Are opening principles correct?

[kasinp]

Funny thing is: both Powerbook2018 and HIARCS book give Bd6 as the best move for black.
A move that doesn't seem to be considered in this thread at all.

PK

[jdart]

I really don't think even deep search from an early opening position is reliable, in terms of choosing the best moves.

What the eval tells you is a rough proxy for what actual game results might be. It is not always accurate because heavy pruning is removing a lot of the search tree and sometimes perceived advantages near the root do not persist through the game. Small eval differences in particular are not reliable.

Here for example is a position I have been looking at recently (Benoni, A65):

[d]rnbq1rk1/pp2ppb1/3p1npp/2pP4/2P1PB2/2N2P2/PP4PP/R2QKBNR b KQ - 0 8

Here .. a6 is probably the most common but .. Qa5 as in Moiseenko-Van Wely, 2005 is recommended by the Hiarcs book. However, a strong engine recently won a game on ICC with .. Nh5, which after Be3 e5 is another known line - this doesn't look good to Stockfish etc. based on search but the game stats from GM games are actually not bad.

--Jon

[zullil]

I have used Contempt too, but results were never that strange for me. Now i would definitely try "Contempt=Off", to see what happens...

Stockfish-dev with Analysis Contempt = Off, Threads = 20, Hash = 32768, MultiPV = 3:

info depth 40 seldepth 55 multipv 1 score cp -3 nodes 36738335623 nps 26830227 hashfull 999 tbhits 0 time 1369289 pv d8f6 d3e2 f8d6 e1g1 g8e7 f1e1 h7h6 b1d2 g4e6 g2g3 e8g8 d2f1 c6b8 f1e3 c7c5 e3g2 b8c6 c1f4 d6f4 g2f4 e7g6 f4g2 b7b6 g2e3 g6e7 g1g2 c5c4 b2b3 b6b5 a2a4 a7a6 b3b4 f8e8 d1d2 e7f5 e3g4 f6d8 g4e5 f5e7 e5c6 e7c6 a4b5 a6b5 a1a8

info depth 40 seldepth 55 multipv 2 score cp -22 nodes 36738335623 nps 26830227 hashfull 999 tbhits 0 time 1369289 pv f8d6 h2h3 g4h5 e1g1 g8e7 f1e1 f7f6 b1a3 a7a6 a3c2 e8g8 b2b3 d8d7 c3c4 b7b6 a2a3 a6a5 c1e3 a5a4 b3a4 d5c4 d3c4 h5f7 c4b5 f7b3 d4d5 d7f5 f3d4 c6d4 c2d4 b3d1 d4f5 e7f5 a1d1 f5e3 e1e3 g7g6 b5c6 a8d8 e3b3 g8g7 d1e1 f6f5 a4a5 b6a5 b3b5 a5a4 b5b7 d6a3 c6a4

info depth 40 seldepth 63 multipv 3 score cp -22 nodes 36738335623 nps 26830227 hashfull 999 tbhits 0 time 1369289 pv d8d7 e1g1 f8d6 b2b4 a7a6 f1e1 g8e7 b1d2 e8g8 d2b3 e7g6 h2h3 g4f5 d3f1 a8e8 c1e3 d7c8 a2a4 c6b8 b3c5 b8d7 d1b3 d7f6 e3g5 f6e4 b3d5 e4c3 d5b7 c8b7 c5b7 d6b4 e1e8 f8e8 g5d2 f5e4 a1e1 e8e7 b7c5 b4c5 d2c3 e4f3 d4c5 e7e1 c3e1 f3c6 f1a6 c6a4

So far, Stockfish-dev prefers Qf6, with an advantage for White of just 3 centipawns.

And now:

info depth 47 seldepth 72 multipv 1 score cp -21 nodes 301211848886 nps 29231619 hashfull 999 tbhits 0 time 10304316 pv f8d6 h2h3 g4h5 e1g1 g8e7 b1d2 e8g8 f1e1 f7f6 d1b3 a8b8 f3h4 g8h8 d2f1 g7g5 h4f5 e7f5 d3f5 c6e7 f5e6 c7c6 a2a4 d6c7 c1d2 h5g6 c3c4 d5c4 b3c4 e7d5 e6d5 d8d5 c4d5 c6d5 f1e3 c7b6 d2c3 b8d8 a4a5 b6c7 c3b4 f8f7 e1c1 f6f5 c1c5 f7d7 b4d2 c7b8 a1c1 f5f4 e3g4 h8g7 b2b3 d8e8 d2c3 g6f7 g4e5 b8e5

info depth 47 seldepth 66 multipv 2 score cp -27 nodes 301211848886 nps 29231619 hashfull 999 tbhits 0 time 10304316 pv d8e7 c1e3 e7d7 h2h3 g4h5 e3f4 d7e6 e1f1 e8c8 b1d2 h5g6 d1e2 g6d3 e2d3 g8e7 g2g3 f7f6 d2b3 e7g6 f4d2 e6e4 d3e4 d5e4 f3g1 a7a5 d2e3 f8d6 f1g2 b7b6 a1e1 c8b7 h3h4 f6f5 b3d2 a5a4 a2a3 g6e7 f2f3 e4f3 g1f3 h8e8 e3g5 h7h6 g5e7 c6e7 d2c4 e7d5 f3e5 d6e5 c4e5 d5f6 e5f7 d8a8 e1e8 a8e8 f7e5

info depth 47 seldepth 69 multipv 3 score cp -31 nodes 301211848886 nps 29231619 hashfull 999 tbhits 0 time 10304316 pv d8d7 a2a4 f8d6 e1g1 g8e7 b1a3 a7a6 a3c2 f7f6 c2e3 g4h5 d3e2 e8g8 f1e1 f8e8 c1d2 d6f4 e3f1 f4d2 f3d2 h5e2 d1e2 e7f5 e2g4 e8e1 a1e1 a8d8 d2b3 b7b6 b3c1 a6a5 c1d3 f5d6 g4d7 d8d7 f1e3 d6c4 g2g4 g8f7 g1g2 d7e7 g2f3 c4e3 e1e3 g7g5 e3e7 c6e7 b2b4 f6f5 b4a5 b6a5 d3c5 f5g4 f3g4 f7f6 c5b3 e7c6

[zullil]

Funny thing is: both Powerbook2018 and HIARCS book give Bd6 as the best move for black.
A move that doesn't seem to be considered in this thread at all.

PK

Now considered: viewtopic.php?f=2&t=69199&start=10#p782011

[MikeB]

Funny thing is: both Powerbook2018 and HIARCS book give Bd6 as the best move for black.
A move that doesn't seem to be considered in this thread at all.

PK

Lc0 likes Bd6 initially and after a while ( 57k nodes later - NN 11261), switches to Qf6, Cerebellum book also likes Qf6. Personally I prefer Bd6 first then Nf6 move and would avoid Qf6, but I'm sure SF will play just fine with Qf6. Openings are not engines' strong suit, perhaps with NN engines that will change.

[Eelco de Groot]

I really don't think even deep search from an early opening position is reliable, in terms of choosing the best moves.

What the eval tells you is a rough proxy for what actual game results might be. It is not always accurate because heavy pruning is removing a lot of the search tree and sometimes perceived advantages near the root do not persist through the game. Small eval differences in particular are not reliable.

Here for example is a position I have been looking at recently (Benoni, A65):

[d]rnbq1rk1/pp2ppb1/3p1npp/2pP4/2P1PB2/2N2P2/PP4PP/R2QKBNR b KQ - 0 8

Here .. a6 is probably the most common but .. Qa5 as in Moiseenko-Van Wely, 2005 is recommended by the Hiarcs book. However, a strong engine recently won a game on ICC with .. Nh5, which after Be3 e5 is another known line - this doesn't look good to Stockfish etc. based on search but the game stats from GM games are actually not bad.

--Jon

Hello Jon,

I think you may have meant ... Nh5 Be3 e6 which is what Kaissa prefers? After ... Nh5 Be3 e5 the score drops alarmingly for Black. Is that really theory? But at least five other moves are perfectly playable, maybe more?

After ... Nh5 Be3 e5


[d]rnbq1rk1/pp3pb1/3p2pp/2pPp2n/2P1P3/2N1BP2/PP4PP/R2QKBNR w KQ e6

Code: Select all

Engine: Kaissa IV (512 MB)
by T. Romstad, M. Costalba, J. Kiiski, G. Linscott

32     6:09   +1.34    3.Qd2 Kh7 4.g4 Nf4 5.Nge2 Nd7 6.h4 a6 
                       7.a3 Qa5 8.Nxf4 exf4 9.Bxf4 Ne5 
                       10.Be2 Bd7 11.Kf2 b5 12.Bxe5 Bxe5 
                       13.f4 Bg7 14.g5 bxc4 15.Bxc4 (471.673.250) 1278 

32     6:09   +1.34    3.Nge2 f5 4.exf5 gxf5 5.Qd2 Na6 
                       6.Bxh6 Qh4+ 7.g3 Nxg3 8.Bg5 Nxf1+ 
                       9.Bxh4 Nxd2 10.Kxd2 Rf7 11.Nb5 Bf8 
                       12.Rhg1+ Kh7 13.Rae1 Bd7 14.Nec3 Bxb5 
                       15.Nxb5 Nc7 16.Nxc7 (471.673.250) 1278 

32     6:09   +1.19    3.g4 Nf4 4.Nge2 h5 5.gxh5 Qh4+ 6.Kd2 Nxh5 
                       7.Qe1 Qxe1+ 8.Rxe1 Rd8 9.Ng3 Nf4 
                       10.h4 Nd7 11.h5 Nf6 12.hxg6 fxg6 
                       13.Nce2 g5 14.Kc2 Kf7 15.Nc3 Bd7 
                       16.Nf5 (471.673.250) 1278 

32     6:09   +1.13    3.Qc1 Nd7 4.Nh3 Kh7 5.Qc2 Nb6 6.Nf2 f5 
                       7.exf5 Bxf5 8.Nfe4 Nd7 9.Bd3 Nf4 
                       10.O-O Nxd3 11.Qxd3 Nf6 12.a3 Rc8 
                       13.b4 b6 14.b5 Qd7 15.a4 Nxe4 
                       16.Nxe4 (471.673.250) 1278 

32     6:09   +1.03    3.Qc2 f5 4.O-O-O f4 5.Bf2 Bd7 6.Kb1 Na6 
                       7.g4 Nf6 8.h4 Nc7 9.Bh3 Rb8 10.g5 Nh5 
                       11.Bg4 b5 12.gxh6 Bxh6 13.cxb5 Nxb5 
                       14.Nxb5 Bxb5 15.Bxh5 gxh5 16.Nh3 (471.673.250) 1278 
_____________________________________________________________

33     7:49   +1.53    3.Qd2 Kh7 4.Nge2 Nd7 5.g4 Nf4 6.h4 a6 
                       7.Nxf4 exf4 8.Bxf4 Ne5 9.Be2 Bd7 
                       10.Kf2 Qe7 11.a4 f5 12.exf5 gxf5 
                       13.Bd3 Nxd3+ 14.Qxd3 Qf6 15.Kg3 Qd4 
                       16.Rad1 (598.765.535) 1276 

33     7:49   +1.26    3.Nge2 f5 4.exf5 gxf5 5.Qd2 Na6 
                       6.Rg1 f4 7.Bf2 Nf6 8.g4 Nc7 9.h4 b5 
                       10.g5 hxg5 11.hxg5 Nh7 12.Ne4 b4 
                       13.g6 Nf6 14.Nxc5 dxc5 15.Bxc5 Bf5 
                       16.Bxb4 (598.765.535) 1276 

33     7:49   +1.24    3.Qc1 Nd7 4.Nh3 Kh7 5.Qc2 Nb6 6.Nf2 Bd7 
                       7.a4 a6 8.a5 Nc8 9.Bd2 Ne7 10.Bd3 f5 
                       11.O-O Nf4 12.Rfe1 fxe4 13.Bxe4 Nf5 
                       14.Bxf5 gxf5 15.Bxf4 exf4 16.Qb3 (598.765.535) 1276 

33     7:49   +1.16    3.g4 Nf4 4.Nge2 h5 5.gxh5 Nxh5 6.Qd2 Nd7 
                       7.O-O-O Ndf6 8.Ng3 Bd7 9.Bg5 Nf4 
                       10.Bxf4 exf4 11.Qxf4 Ne8 12.Qd2 Qa5 
                       13.Kb1 Bxc3 14.Qxc3 Qxc3 15.bxc3 Nf6 
                       16.Bd3 (598.765.535) 1276 

33     7:49   +0.84    3.Qc2 f5 4.O-O-O f4 5.Bf2 Na6 6.Kb1 Bd7 
                       7.g4 Nf6 8.h4 Nc7 9.g5 hxg5 10.hxg5 Nh5 
                       11.Bh3 Bxh3 12.Nxh3 a6 13.Na4 b5 
                       14.Nxc5 dxc5 15.Bxc5 bxc4 16.Bxf8 (598.765.535) 1276
____________________________________________________________

35     14:38  +1.55    3.Nge2 f5 4.exf5 gxf5 5.Qd2 f4 6.Bf2 a6 
                       7.Ne4 Nf6 8.N2c3 Nxe4 9.Nxe4 Bf5 
                       10.Bd3 Qe7 11.O-O Nd7 12.Rfe1 Kh8 
                       13.b3 Bh7 14.a4 Rac8 15.a5 Nf6 
                       16.Nc3 (1.127.954.702) 1283 

35     14:38  +1.45    3.Qd2 Kh7 4.Nge2 Nd7 5.g4 Nf4 6.Nxf4 exf4 
                       7.Bxf4 Ne5 8.Be2 Bd7 9.Bg3 a6 10.O-O g5 
                       11.Rae1 b5 12.Nd1 Rb8 13.cxb5 axb5 
                       14.Ne3 b4 15.Nf5 Ra8 16.Nxg7 (1.127.954.702) 1283 

35     14:38  +1.27    3.g4 Nf4 4.Nge2 h5 5.gxh5 Nxh5 6.Qd2 Nd7 
                       7.Ng3 Ndf6 8.Bg5 Nf4 9.O-O-O Qa5 
                       10.Rg1 Nh7 11.Bxf4 exf4 12.Nge2 a6 
                       13.Nxf4 Bd7 14.Kb1 Bd4 15.Bd3 b5 
                       16.h4 (1.127.954.702) 1283 

35     14:38  +1.13    3.Qc1 Nd7 4.Nh3 Kh7 5.Nf2 Nb6 6.a4 f5 
                       7.Qc2 Bd7 8.a5 Nc8 9.Bd3 Ne7 10.O-O Nf4 
                       11.a6 b6 12.Be2 Nxe2+ 13.Qxe2 Rc8 
                       14.Bd2 Bf6 15.Nb5 Bxb5 16.cxb5 (1.127.954.702) 1283 

35     14:38  +1.06    3.a3 f5 4.exf5 gxf5 5.Qd2 Qh4+ 6.Bf2 Qf6 
                       7.Nh3 Nd7 8.Be2 Nb6 9.O-O Bd7 10.Kh1 Na4 
                       11.Nxa4 Bxa4 12.Rae1 b6 13.b4 Be8 
                       14.Bd3 Bd7 15.Bc2 Rae8 16.bxc5 (1.127.954.702) 1283  

After ... Nh5 Be3, many playable moves, so "Het paard op de rand is geen schand"

[d]rnbq1rk1/pp2ppb1/3p2pp/2pP3n/2P1P3/2N1BP2/PP4PP/R2QKBNR b KQ -

Engine: Kaissa IV (512 MB)
by T. Romstad, M. Costalba, J. Kiiski, G. Linscott

41 40:28 +0.44 2...a6 3.Qd2 Kh7 4.Nge2 Qa5 5.Ng3 Nd7
6.Nxh5 gxh5 7.Be2 b5 8.O-O bxc4 9.f4 Rb8
10.Rac1 Bd4 11.Bf2 Rg8 12.Rb1 Qb6
13.Kh1 Nf6 14.Bh4 Bd7 15.Qc2 Kh8 (3.096.120.759) 1275

41 40:28 +0.52 2...Kh7 3.Qd2 e6 4.Nge2 exd5 5.Nxd5 Nc6
6.Nec3 Bd7 7.Be2 Nd4 8.O-O Nxe2+
9.Nxe2 Be6 10.Rad1 Nf6 11.Nec3 Nxd5
12.Nxd5 b6 13.b3 Rc8 14.Rfe1 Bd7
15.Nf4 g5 (3.096.120.759) 1275

41 40:28 +0.54 2...e6 3.Qd2 exd5 4.Nxd5 Nc6 5.Ne2 Kh7
6.Nec3 Bd7 7.Be2 Nd4 8.O-O Nxe2+
9.Nxe2 Be6 10.Rad1 Nf6 11.Nec3 Nxd5
12.Nxd5 Re8 13.b3 b6 14.Rfe1 Rc8
15.Qd3 Be5 (3.096.120.759) 1275

41 40:28 +0.62 2...Nd7 3.Qd2 Kh7 4.Nge2 Qa5 5.Ng3 a6
6.Nxh5 gxh5 7.Be2 b5 8.O-O bxc4 9.f4 Rb8
10.Rac1 Bd4 11.Bf2 Rg8 12.Rb1 Qb6
13.Kh1 Nf6 14.Bh4 Bd7 15.Qc2 Kh8 (3.096.120.759) 1275

41 40:28 +0.68 2...Nf6 3.Qd2 Kh7 4.Nge2 a6 5.Ng3 e6
6.Be2 exd5 7.cxd5 Nbd7 8.a4 b6 9.O-O Bb7
10.h3 Re8 11.b3 Qe7 12.Kh2 Ng8
13.Rac1 f5 14.Bf4 Ngf6 15.Rfe1 fxe4 (3.096.120.759) 1275
(All scores for White but default contempt)

[jdart]

I actually meant .. e5, and the point is that evals may overestimate White's chances here.

Here is a GM game drawn after .. e5:

[pgn][Event "Moscow op 03rd"] [Site "Moscow"] [Date "2007.02.03"] [Round "8"] [White "Riazantsev, Alexander"] [Black "Amonatov, Farrukh"] [Result "1/2-1/2"] [ECO "E81"] [WhiteElo "2629"] [BlackElo "2559"] [PlyCount "83"] [EventDate "2007.01.27"] [EventType "swiss"] [EventRounds "9"] [EventCountry "RUS"] [SourceTitle "CBM 116 Extra"] [Source "ChessBase"] [SourceDate "2007.03.07"] [SourceVersion "1"] [SourceVersionDate "2007.03.07"] [SourceQuality "1"] 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. f3 O-O 6. Bg5 c5 7. d5 h6 8. Bf4 Nh5 9. Be3 e5 10. Qd2 Kh7 11. Bd3 f5 12. exf5 Bxf5 13. Ne4 Nf6 14. Nh3 b5 15. Nxf6+ Qxf6 16. Nf2 bxc4 17. Bxc4 Nd7 18. O-O Nb6 19. Be2 Qf7 20. Rad1 a5 21. Ne4 Qe7 22. Rde1 a4 23. Nc3 Qh4 24. g4 Bc8 25. Bd3 Ba6 26. Bxa6 Rxa6 27. Ne4 Nc4 28. Qf2 Qxf2+ 29. Rxf2 Rb6 30. Bc1 Rb4 31. Ree2 Rd8 32. Rc2 Nb6 33. Bd2 Rd4 34. Ba5 Rb8 35. Bxb6 Rxb6 36. Rfd2 Rdb4 37. Kf1 Kg8 38. Ke2 Kf7 39. Kd3 Rd4+ 40. Ke2 Rdb4 41. Kd3 Rd4+ 42. Ke2 1/2-1/2 [/pgn]

and here is my program's recent loss, although I am sure its play could be improved:

[pgn][Event "chessclub.com"] [Site "chessclub.com"] [Date "2018.12.09"] [Round "?"] [White "Arasan CCC3-1-8-g6a396a9"] [Black "acgs"] [Result "0-1"] [ECO "A65"] [WhiteElo "2377"] [BlackElo "2641"] [PlyCount "99"] [EventDate "2018.??.??"] [TimeControl "1500+10"] 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 O-O 5. Bg5 d6 6. f3 c5 7. d5 h6 8. Bf4 Nh5 9. Be3 e5 10. Qd2 Kh7 11. Bd3 f5 12. exf5 Bxf5 13. Bxf5 gxf5 14. O-O-O Qh4 15. Nge2 Nd7 16. g4 fxg4 17. fxg4 Qxg4 18. Rhg1 Qf5 19. Nb5 e4 20. Kb1 Be5 21. Nec3 Rf7 22. Rg2 Ndf6 23. Rf2 Qh3 24. Nxe4 Nxe4 25. Rxf7+ Kg6 26. Qd3 Kxf7 27. Qxe4 Nf6 28. Qd3 Rg8 29. Rf1 Qxh2 30. Rf2 Qh1+ 31. Rf1 Qh2 32. Bd2 Ke7 33. Bc3 Rg1 34. Bxe5 dxe5 35. a4 e4 36. Qd1 Rxf1 37. Qxf1 Qe5 38. Qg1 Kf7 39. Qxc5 e3 40. Qd6 Qe4+ 41. Kc1 Qxc4+ 42. Nc3 Qf1+ 43. Kc2 Qf5+ 44. Kb3 h5 45. Qc7+ Kg6 46. Qxb7 h4 47. Qe7 h3 48. Qxe3 h2 49. Qg3+ Ng4 50. Qg2 0-1 [/pgn]

[zullil]

I have used Contempt too, but results were never that strange for me. Now i would definitely try "Contempt=Off", to see what happens...

Stockfish-dev with Analysis Contempt = Off, Threads = 20, Hash = 32768, MultiPV = 3:

info depth 40 seldepth 55 multipv 1 score cp -3 nodes 36738335623 nps 26830227 hashfull 999 tbhits 0 time 1369289 pv d8f6 d3e2 f8d6 e1g1 g8e7 f1e1 h7h6 b1d2 g4e6 g2g3 e8g8 d2f1 c6b8 f1e3 c7c5 e3g2 b8c6 c1f4 d6f4 g2f4 e7g6 f4g2 b7b6 g2e3 g6e7 g1g2 c5c4 b2b3 b6b5 a2a4 a7a6 b3b4 f8e8 d1d2 e7f5 e3g4 f6d8 g4e5 f5e7 e5c6 e7c6 a4b5 a6b5 a1a8

info depth 40 seldepth 55 multipv 2 score cp -22 nodes 36738335623 nps 26830227 hashfull 999 tbhits 0 time 1369289 pv f8d6 h2h3 g4h5 e1g1 g8e7 f1e1 f7f6 b1a3 a7a6 a3c2 e8g8 b2b3 d8d7 c3c4 b7b6 a2a3 a6a5 c1e3 a5a4 b3a4 d5c4 d3c4 h5f7 c4b5 f7b3 d4d5 d7f5 f3d4 c6d4 c2d4 b3d1 d4f5 e7f5 a1d1 f5e3 e1e3 g7g6 b5c6 a8d8 e3b3 g8g7 d1e1 f6f5 a4a5 b6a5 b3b5 a5a4 b5b7 d6a3 c6a4

info depth 40 seldepth 63 multipv 3 score cp -22 nodes 36738335623 nps 26830227 hashfull 999 tbhits 0 time 1369289 pv d8d7 e1g1 f8d6 b2b4 a7a6 f1e1 g8e7 b1d2 e8g8 d2b3 e7g6 h2h3 g4f5 d3f1 a8e8 c1e3 d7c8 a2a4 c6b8 b3c5 b8d7 d1b3 d7f6 e3g5 f6e4 b3d5 e4c3 d5b7 c8b7 c5b7 d6b4 e1e8 f8e8 g5d2 f5e4 a1e1 e8e7 b7c5 b4c5 d2c3 e4f3 d4c5 e7e1 c3e1 f3c6 f1a6 c6a4

So far, Stockfish-dev prefers Qf6, with an advantage for White of just 3 centipawns.

And now:

info depth 47 seldepth 72 multipv 1 score cp -21 nodes 301211848886 nps 29231619 hashfull 999 tbhits 0 time 10304316 pv f8d6 h2h3 g4h5 e1g1 g8e7 b1d2 e8g8 f1e1 f7f6 d1b3 a8b8 f3h4 g8h8 d2f1 g7g5 h4f5 e7f5 d3f5 c6e7 f5e6 c7c6 a2a4 d6c7 c1d2 h5g6 c3c4 d5c4 b3c4 e7d5 e6d5 d8d5 c4d5 c6d5 f1e3 c7b6 d2c3 b8d8 a4a5 b6c7 c3b4 f8f7 e1c1 f6f5 c1c5 f7d7 b4d2 c7b8 a1c1 f5f4 e3g4 h8g7 b2b3 d8e8 d2c3 g6f7 g4e5 b8e5

info depth 47 seldepth 66 multipv 2 score cp -27 nodes 301211848886 nps 29231619 hashfull 999 tbhits 0 time 10304316 pv d8e7 c1e3 e7d7 h2h3 g4h5 e3f4 d7e6 e1f1 e8c8 b1d2 h5g6 d1e2 g6d3 e2d3 g8e7 g2g3 f7f6 d2b3 e7g6 f4d2 e6e4 d3e4 d5e4 f3g1 a7a5 d2e3 f8d6 f1g2 b7b6 a1e1 c8b7 h3h4 f6f5 b3d2 a5a4 a2a3 g6e7 f2f3 e4f3 g1f3 h8e8 e3g5 h7h6 g5e7 c6e7 d2c4 e7d5 f3e5 d6e5 c4e5 d5f6 e5f7 d8a8 e1e8 a8e8 f7e5

info depth 47 seldepth 69 multipv 3 score cp -31 nodes 301211848886 nps 29231619 hashfull 999 tbhits 0 time 10304316 pv d8d7 a2a4 f8d6 e1g1 g8e7 b1a3 a7a6 a3c2 f7f6 c2e3 g4h5 d3e2 e8g8 f1e1 f8e8 c1d2 d6f4 e3f1 f4d2 f3d2 h5e2 d1e2 e7f5 e2g4 e8e1 a1e1 a8d8 d2b3 b7b6 b3c1 a6a5 c1d3 f5d6 g4d7 d8d7 f1e3 d6c4 g2g4 g8f7 g1g2 d7e7 g2f3 c4e3 e1e3 g7g5 e3e7 c6e7 b2b4 f6f5 b4a5 b6a5 d3c5 f5g4 f3g4 f7f6 c5b3 e7c6

Stockfish-dev says its top three are about equally good:

info depth 56 seldepth 83 multipv 1 score cp -24 nodes 2018611697333 nps 29991357 hashfull 999 tbhits 0 time 67306447 pv d8d7 e1g1 f8d6 f1e1 g8e7 b1d2 e8g8 d2f1 g4f5 d3e2 h7h6 f1g3 f5h7 e2d3 h7d3 d1d3 d6g3 h2g3 e7g6 c1d2 a7a6 b2b3 f8e8 a2a4 c6e7 d2c1 e7c6 c1a3 a8d8 d3c2 d8b8 f3d2 e8e1 a1e1 b8e8 e1e8 d7e8 c2d1 e8e6 d2f1 c6e7 f1e3 e7f5 e3c2 f5d6 a3d6 e6d6 c2e3 g6e7 g1f1 d6c6 d1d3 b7b5 a4b5 a6b5 f1e2 c6d7 f2f3 f7f5 e2f2 g8f7 g3g4

info depth 56 seldepth 75 multipv 2 score cp -27 nodes 2018611697333 nps 29991357 hashfull 999 tbhits 0 time 67306447 pv f8d6 e1g1 d8d7 f1e1 g8e7 b1d2 g4f5 d3e2 e8g8 d2f1 h7h6 f1g3 f5h7 e2d3 h7d3 d1d3 d6g3 h2g3 e7g6 c1d2 f8e8 f3h2 a7a5 e1e8 a8e8 d3b5 b7b6 h2f1 g6f8 f1e3 c6e7 b5d7 f8d7 a1e1 c7c6 g3g4 d7f8 g1h2 f8e6 h2g3 e7g6 e3c2 e6c7 f2f4 e8e1 d2e1 g6e7 g3f3 g8f8 f4f5 h6h5 g4h5 e7f5 g2g4 f5d6 e1g3 c7e8 g3d6 e8d6

info depth 56 seldepth 78 multipv 3 score cp -28 nodes 2018611697333 nps 29991357 hashfull 999 tbhits 0 time 67306447 pv d8f6 b1d2 f8d6 e1g1 g8e7 f1e1 e8g8 d3e2 a8e8 d2f1 e7f5 h2h3 g4f3 e2f3 e8e1 d1e1 c6e7 e1e2 c7c6 g2g3 g7g6 h3h4 f5g7 f3g4 e7f5 c1g5 f6e6 e2e6 f7e6 a1e1 f8e8 f1h2 e6e5 d4e5 e8e5 e1e5 d6e5 h2f3 e5d6 g4h3 g8f7 g5d2 g7e6 g1f1 c6c5 h3g2 h7h5 f1e2 c5c4 f3g5 e6g5 d2g5 f7e6 g2h3 b7b5 e2f3 e6f7 g5f4 d6f4 f3f4 f5d6

Terminating this search.

[Michael Sherwin]

I imagine that Stockfish sees Qe7 Be3 Qd7 Bf4 as being the same as Qd7 Bf4 therefore Qe7 and Qd7 are equal with each other. The reason that Qe7 was chosen at that depth over Qd7 is most likely that Qe7 was searched first and Qd7 did not replace it.

[BeyondCritics]

It believes Qe7+ is better than Qd7 even at depth 40 because it seems to believe e3 is bad for the bishop

The main line after Qd7 does not include Be3 but the plan 0-0 Re1 Nbd2 Nf1 Ne3

Here is the main lines at depth 40:Qf6 and Qe7 are considered to be best moves.
I used default contempt in the analysis


FEN: r2qkbnr/ppp2ppp/2n5/3p4/3P2b1/2PB1N2/PP3PPP/RNBQK2R b KQkq - 2 1

Stockfish_18120615_x64_modern:

40/54 2:20:21 24,633,110k 2,925k -0.29 1. ... Nf6 2.0-0 Be7 3.Bf4 0-0 4.Nbd2 Bd6 5.Bg5 h6 6.Bh4 g5 7.Bg3 Bxg3 8.hxg3 Qd6 9.b4 Bh5 10.Qc2 Kg7 11.b5 Ne7 12.Rfe1 c6 13.a4 Rae8 14.Ne5 Ng6 15.Ndf3 Ng4 16.Nxg4 Bxg4 17.Nd2 Bd7 18.Nb3 Rxe1+ 19.Rxe1 b6 20.bxc6 Bxc6 21.a5 Bd7 22.axb6 axb6 23.Nd2 Ne7 24.Qb3 Re8 25.Ra1
40/56 2:20:21 24,633,110k 2,925k -0.15 1. ... Qc8 2.0-0 Bd6 3.Re1+ Nge7 4.Nbd2 0-0 5.Nf1 Bf5 6.Ne3 Bxd3 7.Qxd3 Qd7 8.Bd2 Rae8 9.h3 h6 10.Re2 Bf4 11.Rae1 Ng6 12.Ng4 Bxd2 13.Qxd2 f6 14.Rxe8 Rxe8 15.Rxe8+ Qxe8 16.Qc2 Nce7 17.Ne3 Qd7 18.h4 h5 19.g3 Nf8 20.Kg2 c6 21.a3 b6 22.b3 Qe6 23.c4 Kf7 24.cxd5 Nxd5
40/59 2:20:21 24,633,110k 2,925k -0.09 1. ... Bh5 2.0-0 Bd6 3.Re1+ Nge7 4.Nbd2 h6 5.Nf1 Qd7 6.b4 a6 7.Ne3 0-0 8.h3 f5 9.a4 f4 10.Ng4 Bxg4 11.hxg4 Qxg4 12.Qb3 Qh5 13.Ba3 g5 14.b5 axb5 15.Bxd6 cxd6 16.axb5 Rxa1 17.Rxa1 Nd8 18.Re1 Re8 19.Qa3 Nc8 20.Qa5 Rxe1+ 21.Nxe1 Qe8 22.Nc2 Qd7 23.Kf1 Ne7 24.Nb4 Kh8 25.b6 Qc8 26.Nxd5 Nxd5 27.Qxd5
40/57 2:20:21 24,633,110k 2,925k -0.09 1. ... Qd7 2.0-0 Bd6 3.Re1+ Nge7 4.Nbd2 0-0 5.Nf1 h6 6.Ne3 Be6 7.g3 Ng6 8.Bb5 a6 9.Ba4 Rae8 10.Bd2 Rb8 11.c4 b5 12.Bxb5 axb5 13.cxd5 Bxd5 14.Nxd5 Rfd8 15.Rc1 Bxg3 16.Nf6+ gxf6 17.hxg3 Nxd4 18.Bxh6 Ne6 19.Qxd7 Rxd7 20.Rcd1 Rxd1 21.Rxd1 Ne5 22.Nxe5 fxe5 23.Rd7 b4 24.Bd2 Kg7 25.Kg2 c5
40/58 2:20:21 24,633,110k 2,925k -0.04 1. ... Bd6 2.0-0 Qd7 3.Re1+ Nge7 4.Nbd2 0-0 5.Nf1 h6 6.Ne3 Be6 7.g3 Ng6 8.Bb5 a6 9.Ba4 Rfe8 10.Bd2 b5 11.Bc2 Nce7 12.a4 c6 13.Qe2 Bh3 14.Qd3 a5 15.axb5 cxb5 16.Qe2 a4 17.Qd3 Rab8 18.Kh1 Qb7 19.Qe2 Bd7 20.h4 Nf8 21.Ne5 Bxe5 22.dxe5
40/59 2:20:21 24,633,110k 2,925k 0.00 1. ... Qe7+ 2.Be3 Qd7 3.h3 Bf5 4.Bxf5 Qxf5 5.0-0 Bd6 6.Na3 Bxa3 7.Qb3 Nge7 8.Qxa3 0-0 9.Rae1 Ng6 10.Qa4 h6 11.Re2 a6 12.Qd1 Rfe8 13.Rfe1 Re6 14.Nh2 Rae8 15.Nf1 Na5 16.Ng3 Qf6 17.b3 Nc6 18.c4 Rd8 19.Nh5 Qh4 20.Rd2 Qe7 21.Ng3 Qd7 22.a4 Kh7 23.Nf5 Nce7 24.Ng3
40/55 2:20:21 24,633,110k 2,925k 0.00 1. ... Qf6 2.Nbd2 Qe6+ 3.Qe2 Qxe2+ 4.Bxe2 Bd6 5.0-0 Nf6 6.Nb3 0-0 7.h3 Rfe8 8.Be3 Bh5 9.Rfe1 h6 10.a4 a6 11.a5 g5 12.Nc5 Bxc5 13.dxc5 Bxf3 14.Bxf3 Ne5 15.Be2 Kg7 16.g3 Nc6 17.h4 gxh4 18.gxh4 Rad8 19.Bf3 Ne5 20.Be2

First of all, it does believe 1...Nf6 and 1...Qc8(??!) are the best moves and now it does believe that 1...Qd7 is a tad better than 1..Qe7+. Please clarify.
Second, it got even more weird: Now 1...Qc8 is better than 1...Qd7!??.
I have used Contempt too, but results were never that strange for me. Now i would definitely try "Contempt=Off", to see what happens...

You do not understand the output of arena.

I copied it from arena gui that I use and it show the best move last when I use copy and paste.
evaluation is not from white point of view but from the side to move so Qf6 and Qe7 mean equal position and rest of the moves are advantage for white.

Note that in the screen the order that I see is different

I did not understand the output of arena? Allw me to retort on that one. I don't see what i am supposed to understand about the format you used. For me it is inherently ambigous. If you change the output sign format, all what is supposed to change, is the sign of the value. And if you sort the best move down, all that changes is the sorting. Nobody in this forum should be able to tell from your post, what the alleged best move is and what score it has. If you want that, i personally recommended you to consider to use the same posting format as everyone else.
The majority puts the best move on top (where it belongs) and wants scores from whites point of view.