how much time does stockfish need to see improvement in white's score?

[jtwright]

I expect even with using only piece square table evaluation
that the advantage of white get bigger with a deeper search and I think that the evaluation should encourage trading pieces when I am sure that even if you do not find a forced win of material you should find a force trade of material.

How can the PST evaluation encourage trading pieces? I think that the opposite happens: in absence of forced win of material, the stronger side pieces tend to occupy better squares than the weaker's, since he's stronger, so trading pieces will lower his score.

Alex

While I agree it's not clear-cut whether trading is the best strategy here, one way in which material and/or PSQ in eval can potentially "incentivize" trading when ahead is through the tapered eval. For instance, if the value of all pieces increases as the game phase advances, being up a rook in endgame is worth more points than being up a rook in midgame, so a winning player could increase their score by trading away equal material to advance the game phase.

[Brunetti]

For instance, if the value of all pieces increases as the game phase advances, being up a rook in endgame is worth more points than being up a rook in midgame, so a winning player could increase their score by trading away equal material to advance the game phase.

This occurs only if the phase advance due to the trade increases the score more than the PS score loss. I suspect that it's not often the case, but the research would be a long and slow work... I just tested a simple position with SF 14.1 comparing it with the same one after removing two Bishops:


[fen]1nbk4/1pbppp2/p1p2np1/7p/2P5/2N1PNB1/PPQ2PPP/R3KB1P w Q - 0 1[/fen]

Code: Select all

info string NNUE evaluation using nn-13406b1dcbe0.nnue enabled

 Contributing terms for the classical eval:
+------------+-------------+-------------+-------------+
|    Term    |    White    |    Black    |    Total    |
|            |   MG    EG  |   MG    EG  |   MG    EG  |
+------------+-------------+-------------+-------------+
|   Material |  ----  ---- |  ----  ---- | 18.75 19.58 |
|  Imbalance |  ----  ---- |  ----  ---- |  0.53  0.65 |
|      Pawns |  0.33 -0.36 |  0.85  0.00 | -0.52 -0.36 |
|    Knights |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|    Bishops |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|      Rooks |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|     Queens |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|   Mobility |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|King safety |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|    Threats |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|     Passed |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|      Space |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|   Winnable |  ----  ---- |  ----  ---- |  0.00 -4.48 |
+------------+-------------+-------------+-------------+
|      Total |  ----  ---- |  ----  ---- | 18.76 15.39 |
+------------+-------------+-------------+-------------+

 NNUE derived piece values:
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |   n   |   b   |   k   |       |       |       |       |
|       | -2.78 | -3.01 |       |       |       |       |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |   p   |   b   |   p   |   p   |   p   |       |       |
|       | -0.72 | -3.23 | -0.60 | -0.52 | -0.57 |       |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|   p   |       |   p   |       |       |   n   |   p   |       |
| -0.52 |       | -0.64 |       |       | -3.00 | -0.64 |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |       |       |       |       |       |       |   p   |
|       |       |       |       |       |       |       | -0.52 |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |       |   P   |       |       |       |       |       |
|       |       | +0.66 |       |       |       |       |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |       |   N   |       |   P   |   N   |   B   |       |
|       |       | +3.91 |       | +0.68 | +4.06 | +4.06 |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|   P   |   P   |   Q   |       |       |   P   |   P   |   P   |
| +0.47 | +0.63 | +9.76 |       |       | +0.73 | +0.72 | +0.50 |
+-------+-------+-------+-------+-------+-------+-------+-------+
|   R   |       |       |       |   K   |   B   |       |   P   |
| +5.00 |       |       |       |       | +3.90 |       | +0.60 |
+-------+-------+-------+-------+-------+-------+-------+-------+

 NNUE network contributions (White to move)
+------------+------------+------------+------------+
|   Bucket   |  Material  | Positional |   Total    |
|            |   (PSQT)   |  (Layers)  |            |
+------------+------------+------------+------------+
|  0         |  + 13.97   |  +  2.80   |  + 16.76   |
|  1         |  +  9.60   |  +  3.75   |  + 13.36   |
|  2         |  + 11.31   |  +  5.48   |  + 16.78   |
|  3         |  + 12.21   |  +  2.74   |  + 14.94   |
|  4         |  + 12.51   |  +  3.43   |  + 15.94   |
|  5         |  + 13.21   |  +  4.31   |  + 17.52   |
|  6         |  + 14.48   |  +  3.33   |  + 17.81   | <-- this bucket is used
|  7         |  + 17.16   |  +  1.92   |  + 19.09   |
+------------+------------+------------+------------+


Classical evaluation   +17.23 (white side)
NNUE evaluation        +17.81 (white side)
Final evaluation       +17.23 (white side) [with scaled NNUE, hybrid, ...]

[fen]1nbk4/1p1ppp2/p1p2np1/7p/2P5/2N1PN2/PPQ2PPP/R3KB1P w Q - 0 1[/fen]

Code: Select all

info string NNUE evaluation using nn-13406b1dcbe0.nnue enabled

 Contributing terms for the classical eval:
+------------+-------------+-------------+-------------+
|    Term    |    White    |    Black    |    Total    |
|            |   MG    EG  |   MG    EG  |   MG    EG  |
+------------+-------------+-------------+-------------+
|   Material |  ----  ---- |  ----  ---- | 18.75 19.53 |
|  Imbalance |  ----  ---- |  ----  ---- |  0.46  0.58 |
|      Pawns |  0.33 -0.36 |  0.85  0.00 | -0.52 -0.36 |
|    Knights |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|    Bishops |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|      Rooks |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|     Queens |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|   Mobility |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|King safety |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|    Threats |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|     Passed |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|      Space |  0.00  0.00 |  0.00  0.00 |  0.00  0.00 |
|   Winnable |  ----  ---- |  ----  ---- |  0.00 -5.38 |
+------------+-------------+-------------+-------------+
|      Total |  ----  ---- |  ----  ---- | 18.68 14.36 |
+------------+-------------+-------------+-------------+

 NNUE derived piece values:
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |   n   |   b   |   k   |       |       |       |       |
|       | -2.78 | -3.01 |       |       |       |       |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |   p   |       |   p   |   p   |   p   |       |       |
|       | -0.72 |       | -0.60 | -0.52 | -0.57 |       |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|   p   |       |   p   |       |       |   n   |   p   |       |
| -0.52 |       | -0.64 |       |       | -3.00 | -0.64 |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |       |       |       |       |       |       |   p   |
|       |       |       |       |       |       |       | -0.52 |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |       |   P   |       |       |       |       |       |
|       |       | +0.65 |       |       |       |       |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|       |       |   N   |       |   P   |   N   |       |       |
|       |       | +3.37 |       | +0.68 | +3.43 |       |       |
+-------+-------+-------+-------+-------+-------+-------+-------+
|   P   |   P   |   Q   |       |       |   P   |   P   |   P   |
| +0.46 | +0.63 | +9.65 |       |       | +0.69 | +0.72 | +0.50 |
+-------+-------+-------+-------+-------+-------+-------+-------+
|   R   |       |       |       |   K   |   B   |       |   P   |
| +5.00 |       |       |       |       | +3.47 |       | +0.60 |
+-------+-------+-------+-------+-------+-------+-------+-------+

 NNUE network contributions (White to move)
+------------+------------+------------+------------+
|   Bucket   |  Material  | Positional |   Total    |
|            |   (PSQT)   |  (Layers)  |            |
+------------+------------+------------+------------+
|  0         |  + 13.96   |  +  2.80   |  + 16.75   |
|  1         |  +  9.58   |  +  3.79   |  + 13.37   |
|  2         |  + 11.22   |  +  5.48   |  + 16.69   |
|  3         |  + 12.12   |  +  2.74   |  + 14.85   |
|  4         |  + 12.45   |  +  3.75   |  + 16.20   |
|  5         |  + 13.11   |  +  4.31   |  + 17.42   |
|  6         |  + 14.39   |  +  3.33   |  + 17.72   | <-- this bucket is used
|  7         |  + 17.08   |  +  2.30   |  + 19.38   |
+------------+------------+------------+------------+


Classical evaluation   +16.08 (white side)
NNUE evaluation        +17.72 (white side)
Final evaluation       +16.08 (white side) [with scaled NNUE, hybrid, ...]

The second one is judged inferior not only by PST terms (Material and Imbalance in classical table and PSQT in NNUE) but also by all other terms (but "Winnable", whose weight is too low to influence the final scores). Analyzing the position, Stockfish considers the trade only as its third choice:

Code: Select all

info depth 28 seldepth 35 multipv 1 score cp 1930 nodes 256756384 nps 1859071 hashfull 999 tbhits 0 time 138110 pv f3g5 c7g3 h2g3 d8e8 c2b3 a6a5 c4c5 f6d5 e3e4 b8a6 e4d5 a6c5 b3c4 d7d6 d5c6 c5e6 g5e6 f7e6 c6b7 c8b7 c4e6 b7c6 e6g6 e8d8 g6h5 a5a4 h5f5 d8c7 g3g4 c7d8 g4g5 a4a3 f5a5 d8c8 a5a3
info depth 28 seldepth 38 multipv 2 score cp 1894 nodes 256756384 nps 1859071 hashfull 999 tbhits 0 time 138110 pv c2b3 c7g3 h2g3 d7d6 b3b6 d8d7 f3g5 c6c5 g5f7 b8c6 f1d3 d7e8 d3g6 e8f8 f7g5 f8g7 g6b1 f6e8 b2b3 a6a5 f2f4 e8f6 h1h2 c8d7 b6c7 d7e8 c7b7 g7f8 b7c7 e8d7 e3e4 d7e8 h2h3 e8d7
info depth 28 seldepth 44 multipv 3 score cp 1751 nodes 256756384 nps 1859071 hashfull 999 tbhits 0 time 138110 pv g3c7 d8c7 f3e5 d7d6 e5f7 c8d7 c2g6 d7e8 g6g7 e8f7 g7f7 c7d8 f7f8 f6e8 f8h8 b8d7 h8h5 d7f6 h5a5 e8c7 a5f5 c7e8 e3e4 e8g7 f5a5 d8d7 e1c1 f6g4 a5b6 d7c8 h2h3 g4f6 b6a5 f6e8 a5g5 c8d7 e4e5 g7e6
info depth 28 seldepth 32 multipv 4 score cp 1730 nodes 256756384 nps 1859071 hashfull 999 tbhits 0 time 138110 pv c2a4 d7d6 f3g5 c8e6 e1c1 f6d7 a4c2 d7f8 h2h3 b8d7 a2a3 d7b6 b2b3 d8c8 c1b1 e6f5 e3e4 f5e6 f2f4 f7f6 g5f3 e6f7 b1a1 f8d7 f3d2 g6g5 d1e1 g5f4 g3f4
info depth 27 seldepth 34 multipv 5 score cp 1713 nodes 256756384 nps 1859071 hashfull 999 tbhits 0 time 138110 pv g3h4 d7d6 b2b4 d8e8 h2h3 c6c5 a2a3 b8c6 c2b2 c6e5 f3d2 c5b4 a3b4 e8f8 e1c1 c8e6 h4g3 f8g8 f2f4 e5c6 g3f2 c7b6 b4b5 a6b5 c4b5 c6a5

Alex

[Brunetti]

[fen]1nbk4/1pbppp2/p1p2np1/7p/2P5/2N1PNB1/PPQ2PPP/R3KB1P w Q - 0 1[/fen]

I wrongly put a Pawn instead of a Rook at h1, and SF allows it! After the correction, the previous analysis is still true, but now the trade becomes the 4th choice:

Code: Select all

info depth 28 seldepth 38 multipv 1 score cp 2461 nodes 211840816 nps 1937982 hashfull 1000 tbhits 0 time 109310 pv f3g5 c7g3 g5f7 d8c7 h2g3 d7d5 c4d5 g6g5 f7g5 b7b6 e1c1 c7b7 e3e4 b7a7 c2e2 a7a8 d5c6 b8c6 e4e5 f6d7 h1h5 b6b5 e5e6 d7e5 h5h7 a8b7 b2b3
info depth 28 seldepth 43 multipv 2 score cp 2276 nodes 211840816 nps 1937982 hashfull 1000 tbhits 0 time 109310 pv c2b3 b7b5 c4b5 c7g3 h2g3 c6b5 b3f7 b8c6 f7g6 d7d6 f3g5 c8d7 g6g7 c6e5 a2a3 d7c6 b2b3 c6d7 e3e4 e5g4 a1c1 d7e8 a3a4 b5b4 c3b1
info depth 28 seldepth 39 multipv 3 score cp 2253 nodes 211840816 nps 1937982 hashfull 1000 tbhits 0 time 109310 pv c2a4 d7d6 f3g5 c8e6 e1c1 f6d7 g5e6 f7e6 a4c2 d6d5 f2f4 d7f8 e3e4 h5h4 g3h4 c7f4 c1b1 g6g5 h4f2 f4e5 c3a4 d5d4 c2b3 f8d7 b3b7 e5f6 f2g3 f6e5
info depth 28 seldepth 36 multipv 4 score cp 2215 nodes 211840816 nps 1937982 hashfull 1000 tbhits 0 time 109310 pv g3c7 d8c7 f3g5 d7d6 g5f7 c8d7 c2g6 d7e8 g6g7 e8f7 g7f7 c7d8 f7f8 f6e8 f8h8 b7b5 h8h5 e8f6 h5h8 f6e8 e1c1 b8d7 h8h5 e8f6 h5f5 d7b6 b2b3 d8c7 g2g4 f6d7 f2f4 b5c4 b3c4 d6d5 c4d5
info depth 27 seldepth 30 multipv 5 score cp 2200 nodes 211840816 nps 1937982 hashfull 1000 tbhits 0 time 109310 pv g3h4 d8e8 b2b4 e8f8 e1c1 d7d6 f3d2 a6a5 a2a3 b8d7 b4b5 c6b5 c4b5 d7c5 e3e4 c8e6 h2h3 f8g8 a3a4 c7b6 h4g3 d6d5 f2f3 d5e4 f3e4 g8g7 d1e1
 

Alex

[Uri Blass]

I expect even with using only piece square table evaluation
that the advantage of white get bigger with a deeper search and I think that the evaluation should encourage trading pieces when I am sure that even if you do not find a forced win of material you should find a force trade of material.

How can the PST evaluation encourage trading pieces? I think that the opposite happens: in absence of forced win of material, the stronger side pieces tend to occupy better squares than the weaker's, since he's stronger, so trading pieces will lower his score.

Alex

1)I expect even PST table to see increasing advantage without trading pieces because the better side can improve the square of his pieces.
2)I think that a good classical evaluation should encourage trades for the side with more material regardless of piece square table.

You can multiply the score by 1.1 for every bishop that you trade and when you start from +20 you get +22 after the trade.
Piece square table difference from trading is certainly not going to be +2.

[Brunetti]

1)I expect even PST table to see increasing advantage without trading pieces because the better side can improve the square of his pieces.

Agreed; that's what usually happens, I think.

2)I think that a good classical evaluation should encourage trades for the side with more material regardless of piece square table.

Why? Trading pieces is your strategy, not necessarily the correct one. Just a trivial example:


[fen]4krr1/8/8/3KB3/8/8/8/3R2R1 w - - 0 1[/fen]


The immediate trade draws. Even MOVEI avoids the trade

Code: Select all

FEN: 4krr1/8/8/3KB3/8/8/8/3R2R1 w - - 0 1

MOVEI 0.08.438:
 1	00:00	 224	4k	+4.72	Rg1-h1
 2	00:00	 229	4k	+4.42	Rg1-h1
 2	00:00	 248	4k	+3.92	Rg1-h1 Rg8-g5
 2	00:00	 266	4k	+3.93	Be5-g7
 2	00:00	 284	5k	+4.22	Be5-g7
 2	00:00	 343	6k	+4.45	Be5-g7 Rf8-f4 Rg1-e1+ Ke8-d8
 2	00:00	 508	508k	+4.45	Be5-g7 Rf8-f4 Rg1-e1+ Ke8-d8
 3	00:00	 841	841k	+4.60	Be5-g7 Rf8-f2 Kd5-e6 Rf2-e2+ Ke6-d6
 3	00:00	 2k	809k	+4.60	Be5-g7 Rf8-f2 Kd5-e6 Rf2-e2+ Ke6-d6
 4	00:00	 3k	717k	+4.60	Be5-g7 Rf8-f2 Kd5-e6 Rf2-e2+ Ke6-d6
 4	00:00	 5k	746k	+4.60	Be5-g7 Rf8-f2 Kd5-e6 Rf2-e2+ Ke6-d6
 5	00:00	 8k	812k	+4.30	Be5-g7
 5	00:00	 9k	857k	+3.65	Be5-g7 Rf8-f7 Rg1-e1+ Rf7-e7 Bg7-e5 Rg8-g5
 5	00:00	 10k	806k	+3.66	Rg1-e1
 5	00:00	 10k	828k	+3.95	Rg1-e1
 5	00:00	 22k	784k	+4.53	Rg1-e1 Rf8-f5 Kd5-e6 Rf5-h5 Be5-f6
 5	00:00	 25k	812k	+4.53	Rg1-e1 Rf8-f5 Kd5-e6 Rf5-h5 Be5-f6
 6	00:00	 35k	785k	+4.24	Rg1-e1 Ke8-f7 Kd5-e4 Rg8-g2 Ke4-f5 Rg2-f2+ Kf5-e4
 6	00:00	 45k	789k	+4.25	Rg1-h1
 6	00:00	 57k	787k	+4.54	Rg1-h1
 6	00:00	 81k	793k	+4.59	Rg1-h1 Rf8-f2 Rd1-c1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rc1-c8+ Ke8-f7
 6	00:00	 84k	805k	+4.59	Rg1-h1 Rf8-f2 Rd1-c1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rc1-c8+ Ke8-f7
 7	00:00	 122k	804k	+4.73	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rd2-g2 Rb1-b8+ Ke8-d7 Rh1-h7+ Kd7-c6 Rh7-c7+ Kc6-d5
 7	00:00	 141k	832k	+4.73	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rd2-g2 Rb1-b8+ Ke8-d7 Rh1-h7+ Kd7-c6 Rh7-c7+ Kc6-d5
 8	00:00	 213k	830k	+4.75	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rb1-b8+ Rd2-d8 Rb8-b7
 8	00:00	 263k	865k	+4.75	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rb1-b8+ Rd2-d8 Rb8-b7
 9	00:00	 425k	851k	+4.57	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e4 Rg8-g4+ Be5-f4 Rd2-e2+ Ke4-f5 Rg4-g2 Rb1-b8+ Ke8-d7 Rh1-h7+ Kd7-c6 Rh7-c7+ Kc6-d5
 9	00:00	 588k	916k	+4.57	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e4 Rg8-g4+ Be5-f4 Rd2-e2+ Ke4-f5 Rg4-g2 Rb1-b8+ Ke8-d7 Rh1-h7+ Kd7-c6 Rh7-c7+ Kc6-d5
 10	00:00	 907k	924k	+4.87	Rg1-h1
 10	00:01	 1,021k	939k	+4.87	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rh1-h6 Rg8-f8+ Kf5-e6 Rd2-d8 Be5-d6
 10	00:01	 1,248k	982k	+4.87	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rh1-h6 Rg8-f8+ Kf5-e6 Rd2-d8 Be5-d6
 11	00:01	 1,558k	998k	+4.87	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rh1-h6 Rg8-f8+
 11	00:01	 2,059k	1,060k	+4.87	Rg1-h1 Rf8-f2 Rd1-b1 Rf2-d2+ Kd5-e6 Rg8-g6+ Ke6-f5 Rg6-g8 Rh1-h6 Rg8-f8+
 12	00:03	 3,532k	1,057k	+5.14	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Rf7-f8 Rc1-c8+ Ke8-d7 Rc8-c7+ Kd7-e8 Rh1-d1
 12	00:04	 4,746k	1,101k	+5.14	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Rf7-f8 Rc1-c8+ Ke8-d7 Rc8-c7+ Kd7-e8 Rh1-d1
 13	00:05	 5,923k	1,115k	+5.23	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Ke8-e7 Rc1-c7+ Ke7-d8 Rh1-d1+ Kd8-e8 Rc7-c8+ Ke8-e7 Rd1-b1 Ke7-e6 Rb1-b6+ Ke6-d7
 13	00:07	 8,576k	1,156k	+5.23	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Ke8-e7 Rc1-c7+ Ke7-d8 Rh1-d1+ Kd8-e8 Rc7-c8+ Ke8-e7 Rd1-b1 Ke7-e6 Rb1-b6+ Ke6-d7
 14	00:07	 9,151k	1,159k	+5.53	Rg1-h1
 14	00:11	 14,077k	1,189k	+6.05	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Ke8-e7 Rc1-c7+ Ke7-e8 Rc7-c8+ Ke8-d7 Rh1-h8 Rf5xe5+ Ke4xe5 Rf7-f1
 14	00:15	 18,500k	1,224k	+6.05	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Ke8-e7 Rc1-c7+ Ke7-e8 Rc7-c8+ Ke8-d7 Rh1-h8 Rf5xe5+ Ke4xe5 Rf7-f1
 15	00:23	 28,988k	1,231k	+6.09	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Rf5xe5+ Ke4xe5 Rf7-e7+ Ke5-d6 Re7-d7+ Kd6-c6 Rd7-d2 Rh1-f1 Ke8-d8 Rf1-f7
 15	00:32	 40,522k	1,253k	+6.09	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Be5-d6 Rf7-f5+ Kd5-e4 Rf8-f7 Bd6-e5 Rf5xe5+ Ke4xe5 Rf7-e7+ Ke5-d6 Re7-d7+ Kd6-c6 Rd7-d2 Rh1-f1 Ke8-d8 Rf1-f7
 16	00:51	 64,643k	1,257k	+6.39	Rg1-h1
 16	00:52	 66,381k	1,259k	+7.09	Rg1-h1
 16	01:17	 98,251k	1,262k	+9.37	Rg1-h1 Rf8-f2 Rd1-c1 Ke8-d8 Rh1-h7 Rf2-d2+ Kd5-e4 Rg8-e8 Rc1-a1 Rd2-c2 Ke4-f5 Rc2-f2+
 16	01:25	 109,825k	1,285k	+9.37	Rg1-h1 Rf8-f2 Rd1-c1 Ke8-d8 Rh1-h7 Rf2-d2+ Kd5-e4 Rg8-e8 Rc1-a1 Rd2-c2 Ke4-f5 Rc2-f2+
 17	01:51	 141,737k	1,271k	+9.46	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Rc1-a1 Ke8-d7 Rh1-h6 Rf8-d8 Be5-d6 Rf7-f5+ Kd5-e4 Rf5-g5 Ra1-a7+ Kd7-c8 Ra7-c7+ Kc8-b8 Rc7-g7+ Kb8-c8 Rg7xg5
 17	02:06	 164,157k	1,294k	+9.46	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Rc1-a1 Ke8-d7 Rh1-h6 Rf8-d8 Be5-d6 Rf7-f5+ Kd5-e4 Rf5-g5 Ra1-a7+ Kd7-c8 Ra7-c7+ Kc8-b8 Rc7-g7+ Kb8-c8 Rg7xg5
 18	02:23	 185,493k	1,291k	+9.51	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Rc1-a1 Ke8-d7 Rh1-h6 Rf8-d8 Be5-d6 Rf7-f5+ Kd5-e4 Rf5-f2 Ra1-a7+ Kd7-c8 Ra7-c7+ Kc8-b8 Rc7-f7+ Kb8-c8 Rf7xf2 Rd8-g8
 18	02:57	 234,252k	1,317k	+9.51	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Rc1-a1 Ke8-d7 Rh1-h6 Rf8-d8 Be5-d6 Rf7-f5+ Kd5-e4 Rf5-f2 Ra1-a7+ Kd7-c8 Ra7-c7+ Kc8-b8 Rc7-f7+ Kb8-c8 Rf7xf2 Rd8-g8
 19	04:39	 355,858k	1,275k	+9.81	Rg1-h1
 19	06:34	 492,512k	1,249k	+9.98	Rg1-h1 Rf8-f7 Rd1-c1 Rg8-f8 Rc1-a1 Ke8-d7 Rh1-c1 Rf8-d8 Rc1-c7+ Kd7-e8+ Kd5-e6 Rf7xc7 Be5xc7 Ke8-f8 Bc7xd8 Kf8-g7 Ra1-a7+ Kg7-g6 Bd8-f6 Kg6-h5 Ke6-f5

You can multiply the score by 1.1 for every bishop that you trade and when you start from +20 you get +22 after the trade.
Piece square table difference from trading is certainly not going to be +2.

Well, I commented the "I expect even with using only piece square table evaluation" consideration. Anyway, you can do a lot of things to encourage trades, if you want to encourage trades. The point is that this is not always the correct strategy.

Alex

[Uri Blass]

Here are 2 games Wasp against Stockfish.
mate in 25 for Wasp when Wasp is white.
Mate in 41 for Stockfish when stockfish is white(I do not know why it is written depth 12 in the pgn but it is clear the depth is bigger and I used time control of 5 miunutes per game+3 seconds per move in banksia).

[Event "Computer event"]
[Site "Somewhere on Earth"]
[Date "2021.12.19"]
[Round "1"]
[White "Wasp 4.50"]
[Black "Stockfish 181221"]
[Result "1-0"]
[TimeControl "depth: 12"]
[Time "17:33:51"]
[Board "2"]
[Termination "mate"]
[FEN "1nb1kbn1/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 2"]
[SetUp "1"]

1. e4 {+23.36/20 9754 14633553} Nf6 {+23.23/29 17067 38190508} 2. e5 {+23.94/22 9060 15078059} Nd5 {+23.23/28 7325 15813433}
3. Nc3 {+24.11/22 9132 15293761} Nxc3 {+21.86/26 4303 9258342} 4. dxc3 {+24.32/24 9210 15631332} g6 {+22.00/26 8007 17480418}
5. e6 {+26.51/23 9289 16162227} f6 {+23.69/29 55042 121346175} 6. exd7+ {+27.58/23 9374 16545163} Nxd7 {+23.53/23 4542 9577533}
7. Qd5 {+28.49/23 9463 17031848} Ne5 {+23.61/27 7893 17306072} 8. Qg8 {+28.97/23 9557 17754569} Nf7 {+25.44/27 59898 144875303}
9. Qxh7 {+29.76/23 9652 18460359} Bf5 {+23.76/23 3187 7011244} 10. Bd3 {+30.56/23 9755 19393800} Bxd3 {+23.30/24 22905 54618148}
11. cxd3 {+31.13/21 9541 19182430} Ne5 {+23.50/27 4692 11523306} 12. d4 {+34.13/21 9336 19523209} Nd3+ {+24.18/31 25835 69026850}
13. Kf1 {+35.69/19 9135 20456385} g5 {+25.24/29 57534 148630924} 14. Qxd3 {+39.08/18 8942 20514043} c6 {+25.84/27 14257 39937402}
15. h4 {+43.53/18 8755 21940858} g4 {+26.23/19 7078 20118887} 16. Qg6+ {+49.40/17 8574 20673852} Kd7 {+43.78/25 19224 50872205}
17. Qf7 {M+10/15 8398 18006081} b5 {+75.45/23 23343 55819950} 18. Bf4 {M+8/15 5118 10945379} g3 {M-8/27 1778 3994833}
19. Qxf8 {M+7/13 694 1380279} gxf2 {M-7/27 958 2169825} 20. Kxf2 {M+6/11 88 185125} e5 {M-5/36 1750 4744247}
21. Qf7+ {M+5/9 66 152703} Kd8 {M-4/105 1225 6420232} 22. dxe5 {M+4/7 17 32933} fxe5 {M-3/245 86 668956}
23. Rd1+ {M+3/5 10 3648} Kc8 {M-2/245 14 15868} 24. Bxe5 {M+2/3 10 744} b4 {M-1/245 11 1252}
25. Qc7# {M+1/1 10 324} 1-0


[Event "Computer event"]
[Site "Somewhere on Earth"]
[Date "2021.12.19"]
[Round "1"]
[White "Stockfish 181221"]
[Black "Wasp 4.50"]
[Result "1-0"]
[TimeControl "depth: 12"]
[Time "17:36:50"]
[Board "1"]
[Termination "mate"]
[FEN "1nb1kbn1/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 2"]
[SetUp "1"]

1. g3 {+23.38/29 12842 27986019} Nf6 {+23.08/20 12013 19419076} 2. Bg2 {+23.38/33 13671 29026571} e6 {+23.05/19 12012 19256642}
3. a3 {+23.38/28 3639 7815908} d6 {+23.23/20 12013 19162779} 4. e3 {+23.46/30 5757 12292384} h6 {+23.05/20 12014 19490722}
5. Ne2 {+23.46/29 5502 11699479} c6 {+23.30/20 12015 19729707} 6. O-O {+23.46/26 7079 15285307} Be7 {+23.37/20 12016 19882769}
7. c4 {+23.57/29 7364 15674750} Kf8 {+23.41/21 12017 19547055} 8. Qa4 {+23.53/29 5037 10678912} a6 {+23.65/21 9010 14773606}
9. d3 {+23.46/29 5448 11344700} Bd8 {+23.61/21 12139 19897632} 10. Nd2 {+23.46/32 13774 29093800} e5 {+23.78/20 12142 20120002}
11. b4 {+23.46/27 13221 27729881} Bg4 {+23.74/21 11757 19779382} 12. f3 {+23.46/30 3961 8235842} Be6 {+24.04/20 11389 18998438}
13. Qc2 {+23.42/28 9129 19098509} Nbd7 {+24.25/20 11036 18395011} 14. a4 {+23.38/28 7291 15059487} Kg8 {+23.98/20 10696 17656700}
15. a5 {+23.46/26 6237 12947432} Be7 {+23.92/22 10374 17226995} 16. Rb1 {+23.45/28 16686 35311719} Bd8 {+24.41/20 10062 16913397}
17. f4 {+23.61/26 7650 16295082} Nf8 {+25.07/21 9763 16705136} 18. f5 {+23.61/27 5696 11905893} Bd7 {+25.16/23 7102 11922081}
19. Nb3 {+23.61/27 17003 36275426} Bc7 {+24.26/20 9302 15523762} 20. Nc3 {+23.61/24 5248 11018457} Ne8 {+24.72/22 9313 15695906}
21. g4 {+23.69/25 6429 13624750} Nf6 {+24.60/21 8767 14220531} 22. h3 {+23.76/27 6407 13857931} N8h7 {+24.66/21 6388 10444312}
23. e4 {+24.52/27 5755 12502022} Bc8 {+26.92/20 8924 14975880} 24. Qf2 {+26.46/24 9297 20167959} c5 {+28.88/22 8127 13726070}
25. bxc5 {+27.00/32 5774 12752878} Bd7 {+29.77/22 5926 9767983} 26. cxd6 {+27.02/29 5883 12767666} Bxd6 {+30.70/22 5831 9763999}
27. Nc5 {+27.63/30 5918 12741619} Bc6 {+32.25/22 7662 12864793} 28. Nxb7 {+28.39/34 6079 13684190} Bf8 {+33.56/22 7458 12697026}
29. Nc5 {+28.84/27 5782 13112389} Be8 {+33.81/20 7267 12021106} 30. Nxa6 {+29.92/28 7088 15925616} Bd6 {+35.24/20 7083 11951402}
31. Qb6 {+30.48/33 5865 12893250} Be7 {+38.91/22 7089 11955754} 32. Qa7 {+31.84/28 5818 12946014} Bf8 {+37.17/20 5044 8276410}
33. Rb8 {+32.61/27 6301 14055985} Kh8 {+41.31/22 6641 11047273} 34. Qc7 {+33.87/31 6208 14140089} Ng5 {+48.71/20 11130 20321571}
35. Bxg5 {M+8/55 6192 11266175} Nh7 {M-8/15 914 1631102} 36. Rxe8 {M+6/64 5825 11718282} Kg8 {M-6/13 76 118139}
37. Rb1 {M+5/87 6042 21863269} h5 {M-6/12 134 212719} 38. Qe7 {M+4/154 5900 26554276} hxg4 {M-4/7 11 6304}
39. Rb7 {M+3/245 301 1323216} Nxg5 {M-3/5 10 483} 40. Qxf8+ {M+2/245 14 13325} Kh7 {M-2/3 10 10}
41. Qh8# {M+1/245 13 12528} 1-0