Newbie chess program - Hansel

[okidoki]

Most of the times Hansel beats TSCP (out of 37 games Hansel scored 30.5 and TSCP 6.5)
Here is a loss by Hansel which stumped me.
Hansel has Q-Search and the following evaluation features:
1. Material weights
2. Piece square table weights
3. Penalty for double pawns and Isolated pawns.
4. Bonus for passed pawns.
5. Attackers near king's ring (x-rays passing through rooks and queens)
6. Tapered evaluation (stockfish's ideas and almost the code),
Just to try out, following is being used. Please note this is an older version

Code: Select all

 while (bbWhiteQueenBlack > 0)
            {
                int sqFrom = Utilities.bitscan_forward(bbWhiteQueenBlack);

                ulong occupancyQueenBishop = ~Defs.BB_EMPTY;
                occupancyQueenBishop &= MagicBitBoards.BISHOP_ATTACK_INNER_MASK[sqFrom];
                occupancyQueenBishop *= MagicBitBoards.BishopMagic[sqFrom];
                occupancyQueenBishop >>= 64 - MagicBitBoards.BitForBishopAttack[sqFrom];
                ulong movesQueenAttackingKingRing = (MagicBitBoards.BishopAttacks[sqFrom, occupancyQueenBishop] & occupancyPiecesBlack) & kingRingWhite;

                queenAttackersBlack += Utilities.count_bits(movesQueenAttackingKingRing);

                ulong occupancyQueenRook = ~Defs.BB_EMPTY;
                occupancyQueenRook &= MagicBitBoards.ROOKS_ATTACK_INNER_MASK[sqFrom];
                occupancyQueenRook *= MagicBitBoards.RookMagic[sqFrom];
                occupancyQueenRook >>= 64 - MagicBitBoards.BitForRookAttack[sqFrom];
                ulong movesQueenRookAttackingKingRing = (MagicBitBoards.RookAttacks[sqFrom, occupancyQueenRook] & occupancyPiecesBlack) & kingRingWhite;

                queenAttackersBlack += Utilities.count_bits(movesQueenRookAttackingKingRing);

                bbWhiteQueenBlack &= bbWhiteQueenBlack - 1;

            }

            // Compute the difference.

            int whiteAttackCount = 2 * queenAttackers + 3 * rookAttackers + 4 * bishopAttackers;
            int blackAttackCount = 2 * queenAttackersBlack + 3 * rookAttackersBlack + 4 * bishopAttackersBlack;
            

[pgn][Event "Arena Tournament-1"]
[Site "Local"]
[Date "2022.07.07"]
[Round "17"]
[White "Hansel 64bit"]
[Black "Tscp181"]
[Result "0-1"]
[BlackElo "2200"]
[ECO "D01"]
[Opening "Richter-Veresov, 3.Nf3"]
[Time "12:44:14"]
[Variation "1...d5"]
[WhiteElo "2200"]
[TimeControl "300"]
[Termination "adjudication"]
[PlyCount "103"]
[WhiteType "program"]
[BlackType "program"]

1. Nc3 d5 {(d7d5 d2d4 g8f6 g1f3 b8c6 e2e3 e7e6) 0.00/7 10} 2. d4 Nf6 {(g8f6
e2e3 b8c6 g1f3 e7e6 f1b5 c8d7) 0.00/7 9} 3. Nf3 e6 {(e7e6 e2e3 f8b4 f1d3
b4c3 b2c3 e8g8 e1g1) +0.08/7 9} 4. Bg5 {(Bc1-g5 Bf8-b4 e2-e3 O-O Bf1-d3
Nb8-c6 O-O Rf8-e8 Ra1-c1 Bc8-d7) 0.00/10 10} Bb4 {(f8b4 d1d2 b8c6 g5f6 d8f6
e2e3 b4c3 d2c3) 0.00/8 9} 5. Qd3 {0.00/11 9} O-O {(e8g8 e2e3 b8c6 f1e2 b4c3
d3c3 c8d7 e1c1) +0.05/7 8} 6. O-O-O {(O-O-O Bb4-e7 Rd1-d2 Nb8-c6 Kc1-b1
Nc6-b4 Qd3-b5 Nf6-g4 Bg5xe7 Qd8xe7) 0.00/10 9} Bxc3 {(b4c3 d3c3 f6e4 c3e3
f7f6 g5f4 b8c6) +0.33/7 8} 7. Qxc3 {0.00/11 9} Ne4 {(f6e4 g5d8 e4c3 b2c3
f8d8 e2e3 b8c6 f1d3) +0.31/8 8} 8. Bxd8 {0.00/13 8} Nxc3 {(e4c3 b2c3 f8d8
e2e3 b8c6 f1d3 c8d7 e3e4 d5e4 d3e4) +0.38/8 7} 9. bxc3 {0.00/13 8} Rxd8
{(f8d8 e2e3 b8c6 f1d3 c8d7 d1e1 a7a5 e3e4) +0.35/8 7} 10. e3 {0.00/11 8}
Nc6 {(b8c6 f1d3 f7f6 c3c4 d5c4 d3c4 c8d7) +0.39/7 7} 11. c4 {0.00/11 7} Bd7
{(c8d7 c2c3 c6a5 c4d5 e6d5 f1d3 a5c6) +0.45/7 7} 12. cxd5 {0.00/11 7} exd5
{(e6d5 f1b5 d8e8 b5c6 d7c6 f3e5 c6b5 f2f3) +0.45/8 6} 13. Kb2 {(Kc1-b2
g7-g6 Bf1-e2 a7-a5 h2-h4 Kg8-g7 h4-h5 g6xh5 Rh1xh5 Bd7-e6) 0.00/10 7} Bf5
{(d7f5 f1b5 d8d6 b2b1 c6b4 d1c1 d6e6) +0.70/7 6} 14. Bd3 {0.00/10 7} Bg4
{(f5g4 d3e2 c6b4 b2b1 g4f5 d1c1 d8e8) +0.70/7 6} 15. h3 {0.00/10 6} Bxf3
{(g4f3 g2f3 c6b4 b2b1 d8e8 d1g1 b4d3 c2d3) +0.84/7 6} 16. gxf3 {(g2xf3
a7-a5 Rd1-d2 g7-g6 a2-a4 Kg8-g7 Kb2-c3 Nc6-b4 Bd3-b5 Kg7-f6) +0.87/10 6} a5
{(c6b4 c2c3 b4d3 d1d3 f7f5 b2c2 g8f7) +0.75/6 6} 17. a3 {(a2-a3 g7-g6
Kb2-c3 Kg8-g7 Rd1-d2 Kg7-f6 Rh1-b1 Ra8-a7 Rb1-b5 a5-a4 Rb5-c5) +0.90/11 6}
Ra7 {(a5a4 d1b1 a8a7 d3b5 c6e7 b2c1) +0.74/6 5} 18. Kc3 {(Kb2-c3 a5-a4
h3-h4 g7-g6 h4-h5 Kg8-g7 Rh1-g1 Kg7-f6 h5-h6 Rd8-d7 Rg1-g4) +0.89/11 6} Rd6
{(d8d6 c3d2 c6e7 f3f4 d6b6 d1b1 b6e6) +0.83/7 5} 19. Rb1 {(Rd1-b1 Rd6-f6
f3-f4 Nc6-d8 Rb1-b5 c7-c6 Rb5-b3 Nd8-e6 a3-a4) +1.01/9 5} Rf6 {(d6f6 f3f4
f6e6 c3d2 a5a4 c2c4 d5c4 d3c4) +0.73/7 5} 20. Be2 {(Bd3-e2 Nc6-e7 Rb1-b5
Rf6-c6+ Kc3-d3 b7-b6 Rb5-b3 Rc6-h6 Rb3-c3 a5-a4 Rh1-h2) +0.97/10 5} Rg6
{(f6g6 c3d2 c6e7 c2c4 g6g2 d2e1 f7f5) +0.81/7 5} 21. Rhg1 {(Rh1-g1 Rg6-h6
Rg1-g3 Nc6-e7 Rb1-b5 Kg8-f8 Be2-d3 Rh6-c6+ Rb5-c5 Rc6xc5+ d4xc5 h7-h6)
+1.03/10 5} Rh6 {(g6h6 g1g3 c6e7 c3d2 e7f5 g3g5 h6f6) +0.76/7 5} 22. Rb5
{(Rb1-b5 Nc6-e7 Rg1-h1 f7-f5 Rb5-c5 b7-b6 Rc5-b5 Kg8-f7 a3-a4 Ra7-a6)
+0.87/10 5} Ne7 {(c6e7 g1h1 c7c6 b5b2 h6g6 h1b1 b7b5) +0.93/7 4} 23. Rg3
{(Rg1-g3 f7-f6 Be2-d3 a5-a4 Rb5-c5 b7-b6 Rc5-b5 Kg8-f7 Rb5-b4) +0.86/9 5}
c6 {(c7c6 b5b6 e7f5 g3g5 f5d6 g5g3) +0.94/6 4} 24. Rb1 {(Rb5-b1 Ne7-f5
Rg3-g5 Nf5-d6 Rb1-g1 Rh6xh3 Rg5xg7+ Kg8-f8 Rg1-g5 Kf8-e7 a3-a4 h7-h5)
+0.72/11 5} Nf5 {(e7f5 g3g5 f5d6 g5g3 b7b5 c3d2 a7e7) +0.85/7 4} 25. Rg5
{(Rg3-g5 Nf5-d6 Rb1-g1 g7-g6 Rg1-h1 Kg8-g7 a3-a4 Kg7-f6 Rg5-e5 Nd6-f5)
+0.65/10 4} Nd6 {(f5d6 b1g1 g7g6 g5g3 d6f5 g3g5 f5d6) +0.76/7 4} 26. Rh1
{(Rb1-h1 f7-f6 Rg5-g4 Kg8-f7 h3-h4 g7-g6 Rh1-h3 Nd6-b5+ Be2xb5 c6xb5
Kc3-d3) +0.72/10 4} f6 {(f7f6 g5g2 b7b6 c3d2 a7e7 h1b1) +0.92/6 4} 27. Rg4
{(Rg5-g4 Kg8-f7 h3-h4 g7-g6 Be2-d3 Kf7-e6 Rh1-h3 a5-a4 Kc3-b4 Ke6-f7)
+0.65/10 4} f5 {(f6f5 g4g5 b7b6 e2d3 h6f6 c3d2) +0.88/6 4} 28. Rg2 {(Rg4-g2
Kg8-f7 h3-h4 g7-g6 Be2-d3 Nd6-b5+ Kc3-b3 a5-a4+ Kb3-b4 Nb5-d6 Kb4-c5
Kf7-e6) +0.66/10 4} b5 {(b7b5 c3d2 a7e7 d2e1 h6e6 h1g1) +0.89/6 3} 29. h4
{(h3-h4 Kg8-f7 Be2-d3 Kf7-f6 Rh1-h3 Nd6-c4 Rg2-g5 g7-g6 Bd3xc4 d5xc4)
+0.59/10 4} Re7 {(a7e7 c3d2 h6e6 e2d3 a5a4 h1g1) +0.92/6 3} 30. Bd3
{(Be2-d3 Kg8-f7 Rg2-h2 Kf7-f6 Rh2-h3 Re7-d7 Rh3-g3 Nd6-c4 Rg3-g5) +0.63/9
4} Rf6 {(h6f6 h1g1 a5a4 c3d2 f6e6 d2c1) +0.84/6 3} 31. Rh3 {(Rh1-h3 Kg8-f7
h4-h5 a5-a4 Rh3-h1 Kf7-e6 Rh1-g1 Rf6-h6 Rg2-g5) +0.75/9 3} a4 {(a5a4 c3d2
f6f8 f3f4 f8e8 h3g3) +0.89/6 3} 32. Kb4 {(Kc3-b4 Kg8-f7 Kb4-c5 Nd6-c4
Bd3xc4 d5xc4 h4-h5 Re7-c7 Rg2-g5 Kf7-e6) +0.84/10 3} Rfe6 {(f6e6 b4c3 e7e8
h3g3 e6e7 c3d2) +0.92/6 3} 33. Kc5 {(Kb4-c5 Re7-c7 Kc5-b6 Rc7-c8 Rg2-g5
Re6-f6 Rg5-h5 h7-h6 Kb6-c5) +1.19/9 3} Rc7 {(e7c7 h3g3 g8h8 g3g5 e6f6 h4h5)
+0.86/6 3} 34. Rg5 {(Rg2-g5 g7-g6 Kc5-b6 Rc7-c8 h4-h5 Kg8-g7 h5xg6 Re6xg6
Rg5xg6+) +0.99/9 3} h6 {(h7h6 g5h5 e6f6 f3f4 c7c8 f2f3) +0.92/6 3} 35. Rg2
{(Rg5-g2 Nd6-c4 Bd3xf5 Re6-f6 Bf5-d3 Nc4xa3 Kc5-b6 Rc7-d7 h4-h5) +1.15/9 3}
Kf7 {(g8f7 h4h5 f7g8 g2g6 g8f7 h3g3) +0.68/6 3} 36. h5 {(h4-h5 Re6-f6
Rh3-g3 Nd6-e8 e3-e4 d5xe4 f3xe4 f5xe4 Bd3xe4 Rf6-e6) +1.45/10 3} Kf8 {(f7f8
h3g3 e6f6 g3g6 d6b7 c5b6 f6g6 g2g6) +0.45/6 3} 37. Rhg3 {(Rh3-g3 Nd6-e8
Bd3xf5 Re6-e7 Bf5-g6 Ne8-f6 Rg3-h3 Re7-e6 Rg2-g3 Kf8-e7) +3.08/10 3} Nc4
{(d6c4 d3c4 d5c4 c2c3 f8g8 f3f4 c7e7) +0.15/7 2} 38. Rg6 {(Rg3-g6 Rc7-e7
Rg6xe6 Re7xe6 Bd3xf5 Re6-d6 Rg2-g6 Rd6xg6 Bf5xg6 Nc4xa3 Kc5xc6) +3.78/11 3}
Rxg6 {(e6g6 g2g6 c4a3 g6c6 c7d7 c5b4 a3b1 b4b5) -0.43/8 2} 39. Rxg6
{(Rg2xg6 Nc4xa3 Rg6xc6 Rc7xc6+ Kc5xc6 Na3-c4 Kc6xd5 a4-a3 Bd3xc4 b5xc4
Kd5-e5 a3-a2) +3.64/11 3} Nxa3 {(c4a3 g6c6 c7c6 c5c6 f8f7 c6d5 f7f6 d5c5
f5f4 d3b5 f4e3 f2e3 a3c2) -1.38/8 2} 40. Rxc6 {(Rg6xc6 Rc7xc6+ Kc5xc6
Na3-c4 c2-c3 a4-a3 Bd3-b1 Kf8-e7 Kc6xd5 Ke7-f6 Bb1-d3 Kf6-g5) 0.00/11 2}
Rxc6+ {(c7c6 c5c6 f5f4 c6d5 f4e3 f2e3 f8e7 e3e4 b5b4) -1.59/8 2} 41. Kxc6
{0.00/11 2} Nc4 {(a3c4 c2c3 c4d2 c3c4 a4a3 d3b1 d2b1 c4b5 a3a2) +2.30/9 2}
42. c3 {(c2-c3 Nc4-d2 Bd3-b1 Nd2xb1 c3-c4 b5xc4 Kc6xd5 Nb1-d2 Kd5-e5 a4-a3)
0.00/10 2} Nd2 {(c4d2 c3c4 a4a3 c4b5 a3a2 d3f5 a2a1q c6d5 d2f3) +4.14/8 2}
43. Be4 {(Bd3-e4 d5xe4 d4-d5 e4xf3 Kc6xb5 Nd2-e4 d5-d6 Ne4xd6+ Kb5xa4
Kf8-e7) -8.43/9 2} fxe4 {(f5e4 f3e4 d2e4 c3c4 a4a3 f2f3 e4c3 c4b5 a3a2 e3e4
a2a1q e4d5) +7.36/10 2} 44. fxe4 {0.00/11 2} Nxe4 {(d2e4 c3c4 a4a3 f2f3
e4c3 c4b5 a3a2 e3e4 a2a1q e4d5) +7.36/9 2} 45. c4 {(c3-c4 a4-a3 f2-f3
Ne4-f6 e3-e4 a3-a2 e4-e5 a2-a1Q e5xf6 d5xc4 f6xg7+) -25.62/11 2} dxc4
{(b5c4 c6d5 e4d2 d5c5 c4c3 d4d5 c3c2 c5b5 c2c1q b5a4) +8.86/9 2} 46. f3
{(f2-f3 c4-c3 f3xe4 c3-c2 Kc6xb5 c2-c1Q e4-e5 Qc1xe3 Kb5-c4 a4-a3 Kc4-c5)
0.00/11 2} Nd2 {(e4d2 c6b5 c4c3 b5a4 c3c2 d4d5 c2c1q e3e4 c1c5 e4e5 c5d5)
+10.13/10 2} 47. Kxb5 {(Kc6xb5 c4-c3 Kb5xa4 c3-c2 f3-f4 c2-c1Q Ka4-b5
Nd2-e4 d4-d5 Qc1xe3) -32.30/10 2} c3 {(c4c3 b5a4 c3c2 f3f4 c2c1q a4b5 c1c4
b5b6 c4b3 b6c5 b3e3) +10.46/9 2} 48. Kxa4 {(Kb5xa4 c3-c2 f3-f4 c2-c1Q
Ka4-b5 Nd2-e4 f4-f5 Ne4-d6+ Kb5-b6 Qc1xe3 Kb6-c5) 0.00/10 2} c2 {(c3c2 f3f4
c2c1q a4b4 c1c4 b4a5 c4c3 a5b5 c3e3 d4d5 e3f4) +11.18/9 2} 49. Kb5 {(Ka4-b5
c2-c1Q f3-f4 Nd2-e4 Kb5-b4 Qc1-c3+ Kb4-b5 Ne4-d6+ Kb5-b6 Nd6-c4+ Kb6-c6
Nc4xe3+ Kc6-d6 Qc3xd4+) 0.00/10 2} c1=Q {(c2c1q f3f4 d2e4 b5b4 c1e3 d4d5
e3f4) +11.28/7 1} 50. f4 {(f3-f4 Nd2-e4 Kb5-b4 Qc1xe3 Kb4-c4 Qe3-c3+ Kc4-d5
Ne4-f6+ Kd5-e5 Qc3-e3+ Ke5-d6 Qe3xd4+ Kd6-e6 Nf6xh5) -36.73/10 2} Qc4+
{(c1c4 b5a5 c4d5 a5b4 d5h5 f4f5 h5f3 f5f6 f3e3 f6g7 f8g7) +11.53/6 1} 51.
Kb6 {(Kb5-b6 Qc4-e6+ Kb6-c5 Qe6xe3 f4-f5 Nd2-b3+ Kc5-b4 Nb3xd4 Kb4-c4
Nd4xf5 Kc4-d5 Kf8-e7) -36.84/9 2} Nf1 {(d2f1 d4d5 c4d5 e3e4 d5e4 b6c5 e4f4)
+12.70/6 1} 52. e4 {(e3-e4 Qc4xd4+ Kb6-b5 Qd4xe4 Kb5-c5 Qe4xf4 Kc5-c6
Qf4-f3+ Kc6-d6 Qf3xh5 Kd6-c6 Qh5-f3+ Kc6-d6 Nf1-e3) -39.03/11 1 Arena
Adjudication} 0-1[/pgn]

What stumped me was it was seeing the position as +3.. and from there on it went to lose. From the game, it was quiet evident to play 38. Bc4 and almost after few captures, White gets a passed d-pawn and is winning.
Can somebody explain which feature I need to correct / add to avoid this loss.

[TheBlackPlague]

Hello.

I'm glad you're interested in writing a chess engine in C#. Like you, very recently, I started writing my own chess engine in C#: StockNemo.

StockNemo has been majorly optimized and uses some of the best techniques available. For move generation, StockNemo has pre-generated tables wherever possible including Black Magic Fixed-Shift Tables for Sliding Pieces (Rook, Bishop, Queen). The idea is to move as much processing to compile or startup time to make runtime as fast as possible. This is a continued effort in regards to StockNemo, as I firmly believe it can be even faster. However, with move generation down to 8ns per piece on average, it is already reasonably fast.

For PERFT, to give StockNemo an insane advantage over other PERFT generators and engines, it has a multi-threaded PERFT algorithm. This allows StockNemo to hit higher depths than many other PERFT generators or engines. The algorithm has been optimized to use threads efficiently (ex. not wasting threads on lower depths which can be done faster rather than copying data for parallelization) as well as optimized for stack utilization.

Furthermore, if one wants to hit even higher depths, StockNemo features a Transposition Table designed specifically for PERFT. This allows for it to skip having to recompute PERFT at certain positions and depths.

Combining parallelization and transposition table, StockNemo has PERFT speeds as high as 168.3B nps. It goes to show C# code can be fast.

GitHub: https://github.com/TheBlackPlague/StockNemo

[hgm]

What stumped me was it was seeing the position as +3.. and from there on it went to lose. From the game, it was quiet evident to play 38. Bc4 and almost after few captures, White gets a passed d-pawn and is winning.
Can somebody explain which feature I need to correct / add to avoid this loss.

Well, look at the PV after 39. Rxg6. This PV ends with a3-a2, and that is the position where the evaluation thinks that white has a +3.6 advantage. While in fact it cannot stop the promotion of the Pawn on a2. At the very least it should have evaluated that Pawn as nearly the same value as a minor (because 7th-rank Pawns are always passers, so that you need to dedicate one of your pieces to prevent their promotion). I am not sure if that would have been enough to dissuade the engine from trading the Rooks, though; capturing two Pawns might still seem attractive compared to the rise in value of the passer.

Of course you would have seen that if you had only searched 1 ply deeper, assuming you search promotions in QS. Some engines give an extension when the opponent has a Pawn on the 7th rank (or when a new Pawn reaches the 7th rank), so that you can be sure it cannot ignore the promotion.

But of course you might not search deep enough even with this extension to see the danger. The best solution would indeed be to have an evaluation term that recognizes the danger of an 'unstoppable passer'. Such passers are very common in Pawn endings, and relatively easy to recognize there. Humans use the 'rule of the square' for this, to see whether the King can reach the promotion square in time. That way the evaluation can already take into account that the Pawn is worth about a Queen. Here the situation is a bit more complex, since there still are Bishops. But after Kxc6 the a-pawn has indeed become 'unstoppable'.

Evaluating a highly advance passer is tricky. It usually either becomes a Queen (+8) or you will lose it (-1). That makes a very large difference. Evaluating it in between, like +4, will be very far off in all cases. The situation after Kxc6 is suspect, though, because white only has a Bishop. Which is on the light shade, while there is only a single light square on the path of the Pawn. The King is outside the area where it could prevent promotion. (Even if it were inside, on a crowded board, it might not be able to prevent the promotion because it is blocked and has to take a detour.) If the only other piece was a Knight, it can stop the Pawn on any shade, but, like the King, it can easily be too far away to get there in time.

In this case the Bishop can actually attack a2 in two moves, c3 + Bb1. Problem is that after c3 black plays Nd2, and then there is no square left for the Bishop to attack a2 where it is safe from the Knight. (b3 is also no good.) So there is a lot of relevant tactics here, and you cannot hope that a static evaluation would be able to find out how that ends. You really need the accuracy of search. So the logical solution is to award an extension when in doubt. You assume that a passer outside the reach of the enemy King will be able to promote when there is only a single other piece (by evaluating as +6 or so) that does not block or attack the path to promotion, and give an extension of nearly 1 ply to award extra depth in this game-deciding branch of the tree. This is pretty much what a human would do. As long as you have not determined with certainty whether that passer will promote or not, the fate of all other move sequences is completely irrelevant (unless there would be a checkmate there).

[okidoki]

What stumped me was it was seeing the position as +3.. and from there on it went to lose. From the game, it was quiet evident to play 38. Bc4 and almost after few captures, White gets a passed d-pawn and is winning.
Can somebody explain which feature I need to correct / add to avoid this loss.

Well, look at the PV after 39. Rxg6. This PV ends with a3-a2, and that is the position where the evaluation thinks that white has a +3.6 advantage. While in fact it cannot stop the promotion of the Pawn on a2. At the very least it should have evaluated that Pawn as nearly the same value as a minor (because 7th-rank Pawns are always passers, so that you need to dedicate one of your pieces to prevent their promotion). I am not sure if that would have been enough to dissuade the engine from trading the Rooks, though; capturing two Pawns might still seem attractive compared to the rise in value of the passer.

Of course you would have seen that if you had only searched 1 ply deeper, assuming you search promotions in QS. Some engines give an extension when the opponent has a Pawn on the 7th rank (or when a new Pawn reaches the 7th rank), so that you can be sure it cannot ignore the promotion.

But of course you might not search deep enough even with this extension to see the danger. The best solution would indeed be to have an evaluation term that recognizes the danger of an 'unstoppable passer'. Such passers are very common in Pawn endings, and relatively easy to recognize there. Humans use the 'rule of the square' for this, to see whether the King can reach the promotion square in time. That way the evaluation can already take into account that the Pawn is worth about a Queen. Here the situation is a bit more complex, since there still are Bishops. But after Kxc6 the a-pawn has indeed become 'unstoppable'.

Evaluating a highly advance passer is tricky. It usually either becomes a Queen (+8) or you will lose it (-1). That makes a very large difference. Evaluating it in between, like +4, will be very far off in all cases. The situation after Kxc6 is suspect, though, because white only has a Bishop. Which is on the light shade, while there is only a single light square on the path of the Pawn. The King is outside the area where it could prevent promotion. (Even if it were inside, on a crowded board, it might not be able to prevent the promotion because it is blocked and has to take a detour.) If the only other piece was a Knight, it can stop the Pawn on any shade, but, like the King, it can easily be too far away to get there in time.

In this case the Bishop can actually attack a2 in two moves, c3 + Bb1. Problem is that after c3 black plays Nd2, and then there is no square left for the Bishop to attack a2 where it is safe from the Knight. (b3 is also no good.) So there is a lot of relevant tactics here, and you cannot hope that a static evaluation would be able to find out how that ends. You really need the accuracy of search. So the logical solution is to award an extension when in doubt. You assume that a passer outside the reach of the enemy King will be able to promote when there is only a single other piece (by evaluating as +6 or so) that does not block or attack the path to promotion, and give an extension of nearly 1 ply to award extra depth in this game-deciding branch of the tree. This is pretty much what a human would do. As long as you have not determined with certainty whether that passer will promote or not, the fate of all other move sequences is completely irrelevant (unless there would be a checkmate there).

Thanks for explaining me this. It seemed intuitive although by your reply I agree that search should do the work here smartly by having evaluation as its guidelines.

Hello.

I'm glad you're interested in writing a chess engine in C#. Like you, very recently, I started writing my own chess engine in C#: StockNemo.

StockNemo has been majorly optimized and uses some of the best techniques available. For move generation, StockNemo has pre-generated tables wherever possible including Black Magic Fixed-Shift Tables for Sliding Pieces (Rook, Bishop, Queen). The idea is to move as much processing to compile or startup time to make runtime as fast as possible. This is a continued effort in regards to StockNemo, as I firmly believe it can be even faster. However, with move generation down to 8ns per piece on average, it is already reasonably fast.

For PERFT, to give StockNemo an insane advantage over other PERFT generators and engines, it has a multi-threaded PERFT algorithm. This allows StockNemo to hit higher depths than many other PERFT generators or engines. The algorithm has been optimized to use threads efficiently (ex. not wasting threads on lower depths which can be done faster rather than copying data for parallelization) as well as optimized for stack utilization.

Furthermore, if one wants to hit even higher depths, StockNemo features a Transposition Table designed specifically for PERFT. This allows for it to skip having to recompute PERFT at certain positions and depths.

Combining parallelization and transposition table, StockNemo has PERFT speeds as high as 168.3B nps. It goes to show C# code can be fast.

GitHub: https://github.com/TheBlackPlague/StockNemo

I saw your code, and it seems it is written very nicely. However due to my lack of understanding of C#, it is difficult for me to assimilate.
Like your code, I changed mine to utilize structs and make the move list to be instance based and expected speed gain, however this is not the case.
I would like to understand why is mine slower (single threaded without TT) than yours. If it is not too much of hassle for you, can you go through my code and provide a feedback if it is not too much to ask.

Here is the github link for my code : https://github.com/frankcastle64/Hansel ... StaticImpl

Thank you.

[okidoki]

I have rewritten the program since I ran into many difficulties in my first attempt.

Now my program has the following features.
1. Negamax with alpha-beta pruning. (no internal iterative, no null move, no LMR, no TT).
2. QSearch expanding captures only.
3. Move ordering via MVV/LVA.
4. Evaluation using material values and PSQT. (This is taken from Rofchade version of TSCP)
5. Evaluation uses tapered eval.

Few more details, it on an average gets to 1.5 Mnps+ nodes for processing and reaches depth of 6 in less than 1 sec and 7 in about 3secs.
My aim here is to have a fast search and light eval, allowing engine to make its own decision guided by tapered eval and mobility.

Here are couple of the losses it had with TSCP. Please feel free to analyse them and if possible share some insights where it can be improved.

Please let me know if this is expected or if in your experience the moves played are buggy.
Thank you

[pgn]
[Event "Computer chess game"]
[Site "DESKTOP-I4JNID7"]
[Date "2022.12.13"]
[Round "?"]
[White "Tscp181"]
[Black "Hansel"]
[Result "1-0"]
[BlackElo "2200"]
[ECO "B01"]
[Opening "Scandinavian"]
[Time "13:10:19"]
[Variation "2...Qxd5 3.Nc3"]
[WhiteElo "2200"]
[TimeControl "600"]
[Termination "time forfeit"]
[PlyCount "141"]
[WhiteType "program"]
[BlackType "program"]

1. e4 {(e2e4 e7e5 g1f3 g8f6 b1c3 b8c6 d2d4 d7d6) +0.18/8 20} d5 {(d7-d5
f2-f3 d5xe4 f3xe4 Qd8-d4 Nb1-c3 Ke8-d8) 0.00/7 7} 2. exd5 {(e4d5 g8f6 f1b5
c8d7 b5c4 e7e5 d5e6 d7e6 c4e6 f7e6) +0.55/7 19} Qxd5 {(Qd8xd5 Nb1-c3
Qd5-e6+ Bf1-e2 Nb8-c6 Ke1-f1 Ke8-d8) -0.13/7 6} 3. Nc3 {(b1c3 d5a5 d2d4
e7e5 g1f3 b8c6 f1c4) +0.60/7 18} Qe6+ {(Qd5-e6+ Bf1-e2 Qe6-g6 Ke1-f1 Bc8-f5
Be2-h5 Bf5xc2) -0.41/7 23} 4. Nge2 {(g1e2 b8d7 d2d4 e6b6 a1b1 e7e5 c1e3)
+0.60/6 18} Nf6 {(Ng8-f6 d2-d4 Ke8-d8 Qd1-d2 Qe6-f5 Ke1-d1 Qf5xf2) -0.12/7
13} 5. d4 {(d2d4 b8d7 a2a4 e6d6 f2f4 e7e6 c1e3) +0.51/7 17} Kd8 {(Ke8-d8
f2-f3 Bc8-d7 Ke1-f2 Nb8-c6 Kf2-g1) -0.37/6 3} 6. d5 {(d4d5 e6b6 a1b1 b8d7
c1e3 c7c5 f2f4) +0.97/7 16} Qd7 {(Qe6-d7 g2-g3 e7-e6 Bf1-g2 Nf6xd5 Nc3xd5
e6xd5) +0.10/7 22} 7. Bg5 {(c1g5 c7c6 e2f4 c6d5 f4d5 b8c6 f1c4 f6d5 c3d5)
+1.08/7 16} h6 {(h7-h6 Bg5-e3 e7-e6 d5xe6 f7xe6 Qd1xd7+ Nb8xd7) +0.16/7 15}
8. Bf4 {(g5f4 c7c6 f4b8 a8b8 d5c6 b7c6 d1c1) +1.04/7 15} Na6 {(Nb8-a6
Qd1-d4 e7-e6 d5xe6 f7xe6 O-O-O) -0.20/6 4} 9. Qd4 {(d1d4 d8e8 e1c1 d7d8
f4e5 c8f5) +1.06/6 15} e6 {(e7-e6 O-O-O e6xd5 Nc3xd5 Nf6xd5 Qd4xd5 f7-f5)
-0.31/7 19} 10. dxe6 {(d5e6 f7e6 e1c1 d8e8 e2g3 d7d4 d1d4) +1.22/6 14} fxe6
{(f7xe6 O-O-O Na6-b4 Bf4-e5 Bf8-e7 Kc1-b1 Rh8-e8 Ne2-f4) -0.43/8 7} 11.
O-O-O {(e1c1 a6b4 d4c4 b4d5 f4e5 f8d6 e5d4) +1.17/7 14} Be7 {(Bf8-e7 Qd4-a4
Be7-d6 Qa4xd7+ Kd8xd7 Bf4xd6 c7xd6) -0.48/7 9} 12. Be5 {(f4e5 h8f8 e2f4
a6c5 d4e3 e7d6 f1c4) +1.42/7 13} Nb4 {(Na6-b4 Ne2-f4 Qd7xd4 Rd1xd4+ Nf6-d5
Nc3xd5 e6xd5) -1.09/7 4} 13. Bxf6 {(e5f6 g7f6 a2a3 d7d4 d1d4 b4d5 c3d5 e6d5
d4d5) +1.87/6 13} gxf6 {(g7xf6 a2-a3 Nb4-d5 Nc3xd5 e6xd5 Qd4xd5 Be7-d6
Kc1-b1 Rh8-e8) -0.54/9 24} 14. a3 {(a2a3 d7d4 d1d4 b4d5 c3d5 e6d5 d4d5 c8d7
e2d4) +1.77/7 12} Nd5 {(Nb4-d5 Nc3xd5 e6xd5 Qd4xd5 Rh8-f8 Kc1-b1 Be7-c5
f2-f4) -0.80/8 6} 15. Nxd5 {(c3d5 e6d5 d4d5 d7d5 d1d5 c8d7 e2d4) +1.77/6
12} exd5 {(e6xd5 Qd4xd5 Be7-d6 Kc1-b1 Bd6-e5 f2-f4 Qd7xd5 Rd1xd5+) -0.90/8
3} 16. c4 {(d4d5 d7d5 d1d5 c8d7 e2d4 d8c8 f1d3) +1.62/6 12} Re8 {(Rh8-e8
c4xd5 Qd7-f5 Ne2-c3 Be7-d6 Bf1-d3 Bd6-f4+) -0.57/7 7} 17. Nf4 {(e2f4 f6f5
d4d5 d7d5 d1d5 e7d6 f1d3 c8d7) +1.67/7 11} f5 {(f6-f5 Nf4xd5 c7-c5 Qd4-c3
Be7-g5+ Kc1-b1 f5-f4) -0.79/7 4} 18. Nxd5 {(f4d5 e7g5 f2f4 g5h4 d4c5 c7c6
c5a5) +2.03/6 11} Rf8 {(Re8-f8 Qd4-c3 Kd8-e8 Rd1-e1 Rf8-f7 Nd5-f6+ Rf7xf6)
-2.50/7 10} 19. Qc3 {(d4c3 e7d6 c4c5 f8g8 c3f6 d8e8 c5d6 d7d6) +3.69/6 10}
Bd6 {(Be7-d6 c4-c5 b7-b6 c5xd6 c7xd6 Kc1-b1 Bc8-b7) -3.49/7 14} 20. c5
{(c4c5 d7c6 c5d6 c7d6 f1c4 c8e6) +3.75/6 10} Qc6 {(Qd7-c6 c5xd6 c7xd6
Bf1-c4 Bc8-e6 Rh1-e1 Rf8-e8) -3.59/7 3} 21. Bb5 {(f1b5 c6b5 c5d6 b5e2 d6c7
d8d7 c3c5) +4.43/6 10} Qxc5 {(Qc6xc5 Qc3xc5 Bd6xc5 Nd5-b6+ Bc5-d6 Nb6xa8
Bc8-d7 Bb5-d3) -3.49/8 3} 22. Qxc5 {(c3c5 d6c5 d5b6 d8e7 b6a8 c7c6 h1e1
e7f6 b5c4 c5f2) +5.35/7 9} Bxc5 {(Bd6xc5 Nd5-b6+ Bc5-d6 Nb6xa8 f5-f4 Kc1-b1
a7-a6 Bb5-d3 Bc8-e6) -3.41/9 24} 23. Nb6+ {(d5b6 d8e7 b6a8 c7c6 h1e1 e7f6
b5c4 c5f2 d1d6) +5.35/7 9} Bd6 {(Bc5-d6 Nb6xa8 Bc8-d7 Bb5-c4 Bd7-c6 Bc4-d5
Kd8-d7 Bd5xc6+ Kd7xc6) -3.38/9 16} 24. Nxa8 {(b6a8 c8d7 b5c4 d8c8 c4d5 c8b8
a8c7 b8c7 d5b7 c7b7 d1d6) +4.61/7 9} Bd7 {(Bc8-d7 Bb5-c4 Bd7-c6 Bc4-d5
Kd8-d7 Bd5xc6+ Kd7xc6) -3.38/7 6} 25. Bc4 {(b5c4 d8c8 c4d5 f8d8 c1d2 c8b8
d5b7 b8b7) +4.12/7 8} Kc8 {(Kd8-c8 Bc4-d5 Rf8-d8 Rd1-d3 Kc8-b8 Rd3-b3
Kb8xa8) -2.89/7 6} 26. Kc2 {(c1c2 f8d8 h1e1 c8b8 a8c7 b8c7 h2h4) +3.89/7 8}
Kb8 {(Kc8-b8 Na8xc7 Kb8xc7 Bc4-d5 Bd7-a4+ b2-b3 Ba4-c6) -2.58/7 5} 27. Nxc7
{(a8c7 b8c7 d1d3 d7a4 c2d2 a4c6 f2f3 f8d8) +3.80/7 8} Kxc7 {(Kb8xc7 h2-h4
f5-f4 h4-h5 Bd7-g4 Rd1-e1 f4-f3 g2xf3) -2.57/8 8} 28. f3 {(f2f3 d7a4 b2b3
a4c6 a3a4 f8d8 c4e6 f5f4) +3.70/7 8} Re8 {(Rf8-e8 Bc4-d3 Bd7-a4+ b2-b3
Ba4-d7 Kc2-b2 f5-f4) -2.39/7 5} 29. Kd3 {(c2d3 d7e6 c4b5 e8d8 d3e3 c7c8)
+3.61/6 7} Be5 {(Bd6-e5 Kd3-c2 Bd7-a4+ Bc4-b3 Ba4-c6 Bb3-d5 f5-f4) -2.33/7
5} 30. b3 {(b2b3 d7e6 c4e6 e8e6 d3c4 e6d6 d1c1) +3.49/7 7} f4 {(f5-f4
Bc4-f7 Bd7-f5+ Kd3-c4 Re8-e7 Bf7-d5 Re7-e8) -2.30/7 3} 31. Rc1 {(d1c1 d7f5
d3d2 c7b8 c1d1 e8d8 d2e2 b8c8 c4d5) +3.63/7 7} Bf5+ {(Bd7-f5+ Kd3-d2
Re8-d8+ Kd2-e2 Be5-b2 Rc1-d1 Rd8xd1) -1.85/7 4} 32. Kd2 {(d3d2 c7b8 g2g4
f5e6 c4e6 e8e6 h1e1) +3.59/6 6} Rd8+ {(Re8-d8+ Kd2-e2 Kc7-b6 g2-g4 f4xg3/ep
h2xg3 Rd8-d6) -2.16/7 3} 33. Ke2 {(d2e2 c7b8 g2g4 f5g6 h1d1 d8d1 c1d1)
+3.64/6 6} Kb6 {(Kc7-b6 Rh1-d1 Be5-d4 Ke2-f1 h6-h5 g2-g3 f4xg3) -2.27/7 3}
34. Rhd1 {(h1d1 d8d4 d1d4 e5d4 c4d5 h6h5 e2d2) +3.55/7 6} Bd4 {(Be5-d4
Ke2-f1 Rd8-d6 h2-h4 Bf5-g6 Bc4-e6 Rd6-d8 Rd1-e1) -2.37/8 11} 35. a4 {(a3a4
b6a5 g2g4 f5g6 d1f1 d8e8 e2d2 h6h5) +3.76/7 6} Rd6 {(Rd8-d6 Bc4-f7 Rd6-d7
Ke2-f1 a7-a5 Rc1-c4 Bd4-e5 Bf7-d5) -2.36/8 7} 36. g4 {(g2g4 f4g3 h2g3 d4e5
d1d6 e5d6 f3f4 b6c5) +3.75/7 6} fxg3 {(f4xg3/ep h2xg3 a7-a5 g3-g4 Bf5-g6
Ke2-f1 Bg6-e8 Kf1-g2 Be8-c6) -2.42/9 9} 37. hxg3 {(h2g3 d4e5 d1d6 e5d6 f3f4
d6c5 c4d3 f5e6) +3.95/7 5} h5 {(h6-h5 Rd1-h1 Bf5-g6 Rh1-e1 Bg6-f5 Ke2-f1
Bf5-h3+ Kf1-e2) -2.46/8 6} 38. b4 {(b3b4 a7a6 d1h1 d4b2 c1e1 d6d4 a4a5)
+4.50/7 5} a5 {(a7-a5 b4xa5+ Kb6xa5 Bc4-f7 Rd6-d7 Bf7xh5 Ka5-b4 g3-g4
Rd7-e7+) -2.63/9 5} 39. bxa5+ {(b4a5 b6a5 d1h1 d4b2 c1e1 b2c3 e1c1 d6d2)
+4.31/7 5} Kxa5 {(Kb6xa5 Bc4-f7 Ka5-b6 a4-a5+ Kb6-b5 Rc1-c7 Kb5xa5 Rc7xb7
Ka5-a6) -2.48/9 11} 40. Rh1 {(d1h1 d4b2 c1e1 b2c3 e1g1 d6d4 c4f7) +4.31/7
5} Bg6 {(Bf5-g6 Bc4-b5 Rd6-e6+ Ke2-d1 Re6-e3 Rh1-f1 Ka5-b6 Rc1-c8) -2.01/8
5} 41. Bb5 {(c4b5 d6e6 e2d1 d4f2 g3g4 g6e8 b5e8 e6e8 h1h5 a5a4) +4.14/7 5}
Kb6 {(Ka5-b6 Rc1-c8 Rd6-e6+ Ke2-d2 Re6-d6 g3-g4 h5xg4 f3xg4) -2.18/8 17}
42. f4 {(f3f4 d6e6 e2f1 e6e3 c1d1 b6c5 f1f2 c5d5) +4.39/7 4} Re6+ {(Rd6-e6+
Ke2-f1 Re6-e3 Rc1-d1 Re3-f3+ Kf1-g2 Rf3-f2+) -1.76/7 3} 43. Kf1 {(e2f1 e6e3
c1d1 d4c5 h1h3 h5h4 h3h4 e3g3) +4.19/6 4} Re3 {(Re6-e3 Rh1-h3 Re3-f3+
Kf1-e2 Rf3-e3+ Ke2-d2 Bg6-e4 Rc1-c8) -1.67/8 15} 44. Rd1 {(c1d1 b6c5 h1h3
c5d5 f1g2 g6e4 g2f2 e3d3) +4.29/7 4} Kc5 {(Kb6-c5 Rh1-h3 Re3-f3+ Kf1-e2
Rf3-f2+ Ke2-e1 Rf2-c2 Ke1-f1) -1.69/8 6} 45. Rh3 {(h1h3 e3f3 f1e2 f3a3 d1c1
c5d6 c1c8 a3a2 e2f3 d6e7) +4.21/7 4} Rf3+ {(Re3-f3+ Kf1-e1 Bd4-f2+ Ke1-e2
Rf3xg3 Rh3xg3 Bf2xg3 Rd1-d7) -1.72/8 10} 46. Ke2 {(f1e2 f3a3 d1c1 c5d6 c1f1
a3a2 e2f3 a2a1 f1a1 d4a1) +3.90/6 4} Rf2+ {(Rf3-f2+ Ke2-e1 Rf2-f3 Ke1-d2
Bd4-f2 Rd1-f1 Rf3xg3 Rh3xg3 Bf2xg3) -1.67/9 10} 47. Ke1 {(e2e1 f2g2 f4f5
g6f5 h3h5 d4f2 e1f1 g2g1 f1f2 g1d1 h5f5) +4.04/7 4} Rf3 {(Rf2-f3 Bb5-f1
b7-b6 Bf1-g2 Rf3-a3 Rd1-c1+ Kc5-b4 Bg2-c6) -1.74/8 3} 48. Bf1 {(b5f1 f3b3
f1g2 b3e3 e1f1 g6d3 f1f2 e3e2) +4.26/6 4} b6 {(b7-b6 Bf1-b5 Bd4-f2+ Ke1-e2
Rf3xg3 Rh3xg3 Bf2xg3 Ke2-e3) -1.72/8 5} 49. Rc1+ {(d1c1 f3c3 c1c3 d4c3 e1d1
c5d4 f1g2 c3b2 d1d2) +3.68/6 3} Kd6 {(Kc5-d6 Rc1-c8 Kd6-e6 Bf1-c4+ Ke6-f5
Bc4-d5 Bd4-f2+ Ke1-d2) -1.92/8 4} 50. Bb5 {(f1b5 d4c5 c1d1 d6e6 b5d7 e6e7
d7c8 f3e3 e1d2 g6e4) +4.19/7 3} Bc5 {(Bd4-c5 Rc1-d1+ Kd6-e6 Bb5-c4+ Ke6-f5
Rd1-d8 Rf3-f2) -1.93/7 3} 51. Ke2 {(e1e2 f3f2 e2d1 g6e4 h3h4 c5e3 c1c3)
+4.17/6 3} Re3+ {(Rf3-e3+ Ke2-d2 Re3-b3 Rc1-d1 Bc5-d4 Kd2-e2 Kd6-d5
Bb5-c6+) -1.89/8 6} 52. Kd2 {(e2d2 g6f5 h3h5 e3g3 c1d1 g3g2 d2c3 d6e6 d1d7)
+4.16/6 3} Ra3 {(Re3-a3 Rc1-f1 Bc5-e3+ Kd2-e2 Be3-d4 f4-f5 Ra3-e3+ Ke2-d2)
-1.98/8 10} 53. Rf1 {(c1f1 d6e6 f4f5 g6f5 b5c4 e6e5 h3h5 c5e3) +4.06/6 3}
Be3+ {(Bc5-e3+ Kd2-d1 Bg6-f5 Rh3-h4 Be3-d4 Bb5-e2 Ra3-a1+ Kd1-d2) -1.75/8
5} 54. Kd1 {(d2d1 g6f5 h3h4 f5g4 d1c2 e3d4 b5e8 a3g3 e8h5 g4h5 h4h5)
+3.16/6 3} Bf5 {(Bg6-f5 Rh3-h4 Ra3-c3 a4-a5 Bf5-g4+ Rh4xg4 h5xg4 a5-a6)
-0.84/8 6} 55. Rh4 {(h3h4 a3c3 a4a5 f5g4 h4g4 h5g4 a5a6 d6d5) +1.86/7 3}
Rc3 {(Ra3-c3 Kd1-e2 Bf5-g4+ Rh4xg4 h5xg4 Rf1-d1+ Kd6-c5 Rd1-d8) -0.54/8 10}
56. Ke2 {(d1e2 f5g4 h4g4 h5g4 f1d1 d6c5 f4f5 e3d4 b5d3) +1.82/7 3} Bg4+
{(Bf5-g4+ Rh4xg4 h5xg4 f4-f5 Be3-d4 f5-f6 Rc3xg3 f6-f7) -0.62/8 12} 57.
Rxg4 {(h4g4 h5g4 f1d1 d6c5 d1d7 e3d4 b5d3 c3a3 d7c7) +1.84/7 3} hxg4
{(h5xg4 Rf1-d1+ Kd6-c5 Rd1-d3 Rc3xd3 Bb5xd3 Be3-g1 Bd3-f5 Bg1-h2) -0.82/9
4} 58. Rd1+ {(f1d1 d6c5 d1d3 c3d3 b5d3 e3d4 d3f5 c5b4 f5g4 b4a4) +1.78/7 2}
Kc5 {(Kd6-c5 Rd1-d3 Rc3xd3 Bb5xd3 Be3-d4 Bd3-f5 Kc5-b4 Bf5xg4 Kb4xa4)
-0.88/9 7} 59. Rd3 {(d1d3 c3d3 b5d3 e3d4 d3e4 b6b5 a4a5) +1.83/7 2} Rxd3
{(Rc3xd3 Bb5xd3 Be3-d4 Ke2-e1 Kc5-b4 Bd3-b5 Kb4-a3 Ke1-e2 Ka3-a2 Bb5-d7
Ka2-a3) -0.88/11 4} 60. Bxd3 {(b5d3 e3d4 d3f5 c5b4 f5g4 b4a4 e2d3 d4f2
g4d7) +1.98/9 2} Bd4 {(Be3-d4 Bd3-e4 Kc5-b4 Be4-c6 Bd4-c5 Bc6-d7 Bc5-d6
Bd7xg4 Kb4xa4 Ke2-d3) -1.02/10 7} 61. Bf5 {(d3f5 c5b4 f5g4 b4a4 e2d3 d4f6
g4d1 a4b5 g3g4) +2.03/8 2} Kb4 {(Kc5-b4 Bf5xg4 Kb4xa4 Ke2-d3 Bd4-c5 f4-f5
Bc5-e7 Bg4-d1+ Ka4-b4 Bd1-f3) -1.13/10 5} 62. Bxg4 {(f5g4 b4a4 g4f3 b6b5
e2d3 d4f6 f4f5 b5b4 g3g4) +1.99/9 2} Kxa4 {(Kb4xa4 Ke2-d3 Bd4-b2 Bg4-d1+
Ka4-b4 Bd1-f3 b6-b5 Bf3-d5 Bb2-c3) -0.93/9 3} 63. Kd3 {(e2d3 d4f2 g4d7 b6b5
g3g4 a4a5 f4f5 b5b4 g4g5 b4b3) +2.07/9 2} Bb2 {(Bd4-b2 Bg4-d7+ Ka4-b4 g3-g4
Kb4-c5 g4-g5 Kc5-d6 Bd7-f5 Bb2-c1 Kd3-e4) -1.36/10 13} 64. Bd7+ {(g4d7 b6b5
g3g4 a4a5 g4g5 b5b4 d7e6 b2c1 d3e4) +2.23/8 2} b5 {(b6-b5 g3-g4 Bb2-c1
Kd3-e4 Ka4-a5 g4-g5 b5-b4 g5-g6 b4-b3 g6-g7 b3-b2) -1.50/11 7} 65. g4
{(g3g4 a4b4 g4g5 b4c5 f4f5 b5b4 d7e6 b2e5) +2.04/8 2} Kb4 {(Ka4-b4 g4-g5
Kb4-c5 f4-f5 Bb2-c1 g5-g6 Bc1-b2 Kd3-e4 b5-b4 Bd7-e6) -1.86/10 3} 66. g5
{(g4g5 b4c5 f4f5 b2c1 g5g6 c1b2 d3e4 b5b4 d7e6) +2.38/9 2} Kc5 {(Kb4-c5
f4-f5 Bb2-c1 g5-g6 Bc1-b2 Kd3-e4 b5-b4 Bd7-e6 Kc5-d6 Be6-d5 Bb2-e5)
-1.78/11 9} 67. f5 {(f4f5 b2c1 g5g6 c1b2 d3e4 b5b4 d7e6 b2d4) +2.28/8 2}
Bc1 {(Bb2-c1 g5-g6 Bc1-b2 Bd7-e8 b5-b4 Be8-f7 Bb2-a1 Kd3-e4 Kc5-d6 Bf7-d5
Ba1-e5) -1.78/11 7} 68. g6 {(g5g6 c1b2 d3e4 b5b4 d7e6 b2f6 e6b3 f6g7 b3d5
g7d4) +2.33/10 2} Bb2 {(Bc1-b2 Kd3-c2 Bb2-a1 Kc2-d3 b5-b4 Bd7-e6 Ba1-e5
Kd3-e4 Kc5-d6 Be6-d5) -1.78/10 6} 69. Ke4 {(d3e4 b5b4 d7e6 b2f6 e6g8 f6g7
g8b3 g7d4 b3d5) +2.33/9 1} b4 {(b5-b4 Bd7-e6 Kc5-c6 Ke4-d3 Kc6-c5 Kd3-e3
Kc5-d6 Ke3-e4 Bb2-e5 Be6-d5) -1.78/10 9} 70. Be6 {(d7e6 c5d6 e6f7 b2f6 f7b3
f6g7 b3a2 g7e5 a2d5) +2.33/9 1} Kb5 {(Kc5-b5 Be6-b3 Kb5-c5 Ke4-d3 Kc5-c6
Kd3-e4 Kc6-d6 Bb3-d5 Bb2-e5) -1.78/9 3} 71. Bd5 {(e6d5 b2f6 d5b3 b5c5 b3a2
f6g7 a2d5 g7d4) +2.33/8 1} ... {Black forfeits on time} 1-0
[/pgn]

[pgn][Event "Computer chess game"]
[Site "DESKTOP-I4JNID7"]
[Date "2022.12.13"]
[Round "?"]
[White "Hansel"]
[Black "Tscp181"]
[Result "0-1"]
[BlackElo "2200"]
[ECO "A06"]
[Opening "Reti Opening"]
[Time "13:43:37"]
[Variation "1...d5 2.e3"]
[WhiteElo "2200"]
[TimeControl "600"]
[Termination "normal"]
[PlyCount "66"]
[WhiteType "program"]
[BlackType "program"]

1. Nf3 {(Ng1-f3 d7-d5 e2-e3 Bc8-g4 Bf1-d3 Bg4xf3 Qd1xf3) +0.55/7 4} d5
{(d7d5 d2d4 g8f6 b1c3 b8c6 e2e3 e7e6 f1d3) -0.25/8 20} 2. e3 {(e2-e3 Nb8-c6
Bf1-b5 Qd8-d7 Bb5xc6 Qd7xc6 O-O) +0.66/7 12} e6 {(b8c6 d2d4 e7e6 f1b5 c8d7
e1g1 g8f6) -0.05/7 19} 3. Bb5+ {(Bf1-b5+ Bc8-d7 Bb5xd7+ Qd8xd7 O-O Ke8-d8)
+0.33/6 5} Bd7 {(c8d7 b5e2 g8f6 e1g1 f8d6 d2d4 e8g8 b1c3) -0.10/7 18} 4.
Qe2 {(Qd1-e2 Nb8-c6 d2-d4 a7-a6 Bb5xc6 Bd7xc6 O-O) +0.68/7 4} Nf6 {(c7c6
b5a4 g8f6 e1g1 f8d6 d2d4 e8g8) +0.38/7 18} 5. Bxd7+ {(Bb5xd7+ Qd8xd7 Nf3-e5
Qd7-a4 Ke1-d1 Nf6-e4 Nb1-c3) +0.63/7 8} Nbxd7 {(b8d7 d2d4 f8d6 e1g1 e8g8
b1c3 e6e5 c1d2) +0.18/7 17} 6. O-O {(O-O Bf8-d6 Qe2-b5 Qd8-c8 Nb1-c3 O-O)
-0.09/6 5} e5 {(e6e5 d2d4 e5e4 f3d2 f8d6 c2c4 d8e7) +0.40/7 16} 7. Nc3
{(Nb1-c3 Qd8-e7 Qe2-b5 Qe7-b4 a2-a3 Qb4xb5 Nc3xb5) +0.64/7 10} e4 {(e5e4
f3e1 f8b4 e2b5 b4c3 d2c3 a8b8) +0.40/7 16} 8. Nd4 {(Nf3-d4 c7-c5 Nd4-f5
Qd8-b6 Nc3-a4 Qb6-e6 g2-g4) +0.64/7 8} c5 {(c7c5 d4f5 g7g6 f5g3 f8d6 d2d3
d6g3 f2g3) +0.28/7 15} 9. Nf5 {(Nd4-f5 g7-g6 Nf5-g3 Qd8-e7 d2-d4 O-O-O
Bc1-d2) +0.46/7 4} g6 {(g7g6 f5h4 f8e7 d2d3 e4d3 e2d3 e8g8) +0.31/7 15} 10.
Ng3 {(Nf5-g3 h7-h5 h2-h4 Bf8-g7 d2-d3 O-O d3xe4) +0.46/7 4} h5 {(h7h5 e2b5
h5h4 g3e2 h4h3 g2h3 d8c7) +0.37/7 14} 11. f3 {(f2-f3 h5-h4 Ng3-h1 Ra8-c8
f3xe4 Nf6xe4 Nc3xd5) +0.52/7 7} h4 {(h5h4 g3h1 h4h3 g2g3 d7e5 e2b5 d8d7
b5d7 e5d7 f3e4 d5e4) +0.68/7 14} 12. Nh1 {(Ng3-h1 Bf8-g7 f3xe4 Nf6xe4
Nc3xe4 d5xe4 Nh1-f2 Qd8-b6) +0.32/8 7} Bd6 {(h4h3 g2g3 f8d6 f3e4 d5e4 d2d3
d8e7 d3e4 f6e4 c3e4 e7e4) +0.51/7 13} 13. fxe4 {(f3xe4 d5xe4 Nh1-f2 Qd8-e7
Qe2-b5 O-O-O Qb5-a4 h4-h3) +0.39/8 12} dxe4 {(d5e4 h1f2 d8c7 h2h3 d6h2 g1h1
c7e5 d2d4) +0.38/7 13} 14. Nf2 {(Nh1-f2 Qd8-e7 d2-d3 e4xd3 Nf2xd3 O-O
e3-e4) +0.24/7 12} Qc7 {(d8c7 h2h3 c7c6 c3b5 e8d8 b5d6 c6d6) +0.61/7 12}
15. Nb5 {(Nc3-b5 Bd6xh2+ Kg1-h1 Qc7-g3 d2-d3 e4xd3 Nf2xd3 Nf6-e4) +0.21/8
8} Bxh2+ {(d6h2 g1h1 c7b8 b5c3 b8e5 f2g4 f6g4 e2g4 d7f6 f1f6 e5f6 h1h2)
+1.38/8 12} 16. Kh1 {(Kg1-h1 Qc7-e5 d2-d4 c5xd4 e3xd4 Qe5-h5 Qe2xh5 Rh8xh5)
-0.20/8 4} Qb8 {(c7b8 b5c3 b8e5 f2g4 f6g4 e2g4 d7f6 f1f6 e5f6 h1h2) +1.38/7
12} 17. Qc4 {(Qe2-c4 Nd7-b6 Qc4xc5 Qb8-e5 d2-d4 Qe5xc5 d4xc5) +0.24/7 8} a6
{(a7a6 b5c3 b7b5 c4e2 b8e5 d2d3 e4d3 f2d3) +1.55/7 11} 18. Nc3 {(Nb5-c3
b7-b5 Qc4-e2 O-O Nf2xe4 Nf6xe4 Nc3xe4 Bh2-g3) -0.05/8 5} b5 {(b7b5 c4e2
h4h3 g2h3 h2g3 c3e4 f6e4 f2e4 h8h3) +1.47/7 11} 19. Qe2 {(Qc4-e2 Nf6-h5
Nf2xe4 f7-f5 g2-g4 h4xg3/ep Ne4-g5 O-O) -0.37/8 6} Be5 {(h2e5 c3e4 f6e4
f2e4 h4h3 g2h3 h8h3) +1.57/7 10} 20. Nfxe4 {(Nf2xe4 O-O Ne4xf6+ Nd7xf6
Qe2-f2 Qb8-b6 Kh1-g1 b5-b4) -0.13/8 5} h3 {(h4h3 g2h3 e5c3 e4c3 b5b4 c3a4
h8h3) +1.66/7 10} 21. gxh3 {(g2xh3 b5-b4 Ne4xf6+ Nd7xf6 Nc3-a4 Qb8-c7
Kh1-g1 O-O) +0.18/8 5} b4 {(e5c3 e4c3 h8h3 h1g1 b8g3 e2g2 g3g2 g1g2 h3h5)
+1.62/6 10} 22. Nxf6+ {(Ne4xf6+ Nd7xf6 Nc3-a4 Rh8xh3+ Kh1-g1 Rh3-g3+ Kg1-f2
Nf6-e4+) -0.19/8 8} Nxf6 {(d7f6 e2g2 b4c3 d2c3 b8c8 f1f3 a8b8 h1g1) +1.72/7
9} 23. Na4 {(Nc3-a4 Rh8xh3+ Kh1-g1 Rh3-g3+ Kg1-h1 Qb8-b7+ Kh1-h2 Rg3xe3+)
-0.89/8 11} Rxh3+ {(f6e4 h1g1 e4g3 e2c4 g3f1 c4f1) +2.25/6 9} 24. Kg1
{(Kh1-g1 Qb8-b7 Qe2-c4 Rh3-g3+ Kg1-f2 Nf6-e4+ Kf2-e1 Ra8-c8) -1.06/8 8}
Bh2+ {(e5h2 g1h1 b8b7 e2g2 h2e5 h1g1 h3g3 g2g3 e5g3 f1f6) +2.21/6 9} 25.
Kh1 {(Kg1-h1 Qb8-b7+ e3-e4 Bh2-d6+ Kh1-g1 Rh3-g3+ Kg1-h2) -1.02/7 9} Qb7+
{(b8b7 e2g2 h2e5 h1g1 h3g3 g2g3 e5g3 f1f6) +2.21/6 8} 26. e4 {(e3-e4
Bh2-e5+ Kh1-g1 Rh3-g3+ Kg1-h1 Qb7xe4+ Qe2xe4 Nf6xe4) -1.72/8 3} Be5+ {(h2e5
h1g1 h3g3 g1h1 f6e4 f1f3 e4f2 e2f2 b7f3 f2f3 g3f3 a4c5) +3.51/6 8} 27. Kg1
{(Kh1-g1 Nf6xe4 Rf1-e1 Rh3-h1+ Kg1xh1 Ne4-g3+ Kh1-g1 Ng3xe2+) -2.47/8 4}
Nxe4 {(h3g3 g1f2 b7c6 f1h1 f6g4 f2e1 c6a4 d2d3) +3.48/6 8} 28. Re1 {(Rf1-e1
Rh3-h1+ Kg1xh1 Ne4-g3+ Kh1-g1 Ng3xe2+ Re1xe2) -2.47/7 7} Rg3+ {(h3g3 e2g2
g3g2 g1g2 e4g5 e1e4 b7e4 g2f2 e5d4 f2f1 e4c2) +13.43/6 8} 29. Qg2 {(Qe2-g2
Rg3xg2+ Kg1xg2 Ne4xd2+ Kg2-f2 Nd2-e4+ Kf2-g1 Qb7-d5) -6.96/8 6} Rxg2+
{(g3g2 g1g2 e4g5 e1e4 b7e4 g2g1 e4f3 a4c5 g5h3) +M4/7 4} 30. Kxg2 {(Kg1xg2
Ne4-g5+ Re1-e4 Qb7xe4+ Kg2-f1 Qe4-f3+ Kf1-e1 Be5-g3+) -M500/9 4} Ng5+
{(e4g5 e1e4 b7e4 g2g1 e4f3 a4c5 g5h3) +M3/6 1} 31. Re4 {(Re1-e4 Qb7xe4+
Kg2-f1 Qe4-f3+ Kf1-e1 Be5-g3+) -M500/8 4} Qxe4+ {(b7e4 g2g1 e4f3 a4c5 g5h3)
+M2/5 0} 32. Kf1 {(Kg2-f1 Qe4-f3+ Kf1-e1 Be5-g3+) -M500/8 3} Qf3+ {(e4f3
f1e1 e5g3) +M1/3} 33. Ke1 {(Kf1-e1 Be5-g3+) -M500/9 14} Bg3# {(e5g3) +M0/2}
0-1
[/pgn]