How powerful is the Queen compared to 2 Rooks ?

Discussion of computer chess matches and engine tournaments.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Chessqueen
Posts: 630
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Chessqueen » Thu Jan 31, 2019 4:18 pm

Nordlandia wrote:
Thu Jan 31, 2019 3:49 pm
S-Chess stockfish fork.

PawnValueMg = 166, PawnValueEg = 245,
KnightValueMg = 801, KnightValueEg = 772,
BishopValueMg = 893, BishopValueEg = 856,
RookValueMg = 1313, RookValueEg = 1261,
HawkValueMg = 1954, HawkValueEg = 2172,
ElephantValueMg = 2060, ElephantValueEg = 2556,
QueenValueMg = 2198, QueenValueEg = 2617,


Average value Q= 2407,5 : 265 ~ 9
Average value C= 2308 : 265 ~ 8,70
Average value H= 2063 : 265 ~ 7.785
Average value R= 1287 : 265 ~ 4.9
Average value B 874,5 : 265 ~ 3.3
Average value N 786,5 : 265 ~ 3
Average value P 205,5 : 265 ~ 0.8
When were these pieces values implemented into SF? Are they recent values? Are the value of the pieces equivalent to SF in the latest Komodo version?

User avatar
Nordlandia
Posts: 2471
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Nordlandia » Thu Jan 31, 2019 4:45 pm

Chessqueen: listed values is from a Seirawan Chess stockfish engine (aka fairy stockfish)

Code: Select all

https://github.com/ianfab/Seirawan-Stockfish/blob/dad6e2154ba48f2ff9427fc3923ea2282b74f9e1/src/types.h#L207-L209
Recipe: Middlegame value + Endgame value divided by 2. Then divide the average by appropriate factor. 265 is good but not optimal number.

The number 265 is not optimal but it gives a rough approximation.

lkaufman
Posts: 3739
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: How powerful is the Queen compared to 2 Rooks ?

Post by lkaufman » Thu Jan 31, 2019 5:32 pm

hgm wrote:
Thu Jan 31, 2019 9:07 am
I think that in the Knight-odds example the leveling effect is masked by an even more general principle: when ahead in material, simplify by trading pieces (but conserve Pawns). This principle even applies in cases where all pieces are equal (so that there cannot be any leveling effect), e.g. 4 Knights vs 3. In the usual framework of opening/end-game evaluation interpolated by game phase, the latter measured as total material, this would be implemented as a hefty (but proportionl) increase of all piece values towards the end-game. Intuitively it would make sense to express material avantages as a fraction (white - black)/(white + black) (where even the King could then count for some), but the division is a bit awkward, so we usually take a linear growth like (white - black) - C*(white + black)*(white-black), (where for simplicity I assumed that the weights for determining the game phase (white + black) are the same as the piece values, and all values decrease in the same proportion C towards the opening). The second-order term then is C*(black*black - white*white), i.e. it mainly contains interactions between own pieces. But it doesn't really discriminate between piece types, other than weighting those by their 'intrinsic' value. Basically it says: if you have many pieces, every one of those is somewhat redundant (white army strength = white - C*white*white). No matter how different they move.

Such a 'type-blind' general redundance (as an approximation to expressing the advantage as a relative quantity) should be enough to explain why trading Queens in the face of Knight odds is a bad idea. Sure, the leveling effect was making your own Queen a bit less valuable than his, because she was facing that extra Knight. But that did not take account of the fact that you needed that Queen much harder, because she also had to perform the tasks that otherwise could have been done by that missing Knight. Or, equivalently, the opponent's Queen was to some extent redundant because that extra Knight could sometimes releave her.

So perhaps I should be more precise in my statement: rather than saying that I don't believe in redundancy, I should say I do not believe in type-specific, but think that redunancy should be type-blind, and just express that it is the relative material advantage that matters. But then it can be a pretty big effect. With 6:3:2 weighting of Q:R:minor, the start total of a Knight-odds game would have 38 points, and trading Queens would reduce it to 26. That would be a 46% increase of the relative piece values, so 1.5 Pawn for a Knight's advantage. This will be much larger than the leveling effect of a single Knight on a Queen.
The problem is that all attempts in Rybka, Stockfish, and Komodo to have piece values scale up dramatically towards the endgame have tested poorly. It is pretty clear that best results are obtained with a knight value in the endgame only slightly higher than in the opening, nothing like what your example would suggest. I don't know the reason that such sensible scaling fails.
Komodo rules!

User avatar
hgm
Posts: 23772
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: How powerful is the Queen compared to 2 Rooks ?

Post by hgm » Thu Jan 31, 2019 6:49 pm

I guess that this is because not only the piece values have to be scaled up, but basically everything. The only thing that such a proportional scaling would do is encourage advancing the game phase when you are ahead. (And the more so the more you are ahead.) I suppose this is good. We just agreed that trading Queens when a Knight behind would be a bad thing, and it would be very strange indeed if telling the engine this through penalizing it in the eval would make it weaker.

I guess there is a small caveat here: if the advantage is below the draw margin, simplifying might get you closer to a draw rather than closer to a win. So perhaps it would be better to have something like this:

rawEval = opening + (1 - gamePhase)*(endGame - opening);
eval = rawEval*(1 + (1-gamePhase)*sign(rawEval - 1.5 Pawn));

with the end-game eval roughly similar in magnitude to the opening eval. One further rfinment could be to smooth the sign function a little, e.g. make it a tanh function.

Dann Corbit
Posts: 10186
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Dann Corbit » Thu Jan 31, 2019 8:02 pm

What is S-chess?
Is that suicide chess?
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

User avatar
Nordlandia
Posts: 2471
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Nordlandia » Thu Jan 31, 2019 8:41 pm

Dann Corbit wrote:
Thu Jan 31, 2019 8:02 pm
What is S-chess?
Is that suicide chess?
Seirawan chess. It is played on the standard 8×8 board and uses two new pieces, the hawk (which moves like a knight or a bishop) and the elephant (which moves like a knight or a rook). Source: wikipedia

Seirawan Chess is playable through this website: https://schess.org/#/

Dann Corbit
Posts: 10186
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Dann Corbit » Thu Jan 31, 2019 8:55 pm

Nordlandia wrote:
Thu Jan 31, 2019 8:41 pm
Dann Corbit wrote:
Thu Jan 31, 2019 8:02 pm
What is S-chess?
Is that suicide chess?
Seirawan chess. It is played on the standard 8×8 board and uses two new pieces, the hawk (which moves like a knight or a bishop) and the elephant (which moves like a knight or a rook). Source: wikipedia

Seirawan Chess is playable through this website: https://schess.org/#/
To me, the names don't fit with the other chessmen. (Though they do work with the older names, perhaps).

I would have called the Hawk a Templar, because they were religious knights.
I would have called the Elephant a Garter, because the Order of the Garter guard the castle.
;-)
And besides, it would be funny to say, "I've got your Garter!"
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Chessqueen
Posts: 630
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Chessqueen » Thu Jan 31, 2019 10:20 pm

Chessqueen wrote:
Tue Jan 29, 2019 5:08 pm
lkaufman wrote:
Mon Jan 28, 2019 4:48 pm
With no minor pieces on the board and an average number of pawns (five per side), two rooks are nearly a pawn better than the queen. But on a full board the queen is indeed better than two rooks, as GM Roman Dzindzichashvilli taught me before computers were strong enough to answer the question. On average thruout the game the queen is a bit weaker.
What apply to two GMs does NOT hold true with 2 two powerful programs of equal strength like Komodo vs Komodo or SF10 vs SF10 or Komodo vs SF10. I believe that a Queen can find too many squares to keep the two Rooks busy, but for two Humans GMs is probably a little bit harder to find all the best moves or squares to force a draw. Therefore, there are a lot of pieces situations where what we believe true for two GMs, nowadays very strong programs are proving it to be wrong :shock:


Dann Corbit
Posts: 10186
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Dann Corbit » Thu Jan 31, 2019 10:28 pm

Chessqueen wrote:
Thu Jan 31, 2019 10:20 pm
Chessqueen wrote:
Tue Jan 29, 2019 5:08 pm
lkaufman wrote:
Mon Jan 28, 2019 4:48 pm
With no minor pieces on the board and an average number of pawns (five per side), two rooks are nearly a pawn better than the queen. But on a full board the queen is indeed better than two rooks, as GM Roman Dzindzichashvilli taught me before computers were strong enough to answer the question. On average thruout the game the queen is a bit weaker.
What apply to two GMs does NOT hold true with 2 two powerful programs of equal strength like Komodo vs Komodo or SF10 vs SF10 or Komodo vs SF10. I believe that a Queen can find too many squares to keep the two Rooks busy, but for two Humans GMs is probably a little bit harder to find all the best moves or squares to force a draw. Therefore, there are a lot of pieces situations where what we believe true for two GMs, nowadays very strong programs are proving it to be wrong :shock:

Try the same experiment with a million different board configurations, using the same material.
You cannot draw such a conclusion from a single game.
At least not mathematically.

Is it possible to turn off that horrible nomenclature and have the comments written in PGN?
It's hard to read with all the noise in it.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Chessqueen
Posts: 630
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: How powerful is the Queen compared to 2 Rooks ?

Post by Chessqueen » Thu Jan 31, 2019 10:35 pm

Dann Corbit wrote:
Thu Jan 31, 2019 10:28 pm
Chessqueen wrote:
Thu Jan 31, 2019 10:20 pm
Chessqueen wrote:
Tue Jan 29, 2019 5:08 pm
lkaufman wrote:
Mon Jan 28, 2019 4:48 pm
With no minor pieces on the board and an average number of pawns (five per side), two rooks are nearly a pawn better than the queen. But on a full board the queen is indeed better than two rooks, as GM Roman Dzindzichashvilli taught me before computers were strong enough to answer the question. On average thruout the game the queen is a bit weaker.
What apply to two GMs does NOT hold true with 2 two powerful programs of equal strength like Komodo vs Komodo or SF10 vs SF10 or Komodo vs SF10. I believe that a Queen can find too many squares to keep the two Rooks busy, but for two Humans GMs is probably a little bit harder to find all the best moves or squares to force a draw. Therefore, there are a lot of pieces situations where what we believe true for two GMs, nowadays very strong programs are proving it to be wrong :shock:

Try the same experiment with a million different board configurations, using the same material.
You cannot draw such a conclusion from a single game.
At least not mathematically.

Is it possible to turn off that horrible nomenclature and have the comments written in PGN?
It's hard to read with all the noise in it.
You are correct there are too many board configurations, but in the majority of them you can give white with the Queen to Komodo versus either GM Nakamura or GM MLV or any other top GM with the two Rooks and I will bet money on Komodo. :shock:

Post Reply