Best (computer chess) opening

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

Moderators: hgm, Rebel, chrisw

tpoppins
Posts: 919
Joined: Tue Nov 24, 2015 9:11 pm
Location: upstate

Re: Best (computer chess) opening

Post by tpoppins »

Lyudmil Tsvetkov wrote: My latest observations are that white gets the biggest possible advantage in the (very strangely) completely untried and unresearched opening 1.e4 e5 2. c4!
<...>
or 1.e4 c5 2.c4!
<...>
or, starting with c4, 1.c4 e5 2.Nc3 Nf6 3.e4!
These all are most likely to transpose to some version of either the Maroczy Bind or the Botvinnik System. Where does the "completely untried and unresearched" part come from, then?
Lyudmil Tsvetkov wrote:and the resulting King's Indian Attack (KIA, basically the same as the KID for white with a spare tempo) is very favourable for white.
As far as I remember, Fischer had close to a perfect score with this opening.
I didn't look into it specifically but as far as I can remember he never employed KIA against top-level opposition.

edit: I see that he did employ it once against an international GM - in his second encounter with B. Ivkov in the Second Piatygorsky Cup (1966). It is arguable whether Ivkov could be called a top-level player but at least he played in the Candidates once.

In this particular case psychology - rather than any belief on Fischer's part in the superiority of KIA - may have been the most important factor in the choice of the opening, as Bobby's score against Ivkov until this tournament was a rather miserable +0-2=3. Although he got his first win against the Yugoslav GM several rounds earlier, it is notable that in that game he once again found himself in an inferior position in the opening (as early as move 8, after being shanghaied into the Pirc). Since the mainlines gave him nothing, the "untheoretical" KIA was a logical choice.

It may have been as simple as that.
Dann Corbit
Posts: 12540
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Best (computer chess) opening

Post by Dann Corbit »

As far as brute force computer search goes, a 53 ply search from the root gives 1.d4, which results in this position as best for white (worst for black):

[d]rnbqkbnr/pppppppp/8/8/3P4/8/PPP1PPPP/RNBQKBNR b KQkq - acd 48; bm e6; cce -42; ce -12; pm Nf6 {1004729} d5 {434599} e6 {116089} f5 {63006} d6 {40940} g6 {32622} c5 {15043} c6 {6638} Nc6 {3907} b6 {2255} e5 {1558} b5 {1162} a6 {822} h6 {540} Na6 {446} a5 {334} h5 {316} g5 {233} f6 {171} Nh6 {117}; pv e6 c4 Nf6 Nf3 b6 g3 Be7 Bg2 Bb7 O-O O-O Nc3 Ne4 Qd3 Nxc3 Qxc3 c5 Rd1 Na6 b3 Bf6 Bb2 Rc8 Rac1 d5 Qe3 Nb4 Ne5 cxd4 Bxd4 Nxa2 Ra1 Nb4 Rxa7 Nc2 Qd2 Nxd4 Qxd4 dxc4 Bxb7 Bxe5 Qxd8; white_wins 597411; black_wins 453117; draws 491802; Opening St. George Defense: St. Georgs Gambit. ; CaxtonID: 1623 ECO: A40;

On the other hand, based on percentage for outcomes of actual games, black had the lowest yield of points after white moved 1.b3 (no, I'm not buying it either):

[d]rnbqkbnr/pppppppp/8/8/8/1P6/P1PPPPPP/RNBQKBNR b KQkq - acd 46; bm e5; cce -45; ce -1; pm e5 {12348} d5 {7910} Nf6 {3224} c5 {1898} b6 {515} Nc6 {419} g6 {337} f5 {304} d6 {231} e6 {226} f6 {109} c6 {94} a5 {47} b5 {33} a6 {21} h6 {18} Nh6 {13} h5 {8} g5 {6} Na6 {3}; pv e5 Bb2 Nc6 e3 Nf6 Bb5 Bd6 Ne2 O-O O-O a6 Bxc6 dxc6 d3 a5 Nbc3 Re8 h3 Be6 a4 Qe7 Ng3 Rad8 e4 Bb4 Nce2 Bc8 Qc1 b6 Kh1 Bc5 Ng1 g6 Nf3 Nd7 Nd2 Bb7 Ne2 Qh4 Ba3 f5 Nf3 Qe7 Bxc5 Nxc5 exf5 gxf5 d4 exd4; white_wins 11905; black_wins 9293; draws 4674; Opening Nimzo-Larsen Attack: Symmetrical Variation. ; CaxtonID: 1416 ECO: A01;

On the other, other hand, after analyzing the forward position, a 48 ply search chose 1.e4 as one centipawn better than 1.d4. The root search was much deeper, but sometimes moving a ply forward helps a lot. It is also possible that this position was searched by a newer version of Stockfish or possibly Komodo than the search from the root position.

[d]rnbqkbnr/pppppppp/8/8/4P3/8/PPPP1PPP/RNBQKBNR b KQkq - acd 48; bm e6; cce -38; ce -16; pm c5 {1218905} e5 {412609} e6 {257916} c6 {204175} d5 {78062} d6 {71821} g6 {57457} Nf6 {46929} Nc6 {20542} b6 {4882} a6 {1853} g5 {987} h6 {878} a5 {501} h5 {447} Na6 {410} f6 {336} f5 {276} Nh6 {273} b5 {241}; pv e6 d4 d5 Nd2 c5 Ngf3 Nf6 exd5 exd5 Bb5+ Bd7 Bxd7+ Nbxd7 O-O Be7 dxc5 Nxc5 Nb3 Nce4 Nfd4 O-O Nf5 Re8 Nxe7+ Rxe7 Nd4 Rc8 Re1 Nd6 Rxe7 Qxe7 Bf4 a6 Bxd6 Qxd6 c3 g6 g3 Ne4 Kg2 Qb6 Qb3 Qxb3; white_wins 874667; black_wins 685852; draws 642462; Opening ECO:B00; Opening: King's pawn opening; 1. e4 *;


On the other hand, the WORST move based on game outcomes is Na3:
[d]rnbqkbnr/pppppppp/8/8/8/N7/PPPPPPPP/R1BQKBNR b KQkq - acd 43; acs 81199; bm e5; cce 285; ce 28; id "C.A.P. 445908"; pm e5 {9} Nf6 {5} c5 {5} Nc6 {2} b5 {1} d6 {1} e6 {1} f5 {1} g5 {1} g6 {1} h6 {1}; pv e5 e3 d5 d4 Nd7 Be2 e4 Nh3 Ndf6 Nf4 c6 c4 g5 Nh5 Nxh5 Bxh5 Bb4+ Bd2 Bxd2+ Kxd2 Be6 Be2 Qf6 Qc2 Ne7 Raf1 O-O-O Kc1 Kb8 Kb1 h5 h3 h4 cxd5 Bxd5 Nc4 Qe6 b3 f5 Ne5 Ng6 Nxg6 Qxg6 Rc1 Qd6 Qc5; white_wins 2; black_wins 20; draws 6; Opening Sodium Attack: Celadon Variation. ; CaxtonID: 2348 ECO: A00;

And the WORST move based on centipawn score after computer analysis is 1.g4:
[d]rnbqkbnr/pppppppp/8/8/6P1/8/PPPPPP1P/RNBQKBNR b KQkq - acd 44; acs 74127; bm d5; cce 109; ce 59; id "C.A.P. 445895"; pm d5 {1176} c5 {482} e5 {263} d6 {161} Nc6 {135} e6 {116} c6 {18} h6 {16} h5 {15} g5 {13} g6 {10} Nf6 {8} a5 {8} f6 {7} a6 {5} b5 {5} f5 {5} Nh6 {4} b6 {2} Na6 {1}; pv d5 c4 dxc4 Qa4+ c6 Qxc4 e5 Nf3 Bd6; white_wins 716; black_wins 1315; draws 413; Opening Grob Opening: Grob Gambit. ; CaxtonID: 733 ECO: A00;
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.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Best (computer chess) opening

Post by Lyudmil Tsvetkov »

tpoppins wrote:
Lyudmil Tsvetkov wrote: My latest observations are that white gets the biggest possible advantage in the (very strangely) completely untried and unresearched opening 1.e4 e5 2. c4!
<...>
or 1.e4 c5 2.c4!
<...>
or, starting with c4, 1.c4 e5 2.Nc3 Nf6 3.e4!
These all are most likely to transpose to some version of either the Maroczy Bind or the Botvinnik System. Where does the "completely untried and unresearched" part come from, then?
Lyudmil Tsvetkov wrote:and the resulting King's Indian Attack (KIA, basically the same as the KID for white with a spare tempo) is very favourable for white.
As far as I remember, Fischer had close to a perfect score with this opening.
I didn't look into it specifically but as far as I can remember he never employed KIA against top-level opposition.

edit: I see that he did employ it once against an international GM - in his second encounter with B. Ivkov in the Second Piatygorsky Cup (1966). It is arguable whether Ivkov could be called a top-level player but at least he played in the Candidates once.

In this particular case psychology - rather than any belief on Fischer's part in the superiority of KIA - may have been the most important factor in the choice of the opening, as Bobby's score against Ivkov until this tournament was a rather miserable +0-2=3. Although he got his first win against the Yugoslav GM several rounds earlier, it is notable that in that game he once again found himself in an inferior position in the opening (as early as move 8, after being shanghaied into the Pirc). Since the mainlines gave him nothing, the "untheoretical" KIA was a logical choice.

It may have been as simple as that.
A player who has a positive score against Fischer should not be a weak player, should he?
Ivkov was actually Junior World Champion, whatever that means.

Fischer had 100% score in that opening, and he produced some of his stunningmost games precisely with this opening. I guess he did not employ it all too often, because this is a very complicated opening, and Fischer did not particularly like complications, he wanted things straigth and clear, with minimal risk involved.

If the KID is a good opening, what should the rating of an opening with reversed colours and a tempo more be?

This is my last uncompleted game with Komodo (left it like that, just as a line, on purpose, to focus on the opening stage):

[pgn]
[Event "Blitz 2m+2s"]
[Site "Microsoft"]
[Date "2016.08.19"]
[Round "?"]
[White "Lyudmil Tsvetkov, owner"]
[Black "Komodo 10.1 64-bit"]
[Result "*"]
[ECO "A00"]
[Annotator "owner"]
[PlyCount "42"]
[EventDate "2016.??.??"]
[TimeControl "120+2"]

{512MB, OWNER-PC} 1. g3 {0} e5 {0.00/20 6} 2. Bg2 {2} d5 {-0.29/18 2} 3. d3 {2}
Bd6 {-0.14/20 8} 4. Nf3 {3} Nf6 {-0.18/20 3} 5. O-O {3} O-O {-0.18/21 7} 6. Nc3
{6} c6 {-0.32/19 3} 7. e4 {2} d4 {-0.27/23 6} 8. Ne2 {2} c5 {-0.28/22 4} 9. c4
{3} h6 {-0.34/22 13} 10. h3 {3} Nc6 {-0.36/22 4} 11. Kh1 {4} a6 {-0.39/20 5}
12. Nfg1 {3} b5 {-0.60/21 11} 13. b3 {3} Rb8 {-0.58/22 6} 14. f4 {3} Bd7 {-0.
63/20 6} 15. f5 {3} bxc4 {-0.63/22 6} 16. bxc4 {2} Qe7 {-0.61/24 9} 17. g4 {8}
Bc7 {-0.57/24 4} 18. Ng3 {15} Qd6 {-0.59/21 4} 19. h4 {21} Nh7 {-0.37/18 2} 20.
g5 {4} hxg5 {-0.33/19 4} 21. hxg5 {4} g6 {-0.29/21 5} *

[/pgn]

And this is the final position:

[d]1r3rk1/2bb1p1n/p1nq2p1/2p1pPP1/2PpP3/3P2N1/P5B1/R1BQ1RNK w - - 0 22

Who has advantage above?

Komodo (standard settings, no contempt) gives 30cps black edge, but white simply mates with 22.f6 above.

An opening in which you get a winning position against latest Komodo after some 20 moves should not be that weak, should it?

About the 1.e4 c5 2.c4 opening and transpositions:

When was the last time you saw someone, preferably at the top, play 2.c4 in the Sicilian after 1.e4 c5? Personally, I have not seen that at all.

Or, 3.e4, after 1.c4 e5 2. Nc3 Nf6? Personally, I have not seen that at all, even at the top they do not push e4, instead leave black to do that at some point and, as a result, guess what, black is almost always left with advantage.

Here my last game with Komodo, standard settings, in the line:

[pgn][Event "Blitz 2m+2s"]
[Site "Microsoft"]
[Date "2016.08.19"]
[Round "?"]
[White "Lyudmil Tsvetkov, owner"]
[Black "Komodo 10.1 64-bit"]
[Result "*"]
[ECO "A25"]
[Annotator "owner"]
[PlyCount "40"]
[EventDate "2016.??.??"]
[TimeControl "120+2"]

{512MB, OWNER-PC} 1. c4 {0} e5 {0.03/20 4} 2. Nc3 {2} Nc6 {0.10/21 9} 3. g3 {3}
Bc5 {0.10/20 5} 4. Bg2 {2} Nf6 {0.07/22 9} 5. e4 {2} d6 {-0.26/20 4} 6. h3 {2}
O-O {-0.34/20 5} 7. Nge2 {2} Nd4 {-0.29/21 6} 8. O-O {3} Be6 {-0.22/21 9} 9. d3
{3} a6 {-0.29/20 6} 10. Kh2 {3} c6 {-0.26/19 7} 11. f4 {3} b5 {-0.28/18 4} 12.
f5 {6} Bd7 {-0.42/20 4} 13. b3 {3} b4 {-0.61/22 6} 14. Na4 {18} Nxe2 {-0.56/23
3} 15. Qxe2 {2} Bd4 {-0.54/23 3} 16. Rb1 {11} c5 {-0.66/21 2} 17. g4 {12} h6 {
-0.69/20 3} 18. h4 {57} Nh7 {-0.67/20 4} 19. Qe1 {11} Bc6 {-0.64/20 5} 20. Qg3
{25} f6 {-0.62/22 4} *

[/pgn]

And the final position:

[d]r2q1rk1/6pn/p1bp1p1p/2p1pP2/NpPbP1PP/1P1P2Q1/P5BK/1RB2R2 w - - 0 21

Who has advantage above?

Komodo gives some 60cps black edge, but of course, white is considerably better, as only white has attacking chances on the kingside, the center and queen side are closed.

And this is the position very early on, at move 10:

[d]r2q1rk1/1p3ppp/p1ppbn2/2b1p3/2PnPP2/2NP2PP/PP2N1BK/R1BQ1R2 b - f3 0 11

Ever seen a position like that?
Any reminiscences to the Maroczy Bind or Botvinnik System? If yes, please post games.
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Best (computer chess) opening

Post by Lyudmil Tsvetkov »

jefk wrote:
Lyudmil Tsvetkov wrote: My latest observations are that white gets the biggest possible advantage in the (very strangely) completely untried and unresearched opening 1.e4 e5 2. c4!
after 1.e4 e5 2.c4? there's a hole on d4, and after 2...Nf6
and White has to play 2.d3, blocking his bishop on f1
So i doubt it's a good idea.

1.d3 is playable because it often leads to Kings-Indian (defense)
with colours reversed, and if the Black player has
better positional understanding than White this might
be a good trick. Just with c4 e5 (Sicilian reversed) it
does not mean automatically that White has an advantage.
Bishop goes to g2, of course, revelling on the long diagonal.

If the Sicilian is a very good choice for black, why should not a reversed opening with a tempo more also be?

Of course, it is, people are just scared of the d4 hole, but in the long run it is more than compensated for by sound control of center + early kingside pawn storming. You just have to continue the line.

1.c4 e5 is not good for white, I guess also in terms of statistics, precisely because for some reason (d4 hole scare-devil, someone suggested not playing e4 is good, won a game, and by dint of rut, after hitting the textbooks and supported by an opening expert, everyone later on played solely without e4 for white) white fails to push e4.

1.d3 is not only playable, it is a very good starting choice (possibly with transpositions).
Henk
Posts: 7218
Joined: Mon May 27, 2013 10:31 am

Re: Best (computer chess) opening

Post by Henk »

What if black plays de4: in Kings Indian in forehand. (This is what I encountered many times last year when I played 1. d3 or 1.g3) Then you don't get a closed position. [Getting worse if they can exchange queen too on d1. To prevent that I played Nd2 instead of Nc3. ]
Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 12:41 pm

Re: Best (computer chess) opening

Post by Lyudmil Tsvetkov »

Henk wrote:What if black plays de4: in Kings Indian in forehand. (This is what I encountered many times last year when I played 1. d3 or 1.g3) Then you don't get a closed position. [Getting worse if they can exchange queen too on d1. To prevent that I played Nd2 instead of Nc3. ]
You have gotten a vert good chess player as of late.

Indeed, de4 is a reasonable option.
After Ne4 Ne4 de4, white is simply better, but win is not in sight.

You play in a way that queen on d1 is not exchanged with check and loss of castling. Lots of possibilities to do that.

There are lines in that opening that could altogether avoid any center pawn exchanges with simplification.
Point is, KID/KIA setup with d3,g3,Nf3 and Bg2 is a very good one, and absolutely nowhere in it white should get worse.
carldaman
Posts: 2283
Joined: Sat Jun 02, 2012 2:13 am

Re: Best (computer chess) opening

Post by carldaman »

Lyudmil Tsvetkov wrote:Marco is an Italian, of course it will play an Italian Opening.

My latest observations are that white gets the biggest possible advantage in the (very strangely) completely untried and unresearched opening 1.e4 e5 2. c4!

[d]rnbqkbnr/pppp1ppp/8/4p3/2P1P3/8/PP1P1PPP/RNBQKBNR b KQkq c3 0 2

or 1.e4 c5 2.c4!

[d]rnbqkbnr/pp1ppppp/8/2p5/2P1P3/8/PP1P1PPP/RNBQKBNR b KQkq c3 0 2

or, starting with c4, 1.c4 e5 2.Nc3 Nf6 3.e4!

[d]rnbqkb1r/pppp1ppp/5n2/4p3/2P1P3/2N5/PP1P1PPP/R1BQKBNR b KQkq e3 0 3

All 3 lines are basically part of the same still unnamed opening with transpositions, and my personal opinion is that precisely here white gets the biggest possible advantage with optimal play.

White firmly controls the center and has good chances to start a promising kingside pawn attack earlier, a single black outpost in the center on d4 hardly compensates that.

Another 40 years until SF and Komodo get it. :)

My second choice (second-best white choice) is 1.d3! d5 2. Nf3 Nf6 3.g3

[d]rnbqkb1r/ppp1pppp/5n2/3p4/8/3P1NP1/PPP1PP1P/RNBQKB1R b KQkq - 0 3

and the resulting King's Indian Attack (KIA, basically the same as the KID for white with a spare tempo) is very favourable for white.
As far as I remember, Fischer had close to a perfect score with this opening.

So 1.d3 or possible transpositions is indeed a very good move, though engines tend to eval it with a negative score. Of course, after 1.d3 white has a significant advantage, my Komodo shoot-out only confirmed this, as white scored significantly above 50%.

Another 40 years for engines to understand 1.d3 is very good. :)
Hello Lyudmil,

I like your thoughts on this important issue/topic of chess openings, especially those that can work against both engines and people (since the latter have become over-reliant on computers). It's an area where smart humans can definitely outmaneuver the engines by steering the game into favorable formations early on, in particular when the engines play with a short book (or no book at all).

A minor quibble would be to point out that in the KIA, White typically gets a c3-d4-e5 pawn chain, more akin to a French defense set-up. Probably because of Fischer's success in this opening and otherwise, the KIA had a great reputation as an anti-French weapon, but as theory evolved, and Fischer retired, better plans and resources were found for Black, to the point that some consider it inadvisable to employ the KIA vs a well-prepared French def. player.

However, White can also play (or at least aim for) an actual Reversed KID, where White gets the 'pure' d3-e4-f5 KID pawn chain. I find White's chances in the latter preferable over the KIA set-up (c3-d4-e5), as this is actually an even stronger chain, causing more K-side congestion for Black.

Regards,
CL
Dann Corbit
Posts: 12540
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Best (computer chess) opening

Post by Dann Corbit »

What do you think of 1.e3
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.
Henk
Posts: 7218
Joined: Mon May 27, 2013 10:31 am

Re: Best (computer chess) opening

Post by Henk »

Or a reversed Queens Indian. 1 Nf3 d5 2. b3 and later e3 or d3 e4.

Usually after 1 d3 black also plays g6, d6, Nf6. So you get a symmetrical boring position.
jefk
Posts: 626
Joined: Sun Jul 25, 2010 10:07 pm
Location: the Netherlands
Full name: Jef Kaan

Re: Best (computer chess) opening

Post by jefk »

Dann Corbit wrote:What do you think of 1.e3
comes out well when backsolving large opening books,
as result of transposition to 'closed' type of QGD variations,
with e3 instead of eg Bg5. But then there's the French
exchange, and it's hard to get an advance there for White
in such symmetrical positions; and finally
the strongest Black defense might be a fianchetto
type of position, Kings Indian, whereby e3 has lost
a tempo whereas normally e4 is played. Or Gruenfeld style
example (Gruenfeld, Botvinnik variation)

1. e3 Nf6 0.00/0 0.1 2. Nf3 0.00/0 0.0 g6 0.00/0 0.0 3. d4 0.00/0 0.0 Bg7 0.00/0 0.0 4. c4 0.00/0 0.6 O-O 0.00/0 0.0 5. Nc3 0.00/0 0.6 d5 0.00/0 0.0 6. Qb3 0.00/0 0.5 e6 0.00/0 0.6 7. Be2 0.00/0 0.0 Nc6 0.00/0 0.0 8. cxd5 0.00/0 0.0 exd5 0.06/17 =

So 1.d4, e4, c4 and Nf3 are slightly better, whereby
the White difference for these moves are statistically
completely *in*significant.

Reversed Queens Indian looks equal, for example
1. Nf3 d5 2. b3 Nf6 0.00/0 0.0 3. Bb2 0.00/0 0.0 e6 0.00/0 0.5 4. e3 0.00/0 0.0 Be7 0.00/0 0.0 5. c4 O-O 0.00/0 0.0 6. Nc3 0.00/0 0.0 c5 7. cxd5 0.00/0 0.7 exd5 0.00/0 0.0 8. d4 0.00/0 0.0 Nc6 0.00/0 0.0 9. Be2 =
Also called Reti, Nimzovich-Larsen attack


jef