Push a patch

[Lyudmil Tsvetkov]

I will try my chance with posting one last message - after all I have already retired - although I see very well Roger in green overseeing the forum.

I managed to repel all nice SF contributors ready to push some patches from this forum with my consistent patch-pushing urges.

There was a very nice guy called Arjun Temurnikar - he used to visit this place regularly in the past, but not anymore. I think he is still afraid that, as soon as I take a glimpse of him, I will ask him to push a patch on blocked on 6th.

Same happened to Joerg - he pushed so any patches pursuant to my insistent pleas, that, out of fear that I might ask him to push another patch, he never posts in my threads, preferring to haunt the programming section instead...

Also Lyudmil, my namesake - he was so kind to push some patches for me, but, as soon as their number menacingly grew, he altogether refuses to visit this place.

In spite of that, I will take my chance, and post my values for some SF terms, in no one volunteers to push a patch, then maybe this could be tried in the future, or some people might look at the values.

One way or another, I have come to the conclusion by following closely SF games that SF does not have a single parameter tuned properly, so only on tunig existing parameters SF could gain some 200-500 elo and even more. Meaning, not perfect tuning, but good tuning.

The real problem is that, as SF parameters are all badly tuned, but tuned to the whole, when you tune a single one better, everything breaks, so only infinitesimal steps could be possible.

I know how to tune well many terms starting from scratch, but I have absolutely no clue how to tune 5 terms that are badly tuned separately, but as a whole produce a somewhat working union. The union could be considerably improved from scarcth, but I do not know how to do this with each of the terms badly tuned without breaking the whole.

In any case, a try is always advisable.

Btw., one of the reasons Komodo possibly scales better than SF is that quite certainly the Komodo values for different terms are better tuned than SF terms. Larry should have overseen the whole tuning process, while in SF some parameter values are really fully surrrealistic.

As said, knowledge scales well with long TC, and good tuning is an essential part of knowledge.

[Lyudmil Tsvetkov]

These are my corrected values for the king psqt. I changed only the mg values, as in the eg SF seems to perform much better than in the past, quite probably before of the not long ago introduced term KingonPawn, that makes the king to go after enemy pawns, which usually also involves activisation and space advantage.

SO I am very happy with the eg values, but the mg king psqt values need a general overhaul. The are not disastrous, but could certainly be considerably improved.

S(308, 27), S(332, 81), S(268,108), S(215,116), S(215,116), S(268,108), S(332, 81), S(308, 27),
S(282, 74), S(321,128), S(252,155), S(200,163), S(200,163), S(252,155), S(321,128), S(282, 74),
S(220,111), S(250,165), S(190,192), S(140,200), S(140,200), S(190,192), S(250,165), S(220,111),
S(180,135), S(210,189), S(145,216), S(90,224), S(90,224), S(145,216), S(210,189), S(180,135),
S(140,135), S(170,189), S(110,216), S(70,224), S(70,224), S(110,216), S(170,189), S(140,135),
S(100,111), S(80,165), S(60,192), S( 40,200), S( 40,200), S(60,192), S(80,165), S(100,111),
S(70, 74), S(60,128), S( 50,155), S( 30,163), S( 30,163), S( 50,155), S(60,128), S(70, 74),
S( 50, 27), S(40, 81), S( 30,108), S( 20,116), S( 20,116), S( 30,108), S(40, 81), S( 50, 27)


I would be very happy if someone is able to push a patch on this, you never know, I almost see Arjun returning.

PS. Gosh, I forgot about Vince - he also quit this forum because of me.

[Lyudmil Tsvetkov]

These are my long-thought SF piece values.

They should already pass, as far as it is humanly possible for a similar patch to pass with existing SF ludicrously-looking piece values.

As said, I would have liked to do much more, but having regard to the available values, I do not think much more could be squeezed from that.

PawnValueMg = 190, PawnValueEg = 250,
KnightValueMg = 812, KnightValueEg = 838,
BishopValueMg = 841, BishopValueEg = 865,
RookValueMg = 1278, RookValueEg = 1294,
QueenValueMg = 2517, QueenValueEg = 2554,


As said, this is for volunteers, whenever they would feel like pushing a patch.

[Lyudmil Tsvetkov]

And here my long-thought rook psqt values.

I made non-flat both the mg and eg values, hopefully this passes.

S(-10, -10), S(-7, -7), S(-2, -2), S(0, 0), S(0, 0), S(-2, -2), S(-7, -7), S(-10, -10),
S(-7, -9), S( -4, -6), S(-1, -1), S(3, 3), S(3, 3), S(-1, -1), S( -4, -6), S(-7, -9),
S(0, -7), S( 3, -5), S( 6, 1), S(8, 4), S(8, 4), S( 6, 1), S( 3, -5), S(0, -7),
S(-5, -5), S( -1, -1), S( 2, 2), S(5, 5), S(5, 5), S( 2, 2), S( -1, -1), S(-5, -5),
S(-3, -3), S( 2, 2), S( 4, 4), S(6, 6), S(6, 6), S( 4, 4), S( 2, 2), S(-3, -3),
S( 0, 0), S( 3, 3), S( 6, 6), S(8, 8), S(8, 8), S( 6, 6), S( 3, 3), S( 0, 0),
S( 3, 3), S( 7, 7), S( 11, 11), S(15,15), S(15,15), S( 11, 11), S( 7, 7), S( 3, 3),
S( 1, 1), S( 5, 5), S( 9, 9), S(13, 13), S(13, 13), S( 9, 9), S( 5, 5), S( 1, 1)

So far not a single patch has passed when I wanted it to be so, but just occasionally, maybe this will be the first one.

Lyudmil, are you still around?

[Lanzo]

If I was miraculously endowed with the ability to reverse anything I wanted, I would not reverse my birth and nor I would reverse the Planck Epoch. Instead, I would reverse the parlous departure of our very own Dr. Tsvetkov, whose ideas edified computer chess everlastingly.

[Lyudmil Tsvetkov]

Concerning piece values, I do not know how this could be implemented in SF with its current values, and in the case that you would like to do the most without applying any other additional rules, when starting from scratch the following rules should be true:

- if a pawn is worth 100cps, a knight might be worth 310cps, so more than 3 pawns; I know that in modern engines pawns are scored even lower, somewhere 1 to 4 to a minor piece, but that is mainly due to factors like scoring advanced passers and similar advanced pawn features, if those are not considered, the value of a knight should be somewhere around 310 cps

- the bishop would be worth around 330cps, those are mg values

When we come to the rook and queen values, the following should be more or less true:

- 2 rooks have higher value than a queen, maybe by 2-3%
- 3 average minors=(bishop value + knight value)/2x3 have even higher value than a queen, by maybe 3-5%
- thus, 3 average minors have negligeably higher value than 2 rooks

From this, you could easily compute the rook and queen values for the mg.

Going into the eg, the following rules should be true:

- knight values slightly decreases, by maybe 1-2%
- bishop values increases a bit steeper, by maybe 2-3%
- rook value increases even steeper than bishop value, by maybe 4-5%
- queen values definitely increases, by maybe 3-4%

- for me, if it was not for specific passer features, the pawn value should stay almost the same as in the mg, maybe just a very negligeable increase, by 2-3%, but, as advanced passer features are taken into accoiunt in engines, they score eg pawn value much higher

So, well so good, you can compute everything easily from the above rules, provided they are true, which I think they are, and perfectly apply them in an engine started from scratch, but, is anybody able to tell me what to do in an engine like SF with already very wrongly tuned piece values, and well-tuned at that?

Personally, I have absolutely no clue, as in SF almost none of the above rules is followed, and yet the engine seems to play chess, at least it makes legal moves...

[Lyudmil Tsvetkov]

If I was miraculously endowed with the ability to reverse anything I wanted, I would not reverse my birth and nor I would reverse the Planck Epoch. Instead, I would reverse the parlous departure of our very own Dr. Tsvetkov, whose ideas edified computer chess everlastingly.

Only thing I am still sorry for is that Roger still has not suspended you from this forum - you deserved this very well: you offended Larry, you offend me and everybody alike, you post nonsense, this message of yours is another very clear evidence all you want is to disturb threads and people's peace of mind.

I really do not know why you still have not been suspended: did not anyone already file a complaint woth the mods against you?
Louis, did not you do so?
None of the Komodo team?
Or the SF team?

I can perfectly live with absolutely any other person posting on this forum, absolutely anyone, including the ones I strongly disagree with, but I can not live with you.

Again, if the mods do not supsend your registration, this will be BIG mistake.

[Joerg Oster]

These are my long-thought SF piece values.

They should already pass, as far as it is humanly possible for a similar patch to pass with existing SF ludicrously-looking piece values.

As said, I would have liked to do much more, but having regard to the available values, I do not think much more could be squeezed from that.

PawnValueMg = 190, PawnValueEg = 250,
KnightValueMg = 812, KnightValueEg = 838,
BishopValueMg = 841, BishopValueEg = 865,
RookValueMg = 1278, RookValueEg = 1294,
QueenValueMg = 2517, QueenValueEg = 2554,


As said, this is for volunteers, whenever they would feel like pushing a patch.

Tuning patches seem to have a hard time in the framework these days ...
Nevertheless, I pushed this patch for you.

[Isaac]

Lyudmil, do you wish a single patch to tune the pieces values+psqt tables?
Or 1 patch for the pieces values + 1 patch for the psqt, etc?
I ask this because Joerg is currently trying your values for the piece values but nothing about your psqt values for king middle game nor rook middle game-end game.

[Lyudmil Tsvetkov]

These are my long-thought SF piece values.

They should already pass, as far as it is humanly possible for a similar patch to pass with existing SF ludicrously-looking piece values.

As said, I would have liked to do much more, but having regard to the available values, I do not think much more could be squeezed from that.

PawnValueMg = 190, PawnValueEg = 250,
KnightValueMg = 812, KnightValueEg = 838,
BishopValueMg = 841, BishopValueEg = 865,
RookValueMg = 1278, RookValueEg = 1294,
QueenValueMg = 2517, QueenValueEg = 2554,


As said, this is for volunteers, whenever they would feel like pushing a patch.

Tuning patches seem to have a hard time in the framework these days ...
Nevertheless, I pushed this patch for you.

Many thanks, Joerg!

Of course, now the failure will be common.