Out of the Kai 450 positions, 99 not found by SF

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Uri Blass
Posts: 8533
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: Out of the Kai 450 positions, 99 not found by SF

Post by Uri Blass » Fri Aug 16, 2019 3:02 pm

Ovyron wrote:
Fri Aug 16, 2019 10:45 am
Zenmastur wrote:
Thu Aug 15, 2019 7:06 pm
A VERY good fundamental question. In most “truly” even middle-game positions there can be many “equal” moves. So, there IS NO “BEST” move. I would think that a position that “looks” even under normal analysis but is “in fact” won or at least highly favorable to one side is what is needed.
For any chess position there's ALWAYS a move that is better than another, by some metric. I already suggested some metric (say, with one move it's easier to beat Stockfish Depth 10 than with another; as the level increases, the metric changes, and the best move changes, but it's still clearly better than another one.)
Even with this metric it is possible that 2 moves are the same.
Here is a simple example




What is the best move for white?
I think that even in the middle game it is easy to find positions when 2 moves force the same position so they are equal(and by force I mean that the opponent has no different legal choice).

Zenmastur
Posts: 389
Joined: Sat May 31, 2014 6:28 am

Re: Out of the Kai 450 positions, 99 not found by SF

Post by Zenmastur » Fri Aug 16, 2019 7:23 pm

Uri Blass wrote:
Fri Aug 16, 2019 3:02 pm
Ovyron wrote:
Fri Aug 16, 2019 10:45 am
Zenmastur wrote:
Thu Aug 15, 2019 7:06 pm
A VERY good fundamental question. In most “truly” even middle-game positions there can be many “equal” moves. So, there IS NO “BEST” move. I would think that a position that “looks” even under normal analysis but is “in fact” won or at least highly favorable to one side is what is needed.
For any chess position there's ALWAYS a move that is better than another, by some metric. I already suggested some metric (say, with one move it's easier to beat Stockfish Depth 10 than with another; as the level increases, the metric changes, and the best move changes, but it's still clearly better than another one.)
Even with this metric it is possible that 2 moves are the same.
Here is a simple example




What is the best move for white?
I think that even in the middle game it is easy to find positions when 2 moves force the same position so they are equal(and by force I mean that the opponent has no different legal choice).
This is almost always the case in EQUAL positions. It's a lot less likely to be the case in a position that favors one side or the other. It's possible to find position in which only one move is best (and to prove it's best) to use as test positions. The part I have a problem with is this:



What's the "best" move? This is position #2 on Dann's list of unsolved positions on Kai's list. This is what was given as "best" by Dann

Code: Select all

Ng3 Bd6 Bf1 Qe8 Rxc8 Qxc8 Qa4 a6 Qd1 g6 Kg2 Kg7 Bd3 b5 Qb1 Qe8 a3 Ng8 Qd1 Bc6 b4 Bd7 Qc2 Ne7 Qd1
I did a Multi-PV on this position and found this:

Code: Select all

(  [Cfish 240719 64 BMI2] 63:+0.00 1.Ng3 g6 2.Bf1 Kg7 3.Rxc8 Bxc8 4.Qa4 Bd7 5.Nxd7 Nxd7 6.Bd3 Nf6 7.Qd1 Qb4 8.Qe2 Qa4 9.a3 Bd6 10.Kg2 a5 11.h3 Qb3 12.Ba6 Bc7 13.Qb5 Qxb5 14.Bxb5 g5 15.Kf3 gxf4 16.exf4 Kf8 17.Be2 Bd6 18.Bb5 )
(  [Cfish 240719 64 BMI2] 62:+0.00 1.Rxc8+ Bxc8 2.Bf1 Bd7 3.Nxd7 Qxd7 4.Ng3 g5 5.fxg5 hxg5 6.Bd3 Qd8 7.Qd1 Kf8 8.Kg2 Ke7 9.a3 Qg8 10.Qc2 Bd6 11.a4 Qd8 12.Qc6 Qe8 13.Qc2 )
(  [Cfish 240719 64 BMI2] 62:+0.00 1.a3 Rxc2 2.Qxc2 g6 3.Bf1 Qe8 4.Ng3 Kg7 5.Bd3 Bd6 6.Qd1 Kf8 7.Ne2 Ng8 8.Qc2 Kg7 9.Nc3 Bb8 10.Ne2 )
Then I looked at even more moves:

Code: Select all

1. 44 [+0.00]  1.a3 Rxc2 2.Qxc2 g6 3.Bf1 Bd6 4.Nc3 g5 5.fxg5 hxg5 6.Qa4 Bb8 7.h3 Qd8 8.Nc6 Bxc6 9.Qxc6 Kg7 10.Be2 Qc7 11.Qxc7 Bxc7 12.Kg2 Ne4 13.Bd3 Nd6 14.Kf3 f5 15.Ke2 Kf7 16.a4 Bd8 17.Nb1 Kg6 18.Nc3 Kg7 19.Kf3 Be7 20.Ke2 
2. 44 [+0.00]  1.Rxc8+ Bxc8 2.Bf1 Bd7 3.Nxd7 Qxd7 4.Ng3 g5 5.fxg5 hxg5 6.Qd1 Kf8 7.Bd3 Ke7 8.Kg2 Qd8 9.Qa4 a5 10.Qc6 Qc7 11.Qxc7+ Bxc7 12.a4 Kf8 13.Kf3 Kg7 14.h3 Bd6 15.Be2 Kg6 16.Bd3+ Kh6 17.Kg2 Kg7 18.Kf3 
3. 44 [+0.00]  1.Ng3 g6 2.Bf1 Kg7 3.Rxc8 Bxc8 4.Qa4 Bd7 5.Nxd7 Nxd7 6.Bd3 Nf6 7.Qd1 Qb4 8.Qe2 Qa4 9.a3 Bd6 10.Kg2 a5 11.h3 Qb3 12.Ba6 Bc7 13.Qb5 Qxb5 14.Bxb5 g5 15.Kf3 Kf8 16.fxg5 hxg5 17.Ne2 Kg7 18.Nc3 Bd6 19.Be2 Kg6 20.Nb5 Bb8 21.Kg2 Ne4 22.a4 f5 23.Kf1 Kf6 24.Bd3 Kf7 25.Ke1 
4. 44 [+0.00]  1.Qa4 Rxc2 2.Qxc2 Bd6 3.Bf1 Qc7 4.Qa4 a6 5.Qd1 b5 6.Kg2 Ne4 7.a4 b4 8.Ng3 g6 9.Be2 a5 10.Bd3 Kg7 11.Bb5 Nf6 12.Bd3 Qc8 13.Qb1 Qe8 14.Bb5 Qd8 15.Qd1 h5 16.Ne2 Ne4 17.Ng3 Qc7 18.Ne2 
5. 43 [+0.00]  1.Qd3 Qe8 2.Rxc8 Qxc8 3.Qa3 Qf8 4.Qb3 Bd6 5.Qa4 a6 6.Qb3 Qd8 7.a4 g6 8.Ng3 Kg7 9.Bf1 a5 10.Bd3 Ng8 11.Qd1 Qc8 12.Qb3 
6. 43 [+0.00]  1.Bf3 Rxc2 2.Qxc2 Bd6 3.Ng3 Qc7 4.Qb3 Kf8 5.Kg2 g6 6.a3 Bc6 7.Nxc6 Qxc6 8.Be2 Qc1 9.h3 Kg7 10.Bf3 Bb8 11.Ne2 Qc6 12.Nc3 Qc8 13.Qb5 Bd6 14.Qb3 Kh7 15.Nb5 Bb8 16.Be2 Kg7 17.Bd3 g5 18.fxg5 hxg5 19.Qd1 Qh8 20.Qf3 a6 21.Nc3 b5 22.Kf1 Bd6 23.Kg2 Qc8 24.Kf1 
7. 43 [+0.00]  1.h3 Rxc2 2.Qxc2 Bd6 3.Nc3 a6 4.Qd1 Qc7 5.a4 g6 6.Bf1 Kg7 7.Bd3 Ne4 8.Qc2 Nxc3 9.bxc3 Qe7 10.Kg2 g5 11.Qb2 Qc7 12.Qe2 Qxc3 13.fxg5 Bxe5 14.gxh6+ Kxh6 15.dxe5 Qxe5 16.Bxa6 Bxa6 17.Qxa6 Qg7+ 18.Kf3 Qf6+ 19.Kg2 
8. 43 [+0.00]  1.Bf1 Rxc2 2.Qxc2 Ne4 3.Qa4 Nd6 4.Qc2 Qc7 5.Nc3 g5 6.fxg5 hxg5 7.f4 f6 8.Bh3 Nf5 9.Bxf5 exf5 10.Qxf5 fxe5 11.Qg6+ Kf8 12.Qf6+ Ke8 13.Qh8+ Kd7 14.Qh3+ Ke7 15.fxe5 Ke8 16.Qe6+ Kf8 17.Qf6+ Ke8 
9. 43 [+0.00]  1.Bh3 Rxc2 2.Qxc2 Qe8 3.Ng3 Bd6 4.Bf1 g6 5.Be2 Kg7 6.Bd3 Ng8 7.a3 a6 8.Qd1 b5 9.Kg2 Ne7 10.h3 Bc8 11.Qg4 Ng8 12.Qe2 Ne7 
10. 43 [+0.00]  1.Rc3 Rxc3 2.Nxc3 Ne8 3.Qa4 Nd6 4.Nc6 Bxc6 5.Qxc6 g6 6.Bf1 Qc7 7.Qxc7 Bxc7 8.Kg2 g5 9.fxg5 hxg5 10.Bd3 Kg7 11.a4 f5 12.h3 Bd8 13.Be2 Be7 14.Kf3 Kf7 15.Nb5 Nxb5 16.Bxb5 Bf6 17.Bd3 Ke7 18.Kg3 Kf7 
With a large fraction of the legal moves judged "best/equal" by CFISH what good is this position as a test? (I'm not saying that "fish" programs are perfect, they're not, but it's a good indication of what direction the truth may be.) And I didn't do further analysis on the lines of play that it gave, mostly because I believe the lines are accurate enough to show that this position is "equal" and there is NO BEST MOVE!

To me, that's a problem for this position.

Regards,

Zenmastur
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

zullil
Posts: 5562
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: Out of the Kai 450 positions, 99 not found by SF

Post by zullil » Fri Aug 16, 2019 9:58 pm

Zenmastur wrote:
Fri Aug 16, 2019 7:23 pm
The part I have a problem with is this:



What's the "best" move? This is position #2 on Dann's list of unsolved positions on Kai's list. This is what was given as "best" by Dann

Code: Select all

Ng3 Bd6 Bf1 Qe8 Rxc8 Qxc8 Qa4 a6 Qd1 g6 Kg2 Kg7 Bd3 b5 Qb1 Qe8 a3 Ng8 Qd1 Bc6 b4 Bd7 Qc2 Ne7 Qd1
I did a Multi-PV on this position and found this:

Code: Select all

(  [Cfish 240719 64 BMI2] 63:+0.00 1.Ng3 g6 2.Bf1 Kg7 3.Rxc8 Bxc8 4.Qa4 Bd7 5.Nxd7 Nxd7 6.Bd3 Nf6 7.Qd1 Qb4 8.Qe2 Qa4 9.a3 Bd6 10.Kg2 a5 11.h3 Qb3 12.Ba6 Bc7 13.Qb5 Qxb5 14.Bxb5 g5 15.Kf3 gxf4 16.exf4 Kf8 17.Be2 Bd6 18.Bb5 )
(  [Cfish 240719 64 BMI2] 62:+0.00 1.Rxc8+ Bxc8 2.Bf1 Bd7 3.Nxd7 Qxd7 4.Ng3 g5 5.fxg5 hxg5 6.Bd3 Qd8 7.Qd1 Kf8 8.Kg2 Ke7 9.a3 Qg8 10.Qc2 Bd6 11.a4 Qd8 12.Qc6 Qe8 13.Qc2 )
(  [Cfish 240719 64 BMI2] 62:+0.00 1.a3 Rxc2 2.Qxc2 g6 3.Bf1 Qe8 4.Ng3 Kg7 5.Bd3 Bd6 6.Qd1 Kf8 7.Ne2 Ng8 8.Qc2 Kg7 9.Nc3 Bb8 10.Ne2 )
Then I looked at even more moves:

[snipped]

With a large fraction of the legal moves judged "best/equal" by CFISH what good is this position as a test? (I'm not saying that "fish" programs are perfect, they're not, but it's a good indication of what direction the truth may be.) And I didn't do further analysis on the lines of play that it gave, mostly because I believe the lines are accurate enough to show that this position is "equal" and there is NO BEST MOVE!

To me, that's a problem for this position.

Regards,

Zenmastur
I agree that including this position in a test set is problematic. That said, Lc0 with network 320x24.J13-swa-410000 and 100M nodes prefers Rxc8+. The network says there's about a 22% chance that this is the "best" move in this position.

Nodes: 100M, N/s: 39.5k, Time: 42m 9s (limit met)

49.5% Rxc8+ Bxc8 Nc3 g6 Qa4 Bb7 Bf1 Kg7 Bb5 Qd8 Bc6 Bxc6 Nxc6 Qc7 Ne5 g5 Qd1 gxf4 exf4 Qc8 Qf3 b5 Qg3+ Kf8 Qh4 Kg7 Qg3+ Kf8 Qh3 Kg7 Qg2+ Kf8 Qh3 (N: 96.3M, P: 21.87%)

48.6% Ng3 Rxc2 Qxc2 g6 Bf1 Kg7 Bd3 Qe8 b4 a6 a4 Bd6 b5 a5 Qd1 Qe7 Qc2 Qe8 Qd1 (N: 938k, P: 8.84%)
48.8% Bf1 Rxc2 Qxc2 Qe8 Ng3 g6 Bd3 Kg7 Kg2 Bd6 Ne2 Ng8 Nc3 Ne7 Nb5 Bb8 Nc7 Qc8 Nb5 Qxc2 (N: 753k, P: 5.43%)
47.8% Qa4 Rxc2 Qxc2 g6 Bf1 Kg7 Ng3 Qe8 Bd3 Bd6 a3 Ng8 Qd1 Ne7 Qg4 Kf8 h4 (N: 657k, P: 11.16%)
47.5% a3 Qe8 Bf1 Ne4 Ng3 Nd6 h4 Rxc2 Qxc2 Qc8 Qd1 Nc4 Qe2 Ba6 Qc2 Nxe5 Qxc8+ (N: 147k, P: 2.97%)
47.2% h3 Qe8 Rxc8 Qxc8 Nc3 Ne4 Nxe4 dxe4 Qa4 g5 fxg5 hxg5 Bf1 Kg7 Qd1 f6 Nc4 (N: 128k, P: 2.95%)
47.3% Bf3 Rxc2 Qxc2 g6 Ng3 Kg7 Be2 Qc7 Qb1 Qc8 Bd3 Qe8 Qc2 Ng8 Qd1 (N: 115k, P: 2.63%)
47.3% Qd3 Qe8 Ng3 Bd6 Bf1 g6 Rxc8 Bxc8 Qc2 Kg7 Bd3 Bd7 Qe2 Ba4 b3 (N: 106k, P: 2.37%)
46.9% Nc3 Ne8 Ne2 Rxc2 Qxc2 Nd6 Ng3 Qc7 Qd1 Qc8 Bf1 Nc4 b3 Nd6 Bd3 Ne4 (N: 105k, P: 2.82%)
47.4% Rc3 Qe8 Qc2 Rxc3 Nxc3 g6 Bf1 Kg7 Bd3 Ng8 Ne2 Ne7 Ng3 f6 Nf3 Bc6 h4 (N: 102k, P: 2.18%)

jp
Posts: 708
Joined: Mon Apr 23, 2018 5:54 am

Re: Out of the Kai 450 positions, 99 not found by SF

Post by jp » Sat Aug 17, 2019 7:42 am

Laskos wrote:
Tue Aug 13, 2019 9:02 am
While about my positional test-suites I had great doubts myself, about positional superiority of Lc0 on a reasonable GPU with a strong net, I have few doubts. Maybe I don't understand what "positional" means. What is obvious, from games, test-suites, is that Lc0 is clearly weaker tactically compared to not only Stockfish, but even to much weaker modern AB engines with a regular eval. It is again obvious to me that to be the strongest engine on my PC from regular openings, Lc0 compensates by its very strong "positional" play, maybe in my wrong understanding of the notion of "positional", as some sort of conjugate of "tactical". I am curious, do you have some confidence that a strong Lc0 is superior "positionally" to a strong regular eval AB engine?
I'm open to the possibility, but will have to say something about this idea of "positional" vs "tactical" in another post. Apart from that, I'm not sure that they are really giving different evaluations and PVs most of the time, when you give a lot of computing time.

e.g.
zullil wrote:
Tue Aug 13, 2019 5:38 pm
Here's Lc0 (320x24b J13-410 net) after about one hour searching the first position in Dann's list. Note that this Lc0 agrees with Cfish-dev as to what the position should be six plies in the future:

info depth 26 seldepth 84 time 3604808 nodes 74943920 score cp 88 hashfull 337 nps 20789 tbhits 0 pv a1c1 g4h5 f1g2 h5g6 d2c4 f7f6 c2b3 g6f7 b3b1 b8a7 e3e4 f5e4 g2e4 g8g7 e4h7 e8e1 c1e1 c8g4 c4e3 d5e3 f2e3 d8e8 h7f5 g4f3 f5e4 f3g4 e4f5 g4f3 f5e4 f3g4 e4g2 f7g6 b1b2 g4c8 e1f1 a7c5 b4c5 g6d3 f1e1 d3e4 a3a4 c8d7 a4a5 e8e7 b2f2 e4g2 g1g2 e7e4 h2h3 d7d5 g2h2 d5e6 g3g4 e6b3 f2d2 g7f7 h2g3 e4e8 g3f2 f7g7

(Cfish-dev suggested a1c1 g4h5 d2c4 h5g6 f1g2 f7f6 c2b2 b8a7 a3a4 g6f7 c5d3 c8d7 c4a5 d7c7 d3c5 d8b8 a5c4 b8d8 b2d2 d5e7 c4a5 d8b8 c1b1 g8g7 e1c1 e8c8 a5c4 c8d8 d2e2 h7h5 b4b5 a6b5 a4b5 a7c5 d4c5 f7c4 c1c4 c7e5 h2h4 g7g6 h4g5 f6g5 f2f4 g5f4 g3f4 e5e6 b5c6 e7c6 g2c6 b7c6 b1b8 d8b8 c4d4 g6g7 d4d6 b8b1 g1h2.)
That's not the most extreme example. We were looking at a position in another thread and Leela & SF agreed to 14 or 20 plies, when we are not sure either of them was "right" or "understood" the position.

User avatar
Ovyron
Posts: 2348
Joined: Tue Jul 03, 2007 2:30 am

Re: Out of the Kai 450 positions, 99 not found by SF

Post by Ovyron » Sat Aug 17, 2019 9:58 am

Uri Blass wrote:
Fri Aug 16, 2019 3:02 pm
Even with this metric it is possible that 2 moves are the same.
Here is a simple example

At least you didn't post a position with just two naked Kings :roll:

So, can white win here with some series of black moves? Then, assume black wants to get checkmated, and it cooperates with white to achieve this. What's the white moves that allow white to mate black the fastest? Otherwise, assume black is a random player, then white plays the moves that increase the chance that black plays those random moves that allow white to mate them.

In all these cases, white can rank its moves, and the one at the top is the best one. They're not equal, equality assumes best play (black can secure a draw no matter what white plays), but you don't know if your opponent will play those best moves (there's a non-zero chance that they have a mouse slip and give up the game), so there's a move that maximizes your chances of winning.
Zenmastur wrote:
Fri Aug 16, 2019 7:23 pm
And I didn't do further analysis on the lines of play that it gave, mostly because I believe the lines are accurate enough to show that this position is "equal" and there is NO BEST MOVE!
And THIS is the reason talking about "best moves" makes no sense whatsoever, unless there's a game happening from that position, white is someone, black is someone, and white is trying to win. Then and only then, this person controlling the black pieces has some strength, and there's a series of moves white can play that defeat this strength, and from those there's a string that gets black mated in the shortest amount of moves.

Because black can't play perfectly.

So, let's say you know for a fact that the black player will play Stockfish Depth 22 moves, single threaded, at MultiPV=4. You can predict all its moves. Aren't you able to find a game continuation that ensures that you as white, win? I think you could if you tried. And then you'll find that this is easier to do with some moves than with others, and then you'd rank the moves, and find one that is clearly best. And if you were black, chances are someone in the world would be able to find a string of moves that would defeat any defense you can come up with.

At least, if some alien race came from space and told humanity that they have to defeat you as white from that position, and everyone gathered their resources to beat you, they could find the string that beats your strength, and a white move would be clearly best for this, over the others. That's THE BEST MOVE.

In this game I posted my opponent played a move weaker than Stockfish Depth 22:



19/27 0:01 -0.02 17...Ned7 18.Ng3 Nh7 19.Bf4 Bxc3 20.Qxc3 Qxc3+ 21.bxc3 Ne5 22.Ne4 Rad8 23.Kb2 Kg7 24.Kb3 b6 25.Rde1 a5 26.a4 f6 27.Nf2 Bf5 28.h4 Kf7 29.Ng4 Ke7 (5.778.796) 3694
19/09 0:01 0.00 17...a6 18.Ne4 Qxd2+ 19.Nfxd2 Nxe4 20.Nxe4 Nd7 21.Bd3 Ne5 (5.778.796) 3694
19/39 0:01 +0.05 17...Rab8 18.Ne4 Qxd2+ 19.Nfxd2 Nxe4 20.Nxe4 Nd7 21.Bf4 Bf5 22.Bxd6 Rbd8 23.Rge1 Bxe4 24.fxe4 Rxe4 25.Bd3 Ree8 26.Bg3 Be5 27.Bc2 Bxg3 28.hxg3 Nf6 29.Kd2 Rxe1 30.Rxe1 (5.778.796) 3694
19/28 0:01 +0.16 17...b5 18.cxb5 c4 19.Bxf6 Bxf6 20.Ne4 Qd8 21.Nxf6+ Qxf6 22.Ng3 Nd3+ 23.Kb1 Re3 24.Ne4 Qd4 25.Nxd6 Qxb2+ 26.Qxb2 Nxb2 27.Kxb2 Rxe2+ 28.Kc3 Rxa2 29.Nxc4 Rc8 30.Kb4 (5.778.796) 3694

20/29 0:02 -0.20 17...Ned7 18.Ng3 Nh7 19.Bf4 Bxc3 20.Qxc3 Qxc3+ 21.bxc3 Ne5 22.Ne4 Rad8 23.Kb2 Kg7 24.Kb3 b6 25.Rde1 Bf5 26.Ng3 Bd7 27.h4 Nf6 28.Bg5 b5 29.Bxf6+ Kxf6 30.Ne4+ Ke7 (9.888.797) 3576
20/09 0:02 0.00 17...a6 18.Ne4 Qxd2+ 19.Nfxd2 Nxe4 20.Nxe4 Nd7 21.Bd3 Ne5 (9.888.797) 3576
20/39 0:02 +0.06 17...b5 18.cxb5 c4 19.Bxf6 Bxf6 20.Ne4 Qd8 21.Ne3 Bg7 22.Kb1 Bh6 23.Qc3 Bxe3 24.Qxe3 Bf5 25.Ka1 Bxe4 26.fxe4 Qa5 27.Rdf1 Qxb5 28.Qh6 Rab8 (9.888.797) 3576
20/36 0:02 +0.11 17...Bxf1 18.Rdxf1 Ned7 19.Kb1 Nh7 20.Bf4 Bxc3 21.bxc3 Nhf6 22.Bd3 Nh5 23.Bxg6 fxg6 24.Rxg6+ Kf7 25.Rfg1 Ndf6 26.Bh6 Re5 27.Qc2 Ke7 28.Bg5 Rh8 29.Ka1 (9.888.797) 3576

21/30 0:06 -0.51 17...Ned7 18.Ng3 a6 19.Bh6 Bh8 20.Kb1 b5 21.Nce4 Qxd2 22.Rxd2 Nxe4 23.Nxe4 Bf5 24.cxb5 Nf6 25.bxa6 Kh7 26.Bg5 Nxe4 27.fxe4 Rxe4 28.Bd3 Rb4 29.Bxf5 gxf5 30.Be3 Rxa6 (20.787.966) 3461
21/36 0:06 +0.08 17...b5 18.cxb5 c4 19.Bxf6 Bxf6 20.Ne4 Qd8 21.Nxf6+ Qxf6 22.Ng3 Nd3+ 23.Bxd3 cxd3 24.Qxd3 Qf4+ 25.Kb1 Re3 26.Qd4 Rxf3 27.Ka1 Bf5 28.Qxf4 Rxf4 29.Nxf5 Rxf5 30.a4 Kg7 (20.787.966) 3461
21/41 0:06 +0.17 17...a6 18.Ne4 Qxd2+ 19.Nfxd2 Nxe4 20.Nxe4 Nd7 21.Bf4 Bf5 22.Bxd6 Bxe4 23.fxe4 Rxe4 24.Bd3 Re3 25.Rg3 Rxg3 26.Bxg3 Re8 27.Kc2 Be5 28.Bxe5 Nxe5 29.d6 Rd8 30.Bf1 Nf3 (20.787.966) 3461
21/23 0:06 +0.49 17...Nh5 18.Bh6 b5 19.cxb5 Qb4 20.Bxg7 Kxg7 21.Ng3 Nf4 (20.787.966) 3461

22/31 0:09 -0.78 17...Ned7 18.Ng3 a6 19.Bh6 Bh8 20.Bf4 b5 21.a3 b4 22.Nce4 Qa4 23.Qc2 Qxc2+ 24.Kxc2 b3+ 25.Kc1 Nxe4 26.Nxe4 Bf5 27.Bxd6 Bxe4 28.fxe4 Rxe4 29.Bd3 Re3 30.Rg3 Bd4 (32.345.354) 3412
22/35 0:09 0.00 17...a6 18.Ne4 Qxd2+ 19.Nfxd2 Nxe4 20.Nxe4 Nd7 21.Bf4 Bf5 22.Ng3 Bh3 23.Ne4 (32.345.354) 3412
22/31 0:09 +0.23 17...b5 18.cxb5 c4 19.Bxf6 Bxf6 20.Ne4 Qd8 21.Nxf6+ Qxf6 22.Ng3 Nd3+ 23.Kb1 Re3 24.Rc1 Qd4 25.Ne4 Bf5 26.Rxc4 Qxb2+ 27.Qxb2 Nxb2 28.Kxb2 Rxe2+ 29.Ka1 Bxe4 30.fxe4 Rxh2 (32.345.354) 3412
22/32 0:09 +0.25 17...Nh5 18.Bh6 b5 19.cxb5 Qb4 20.Bxg7 Kxg7 21.Ng3 Nf4 22.Kb1 Nxe2 23.Qxe2 Nc4 24.Nce4 Re5 25.Rd3 Bd7 26.Rb3 Qa4 27.f4 Rxe4 28.Nxe4 Re8 (32.345.354) 3412

Opponent plays 17...Nfd7?? and loses the game. I'll claim that in this game I played the best moves up to this point, despite what engines think (they'll all like the black side up to this point, or think it's equal.) After this point I don't know, I perhaps could have beaten them faster with some other moves, I chose what was the clearest victory, not the fastest one, but before this, I ranked my moves and played the BEST MOVE up to this position where the blunder was played, and I won.

Conclusion:

If the position is being played by white and black, and black is not a perfect opponent, then there's a probability that the black player makes a blunder, and loses the game, depending on their strength. THE BEST MOVE in the position is the one that maximizes those chances to the point at which blunder is unavoidable, so black will play it, and white will win.

If black is nobody and there's no game being played, then analyzing a chess position makes no sense, because black will not play a move, and they can't lose. We'd say the position is draw with best play, most moves are the same, and we move on with our life.

But that's like saying 1.f3 in the opening position is like the other moves, because it's draw with best play. Saying 1.d4/1.e4/1.c4/1.Nf3 are better only makes sense if black is someone you want to beat. But if black is someone you want to beat, and it's an imperfect opponent, then there's a BEST MOVE in the opening position, the one that is the shortest string of moves that allows you to mate them.
Make someone happy today.

zullil
Posts: 5562
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: Out of the Kai 450 positions, 99 not found by SF

Post by zullil » Sat Aug 17, 2019 11:08 am

Ovyron wrote:
Sat Aug 17, 2019 9:58 am

I'll claim that in this game I played the best moves up to this point, despite what engines think (they'll all like the black side up to this point, or think it's equal.) After this point I don't know, I perhaps could have beaten them faster with some other moves, I chose what was the clearest victory, not the fastest one, but before this, I ranked my moves and played the BEST MOVE up to this position where the blunder was played, and I won.
I examined your game, the one that you claimed just earned you the ICCF IM norm. Your claim that each of your first 17 moves was the BEST MOVE shows me that your definition of BEST MOVE is unusable and/or you are delusional. To rule out the latter possibility, would you please post the full info for that ICCF game? That way, we can at least verify that it was played.

Zenmastur
Posts: 389
Joined: Sat May 31, 2014 6:28 am

Re: Out of the Kai 450 positions, 99 not found by SF

Post by Zenmastur » Sat Aug 17, 2019 11:21 am

jp wrote:
Sat Aug 17, 2019 7:42 am

That's not the most extreme example. We were looking at a position in another thread and Leela & SF agreed to 14 or 20 plies, when we are not sure either of them was "right" or "understood" the position.
Which thread and position?

Regards,

Zenmastur
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

Uri Blass
Posts: 8533
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: Out of the Kai 450 positions, 99 not found by SF

Post by Uri Blass » Sat Aug 17, 2019 12:08 pm

zullil wrote:
Sat Aug 17, 2019 11:08 am
Ovyron wrote:
Sat Aug 17, 2019 9:58 am

I'll claim that in this game I played the best moves up to this point, despite what engines think (they'll all like the black side up to this point, or think it's equal.) After this point I don't know, I perhaps could have beaten them faster with some other moves, I chose what was the clearest victory, not the fastest one, but before this, I ranked my moves and played the BEST MOVE up to this position where the blunder was played, and I won.
I examined your game, the one that you claimed just earned you the ICCF IM norm. Your claim that each of your first 17 moves was the BEST MOVE shows me that your definition of BEST MOVE is unusable and/or you are delusional. To rule out the latter possibility, would you please post the full info for that ICCF game? That way, we can at least verify that it was played.
I found the game here

https://www.iccf.com/game?id=1019319

User avatar
Ovyron
Posts: 2348
Joined: Tue Jul 03, 2007 2:30 am

Re: Out of the Kai 450 positions, 99 not found by SF

Post by Ovyron » Sat Aug 17, 2019 12:11 pm

zullil wrote:
Sat Aug 17, 2019 11:08 am
Your claim that each of your first 17 moves was the BEST MOVE shows me that your definition of BEST MOVE is unusable
It's only "unusable" for silly "chess analysis" where all moves are equal because all of them go from a position that is drawn with best play to another that is drawn with best play. But very usable when you're against an imperfect opponent that might blunder, and since chess isn't solved, all opponents are such an opponent, so for any given game, positions have a BEST MOVE.

You don't understand because you don't contend, if you contended you'd realize that, just because you can easily find the moves in a chess position that defend it properly, it doesn't mean that your opponents can, so if you're superior to your opponents, and you know they're going to blunder in some positions, of course there's some moves that increase their chances more than others, and the ones that increase them the most are THE BEST MOVE.

I don't mind being called delusional by ignorant people that can't find a game, specially if the info I provide turns out to be hard to find, so if we played a game you couldn't find my games to prepare against me or whatever, so proving what I say is against my own interests :)

EDIT - And then Uri finds and posts it. The toothpaste is out of the bottle, oh well...

But I guess I can agree that there's NO BEST MOVE in positions that are drawn with best play where many moves lead to others that are drawn with best play. But that's not chess as in chess best play isn't known, so if THE BEST MOVES have been found in a testing suite and Stockfish is unable to find them in 99 of them, that'd mark a good path to improve it, at least on average, so some positions where THE BEST MOVE can be sacrificed to allow it to find it in others, would make the engine stronger overall.
Make someone happy today.

zullil
Posts: 5562
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: Out of the Kai 450 positions, 99 not found by SF

Post by zullil » Sat Aug 17, 2019 12:21 pm

Uri Blass wrote:
Sat Aug 17, 2019 12:08 pm
zullil wrote:
Sat Aug 17, 2019 11:08 am
Ovyron wrote:
Sat Aug 17, 2019 9:58 am

I'll claim that in this game I played the best moves up to this point, despite what engines think (they'll all like the black side up to this point, or think it's equal.) After this point I don't know, I perhaps could have beaten them faster with some other moves, I chose what was the clearest victory, not the fastest one, but before this, I ranked my moves and played the BEST MOVE up to this position where the blunder was played, and I won.
I examined your game, the one that you claimed just earned you the ICCF IM norm. Your claim that each of your first 17 moves was the BEST MOVE shows me that your definition of BEST MOVE is unusable and/or you are delusional. To rule out the latter possibility, would you please post the full info for that ICCF game? That way, we can at least verify that it was played.
I found the game here

https://www.iccf.com/game?id=1019319
Thanks. I can't believe how horrible the quality of the play is. At a time control of 10 moves in 50 days? With open access to chess engines?

Post Reply