Capablanca's Theorem Q+N > Q+B

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.
User avatar
Nordlandia
Posts: 2588
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Capablanca's Theorem Q+N > Q+B

Post by Nordlandia » Sun Sep 10, 2017 7:01 am

Lyudmil Tsvetkov wrote:
Nordlandia wrote:Lyudmil Tsvetkov: you're right, my example positions isn't reliable for this kind of experiment.

May i ask if you can find eligible positions :)
I can't.

you are much better at that stuff. :)

I guess none of existing chess software has the necessary filtering options to successfully mask out an acceptable set, so maybe we should just test with engine games from a neutral position.
In that case, engine need to know Q+N duo is slightly stronger than Q+B. Otherwice testing is futile. Don't you agree?

If engine value Q+N and Q+B as equal, there is nothing to test at the moment.

Wasn't there submitted recently a patch for tweaking material values for Q+N vs Q+B in the endgame, is the patch likely to pass?

http://tests.stockfishchess.org/tests/v ... 16ff64aff4

User avatar
Nordlandia
Posts: 2588
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Capablanca's Theorem Q+N > Q+B

Post by Nordlandia » Sun Sep 10, 2017 7:10 am

Lyudmil: shouldn't surprise me if you don't agree to the cliché of Rook and Bishop is better than Rook and Knight either.

I'm not here to argue about well-known fact.

whereagles
Posts: 562
Joined: Thu Nov 13, 2014 11:03 am

Re: Capablanca's Theorem Q+N > Q+B

Post by whereagles » Sun Sep 10, 2017 8:24 am

Since "better" here refers obviously to the amount of QN/QB positions (plus king and pawns of course) that are won/drawn/loss, I suggest someone randomly generates a zillion of those and make an engine play them out.

That should give us some hint as to whether the so-called "theorem" is true or not. But more important would be to try and understand what type of positions favour QN/QB, because those tips are of practical value for human play.

User avatar
Graham Banks
Posts: 33972
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: Capablanca's Theorem Q+N > Q+B

Post by Graham Banks » Sun Sep 10, 2017 8:28 am

From my observations of many engine v engine games, I'd say that the theory of Q+N generally being stronger than Q+B holds true.
My email addresses:
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz

User avatar
Nordlandia
Posts: 2588
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Capablanca's Theorem Q+N > Q+B

Post by Nordlandia » Sun Sep 10, 2017 9:22 am

Graham Banks wrote:From my observations of many engine v engine games, I'd say that the theory of Q+N generally being stronger than Q+B holds true.
Ceteris paribus, the advantage is comparable to home-field advantage in football.

Overall Q+N scores 5-7% better than Q+B.

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

Re: Capablanca's Theorem Q+N > Q+B

Post by Lyudmil Tsvetkov » Mon Sep 11, 2017 3:45 am

Nordlandia wrote:
Lyudmil Tsvetkov wrote:
Nordlandia wrote:Lyudmil Tsvetkov: you're right, my example positions isn't reliable for this kind of experiment.

May i ask if you can find eligible positions :)
I can't.

you are much better at that stuff. :)

I guess none of existing chess software has the necessary filtering options to successfully mask out an acceptable set, so maybe we should just test with engine games from a neutral position.
In that case, engine need to know Q+N duo is slightly stronger than Q+B. Otherwice testing is futile. Don't you agree?

If engine value Q+N and Q+B as equal, there is nothing to test at the moment.

Wasn't there submitted recently a patch for tweaking material values for Q+N vs Q+B in the endgame, is the patch likely to pass?

http://tests.stockfishchess.org/tests/v ... 16ff64aff4
endgame position will be sufficiently easy to decide without sophisticated knowledge.

the eval is more valuable in taking pruning decisions earlier on.

patch will not pass, maybe the much smaller value Stefan is trying right now might do better.

Jack Lad
Posts: 8555
Joined: Wed Mar 08, 2006 9:01 pm
Location: UK

Re: Capablanca's Theorem Q+N > Q+B

Post by Jack Lad » Thu Feb 27, 2020 1:37 pm

Arpad Rusz wrote:
Wed Sep 06, 2017 3:34 pm
Nordlandia wrote:Likewise the queen + knight tend to work better than queen + bishop in the endgame because the queen + bishop duo can sometimes be redundant.
This might be true but it doesn't hold against two rooks: QN vs RR is a general draw while QB vs RR is a general win. The rook pair's main resource is the third rank defence which can be broken only by the Q+B duo. The key to success seems to be the bishop's ability to attack the first rank. An example:
[D]8/8/8/8/6kb/R4R2/3q4/5K2 b - -
The shortest win is 1...Bf2! -+
This QN v RR endgame is a win for black in 127 moves:
[d] 8/8/R3K3/8/8/8/3q4/R1nk4 b - - 0 1
1...Ke2! wins but perhaps is still a draw if the 50 move rule limit applies. :roll:
[d] 5R2/8/8/8/6q1/5n2/5k1K/R7 w - - 0 1
After exactly 100 moves white is now forced to take the knight on f3 but maybe he claim a draw just in time if there is a 100 move rule instead. :?
All of the basic checkmates can be accomplished in well under 50 moves. However, in the 20th century it was discovered that certain endgame positions are winnable but require more than 50 moves (without a capture or a pawn move). The rule was therefore changed to allow certain exceptions in which 100 moves were allowed with particular material combinations. However, winnable positions that required even more moves were later discovered, and in 1992, FIDE abolished all such exceptions and reinstated the strict 50-move rule.
https://en.wikipedia.org/wiki/Fifty-move_rule
Now cracks a noble heart.—Good night, sweet Princess, And flights of angels sing thee to thy rest!

Post Reply