Longer chain pawn again

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

Longer chain pawn again

Post by Lyudmil Tsvetkov » Wed Feb 25, 2015 4:58 am

I see SF managed to successfully implement one more element of chain evaluation.

That is great, and bodes well in terms of positional play.

However, going back to longer chains, the so called longer chain pawn, not long ago I posted such a thread, anyone aving a clue what happens if a longer chain pawn (basically a pawn defended by a pawn and defending another own pawn at the same time) is scored not uniformly, as I suggested initially, being afraid of possible redundancies, but in terms of files and ranks, using the connected pawns/candidate array?

I think this is the working approach, as it is important where precisely the pawn is to be found, just as with all other pawns.

[d]6k1/8/8/8/3P4/2P5/1P6/6K1 w - - 0 1
We give here additional bonus to c3, a defended and defending pawn at the same time, or otherwise a longer chain pawn. c3 is defended by b2 and defends d4

[d]6k1/8/8/3P4/4P3/5P2/6P1/6K1 w - - 0 1
We give here additional bonus to the f3 and e4 pawns - longer chain pawns. Both are defended and defending pawns at the same time. f3 is defended by g2 and defends e4, while e4 is defended by f3 and defends d5

[d]6k1/8/5P2/4P3/3P4/2P5/1P6/6K1 w - - 0 1
We give here additional bonus to c3,d4 and e5 pawns - defended and defending pawns, longer chain pawns. c3 is defended by b2 and defends d4, d4 is defended by c3 and defends e5, and e5 is defended d4 and defends f6

I think, based on the connected pawns/candidates array, longer chain pawns might score somewhere between 1/2 to 1/10 the value of a connected pawn, who knows?

Longer chain pawns are relatively rare, even rarer than so called apex pawns (a pawn defended by 2 own pawns), but really much more important, so I think a good engine should have those implemented.

What do you think of this kind of implementation?
Anyone of the SF team reading this? Joerg? :D

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

Re: Longer chain pawn again

Post by Lyudmil Tsvetkov » Wed Feb 25, 2015 5:04 am

Just to mention briefly that longer chain pawns and pointed chains are 2 completely different things.

Chain eval has a multitude of different elements.

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

Re: Longer chain pawn again

Post by Lyudmil Tsvetkov » Wed Feb 25, 2015 12:58 pm

OK, no one interested in longer chain pawns, people like to mostly tune features like mobility, attacks and isolated pawns to death.

I do not know how to implement this, I am almost certain a file and rank based approach will work, after sufficient tuning.

What there is absolutely no doubt however, is that those pawns connecting longer chains are a very real and very sound eval term with undoubted added value.

Longer chains represent considerable advantage, mainly because of 2 factors:

- they are much more efficient in restraining enemy pieces than smaller chains, and
- they are also mobile in a way that does not quite undermine their solidity

So that, no matter what Arjun might think, longer chains are more than the sum of their parts, and definitely deserve a bonus apart.

OK, next time I am going to post about some major clone issue, or even better, a poll about the validity of certain mod decisions.
That certainly is going to attract crowds of people. :shock:

nionita
Posts: 161
Joined: Fri Oct 22, 2010 7:47 pm
Location: Austria

Re: Longer chain pawn again

Post by nionita » Wed Feb 25, 2015 8:30 pm

Lyudmil Tsvetkov wrote:no one interested in longer chain pawns
Lyudmil, I would really try this idea in my engine (Barbarossa) but my pawn evaluation is still rudimentar (passed, doubled, isolated pawns, but not even backward pawns yet). So I guess you are expecting someone with a stronger engine to try it.

Do you think that after backward pawns this could be the next thing to try? If yes, I could try to implement those 2 next, but it will for sure need 2-3 weeks until it is tuned and I can report elo differences...

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

Re: Longer chain pawn again

Post by Lyudmil Tsvetkov » Thu Feb 26, 2015 9:27 am

nionita wrote:
Lyudmil Tsvetkov wrote:no one interested in longer chain pawns
Lyudmil, I would really try this idea in my engine (Barbarossa) but my pawn evaluation is still rudimentar (passed, doubled, isolated pawns, but not even backward pawns yet). So I guess you are expecting someone with a stronger engine to try it.

Do you think that after backward pawns this could be the next thing to try? If yes, I could try to implement those 2 next, but it will for sure need 2-3 weeks until it is tuned and I can report elo differences...
Many thanks, Nicu!

I am delighted by your interest.

Yeah, I was specifically targeting SF developers, but they would not even look here, whenever they see the word chain, they quickly switch off their browsers. :shock:

I have absolutely no clue what will work and what not in an engine, it should be a matter of tuning, you know this better.

What I certainly know is that the term is based on solid chess knowledge, and that usually longer chains, at least in top human chess, are used by the stronger players.

Weaker human chess players tend to rely more on terms like isolated and doubled pawns, and I suppose it is the same with engines.

Which does not really mean of course that an engine that is not quite the strength of SF would be unable to implement this term...

Many thanks again for the interest, will be expecting some feedback if available.

User avatar
lantonov
Posts: 216
Joined: Sun Apr 13, 2014 3:19 pm

Re: Longer chain pawn again

Post by lantonov » Fri Feb 27, 2015 2:47 pm

Lyudmil Tsvetkov wrote:OK, no one interested in longer chain pawns, people like to mostly tune features like mobility, attacks and isolated pawns to death.

OK, next time I am going to post about some major clone issue, or even better, a poll about the validity of certain mod decisions.
That certainly is going to attract crowds of people. :shock:
Stefan Geschwentner has scheduled a test in two variants implementing your idea for giving bonus to the inner pawns of a longer chain.

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

Re: Longer chain pawn again

Post by Lyudmil Tsvetkov » Fri Feb 27, 2015 3:25 pm

lantonov wrote:
Lyudmil Tsvetkov wrote:OK, no one interested in longer chain pawns, people like to mostly tune features like mobility, attacks and isolated pawns to death.

OK, next time I am going to post about some major clone issue, or even better, a poll about the validity of certain mod decisions.
That certainly is going to attract crowds of people. :shock:
Stefan Geschwentner has scheduled a test in two variants implementing your idea for giving bonus to the inner pawns of a longer chain.
I am following it quite closely, Lyudmil, thanks.

It would be as if I would miss the final match of the Football World Cup. :)

I hope Stefan, a very experienced programmer, will also try some smaller values, like 1/12, 1/16, it seems that smaller does better until now.

Thumbs up.

User avatar
lantonov
Posts: 216
Joined: Sun Apr 13, 2014 3:19 pm

Re: Longer chain pawn again

Post by lantonov » Fri Feb 27, 2015 11:02 pm

Lyudmil Tsvetkov wrote:

I hope Stefan, a very experienced programmer, will also try some smaller values, like 1/12, 1/16, it seems that smaller does better until now.

Thumbs up.
Do you have this in mind ?

if(innerLongChain & s)
score += Connected[opposed][phalanx][relative_rank(Us, s)] / 8;

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

Re: Longer chain pawn again

Post by Lyudmil Tsvetkov » Fri Feb 27, 2015 11:51 pm

lantonov wrote:
Lyudmil Tsvetkov wrote:

I hope Stefan, a very experienced programmer, will also try some smaller values, like 1/12, 1/16, it seems that smaller does better until now.

Thumbs up.
Do you have this in mind ?

if(innerLongChain & s)
score += Connected[opposed][phalanx][relative_rank(Us, s)] / 8;
That is his current attempt that jsut failed (we lost the final match ) :(

I meant same as above, but /16

Lyudmil, can you ask him on github to possibly push such a patch.
Temporarily, I do not have/remember my password for there.

I should probably renew my password there, but, having a bit of experience from lst year, I know it is no good to bother the programmers.

They usually know best, push some preliminary tests, etc.

Anyway, would be grateful if you could ask him.

It is a matter of implementation, so that redundancies are avoided.
A couple of attempts might fail, but the term itself is useful, very useful, I would say.

User avatar
lantonov
Posts: 216
Joined: Sun Apr 13, 2014 3:19 pm

Re: Longer chain pawn again

Post by lantonov » Sat Feb 28, 2015 12:26 am

Lyudmil Tsvetkov wrote:
That is his current attempt that jsut failed (we lost the final match ) :(

I meant same as above, but /16

Lyudmil, can you ask him on github to possibly push such a patch.
Temporarily, I do not have/remember my password for there.

I should probably renew my password there, but, having a bit of experience from lst year, I know it is no good to bother the programmers.

They usually know best, push some preliminary tests, etc.

Anyway, would be grateful if you could ask him.

It is a matter of implementation, so that redundancies are avoided.
A couple of attempts might fail, but the term itself is useful, very useful, I would say.
Sure, I will ask him on your behalf, and maybe he will listen, because you are an authority on chess matters. BTW, the FishTest forum (FishCooking) is a more direct way than TalkChess for suggestions to SF programmers. It is watched constantly by everyone who matters in SF testing. If you don't object (no adverse experience?) you can post there, and your excellent knowledge and experience will be implemented faster and fuller. But it is up to you to decide ...

Ok, going to github now.

Post Reply