Zangdar 7.0 (something is wrong, game example)

[Frank Quisinsky]

[pgn][Event "5 Minutes/Game + 2 Seconds/Move"]
[Site "etoc-g_t193, F-4"]
[Date "2026.08.08"]
[Round "1.26"]
[White "Spark 1.0 HCE"]
[Black "Zangdar 7.0"]
[Result "1-0"]

1. e4 {book 0s} c6 {book 0s} 2. d4 {book 0s} d5 {book 0s}
3. Nd2 {book 0s} dxe4 {book 0s} 4. Nxe4 {book 0s} Bf5 {book
0s} 5. Ng3 {book 0s} Bg6 {book 0s} 6. h4 {book 0s} h6 {book
0s} 7. Nf3 {book 0s} Nd7 {book 0s} 8. h5 {book 0s} Bh7
{book 0s} 9. Bd3 {book 0s} Bxd3 {book 0s} 10. Qxd3 {book
0s} e6 {book 0s} 11. Bf4 {book 0s} Bb4+ {book 0s} 12. c3
{book 0s} Be7 {book 0s} 13. O-O-O {+0.47/17 9s} Ngf6
{+0.62/35 15s} 14. Kb1 {+0.38/17 22s} Qa5 {+0.24/32 10s
(a5)} 15. Ne5 {+0.54/16 10s} O-O {+0.19/30 6s} 16. Nxd7
{+0.47/15 8s} Nxd7 {+0.54/31 8s} 17. Ka1 {+0.40/16 20s} Nf6
{0.00/34 12s} 18. Ne4 {+0.34/16 9s} Nxe4 {+0.01/33 7s}
19. Qxe4 {+0.24/17 7s} Rad8 {0.00/34 8s} 20. g4 {+0.23/17
14s} Rd7 {+0.34/32 20s (Bd6)} 21. Rhg1 {+0.27/16 11s} Rfd8
{0.00/33 7s (Qd5)} 22. Rd3 {+0.23/14 7s} Qd5 {-0.44/30 7s
(Qb5)} 23. Qxd5 {-0.03/18 7s} Rxd5 {-0.35/33 6s} 24. Rgd1
{-0.21/18 8s} c5 {-0.27/34 15s} 25. Bc7 {0.00/21 7s} R8d7
{-0.67/35 13s (Rc8)} 26. Be5 {-0.18/20 5s} c4 {-0.42/33 7s
(b5)} 27. R3d2 {-0.21/20 5s} Bf6 {-0.55/36 15s (f6)}
28. Bxf6 {-0.09/20 5s} gxf6 {-0.34/32 4s} 29. Rg1 {+0.01/21
8s} b5 {-0.47/32 4s (e5)} 30. Kb1 {+0.14/19 6s} a5
{-0.57/31 6s (Kh7)} 31. Kc2 {+0.13/19 15s} b4 {-0.45/32 9s
(Rd8)} 32. Re2 {+0.22/18 8s} Rb7 {-0.41/33 23s (Rd8)}
33. f4 {+0.23/16 4s} Kf8 {-0.33/31 4s (Kh7)} 34. Rf1
{+0.27/15 4s} Rb8 {-0.43/36 22s (Rbb5)} 35. Rff2 {+0.24/18
25s} Rbb5 {-0.39/32 12s (Ke7)} 36. Rf1 {+0.22/17 16s} Rb8
{-0.35/35 7s} 37. Rff2 {+0.27/18 4s} Rb6 {-0.25/35 16s
(Ke7)} 38. Re1 {+0.33/18 16s} Rdb5 {-0.13/32 8s (Rbb5)}
39. Kd2 {+0.51/16 6s} Rd5 {-0.06/31 6s (Rb8)} 40. Ke3
{+0.43/16 3s} a4 {-0.59/29 4s (bxc3)} 41. Rc1 {+0.24/17 7s}
f5 {-0.26/29 3s (Ke7)} 42. g5 {+0.21/17 3s} Kg7 {0.00/29
4s} 43. Rg2 {+0.11/16 4s} Rdb5 {-0.17/31 4s} 44. b3
{+0.11/16 3s} cxb3 {-0.46/30 3s} 45. c4 {0.00/17 8s} bxa2
{-1.73/29 2s} 46. cxb5 {0.00/17 3s} b3 {0.00/42 2s}
47. gxh6+ {0.00/16 3s} Kh7 {0.00/41 2s (Kxh6)} 48. Rg7+
{+14.36/17 2s} Kxh6 {0.00/43 2s} 49. Rxf7 {+35.76/19 8s} e5
{0.00/43 2s (a1Q)} 50. dxe5 {+35.76/14 7s} a1=Q {+20.20/34
24s (Rg6)} 51. Rxa1 {+M11/15 7s} b2 {+26.70/24 2s} 52. Rg1
{+M9/14 4s} b1=Q {+92.16/29 9s} 53. Rxb1 {+M8/14 2s} Rb8
{+M16/25 5s (a3)} 54. Rg1 {+M7/12 1s} Rc8 {+M6/30 4s
(Rxb5)} 55. Rg6+ {+M6/10 0s} Kxh5 {+M5/41 1s} 56. Rff6
{+M5/8 0s} Rc3+ {+M4/51 1s} 57. Kd4 {+M4/6 0s} Rc4+ {+M3/72
8s (Kh4)} 58. Kxc4 {+M3/4 0s} a3 {+M2/128 0s} 59. Rg5+
{+M2/2 0s} Kh4 {+M1/128 0s} 60. Rh6# {+M1/1 0s} 1-0[/pgn]

[Carbec]

Hi,

The move 47... Kh7 0.00/41 2s is indeed losing very quickly. I can't reproduce the problem, as Zangdar play the right move. I loaded the game in Arena, and made Zangdar analyse the move.
Well, Im lost for the moment.

[Frank Quisinsky]

Hi,

I have no idea either. I looked up the game in the database and wondered why Spark won. Maybe it's because of the 5-pieces syzygy and possibly the accesses given the short response time. I don't know.

Best
Frank

[Tibono]

Hi,
I copy/pasted the moves into a spreadsheet and summed black's thinking times. Of course it is not very accurate as only seconds are reported, yet on average this might give a reasonnable clue. I found 4mins and 59secs used after move #46 and 5mins and 1sec for next move 47...Kh7.
Game pace was 5 Minutes/Game + 2 Seconds/Move; maybe the weird occurrence relates to that very moment.
And maybe the issue lies in the GUI (Shredder 12, if I am correct); or a misfit between the GUI and the engine.

[Frank Quisinsky]

Good job!
That might be the problem.

So it would make sense to check other lost games from the ongoing test run. I think that should be done in a day and a half. I'll check this using the games Zandar played.

ETOC-G detail site:
https://www.amateurschach.de/main/_etoc-g.htm
Conditions or "parameters", later the games can be found here ...
Yes, Shredder 12 GUI.

Best
Frank

[Frank Quisinsky]

Hi,

I am looking in the lost games.
Difficult!

4 or 5 lost games I saw a consistant negative eval from Zangdar, and suddenly 0,00 after few seconds in moves after. Again after the eval goes higher for the opponent.

Really not easy, means to search and find here a bit.

Again, in good a day the test-run is over and you can download the games.

Need a bit, at the moment 4 engines are playing simultaneous.
4 engines x 4 games = 16 games are still running. So it need a bit longer.

Best
Frank

[Frank Quisinsky]

Zangdar 7.0:
After 0.860 out of 1.300 games.

https://www.amateurschach.de/download/g ... dar_70.zip

Note:
One of the ETOC-G rules:
Replay games: Every draw less than 20 moves, every game with more than 299 moves

Possible that such games are inside the database, not selected out (the database with 860 games).

Best
Frank

[Carbec]

Hi,
I looked in the time/move :

Code: Select all

   Spark 1.0 HCE - Zangdar 7.0   1-0   (etoc-g_t193.pgn, partie 27, ronde 1.26)
  Cadence 5'+2"   dispo = pendule avant le coup   t = temps du coup   S = cumul   (s)

    #  Spark   dispo   t    S    ev/prof   Zangdar dispo   t    S    ev/prof
  --------------------------------------------------------------------------
    1  e4        300   0    0       book   c6        300   0    0       book
    2  d4        302   0    0       book   d5        302   0    0       book
    3  Nd2       304   0    0       book   dxe4      304   0    0       book
    4  Nxe4      306   0    0       book   Bf5       306   0    0       book
    5  Ng3       308   0    0       book   Bg6       308   0    0       book
    6  h4        310   0    0       book   h6        310   0    0       book
    7  Nf3       312   0    0       book   Nd7       312   0    0       book
    8  h5        314   0    0       book   Bh7       314   0    0       book
    9  Bd3       316   0    0       book   Bxd3      316   0    0       book
   10  Qxd3      318   0    0       book   e6        318   0    0       book
   11  Bf4       320   0    0       book   Bb4+      320   0    0       book
   12  c3        322   0    0       book   Be7       322   0    0       book
   13  O-O-O     324   9    9   +0.47/17   Ngf6      324  15   15   +0.62/35
   14  Kb1       317  22   31   +0.38/17   Qa5       311  10   25   +0.24/32
   15  Ne5       297  10   41   +0.54/16   O-O       303   6   31   +0.19/30
   16  Nxd7      289   8   49   +0.47/15   Nxd7      299   8   39   +0.54/31
   17  Ka1       283  20   69   +0.40/16   Nf6       293  12   51    0.00/34
   18  Ne4       265   9   78   +0.34/16   Nxe4      283   7   58   +0.01/33
   19  Qxe4      258   7   85   +0.24/17   Rad8      278   8   66    0.00/34
   20  g4        253  14   99   +0.23/17   Rd7       272  20   86   +0.34/32
   21  Rhg1      241  11  110   +0.27/16   Rfd8      254   7   93    0.00/33
   22  Rd3       232   7  117   +0.23/14   Qd5       249   7  100   -0.44/30
   23  Qxd5      227   7  124   -0.03/18   Rxd5      244   6  106   -0.35/33
   24  Rgd1      222   8  132   -0.21/18   c5        240  15  121   -0.27/34
   25  Bc7       216   7  139    0.00/21   R8d7      227  13  134   -0.67/35
   26  Be5       211   5  144   -0.18/20   c4        216   7  141   -0.42/33
   27  R3d2      208   5  149   -0.21/20   Bf6       211  15  156   -0.55/36
   28  Bxf6      205   5  154   -0.09/20   gxf6      198   4  160   -0.34/32
   29  Rg1       202   8  162   +0.01/21   b5        196   4  164   -0.47/32
   30  Kb1       196   6  168   +0.14/19   a5        194   6  170   -0.57/31
   31  Kc2       192  15  183   +0.13/19   b4        190   9  179   -0.45/32
   32  Re2       179   8  191   +0.22/18   Rb7       183  23  202   -0.41/33
   33  f4        173   4  195   +0.23/16   Kf8       162   4  206   -0.33/31
   34  Rf1       171   4  199   +0.27/15   Rb8       160  22  228   -0.43/36
   35  Rff2      169  25  224   +0.24/18   Rbb5      140  12  240   -0.39/32
   36  Rf1       146  16  240   +0.22/17   Rb8       130   7  247   -0.35/35
   37  Rff2      132   4  244   +0.27/18   Rb6       125  16  263   -0.25/35
   38  Re1       130  16  260   +0.33/18   Rdb5      111   8  271   -0.13/32
   39  Kd2       116   6  266   +0.51/16   Rd5       105   6  277   -0.06/31
   40  Ke3       112   3  269   +0.43/16   a4        101   4  281   -0.59/29
   41  Rc1       111   7  276   +0.24/17   f5         99   3  284   -0.26/29
   42  g5        106   3  279   +0.21/17   Kg7        98   4  288    0.00/29
   43  Rg2       105   4  283   +0.11/16   Rdb5       96   4  292   -0.17/31
   44  b3        103   3  286   +0.11/16   cxb3       94   3  295   -0.46/30
   45  c4        102   8  294    0.00/17   bxa2       93   2  297   -1.73/29
   46  cxb5       96   3  297    0.00/17   b3         93   2  299    0.00/42
   47  gxh6+      95   3  300    0.00/16   Kh7        93   2  301    0.00/41			<<<<<
   48  Rg7+       94   2  302  +14.36/17   Kxh6       93   2  303    0.00/43
   49  Rxf7       94   8  310  +35.76/19   e5         93   2  305    0.00/43
   50  dxe5       88   7  317  +35.76/14   a1=Q       93  24  329  +20.20/34
   51  Rxa1       83   7  324    +M11/15   b2         71   2  331  +26.70/24
   52  Rg1        78   4  328     +M9/14   b1=Q       71   9  340  +92.16/29
   53  Rxb1       76   2  330     +M8/14   Rb8        64   5  345    +M16/25
   54  Rg1        76   1  331     +M7/12   Rc8        61   4  349     +M6/30
   55  Rg6+       77   0  331     +M6/10   Kxh5       59   1  350     +M5/41
   56  Rff6       79   0  331      +M5/8   Rc3+       60   1  351     +M4/51
   57  Kd4        81   0  331      +M4/6   Rc4+       61   8  359     +M3/72
   58  Kxc4       83   0  331      +M3/4   a3         55   0  359    +M2/128
   59  Rg5+       85   0  331      +M2/2   Kh4        57   0  359    +M1/128
   60  Rh6#       87   0  331      +M1/1
  --------------------------------------------------------------------------
       TOTAL      89      331                TOTAL      59      359

The interesting point is the losing move, 47…Kh7 (instead of 47…Kxh6):

- Zangdar had 93 s on the clock;
- it played in 2 s, at depth 41, with an evaluation of 0.00.

Position after 47.gxh6+:
8/5pk1/1r2p2P/1P3p1P/p2P1P2/1p2K3/p5R1/2R5 b - - 0 47

So this is not a time-management failure in the usual sense: the engine did not run out of seconds, it had no reason to spend any. From move 45 onward it settles at 93 s and gives back exactly the increment on every move — the "position judged drawn, nothing to dig into" regime. Every successive root search returns 0.00 at depth 41-43, with no fail-low and no best-move instability, so none of the triggers that normally extend thinking time ever fires.

What was missing is horizon: the refutation of 47…Kh7 lies roughly 7 plies beyond what the search could see, and the evaluation stays flat until it is too late (the score only flips at move 50, when Spark is already announcing +35). An engine that had spent 30 s instead of 2 s on that move would most likely have seen nothing more: the expected depth gain (1 to 2 plies) does not cover the gap.

[Frank Quisinsky]

Philippe,

you're going to a lot of trouble over my post. It reminds me of the glorious Winboard days in chess forums. The proud programmers almost always wanted to fix every bug.

Maybe I shouldn't have posted that after all. Zangdar plays some really great games here, and your engine is constantly improving. The post obscures that fact, it wasn't a good idea on my part.

Spark 1.0 HCE - Zangdar 7.0=100000000000000000000000000000000X
Excerpt from the LIVE *.sto file (Shredder's tournament organization)
X = game is still running!

What Zangdar do with Spark 1.0 HCE. So the first game looks strange!

To start the thread was a bad idea by myself, sorry for that!

That's a really nice explanation of the problem, thank you so much!

Best
Frank