jomega
STS(v10.0) Simplification.002
To exchange or not exchange; that is the question.A continuation of the discussion started here:
jomega's Blog • Strategic Test Suite (STS): The EPD file's best and alternate best moves. • lichess.org
Another interesting position for which Stockfish 14 failed to find the best move in the 3 second (depth 24) test. I think Stockfish is correct in not exchanging Knights here.
The STS web page says of this category of positions:
Simplification: A strategy of exchanging pieces of equal value. Simplification can be used defensively to reduce the size of an attacking force. It can also be used by a player with an advantage to amplify that advantage or reduce the opponent's counterplay. Simplification is also used as an attempt to obtain a draw, or as an attempt to gain an advantage by players who are strong in endgame play with simplified positions.
* Tests your engine's knowledge on initiatives, good trades, value of pieces, understanding of the position material-wise and positional-wise, and ability to distinguish between good, average and bad pieces.
The Human Perspective
Material is equal. Both sides have some weak pawns; g7, e5, e4, and b2. Black has more space and slightly better mobility. White's has, by definition, a bad Bishop. Black would like to put another attacker on the e4-pawn; for example, ...Rd4, but White's Knight guards that square, and as the study shows, the tactics don't work out to play ...Rd4 immediately.
Except for ...Rge8, which guards the e5-pawn, the other Rook moves seem to be a waste of time. So do the other choices. Hence, the choice seems to be either 1...Nxf3 or 1...Rge8. However, 1...Nxf3 allows 2.gxf3 and now White has a pawn protector of e4. Hence, 1...Rge8 is better. Black can mount an attack on the e4-pawn.
Stockfish's Perspective
Stockfish looks deep enough that it can see that the Knight exchange results in White having a pawn protector of the once weak e4-pawn. Black can get a Rook to d4, but Stockfish evaluates the position at that point as -0.4/23. Instead, Stockfish protects the e5-pawn, and then retreats the g5-Knight, with the idea to later use it to mount an attack on White's weak e4-pawn. The evaluation of the resulting position after 2...Nf7 is -1.4/22. The reason for that evaluation is that Stockfish sees White's best chance as a line where White gives up a pawn on the queenside to avoid moving his c3-Knight; which is protecting the e4-pawn.
EPD Changes Needed
The scores in the EPD should be Rge8=10, Bc6=7, Rh8=4, Rgf8=4, Ba8=4. Those scores of 4 are an indication of the engine playing "cat-and-mouse". The move 1...Nxf3 is clearly worse than 1...Rge8!
Links
- The Strategic Test Suite (STS) home page.
https://sites.google.com/site/strategictestsuite/
- The STS-rating code.
https://github.com/fsmosca/STS-Rating
