the d4 and e5 pawns additionally to other bonus points. I suppose the bonus should be uniform, as ranks are not that important here, but rather the mere fact of the length of the chain.

Once you have a bonus for defended and defending pawns, and LCP pawns, you could very easily implement the concept of a chain aimed at the enemy king, the so called pointy chain, according to Carl's definition.
Such pointy chains are very useful in KID structures, where engines are very weak, precisely because they do not understand the power of the pointy chain.

I think that, even if you are unable to do LCPs, you should try using the concept to implement bonus for a pointy chain, where all engines fail.

The implementation using LCPs should be very easy and straightforward.

There are 2 conditions when you give a bonus for a pointy chain, maybe some 30cps:

1. you find an LCP pawn and that LCP pawn is a storming pawn at the same time, or
2. you find an LCP pawn that defends a storming pawn

[d]6k1/8/3p4/3Pp3/4Pp2/5PpP/6P1/6K1 w - - 0 1
here condition 1 is valid
you find an LCP pawn, f4, that is a storming pawn at the same time, so you give this pawn some 20cps bonus for being LCP plus some 30cps bonus for representing a pointy chain
This is a typical KID structure, where engines struggle precisely because they do not know what a pointy chain is

[d]6k1/2p5/2Pp4/3Pp3/4Pp2/5P2/6PP/6K1 w - - 0 1
another typical KID structure

here condition 2 is valid
you find an LCP, e5, that defends at the same time a storming pawn, f4. So you give the bonus for e5 being an LCP, as well as the bonus for a pointy chain.
Please note, that while white e4 and d5 pawns are also due an LCP bonus, the white chain points away from the enemy king, so it does not receive a pointy chain bonus. So, in spite of the fact that the black chain is shorter above than the white one, it performs better, as it additionally gets the pointy chain bonus.

How stupid you think such an idea might be?

jorose
Posts: 267
Joined: Thu Jan 22, 2015 2:21 pm
Location: Zurich, Switzerland
Full name: Jonathan Rosenthal

Re: LCP

Post by jorose » Mon Feb 09, 2015 10:58 am

Hi, unfortunately I'm on vacation so I don't have time to write longer =(

I was just curious if I understood your point correctly about a pointy pawn chain: A pointy pawn chain is a pawn chain where the end (= defended part) of the chain is near the enemy king?

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: LCP

Post by Lyudmil Tsvetkov » Mon Feb 09, 2015 11:39 am

jorose wrote:Hi, unfortunately I'm on vacation so I don't have time to write longer =(

I was just curious if I understood your point correctly about a pointy pawn chain: A pointy pawn chain is a pawn chain where the end (= defended part) of the chain is near the enemy king?
Fully correct.

Ask Carl about further details. :)

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: LCP

Post by Lyudmil Tsvetkov » Mon Feb 09, 2015 12:16 pm

Now, there is no doubt that longer pawn chains deserve an additional bonus.

Consider the belo diagram:

[d]6k1/7p/p3p1p1/1p2P3/2pP4/2P5/1P3PP1/6K1 w - - 0 1

White has 3 chain/defended pawns, c3,d4 and e5, black also has 3 defended pawns, c4,b5 and g6, so you could suppose equality here in terms of bonus points. Moreover, the pawns for both sides are on same ranks, c4 equals e5 on the 5th rank, b5 equals d4 on the 4th rank, and g6 equals c3 on the 2nd rank.

However, the white chain is much stronger than the 2 separate shorter black chains. You should somehow take account of this fact.
Besides, 4 white pawns achieve the same effect of 5 black pawns, so longer chains are very efficient.

white has one defended and defending pawn on c3, and one LCP pawn on d4 that should get additional bonus, while black has only one defended and defending pawn on b5 that should get an over bonus
so white should be better here in terms of longer chains by some 20cps

[d]6k1/8/6p1/1P2p2p/2Pp4/3P4/8/6K1 w - - 0 1
another example
the white chain is stronger than the 2 small black chains, although the respective defended pawns are placed on same ranks

So the connecting, defended and defending pawn is very important.

here you have a defended and defending pawn on c4 for white, as well as a defended pawn on b5, while black only has 2 defended pawns on d4 and h5

so white must be stronger here in terms of long chains by some 10cps, nothing special, less than in the first example, but still evident.

The conclusion from the above examples could be that, with similarly placed pawns, it is better to have one chain of 3 pawns instead of 2 separate chains of 2 pawns each, but even better to have one longer chain of 4 pawns instead of 2 separate chains of 2 and 3 pawns. So the longer the chain grows, the more visible the effect.

Concerning the pointy chain, you might try to give a small bonus for storming pawns that are defended at the same time, or to defended pawns that defend another own storming pawn at the same time, but I am afraid this could backfire on a number of occasions.
I think an LCP definition for a longer chain is essential in capturing the essence of KID structures, besides such a definition is highly unlikely to fail.

So my first try for a pointy chain bonus would be considering one of the 2 below conditions:

- LCP pawn that is a storming pawn at the same time, or
- LCP pawn that defends a stroming pawn at the same time

jorose
Posts: 267
Joined: Thu Jan 22, 2015 2:21 pm
Location: Zurich, Switzerland
Full name: Jonathan Rosenthal

Re: LCP

Post by jorose » Mon Feb 09, 2015 3:02 pm

While intuitively I agree that a longer pawn chain should give some bonus it is very hard to come up with a good concrete example.

[d]6k1/7p/p3p1p1/1p2P3/2pP4/2P5/1P3PP1/6K1 w - - 0 1

The problem with this one is that it is very imbalanced, black has 3 pawn islands to whites 1 and an argument could easily be made that white's structure would be even better with the e-pawn on e4 instead of e5.

[d]6k1/8/6p1/1P2p2p/2Pp4/3P4/8/6K1 w - - 0 1

This position doesn't illustrate your point very well either due to other imbalances namely

[d]6k1/8/8/4p3/2Pp4/3P4/8/6K1 w - - 0 1

Is also a favorable structure for white, with a protected passed pawn vs a backward pawn on an open file.

That's not to say the idea is per se bad, simply that it's hard to find good examples to illustrate the point.

I find your idea of "pointy" pawn chains interesting as I find the KID to be a very difficult opening for engines to play well. In that case I'm trying to think of other examples outside of the KID where it may occur, but it's not so easy to come up with an example.

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: LCP

Post by Lyudmil Tsvetkov » Mon Feb 09, 2015 4:06 pm

jorose wrote:While intuitively I agree that a longer pawn chain should give some bonus it is very hard to come up with a good concrete example.

[d]6k1/7p/p3p1p1/1p2P3/2pP4/2P5/1P3PP1/6K1 w - - 0 1

The problem with this one is that it is very imbalanced, black has 3 pawn islands to whites 1 and an argument could easily be made that white's structure would be even better with the e-pawn on e4 instead of e5.

[d]6k1/8/6p1/1P2p2p/2Pp4/3P4/8/6K1 w - - 0 1

This position doesn't illustrate your point very well either due to other imbalances namely

[d]6k1/8/8/4p3/2Pp4/3P4/8/6K1 w - - 0 1

Is also a favorable structure for white, with a protected passed pawn vs a backward pawn on an open file.

That's not to say the idea is per se bad, simply that it's hard to find good examples to illustrate the point.

I find your idea of "pointy" pawn chains interesting as I find the KID to be a very difficult opening for engines to play well. In that case I'm trying to think of other examples outside of the KID where it may occur, but it's not so easy to come up with an example.
Pawn islands are pawn islands, and longer chains are longer chains.
Everything has its bonus and penalty points in the eval.

A passer may be doubled, isolated, connected, etc.
Same with chain pawns.

Longer chains occur absolutely everywhere, just look more carefully at some games.
Sometimes the chains are blocked, sometimes not, and sometimes partially blocked.

In my games I get such structures from absolutely every opening, French, Ruy Lopez, English, Caro-Cann, Modern Defence, etc.
With white, or with black.
On the queen, or on the king side.

It is a matter of understanding and choice of style. But for me there is no doubt that longer chains matter.

jorose
Posts: 267
Joined: Thu Jan 22, 2015 2:21 pm
Location: Zurich, Switzerland
Full name: Jonathan Rosenthal

Re: LCP

Post by jorose » Mon Feb 09, 2015 4:53 pm

As I said, intuitively it would make sense. The only simple way to tell is to try it out!

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: LCP

Post by Lyudmil Tsvetkov » Mon Feb 09, 2015 5:35 pm

I just found the time to play some more blitz games with SF.

SF plays on 4 cores, ponder on, time control is 5 min. + 3 sec.

I managed some wins that do not quite concern us, so I post the game that I lost in a KID structure, which however was completely won for me.

[pgn][PlyCount "88"]
[Event "Blitz 5m+3s"]
[Site "Sofia"]
[Date "2015.02.09"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish 6 64 POPCNT"]
[Result "0-1"]
[ECO "A08"]
[TimeControl "300+3"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "44"]
[MLFlags "000100"]

{512MB, Dell XPS 4Cores} 1. Nf3 {[%emt 0:00:00]} 1... d5 {[%emt 0:00:34]} 2. g3
{[%emt 0:00:02]} 2... c5 {[%emt 0:00:22]} 3. Bg2 {[%emt 0:00:02]} 3... Nf6
{[%emt 0:00:17]} 4. O-O {[%emt 0:00:02]} 4... Nc6 {[%emt 0:00:12]} 5. d3
{[%emt 0:00:02]} 5... e6 {[%emt 0:00:10]} 6. Nc3 {[%emt 0:00:02]} 6... Be7
{[%emt 0:00:08]} 7. e4 {[%emt 0:00: 03]} 7... O-O {[%emt 0:00:06]} 8. Bd2
{[%emt 0:00:02]} 8... d4 {[%emt 0:00:20]} 9. Ne2 {[%emt 0:00:03]} 9... e5
{[%emt 0:00:07]} 10. h3 {[%emt 0:00:03]} 10... Nd7 {[%emt 0:00:13]} 11. Ne1
{[%emt 0:00:06]} 11... b6 {[%emt 0:00:07]} 12. f4 {[%emt 0:00:02]} 12... Ba6
{[%emt 0:00:08]} 13. b3 {[%emt 0:00:04]} 13... Bd6 {[%emt 0:00:12]} 14. f5
{[%emt 0: 00:03]} 14... h6 {[%emt 0:00:07]} 15. g4 {[%emt 0:00:01]} 15... Bb7
{[%emt 0:00:22]} 16. Nf3 {[%emt 0:00:06]} 16... Be7 {[%emt 0:00:00]} 17. a3
{[%emt 0:00:10]} 17... b5 {[%emt 0: 00:12]} 18. Qe1 {[%emt 0:00:09]} 18... c4
{[%emt 0:00:08]} 19. h4 {[%emt 0:00:03]} 19... Kh8 {[%emt 0:00:11]} 20. g5
{[%emt 0:00:02]} 20... Rg8 {[%emt 0:00:09]} 21. Qg3 {[%emt 0:00:04]} 21... Qe8
{[%emt 0:00:07]} 22. Nh2 {[%emt 0:00:31]} 22... f6 {[%emt 0:00: 21]} 23. g6
{[%emt 0:00:11]} 23... Qf8 {[%emt 0:00:23]} 24. Qg4 {[%emt 0:00:34]} 24... cxd3
{[%emt 0:00:34]} 25. cxd3 {[%emt 0:00:14]} 25... Nb4 {[%emt 0:00:00]} 26. axb4
{[%emt 0:00:11]} 26... Bxb4 {[%emt 0:00:03]} 27. Bc1 {[%emt 0:00:34]} 27... Nc5
{[%emt 0:00: 11]} 28. Qh5 {[%emt 0:00:35]} 28... Nxb3 {[%emt 0:00:00]} 29. Bxh6
{[%emt 0:00:24]} 29... gxh6 {[%emt 0:00:01]} 30. Rad1 {[%emt 0:00:06]} 30... Bd2
{[%emt 0:00:26]} 31. Rxd2 {[%emt 0:00:09]} 31... Nxd2 {[%emt 0:00:00]} 32. Ng4
{[%emt 0:00:03]} 32... Nxf1 {[%emt 0: 00:04]} 33. Bxf1 {[%emt 0:00:59]} 33...
Rc8 {[%emt 0:00:00]} 34. Nxf6 {[%emt 0:00:18]} 34... Rc7 {[%emt 0:00:00]} 35.
Nxg8 {[%emt 0:00:45]} 35... Kxg8 {[%emt 0:00:04]} 36. Ng3 {[%emt 0:00:02]} 36...
b4 {[%emt 0:00:03]} 37. Qg4 {[%emt 0:00:04]} 37... Bc6 {[%emt 0:00: 06]} 38. Nh5
{[%emt 0:00:04]} 38... b3 {[%emt 0:00:11]} 39. f6 {[%emt 0:00:01]} 39... Be8
{[%emt 0:00:03]} 40. f7+ {[%emt 0:00:02]} 40... Bxf7 {[%emt 0:00:02]} 41. gxf7+
{[%emt 0:00:01]} 41... Kxf7 {[%emt 0:00:05]} 42. Qf5+ {[%emt 0:00:03]} 42... Kg8
{[%emt 0: 00:06]} 43. Qg6+ {[%emt 0:00:04]} 43... Kh8 {[%emt 0:00:04]} 44. Bh3
{[%emt 0:00:06]} 44... b2 {[%emt 0:00:05]} 0-1
[/pgn]

[d]r1bq1rk1/pp2bppp/2n2n2/2p1p3/3pP3/3P1NP1/PPPBNPBP/R2Q1RK1 w - - 0 10
This is some kind of King's Indian attack that only me and SF play.

[d]r2q1r1k/pb1nbpp1/2n4p/1p2pP2/2ppP1PP/PP1P1N2/2PBN1B1/R3QRK1 w - - 0 20
Both sides attack, but white attacks on the more important king side of the board.
Also look at the powerful c2-f5 white chain, pointed towards the enemy king.

[d]r4qrk/pb1nb1p1/2n2pPp/1p2pP2/2ppP1QP/PP1P4/2PBN1BN/R4RK1 b - - 0 24
On move 24 SF is completely lost.
White threatens Qh5, followed by Ng4, sacrifice on h6 with mate. Against which there is no defence.

[d]r4qrk/pb1nb1p1/5pPp/1p2pP2/1n1pP1QP/PP1P4/3BN1BN/R4RK1 w - - 0 26
SF sacrifices a knight with Nb4, which I, inexplainably why, capture. Following the abovementioned plan with Qh5 and Ng4 still mates.
I can not explain how I could have lost that even after I missed the easy best move. But when I play with 5 instead of 2 minutes, I apply strictly the time control, so something threw me off balance, presumably the joy I will score a good win against SF, as I still had sufficient time.

The rest of the game is not important. Probably I could have won a couple of times again along the way, but at the end it was SF that triumphed.

My point is that, look at the above position.

[d]r4qrk/pb1nb1p1/2n2pPp/1p2pP2/2ppP1QP/PP1P4/2PBN1BN/R4RK1 b - - 0 24

What kind of position is this to play for an engine the level of SF?
Sf simply plays very bad KID structures, as all other engines, and it completely misses the tremendous power of the c2-g6 pointy chain, that has already grown to a 5-member chain.
That chain is worth gold, as it prevents the enemy counterplay on the one hand, and helps with the own attack, on the other.
So the pointy chain is a very prophylactic, very positional and very true concept.

Damned time!
If it had not been for scrupulously aplying the time control, I would have won that game. Or, if the game was at longer time control.

Nevermind.

I have already abandoned all hope some engine might possibly implement a longer/pointy chain concept, although this could earn the respective engine a lot of elo and a nice positional style.

carldaman
Posts: 1717
Joined: Sat Jun 02, 2012 12:13 am

Re: LCP

Post by carldaman » Mon Feb 09, 2015 7:32 pm

Lyudmil Tsvetkov wrote:
jorose wrote:Hi, unfortunately I'm on vacation so I don't have time to write longer =(

I was just curious if I understood your point correctly about a pointy pawn chain: A pointy pawn chain is a pawn chain where the end (= defended part) of the chain is near the enemy king?
Fully correct.

Ask Carl about further details. :)
Well, I also like to stipulate that the defending King be on the shorter (unshielded) side of the board relative to the pointy chain. For example, with a c3-d4-e5-f6/f7-e6-d5 chain, the Black King is just as close to f6 on e8 as it is on g8, but the latter is far less safe since it's on the short, exposed side of the chain (in other words, the chain is pointing towards the King) - whereas on e8 the King is shielded by the chain itself.

Please note that I like to take into account both the White AND corresponding Black pawns that form the chain (c3-d4-e5-f6/f7-e6-d5 in this example) , which really make up a compound unit together. I prefer the term diagonal pawn barrier to refer to these chains taken together as one unit.

Hope this distinction all makes sense. :)

Regards,
CL

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: LCP

Post by Lyudmil Tsvetkov » Mon Feb 09, 2015 8:38 pm

Next time I play the same setup against SF, I will not miss it.

[pgn][PlyCount "69"]
[Event "Blitz 5m+3s"]
[Site "Sofia"]
[Date "2015.02.09"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish 6 64 POPCNT"]
[Result "1-0"]
[ECO "A07"]
[TimeControl "300+3"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "34"]
[MLFlags "000100"]

{512MB, Dell XPS 4Cores} 1. Nf3 {[%emt 0:00:00]} 1... Nf6 {[%emt 0:00:11]} 2.
g3 {[%emt 0:00:02]} 2... d5 {[%emt 0:00:33]} 3. Bg2 {[%emt 0:00:03]} 3... e6
{[%emt 0:00:08]} 4. O-O {[%emt 0:00:03]} 4... Be7 {[%emt 0:00:02]} 5. d3
{[%emt 0:00:03]} 5... O-O {[%emt 0:00:12]} 6. Nc3 {[%emt 0:00:03]} 6... c5
{[%emt 0:00:08]} 7. e4 {[%emt 0:00:02]} 7... Nc6 {[%emt 0:00:05]} 8. Bd2
{[%emt 0:00:03]} 8... d4 {[%emt 0:00:10]} 9. Ne2 {[%emt 0:00:03]} 9... e5
{[%emt 0:00:08]} 10. h3 {[%emt 0:00:03]} 10... Nd7 {[%emt 0:00:10]} 11. Ne1
{[%emt 0:00:05]} 11... b6 {[%emt 0:00:09]} 12. f4 {[%emt 0:00:03]} 12... Ba6
{[%emt 0: 00:10]} 13. b3 {[%emt 0:00:06]} 13... f6 {[%emt 0:00:07]} 14. f5
{[%emt 0:00:02]} 14... Bb7 {[%emt 0:00:07]} 15. g4 {[%emt 0:00:02]} 15... a6
{[%emt 0:00:43]} 16. h4 {[%emt 0: 00:05]} 16... Rb8 {[%emt 0:00:11]} 17. Nf3
{[%emt 0:00:05]} 17... b5 {[%emt 0:00:14]} 18. Ng3 {[%emt 0:00:25]} 18... c4
{[%emt 0:00:08]} 19. Kh1 {[%emt 0:00:13]} 19... c3 {[%emt 0: 00:14]} 20. Bc1
{[%emt 0:00:03]} 20... Nc5 {[%emt 0:00:37]} 21. a3 {[%emt 0:00:04]} 21... h6
{[%emt 0:00:10]} 22. Bh3 {[%emt 0:00:17]} 22... Nd7 {[%emt 0:00:21]} 23. Qe2
{[%emt 0:00:06]} 23... Kh8 {[%emt 0:00:10]} 24. Nh5 {[%emt 0:00:10]} 24... Qe8
{[%emt 0:00: 05]} 25. Rg1 {[%emt 0:00:03]} 25... Qf7 {[%emt 0:00:17]} 26. Ng3
{[%emt 0:00:02]} 26... Bd6 {[%emt 0:00:17]} 27. Rg2 {[%emt 0:00:27]} 27... Rg8
{[%emt 0:00:08]} 28. g5 {[%emt 0: 00:01]} 28... Bf8 {[%emt 0:00:17]} 29. g6
{[%emt 0:00:21]} 29... Qe7 {[%emt 0:00:00]} 30. Nh2 {[%emt 0:00:12]} 30... b4
{[%emt 0:00:12]} 31. a4 {[%emt 0:00:04]} 31... Rd8 {[%emt 0: 00:07]} 32. Qh5
{[%emt 0:00:08]} 32... Bc8 {[%emt 0:00:04]} 33. Ng4 {[%emt 0:00:04]} 33... Nc5
{[%emt 0:00:00]} 34. Nxh6 {[%emt 0:01:18]} 34... gxh6 {[%emt 0:00:03]} 35. Bxh6
{[%emt 0:00:01]} 1-0
[/pgn]
My conclusion is that SF did not budge an inch from SF 5 in its understanding of closed positions, KID structures and pointy chains.

Post Reply