Asperger only

Henk · Post by **Henk** » Mon May 16, 2016 11:01 am

Posted this position before. But I don't understand why my chess program plays a2-a1 and not Rg2. Or is a2-a1 better ? So first question: Is a2-a1 better.

[d] R7/7k/1b5p/4P3/6P1/3K4/p3r3/8 b - - 3 60

Code: Select all

Depth  Value   Time&#40;seconds&#41;   Nodes
 28    3.2093      61.05    57935272   a2-a1 Ra8xa1 Re2xe5 Ra1-b1 
 27    3.2093      71.03    67471208   a2-a1 Ra8xa1 Re2xe5 Ra1-b1 Bb6-a5 Rb1-c1 Re5-b5 Kd3-e4 Ba5-b6 Rc1-f1 Kh7-g7 Ke4-f3 Rb5-b2 Kf3-e4 
 26    3.2130      23.39    22498571   a2-a1 Ra8xa1 Re2xe5 Ra1-a6 
 25    3.2130      11.47    11216382   a2-a1 Ra8xa1 Re2xe5 Ra1-a6 
 24    3.2130      12.73    12734384   a2-a1 Ra8xa1 Re2xe5 Ra1-a6 Bb6-e3 Kd3-e2 Re5-e7 Ke2-d3 Kh7-g7 Ra6-a2 Be3-a7 
 23    3.1855       2.53     2353502   a2-a1 Ra8xa1 Re2xe5 
 22    3.1855       5.08     4722526   a2-a1 Ra8xa1 Re2xe5 Ra1-a2 Kh7-g6 Ra2-a6 Re5-b5 Kd3-c4 Rb5-b1 Ra6-a2 Bb6-e3 Ra2-h2 
 21    3.1769       0.75      875752   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-a2 Kh7-h8 Kd2-d3 Bb6-e3 
 20    3.1715       1.06      614016   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-a2 Kh7-h8 Kd2-d3 
 19    3.1629       0.40      472133   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-a2 Kh7-h8 Kd2-d3 
 18    3.1661       0.23      264340   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-a2 Kh7-h8 Kd2-d3 
 17    3.1672       0.15      167356   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-a2 Kh7-h8 
 16    3.1571       0.10      115015   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-a2 Kh7-h8 
 15    3.1665       0.06       72769   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-a2 Kh7-h8 
 14    3.1326       0.03       37595   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-f1 
 13    3.1476       0.03       31776   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-f1 
 12    3.1330       0.02       22758   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 Ra1-f1 
 11    3.1664       0.01       14482   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 
 10    3.1707       0.01        7495   a2-a1 Ra8xa1 Re2-e3 
  9    3.1707       0.00        5396   a2-a1 Ra8xa1 Re2-e3 
  8    3.1707       0.00        4497   a2-a1 Ra8xa1 Re2-e3 Kd3-d2 Re3xe5 
  7    3.1568       0.00        1087   a2-a1 
  6    3.1568       0.00        2249   a2-a1 Ra8xa1 Re2xe5 Ra1-f1 
  5    3.2624       0.00         967   Re2-e3 
  4    3.2624       0.00         666   Re2-e3 
  3    3.2624       0.00        1667   Re2-e3 Kd3-c4 Re3xe5 
  2    2.9166       0.00         332   a2-a1 
  1    3.2740       0.00         124   Re2xe5

AndrewGrant · Post by **AndrewGrant** » Mon May 16, 2016 11:56 am

Ran it through my engine quickly... results...

info depth 1 score cp 328 time 0 nodes 139 pv e2f2
info depth 2 score cp 328 time 0 nodes 518 pv e2g2
info depth 3 score cp 338 time 0 nodes 1022 pv e2g2
info depth 4 score cp 338 time 0 nodes 2958 pv e2g2
info depth 5 score cp 341 time 0 nodes 7461 pv e2g2
info depth 6 score cp 343 time 0 nodes 24002 pv e2g2
info depth 7 score cp 340 time 0 nodes 50899 pv e2g2
info depth 8 score cp 346 time 0 nodes 149695 pv e2g2
info depth 9 score cp 346 time 1000 nodes 234630 pv e2g2
info depth 10 score cp 397 time 1000 nodes 694416 pv e2g2
info depth 11 score cp 407 time 1000 nodes 1242625 pv e2g2
info depth 12 score cp 407 time 1000 nodes 2189437 pv e2g2
info depth 13 score cp 417 time 1000 nodes 4116085 pv e2g2
info depth 14 score cp 407 time 2000 nodes 7943049 pv e2g2

If I had to take a guess at the issue, it would be this.

Your engine does not understand that promoting that pawn is worth (QueenValue - PawnValue).

Henk · Post by **Henk** » Mon May 16, 2016 12:06 pm

Yes probably somehow it thinks e5-e6 > a2-a1. Will be tough debugging session (as usual).

Joost Buijs · Post by **Joost Buijs** » Mon May 16, 2016 12:46 pm

Nightmare also thinks about Rg2, but after 10 seconds it switches to Rxe5 which seems to lead to a forced mate in 31 moves.

Code: Select all

info depth 11 score cp 508 time 83 nodes 176430 nps 2125662 tbhits 252 pv e2g2 e5e6 b6d8 d3c4 d8f6 e6e7 g2g4 c4d3 f6e7 d3c2 h7g6
info depth 12 score cp 543 time 249 nodes 769381 nps 3089883 tbhits 1281 pv e2g2 a8a6 g2g3 d3c4 g3g4 c4d5 g4g2 d5c4 h6h5 c4b3 b6d4 a6a2
info depth 13 score cp 581 time 446 nodes 1403255 nps 3146311 tbhits 3380 pv e2g2 d3c4 g2g4 c4d5 g4g2 d5c4 g2e2 c4d5 b6c7 e5e6 c7d8 d5c4 d8f6
info depth 14 score cp 618 time 771 nodes 2259647 nps 2930800 tbhits 6479 pv e2g2 a8a6 g2g3 d3c4 g3g4 c4d5 g4g2 d5c4 g2e2 c4d5 h7g7 a6a3 g7f7 a3f3
info depth 15 score mate 32 time 9476 nodes 71295348 nps 7523780 tbhits 136535 pv e2e5 a8a2 e5e3 d3d2 e3e4 a2a6 b6g1 d2d1 e4g4 d1c2
info depth 16 score mate 31 time 15517 nodes 115483075 nps 7442358 tbhits 261217 pv e2e5 a8a2 e5e3 d3d2 e3e4 a2a6 b6d8 d2d3 e4g4 a6h6
info depth 17 score mate 31 time 27178 nodes 216671543 nps 7972313 tbhits 519743 pv e2e5 a8a2 e5e3 d3d2 e3e4 a2a6 b6d8 d2d3 e4g4 a6h6
info depth 18 score mate 31 time 48595 nodes 419645326 nps 8635565 tbhits 1072813 pv e2e5 a8a2 e5e3 d3d2 e3e4 a2a6 b6d8 d2d3 e4g4 a6h6
info depth 19 score mate 31 time 89689 nodes 821254035 nps 9156686 tbhits 2276752 pv e2e5 a8a2 e5e3 d3d2 e3e4 a2a6 b6d8 d2d3 e4g4
info depth 20 score mate 31 time 181204 nodes 1810365416 nps 9990758 tbhits 5093848 pv e2e5 a8a2 e5e3 d3d2 e3e4 a2a6 b6d8 d2d3 e4g4

Robert Pope · Post by **Robert Pope** » Mon May 16, 2016 3:38 pm

Abbess bounces around several e2 moves, with a score of 350-425. If I force a1=Q, it scores it at 331, so not much weaker. I don't think it would take a lot to get piece square tables to make you prefer one over the other.

If you force Rg2 or RxP, what does your engine evaluate on the next move?

Daniel Anulliero · Post by **Daniel Anulliero** » Mon May 16, 2016 3:44 pm

Isa 1.7.41
It like e2g2 but it change to e2xe5 ,just like Nightmare

But I don't know if Isa can find the Mate in a raisonable time !

Code: Select all

P Eval	T Nodes	PV
------------------------------------
1 +3.51 0 106   	e2b2
2 +3.24 0 184   	e2b2 d3c3
3 +3.53 0 1311   	e2f2 a8a6 f2b2
4 +3.30 0 5206	e2g2 e5e6 g2g3 d3d2 h7g6 a8a2 g3g4
5 +3.24 0 9591	e2g2 a8a6 b6d8 d3c3 h7g7
6 +3.32 0 46074	e2f2 e5e6 b6c5 a8a5 c5d6 d3d4
7 +3.38 0 60546	e2f2 e5e6 b6c5 d3c4 c5d6 c4b3 f2f4
8 +3.45 0 442923   	e2g2 a8a6 b6g1 a6a4 h7g6 d3c3 g6g5 a4a6 g5g4 a6h6
9 +3.67 1 523001	e2g2 a8a6 b6g1 a6a4 h7g6 d3c3 g6g5 a4a6 g1e3
10 +4.73 3 1390733	e2e5 a8a2 e5e3 d3c4 e3g3 c4d5 g3g4 a2e2 b6d4 e2d2 d4e3
11 +4.71 7 3049942	e2e5 a8a2 e5e3 d3c4 e3g3 c4d5 g3g4 a2e2 g4d4 d5e5 h7g6 e2g2 g6h5 g2g7
12 +4.74 17 7365418	e2e5 a8a2 e5e3 d3c4 e3g3 c4d5 g3g4 a2b2 g4g5 d5e4 b6c5 b2h2 h7g6 h2h1
13 +4.76 58 27920056e2e5 a8a2 e5e3 d3c4 e3g3 c4d5 g3g4 a2b2 g4g5 d5e4 b6c5 b2a2 h7g6 a2a6 g6g7 a6a1
-----------------------------------------------------------------------------------------------------------------------

Joost Buijs · Post by **Joost Buijs** » Tue May 17, 2016 12:50 pm

Nightmare finds a forced line to a table-base position which is mate in xxx, without table-base it will never find it.
It is using 6 men Nalimov on SSD and probes at the leaves of the tree, but not in quiescence.
With the current crop of fast SSD's this is doable without much speed degradation.
With Nalimov on HD it usually probes 6 to 8 ply from the leaves, in this case the problem is the bad seek-time of mechanical disks.

Henk · Post by **Henk** » Tue May 17, 2016 1:15 pm

Maybe Re2-f2 is because of overrated rook mobility. (LMR disabled)

Code: Select all

Depth  Value   Time&#40;seconds&#41;   Nodes
 15    3.9792      80.35    77110497   Re2-h2 Kd3-c4 h6-h5 g4-g5 Rh2-b2 Kc4-c3 Rb2-e2 Ra8-a4 Kh7-g7 Kc3-b3 Re2xe5 Kb3xa2 Bb6-c5 Ka2-a1 Re5xg5 
 14    3.9792      20.62    19409147   Re2-h2 Kd3-c4 h6-h5 g4-g5 Rh2-e2 Kc4-b3 Re2xe5 Ra8xa2 Re5xg5 Ra2-a4 Bb6-e3 Kb3-c2 Kh7-h8 Kc2-b3 
 13    3.9792      30.89    30320030   Re2-h2 Kd3-c4 h6-h5 g4-g5 Rh2-e2 Kc4-b3 Re2xe5 Ra8xa2 Re5xg5 Ra2-a4 Bb6-e3 Kb3-c2 Kh7-h8 
 12    3.9600       5.68     5828059   Re2-g2 Kd3-c4 h6-h5 g4xh5 Rg2-h2 Kc4-b3 Rh2xh5 e5-e6 Rh5-e5 Kb3xa2 Re5xe6 Ka2-a1 
 11    3.9536       1.75     1681378   Re2-h2 Kd3-c3 h6-h5 g4-g5 Rh2-e2 Kc3-b3 Re2xe5 Kb3xa2 Bb6-c5 Ra8-d8 Re5xg5 
 10    3.9536       0.90     1019693   Re2-h2 Kd3-c3 h6-h5 g4-g5 Rh2-e2 Kc3-b3 Re2xe5 Kb3xa2 Bb6-c5 Ra8-d8 
  9    4.0496       0.25      313421   Re2-h2 Kd3-c3 h6-h5 g4xh5 Rh2xh5 Ra8xa2 Rh5-e5 Ra2-h2 Kh7-g6 Kc3-b2 
  8    4.0496       0.63      771280   Re2-h2 Kd3-c3 h6-h5 g4xh5 Rh2xh5 Ra8xa2 Rh5-e5 Kc3-b2 
  7    4.9328       0.09      105446   Re2-f2 e5-e6 Bb6-d8 Kd3-e3 Rf2-b2 Ra8-a7 Kh7-h8 Ke3-d3 
  6    4.9712       0.02       25745   Re2-f2 e5-e6 Bb6-c5 Kd3-c3 Bc5-e7 Kc3-c4 
  5    4.9840       0.01       10574   Re2-f2 e5-e6 Bb6-c5 Kd3-e4 Rf2-b2 
  4    4.9712       0.01        6475   Re2-f2 e5-e6 Bb6-c5 Kd3-e4 
  3    4.9712       0.00        2128   Re2-f2 e5-e6 Rf2-b2 
  2    4.9712       0.00         924   Re2-f2 e5-e6 
  1    5.2656       0.05         203   Re2-f2

Daniel Anulliero · Post by **Daniel Anulliero** » Tue May 17, 2016 1:28 pm

Joost Buijs wrote:Nightmare finds a forced line to a table-base position which is mate in xxx, without table-base it will never find it.
It is using 6 men Nalimov on SSD and probes at the leaves of the tree, but not in quiescence.
With the current crop of fast SSD's this is doable without much speed degradation.
With Nalimov on HD it usually probes 6 to 8 ply from the leaves, in this case the problem is the bad seek-time of mechanical disks.

Aah ok , thanks for the explanations

Asperger only

Asperger only

Re: Asperger only

Re: Asperger only

Re: Asperger only

Re: Asperger only

Re: Asperger only

Re: Asperger only

Re: Asperger only

Re: Asperger only