SFNNUE Fails!! In this Positions!!

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

Moderators: Harvey Williamson, Dann Corbit, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Master Om
Posts: 412
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: SFNNUE Fails!! In this Positions!!

Post by Master Om » Wed Nov 11, 2020 6:13 am

Ovyron wrote:
Tue Nov 10, 2020 8:24 pm
Master Om wrote:
Fri Oct 30, 2020 2:27 pm
Irrelevant for the context of finding the BEST move in a position.
It's relevant because you haven't shown WHY is a shorter game better than a longer game.

And you haven't shown WHY is a shorter game better than a longer one ONLY if you're winning the game, BUT if you're losing the game then making it longer is suddenly better? It's all arbitrary!
Strawman argument.
In the post I mentioned about finding the best move and not about any game. Its moronic to think its always between best between wins. Its most of time a decision between getting the chance of a win. It can be in quiet positions.
[d]2b1rk2/5p2/p1P5/2p2P2/2p5/7B/P7/2KR4 w - - 0 1
For example at this position
SF NNUUE latest dev from abrok still cant find the best move to win here. And the eval is so misleading even to depth 50 even.

Code: Select all

 [+0.00]  d=42  1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rh8 4.Rd5 Bb7 5.Rd1(0:04:32) 90670kN
 [+0.00]  d=41  1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rh8 4.Rd5 Bb7 5.Rd1(0:04:12) 85999kN
 [+0.00]  d=40  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8(0:04:01) 81843kN
 [+0.00]  d=39  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8 4.Rd1(0:04:01) 81836kN
 [+0.00]  d=38  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:01:26) 38398kN
 [+0.00]  d=37  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.Rd5(0:01:23) 37411kN
 [+0.00]  d=36  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:55) 25807kN
 [+0.00]  d=35  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:46) 22404kN
 [+0.00]  d=34  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:35) 17784kN
 [+0.00]  d=33  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:28) 14949kN
 [+0.00]  d=32  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8(0:00:13) 9003kN
 [+0.00]  d=31  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8(0:00:10) 7956kN
 [+0.00]  d=30  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.Rd5(0:00:07) 6229kN
 [+0.00]  d=29  1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rh8 4.Rd5 Bb7 5.Rd2 Bc8 6.Rd1 Bxf5 7.Bxa6 Ra8 8.Bb7 Rh8 9.Ba6(0:00:03) 3819kN
 [+0.00]  d=28  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5(0:00:03) 3606kN
 [+0.00]  d=27  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:01) 2737kN
 [+0.00]  d=26  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.Rd5(0:00:01) 2195kN
 [+0.00]  d=25  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8(0:00:00) 1257kN
 [+0.00]  d=24  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8 4.Rd1(0:00:00) 1255kN
 [+0.00]  d=23  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2(0:00:00) 1131kN
Always Expect the Unexpected

User avatar
MikeB
Posts: 4500
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: SFNNUE Fails!! In this Positions!!

Post by MikeB » Thu Nov 12, 2020 4:28 am

Master Om wrote:
Wed Nov 11, 2020 6:13 am
Ovyron wrote:
Tue Nov 10, 2020 8:24 pm
Master Om wrote:
Fri Oct 30, 2020 2:27 pm
Irrelevant for the context of finding the BEST move in a position.
It's relevant because you haven't shown WHY is a shorter game better than a longer game.

And you haven't shown WHY is a shorter game better than a longer one ONLY if you're winning the game, BUT if you're losing the game then making it longer is suddenly better? It's all arbitrary!
Strawman argument.
In the post I mentioned about finding the best move and not about any game. Its moronic to think its always between best between wins. Its most of time a decision between getting the chance of a win. It can be in quiet positions.
[d]2b1rk2/5p2/p1P5/2p2P2/2p5/7B/P7/2KR4 w - - 0 1
For example at this position
SF NNUUE latest dev from abrok still cant find the best move to win here. And the eval is so misleading even to depth 50 even.

Code: Select all

 [+0.00]  d=42  1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rh8 4.Rd5 Bb7 5.Rd1(0:04:32) 90670kN
 [+0.00]  d=41  1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rh8 4.Rd5 Bb7 5.Rd1(0:04:12) 85999kN
 [+0.00]  d=40  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8(0:04:01) 81843kN
 [+0.00]  d=39  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8 4.Rd1(0:04:01) 81836kN
 [+0.00]  d=38  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:01:26) 38398kN
 [+0.00]  d=37  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.Rd5(0:01:23) 37411kN
 [+0.00]  d=36  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:55) 25807kN
 [+0.00]  d=35  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:46) 22404kN
 [+0.00]  d=34  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:35) 17784kN
 [+0.00]  d=33  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:28) 14949kN
 [+0.00]  d=32  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8(0:00:13) 9003kN
 [+0.00]  d=31  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8(0:00:10) 7956kN
 [+0.00]  d=30  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.Rd5(0:00:07) 6229kN
 [+0.00]  d=29  1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rh8 4.Rd5 Bb7 5.Rd2 Bc8 6.Rd1 Bxf5 7.Bxa6 Ra8 8.Bb7 Rh8 9.Ba6(0:00:03) 3819kN
 [+0.00]  d=28  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5(0:00:03) 3606kN
 [+0.00]  d=27  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2 Bxf5 12.Bxa6(0:00:01) 2737kN
 [+0.00]  d=26  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.Rd5(0:00:01) 2195kN
 [+0.00]  d=25  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8(0:00:00) 1257kN
 [+0.00]  d=24  1.c7 f6 2.Bg2 Ke7 3.Re1 Kf8 4.Rd1(0:00:00) 1255kN
 [+0.00]  d=23  1.Rd5 f6 2.Kd2 Re4 3.Rd8 Re8 4.c7 Ke7 5.Kc3 Rxd8 6.cxd8Q Kxd8 7.Kxc4 Kc7 8.Kxc5 Bb7 9.Bg4 Bg2 10.Bd1 Be4 11.Be2(0:00:00) 1131kN
Black Diamond using the Night Nurse 0.3 net find sit it less than 15 seconds on an Android Phone.

Will release this tomorrow on the Android ...
Right click and select open in new window to see full size screenshot
Image
Image

User avatar
Master Om
Posts: 412
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: SFNNUE Fails!! In this Positions!!

Post by Master Om » Thu Nov 12, 2020 7:44 am

Hi Mike ,
Your program is a gem. Wish if u could make a Bluefish XII FD XR2 NNUE.
Also try modern dev of SF of Cfish they dont find this move.
Always Expect the Unexpected

User avatar
Ovyron
Posts: 4405
Joined: Tue Jul 03, 2007 2:30 am

Re: SFNNUE Fails!! In this Positions!!

Post by Ovyron » Tue Nov 17, 2020 5:36 am

Master Om wrote:
Wed Nov 11, 2020 5:42 am
There are trolls like that.
All the people that continue playing the game when they're getting mated are trolls.

You're the one building straw men, this is my actual statement:

There's no objective argument of why a mate in 10 is better than a mate in 11.

In that situation, the losing player is a troll who should resign, the winning player isn't required to follow any etiquette.

I don't know how your other comments relate to this, in my statement the moves played still mate.

Jouni
Posts: 2251
Joined: Wed Mar 08, 2006 7:15 pm

Re: SFNNUE Fails!! In this Positions!!

Post by Jouni » Tue Nov 17, 2020 2:26 pm

Crystal territory here:

Analysis by Crystal 110920:

..
1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rf8 4.Rd5 Bb7 5.Rd2 Bc8
= (0.00) Depth: 27/35 00:00:07 74350kN, tb=209750
1.f6
= (0.08 ++) Depth: 28/39 00:00:08 86808kN, tb=288821
1.f6
= (0.16 ++) Depth: 28/39 00:00:08 88172kN, tb=300312
1.f6
= (0.28 ++) Depth: 28/39 00:00:08 89368kN, tb=311017
1.f6
+/= (0.46 ++) Depth: 28/41 00:00:08 90947kN, tb=322611
1.f6
+/- (0.71 ++) Depth: 28/41 00:00:08 92366kN, tb=333707
1.f6
+/- (1.03 ++) Depth: 28/45 00:00:08 94914kN, tb=365312
1.f6
+/- (1.47 ++) Depth: 28/48 00:00:09 98525kN, tb=404705
1.f6
+- (2.03 ++) Depth: 28/48 00:00:09 102mN, tb=442457
1.f6
+- (2.75 ++) Depth: 28/48 00:00:09 106mN, tb=490159
1.f6
+- (3.68 ++) Depth: 28/48 00:00:10 114mN, tb=555530
1.f6
+- (4.87 ++) Depth: 28/48 00:00:11 134mN, tb=703599
1.f6
+- (6.37 ++) Depth: 28/59 00:00:15 180mN, tb=1113719
1.f6
+- (8.26 ++) Depth: 28/78 00:00:30 389mN, tb=2693154
Jouni

User avatar
Master Om
Posts: 412
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: SFNNUE Fails!! In this Positions!!

Post by Master Om » Fri Nov 20, 2020 12:12 pm

Jouni wrote:
Tue Nov 17, 2020 2:26 pm
Crystal territory here:

Analysis by Crystal 110920:

..
1.c7 f6 2.Bf1 Ke7 3.Bxc4 Rf8 4.Rd5 Bb7 5.Rd2 Bc8
= (0.00) Depth: 27/35 00:00:07 74350kN, tb=209750
1.f6
= (0.08 ++) Depth: 28/39 00:00:08 86808kN, tb=288821
1.f6
= (0.16 ++) Depth: 28/39 00:00:08 88172kN, tb=300312
1.f6
= (0.28 ++) Depth: 28/39 00:00:08 89368kN, tb=311017
1.f6
+/= (0.46 ++) Depth: 28/41 00:00:08 90947kN, tb=322611
1.f6
+/- (0.71 ++) Depth: 28/41 00:00:08 92366kN, tb=333707
1.f6
+/- (1.03 ++) Depth: 28/45 00:00:08 94914kN, tb=365312
1.f6
+/- (1.47 ++) Depth: 28/48 00:00:09 98525kN, tb=404705
1.f6
+- (2.03 ++) Depth: 28/48 00:00:09 102mN, tb=442457
1.f6
+- (2.75 ++) Depth: 28/48 00:00:09 106mN, tb=490159
1.f6
+- (3.68 ++) Depth: 28/48 00:00:10 114mN, tb=555530
1.f6
+- (4.87 ++) Depth: 28/48 00:00:11 134mN, tb=703599
1.f6
+- (6.37 ++) Depth: 28/59 00:00:15 180mN, tb=1113719
1.f6
+- (8.26 ++) Depth: 28/78 00:00:30 389mN, tb=2693154
Absolutely..... Awesome engine that is...
Always Expect the Unexpected

User avatar
Master Om
Posts: 412
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: SFNNUE Fails!! In this Positions!!

Post by Master Om » Fri Nov 20, 2020 12:56 pm

Ovyron wrote:
Tue Nov 17, 2020 5:36 am
All the people that continue playing the game when they're getting mated are trolls.
Irrelevant with respect to the topic posted.
You're the one building straw men, this is my actual statement:
The one who is replying with a bogus diverted argument is the strawman. I stand corrected what i have posted.
There's no objective argument of why a mate in 10 is better than a mate in 11.
Already told. The diff is not between 10 and 11. The difference is a forced win and and trolling for 50 moves.
Even if u miss 1 move means its your inability to find shortest path.
In that situation, the losing player is a troll who should resign, the winning player isn't required to follow any etiquette
.

There is no such rule. One can play as he wishes. If u don't have patience then CC is not for u. use conditional moves to optimum level.
When u have forced mate in 10 ( which starts from finding the BEST MOVE which leads to SHORTEST WIN) u cant be trolled to 50 moves as its forced.
I don't know how your other comments relate to this, in my statement the moves played still mate.
The OP was to find the BEST Move which wins INSTANTLY. Thats it. And best is a single move which outnumbers others in eval.
Always Expect the Unexpected

User avatar
hgm
Posts: 25423
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: SFNNUE Fails!! In this Positions!!

Post by hgm » Fri Nov 20, 2020 2:09 pm

Going for the fastest mate is not always the best. I remember that from when I still played Chess myself (which must be 40 years ago!). I was often criticized by my team mates for my end-game play: "Why did you play that? Now it will take much longer to win!" When I then sat down at the board again the opponent usually resigned immediately. The path to the fastest mate is often wild and treacherous, which gives your opponent hope. The trick is to show that you won't allow them any chances.

For computers going for the fastest mate is a good way to make sure you actually reach a mate. Going for the highest heuristic eval... well, that is only as good as your eval. High eval doesn't necessarily mean fast win. KBNK has a pretty high eval, but can take over 30 moves.

When you are losing, going for the longest game is often a fatal mistake. If you can postpone being checkmated by one move by a spite check like Qxh7+, this will actually be the worst thing you could do. The opponent will not overlook he can reacapture the Queen with Kxh7. (It could actually be the only legal move he has.) After that the mate is as close as it was, and it will be easier to find, as the branching ratio of the tree is smaller because you no longer have a Queen. If the opponent was a computer. If it was a human you have alerted him to the fact that he can checkmate you. And when this did not help enough to make the opponent see the mate that he might otherwise have overlooked, the game is decided in his advantage anyway.
Get rid of the shit: vote for SHID!

User avatar
Master Om
Posts: 412
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: SFNNUE Fails!! In this Positions!!

Post by Master Om » Fri Nov 20, 2020 2:58 pm

hgm wrote:
Fri Nov 20, 2020 2:09 pm
Going for the fastest mate is not always the best. I remember that from when I still played Chess myself (which must be 40 years ago!). I was often criticized by my team mates for my end-game play: "Why did you play that? Now it will take much longer to win!" When I then sat down at the board again the opponent usually resigned immediately. The path to the fastest mate is often wild and treacherous, which gives your opponent hope. The trick is to show that you won't allow them any chances.

For computers going for the fastest mate is a good way to make sure you actually reach a mate. Going for the highest heuristic eval... well, that is only as good as your eval. High eval doesn't necessarily mean fast win. KBNK has a pretty high eval, but can take over 30 moves.

When you are losing, going for the longest game is often a fatal mistake. If you can postpone being checkmated by one move by a spite check like Qxh7+, this will actually be the worst thing you could do. The opponent will not overlook he can reacapture the Queen with Kxh7. (It could actually be the only legal move he has.) After that the mate is as close as it was, and it will be easier to find, as the branching ratio of the tree is smaller because you no longer have a Queen. If the opponent was a computer. If it was a human you have alerted him to the fact that he can checkmate you. And when this did not help enough to make the opponent see the mate that he might otherwise have overlooked, the game is decided in his advantage anyway.
Going for the fastest mate is not always the best. I remember that from when I still played Chess myself (which must be 40 years ago!).
Not in Correspondence chess may be as time is a big factor.
For computers going for the fastest mate is a good way to make sure you actually reach a mate. Going for the highest heuristic eval... well, that is only as good as your eval. High eval doesn't necessarily mean fast win. KBNK has a pretty high eval, but can take over 30 moves.
Agree
Always Expect the Unexpected

User avatar
Master Om
Posts: 412
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: SFNNUE Fails!! In this Positions!!

Post by Master Om » Fri Nov 20, 2020 3:07 pm

hgm wrote:
Fri Nov 20, 2020 2:09 pm
When you are losing, going for the longest game is often a fatal mistake.
I can say my school and college days where i have won lost games only by swindling and delaying (and when i was losing side) and my opponent most of the time broke down in time trouble. I have also won while delaying in won positions as a kid i was Dr Lasker fan and i always wanted to play the move that tortures the opponent rather than play for win. In CC thats not the case.
Not once but many times in tournaments.
Always Expect the Unexpected

Post Reply