Does your engine blunder here?

[peter]

Good morning!

But my impression is, 18. ..Rb8 isn't a good move. You can see score drops in favor of W within few plies. 18. ..b4 could have been better. E.g. 19.axb4 Qxb4 20.e5 Bb7 21.Qf4 d5

Again I could let McBrain do the job for me overnight, after 20.Kb1 instead of e5 in your new line I started all over again, this time even against the newer asmFish, all the other conditions the same, 12 threads and 8G hash each engine:

[pgn]
[Event "SCHWARZERBLITZ, Lang 180m+30s"]
[Site "SchwarzerBlitz"]
[Date "2017.07.05"]
[Round "1"]
[White "SF-McBrain 2.6 64 POPCNT"]
[Black "asmFishW_2017-07-04_popcnt"]
[Result "1-0"]
[Annotator "1.08;0.93"]
[ECO "B99"]
[TimeControl "10800+30"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Bg5 e6 7.f4 Be7 8.Qf3 Qc7 9.O-O-O Nbd7 10.g4 b5 11.Bxf6 Nxf6 12.g5 Nd7 13.f5 Nc5 14.f6 gxf6 15.gxf6 Bf8 16.Rg1 h5 17.a3 Qb6 18.Rg7 b4 19.axb4 Qxb4 20.Kb1 Rb8 { [%eval 93, 38] [%emt 0:08:39] } 21.b3 { [%eval 108,37] [%emt 0:10:54] } 21...Bd7 { [%eval 119,41] [%emt 0:08:07] } 22.Bh3 { [%eval 136,38] [%emt 0:01:24] } 22...Rb7 { [%eval 129,40] [%emt 0:17:17] (Lxg7) } 23.Bf5 { [%eval 157,38] [%emt 0:09:03] } 23...Bxg7 { [%eval 191,40] [%emt 0:17:21] } 24.fxg7 { [%eval 192,43] [%emt 0:00:00] } 24...Rg8 { [%eval 181,40] [%emt 0:01:10] } 25.e5 { [%eval 218,41] [%emt 0:03:44] } 25...dxe5 { [%eval 193,42] [%emt 0:04:03] } 26.Nxe6 { [%eval 220,42] [%emt 0:00:56] } 26...fxe6 { [%eval 181,40] [%emt 0:00:20] } 27.Qxh5+ { [%eval 233,44] [%emt 0:05:06] } 27...Kd8 { [%eval 289,47] [%emt 0:13:28] } 28.Qh8 { [%eval 263,50] [%emt 0:00:00] } 28...Qxc3 { [%eval 302,40] [%emt 0:01:04] } 29.Qxg8+ { [%eval 263,51] [%emt 0:05:40] } 29...Kc7 { [%eval 337,45] [%emt 0:06:35] } 30.Rxd7+ { [%eval 270,52] [%emt 0:00:00] } 30...Kxd7 { [%eval 339,40] [%emt 0:01:06] } 31.Bxe6+ { [%eval 270,55] [%emt 0:08:46] } 31...Nxe6 { [%eval 365,46] [%emt 0:00:57] } 32.Qf7+ { [%eval 270,55] [%emt 0:02:08] } 32...Kd6 { [%eval 365,47] [%emt 0:00:00] } 33.Qxb7 { [%eval 277,50] [%emt 0:04:04] } 33...Nxg7 { [%eval 380,45] [%emt 0:03:18] } 34.Qxg7 { [%eval 285,50] [%emt 0:00:00] } 34...Qe1+ { [%eval 406,49] [%emt 0:26:55] (Df3) } 35.Kb2 { [%eval 303,52] [%emt 0:03:42] } 35...Qh4 { [%eval 414,44] [%emt 0:02:59] } 36.Qg3 { [%eval 303,50] [%emt 0:00:29] } 36...Qd4+ { [%eval 421,39] [%emt 0:00:26] } 37.Ka2 { [%eval 318,53] [%emt 0:17:10] } 37...Ke6 { [%eval 447,45] [%emt 0:05:01] (Df4) } 38.h4 { [%eval 398,43] [%emt 0:09:22] } 38...e4 { [%eval 483,47] [%emt 0:02:07] (De4) } 39.h5 { [%eval 435,51] [%emt 0:03:16] } 39...e3 { [%eval 985,48] [%emt 0:08:38] (Df6) } 40.h6 { [%eval 1593,50] [%emt 0:20:11] } 40...Qe4 { [%eval 1145,54] [%emt 0:02:03] (e2) } 41.Qc7 { [%eval 1934,47] [%emt 0:03: 10] } 41...e2 { [%eval 1233,49] [%emt 0:04:18] } 42.h7 { [%eval 2009,50] [%emt 0:00:00] } 1-0
[/pgn]
Quote to see the evals too. BTW does anybody know, why they (the evals and times) are to be seen in [pgn]-window here sometimes and then again only when quoting the posting
[/pgn]

[peter]

And to make it a little more complete showing what I think about the whole hog, this .pgn I made during the last day and today already having known before, that it's a bad position for Black after White's 16th move, you remember my first corr.-database statistics about it?

And there are some suboptimal moves in Cerebellum and Noomen.ctg (the one that I downloaded a while ago from Rebel- site. Is there a newer version already? Jeroen? BTW thanks again for your Mar del Plata, I'm still by far not through with it ).

As for 16...b4 (?! to ?) Black is as well as lost, you're right, but that's known to Cerebellum already, the main line of yours you gave to ...b4 is up to move 29 in there:

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Be7 8. Qf3
Qc7 9. O-O-O Nbd7 10. g4 b5 11. Bxf6 Nxf6 12. g5 Nd7 13. f5 Nc5 14. f6 gxf6 15.
gxf6 Bf8 16. Rg1 b4 17. Nd5 exd5 18. exd5 Nd7 19. Nc6 Bb7 20. Bh3 Ne5 21. Rge1
Rg8 22. Qf4 Bxc6 23. dxc6 Rg6 24. Kb1 Bh6 25. Qxb4 Rxf6 26. Bd7+ Kd8 27. Bh3
Ke8 28. Bd7+ Kd8 29. Rxe5 dxe5 30. Be6+ Ke8 31. Bd7+ Kd8 32. a4 Rb8 33. Qc5 Rd6
and after 34. Qxd6 Black's rather lost too.


I didn't have much more care about this rather lost lines, you did that already and I guess your tree is rather complete, doesn't matter that much if we agree about thath anyhow.

But especially after 16...h5 17a3 Qb6 (that's by far not the only "playable" move here for Black of course, but I did stick to this one at first glance, it's Cerebellum's choice there too (the one from last week).

The cont.18.Kb1 is fully playable, but 18...b4 after that is a ?! at least, I'd say, cause after the Noomen+Cereb.- moves 19.axb4 Qxb4 20Rg7! is quite a bit better than 20.e5 (?) there in the books I talked about, and I think, Black's already lost there again.

In this .pgn I collected some more lines. For those who don't know this kind of commentaries in my .pgns I did post here now and then:
The evals at line- endings are so called end- evals, they are pasted together with the move they stand for in the also stored depth, they aren't simoply the endings of longer output- lines, the lines are all made interactively by Forward- Backward- Forward with full hash (engines normally don't delete their hash- entries by moves you let or make them play, but I do know by now, Rodolfo, what you mean by persistent hash, that's a story of it's own for some more postings about )
Engine is normally SF dev., in this case it's McBrain 2.6.

[pgn]
[Event "B99y analysis"]
[Site "UTENTE-PC"]
[Date "2017.06.16"]
[Round "?"]
[White "Leoni"]
[Black "Critter_1.6a_64bit"]
[Result "0-1"]
[Annotator "Martan,Peter"]
[ECO "B99"]
[EventDate "2017.??.??"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Bg5 e6 7.f4 Be7 8.Qf3 Qc7 9.O-O-O Nbd7 10.g4 b5 11.Bxf6 Nxf6 12.g5 Nd7 13.f5 Nc5 14.f6 gxf6 15.gxf6 Bf8 16.Rg1 h5 ( 16...b4 $5 17.Nd5 { [%eval 45,30] } 17...exd5 { [%eval 1,0] } 18.exd5 Nd7 ( 18...Bb7 19.Bh3 { [%eval 308,30] } ) 19.Nc6 Bb7 20.Bh3 ( 20.Qe4+ $6 Ne5 21.Kb1 Bxc6 22.dxc6 Qxc6 23.Rd5 O-O-O 24.Rxe5 Qxe4 25.Rxe4 d5 { [%eval 54,30] } ) 20...Ne5 21.Rge1 Rg8 22.Qf4 Bxc6 23.dxc6 Rg6 24.Kb1 Bh6 25.Qxb4 Rxf6 26.Bd7+ Kd8 27.Bh3 Ke8 28.Bd7+ Kd8 29.Rxe5 ( 29.a4 { RL } 29...Rb8 { [%eval 136,38] } ) 29...dxe5 30.Be6+ Ke8 31.Bd7+ Kd8 32.a4 Rb8 33.Qc5 Rd6 { [%eval 1,0] } 34.Qxd6 Qxd6 { [%eval 181,33] } ) 17.a3 ( 17.Kb1 { [%eval 31,28] } 17...b4 { [%eval 1,0] } 18.Nd5 exd5 19.exd5 Bg4 20.Re1+ Kd8 21.Qf2 Kc8 22.h3 Bd7 23.Rg7 Bxg7 24.fxg7 Rg8 25.Qxf7 Qd8 26.Bg2 { [%eval 1,0] } 26...Kb7 { [%eval 0,35] } ) ( 17.Rg7 $6 { RL? } 17...b4 18.Nd5 exd5 19.exd5 Nd7 20.Nc6 Bb7 21.Bh3 Bxc6 22.dxc6 Ne5 23.Bd7+ Kd8 24.Qe4 Qb6 25.Be6 Kc7 26.Bxf7 { O.G.? Engine- variation Jon? } 26...Nxc6 { [%eval 0,35] } ) 17...Qb6 $5 { [%eval 0,26] } 18.Rg7 ( 18.Kb1 $5 { [%eval 1,0] } 18...Nd7 { [%eval 88,30] } ( 18...b4 $2 19.axb4 Qxb4 20.Rg7 { [%eval 158,33] } ( 20.e5 $2 Bb7 { [%eval 1,0] } 21.Qf2 { [%eval 31,31] } ) ) ) 18...Rb8 ( 18...b4 19.axb4 Qxb4 20.Kb1 ( 20.e5 $6 Bb7 21.Qf4 d5 { RL } ) 20...Rb8 ( 20...Nd7 21.Rg5 Qb7 ( 21...Nc5 22.Bg2 { [%eval 168,30] } ) 22.Bc4 { [%eval 162,31] } ) 21.b3 Bd7 ( 21...Bb7 22.Nd5 { [%eval 166,30] } ) 22.Bh3 Rb7 ( 22...Qa3 23.e5 { [%eval 172, 30] } ) ( 22...Ba4 23.Nd5 { [%eval 185,30] } ) ( 22...Bxg7 23.fxg7 { [%eval 150,32] } ) 23.Bf5 Bxg7 24.fxg7 Rg8 25.e5 dxe5 26.Nxe6 fxe6 27.Qxh5+ Kd8 28.Qh8 Qxc3 29.Qxg8+ Kc7 30.Rxd7+ Kxd7 31.Bxe6+ Nxe6 32.Qf7+ Kd6 33.Qxb7 Nxg7 34.Qxg7 Qe1+ 35.Kb2 Qh4 36.Qg3 Qd4+ 37.Ka2 Ke6 38.h4 e4 39.h5 e3 40.h6 Qe4 41.Qc7 e2 42.h7 { 1-0 (42) SF-McBrain 2.6-asmFishW_2017-07-04 } ) 19.Bh3 ( 19.Nd5 $2 { N } 19...exd5 20.exd5 { N } 20...Nd7 { [%eval -46,31] } ) 19...b4 20.Nd5 exd5 21.exd5 Nd7 22.Bxd7+ Bxd7 23.a4 Kd8 24.Rxf7 Rb7 25.Qe4 Kc8 26.b3 Bh6+ 27.Kb1 Bg4 28.a5 Qc5 29.Ne6 Qxa5 30.Rxb7 Kxb7 31.Rd4 Rc8 32.Qh7+ Kb8 33.f7 Qb5 34.Qxh6 Qf1+ 35.Qc1 Qxf7 36.Rxb4+ Ka8 37.Rb6 { 1-0 (37) SF-McBrain 2. 6-asmFishW_2017-06-28 } 0-1
[/pgn]

[peter]

But especially after 16...h5 17a3 Qb6 (that's by far not the only "playable" move here for Black of course,

Not to leave that fully discontinued, let's have a look at one of some other book moves here, e.g. Hiarcs has rated better than 17...Qb6.

17...Tb8 is top of list in Hiarcs online GM++, yet only one of two (together with 17...Qb6) with ! and alomost equal number of games and play- probability in HCE- GUI.

Not to come along with the next .pgn, I just went a little forward and backward with McBrain in lines from Hiarcs and Noomen.ctg, in the latter 20.e5 is top of list as for probability but has a !?, I'd give it a ?! at least, guess it leads to quick exchange of Queens and equalization, but
20.Re1! (funny enough the end of Cerebellum here already ) gives this output with full hash after some more Forward- Backward:


[d]1rb1kb1r/2q2p2/p2ppP2/1pn4p/3NP3/P1N2Q2/1PP4P/2K1RBR1 b k - 0 1

Analysis by SF-McBrain 2.6 64 POPCNT:

18...Na4 19.Nd5 Qa7 20.Qd1 Qc5 21.Rg7 e5 22.Nb3 Qa7 23.Na5 Bd7 24.Be2 Rc8 25.Kb1 Be6 26.Reg1 Qf2 27.R7g2 Qd4 28.Qxd4 exd4 29.Nb3 Bxd5 30.exd5 Rh6 31.Nxd4 Rxf6 32.Nc6 Kd7 33.Nb4 Nb6 34.Nxa6 Nxd5 35.Bxb5+ Ke7 36.Rd2 Nc7 37.Nxc7 Rxc7 38.a4 Rc5 39.Rdg2 Bh6 40.Rd1 Re5 41.c4 Rg6 42.Rxg6 fxg6 43.a5 Bg5
+- (1.62) Depth: 38/67 00:08:25 10190MN

Position reached by 16...h5 17.a3 Rb8 (!?, same comment as well as for 17...Qb6!? too)
18.Re1

[Rodolfo Leoni]

Good morning!

But my impression is, 18. ..Rb8 isn't a good move. You can see score drops in favor of W within few plies. 18. ..b4 could have been better. E.g. 19.axb4 Qxb4 20.e5 Bb7 21.Qf4 d5

Again I could let McBrain do the job for me overnight, after 20.Kb1 instead of e5 in your new line I started all over again, this time even against the newer asmFish, all the other conditions the same, 12 threads and 8G hash each engine:

[pgn]
[Event "SCHWARZERBLITZ, Lang 180m+30s"]
[Site "SchwarzerBlitz"]
[Date "2017.07.05"]
[Round "1"]
[White "SF-McBrain 2.6 64 POPCNT"]
[Black "asmFishW_2017-07-04_popcnt"]
[Result "1-0"]
[Annotator "1.08;0.93"]
[ECO "B99"]
[TimeControl "10800+30"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Bg5 e6 7.f4 Be7 8.Qf3 Qc7 9.O-O-O Nbd7 10.g4 b5 11.Bxf6 Nxf6 12.g5 Nd7 13.f5 Nc5 14.f6 gxf6 15.gxf6 Bf8 16.Rg1 h5 17.a3 Qb6 18.Rg7 b4 19.axb4 Qxb4 20.Kb1 Rb8 { [%eval 93, 38] [%emt 0:08:39] } 21.b3 { [%eval 108,37] [%emt 0:10:54] } 21...Bd7 { [%eval 119,41] [%emt 0:08:07] } 22.Bh3 { [%eval 136,38] [%emt 0:01:24] } 22...Rb7 { [%eval 129,40] [%emt 0:17:17] (Lxg7) } 23.Bf5 { [%eval 157,38] [%emt 0:09:03] } 23...Bxg7 { [%eval 191,40] [%emt 0:17:21] } 24.fxg7 { [%eval 192,43] [%emt 0:00:00] } 24...Rg8 { [%eval 181,40] [%emt 0:01:10] } 25.e5 { [%eval 218,41] [%emt 0:03:44] } 25...dxe5 { [%eval 193,42] [%emt 0:04:03] } 26.Nxe6 { [%eval 220,42] [%emt 0:00:56] } 26...fxe6 { [%eval 181,40] [%emt 0:00:20] } 27.Qxh5+ { [%eval 233,44] [%emt 0:05:06] } 27...Kd8 { [%eval 289,47] [%emt 0:13:28] } 28.Qh8 { [%eval 263,50] [%emt 0:00:00] } 28...Qxc3 { [%eval 302,40] [%emt 0:01:04] } 29.Qxg8+ { [%eval 263,51] [%emt 0:05:40] } 29...Kc7 { [%eval 337,45] [%emt 0:06:35] } 30.Rxd7+ { [%eval 270,52] [%emt 0:00:00] } 30...Kxd7 { [%eval 339,40] [%emt 0:01:06] } 31.Bxe6+ { [%eval 270,55] [%emt 0:08:46] } 31...Nxe6 { [%eval 365,46] [%emt 0:00:57] } 32.Qf7+ { [%eval 270,55] [%emt 0:02:08] } 32...Kd6 { [%eval 365,47] [%emt 0:00:00] } 33.Qxb7 { [%eval 277,50] [%emt 0:04:04] } 33...Nxg7 { [%eval 380,45] [%emt 0:03:18] } 34.Qxg7 { [%eval 285,50] [%emt 0:00:00] } 34...Qe1+ { [%eval 406,49] [%emt 0:26:55] (Df3) } 35.Kb2 { [%eval 303,52] [%emt 0:03:42] } 35...Qh4 { [%eval 414,44] [%emt 0:02:59] } 36.Qg3 { [%eval 303,50] [%emt 0:00:29] } 36...Qd4+ { [%eval 421,39] [%emt 0:00:26] } 37.Ka2 { [%eval 318,53] [%emt 0:17:10] } 37...Ke6 { [%eval 447,45] [%emt 0:05:01] (Df4) } 38.h4 { [%eval 398,43] [%emt 0:09:22] } 38...e4 { [%eval 483,47] [%emt 0:02:07] (De4) } 39.h5 { [%eval 435,51] [%emt 0:03:16] } 39...e3 { [%eval 985,48] [%emt 0:08:38] (Df6) } 40.h6 { [%eval 1593,50] [%emt 0:20:11] } 40...Qe4 { [%eval 1145,54] [%emt 0:02:03] (e2) } 41.Qc7 { [%eval 1934,47] [%emt 0:03: 10] } 41...e2 { [%eval 1233,49] [%emt 0:04:18] } 42.h7 { [%eval 2009,50] [%emt 0:00:00] } 1-0
[/pgn]
Quote to see the evals too. BTW does anybody know, why they (the evals and times) are to be seen in [pgn]-window here sometimes and then again only when quoting the posting
[/pgn]

An interesting game. I'd need some time to make an analysis, I think I'll do this evening.

[peter]

Hi Rodolfo!

An interesting game. I'd need some time to make an analysis, I think I'll do this evening.

Good to see your on it again too.

If you haven't done so already, look at the follow- up postings about this Sicilian Whole Hog Variation I made in the meantime too, the single one engine- game here was only for fun.

Especially the .pgn with the other lines after 17...Qb6(!?) here,

http://www.talkchess.com/forum/viewtopi ... 57&t=64510

,I'd really like to have your opinion about.
Have fun!

[Rodolfo Leoni]

Hi Rodolfo!

An interesting game. I'd need some time to make an analysis, I think I'll do this evening.

Good to see your on it again too.

If you haven't done so already, look at the follow- up postings about this Sicilian Whole Hog Variation I made in the meantime too, the single one engine- game here was only for fun.

Especially the .pgn with the other lines after 17...Qb6(!?) here,

http://www.talkchess.com/forum/viewtopi ... 57&t=64510

,I'd really like to have your opinion about.
Have fun!

I only have the time to be impressed by your McBrain-AsmFish game. Scores say W is very strong in that game. At first glance, I don't understand 22. ..Rb7. W reply is as strong as spectacular. I'd start an analysis of 22. ..Qa3... this evening! See later!

Rodolfo

[Nordlandia]

Najdorf Variation. Goteborg Argentine

13...Kg7 side variation is losing ->

SF at end: +1.4

[pgn][Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "51"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Be7 8. Qf3
h6 9. Bh4 g5 10. fxg5 Nfd7 11. Nxe6 fxe6 12. Qh5+ Kf8 13. Bb5 Kg7 14. O-O Ne5
15. Bg3 Ng6 16. gxh6+ Rxh6 17. Rf7+ Kxf7 18. Qxh6 axb5 19. Rf1+ Ke8 20. Qxg6+
Kd7 21. Rf7 b4 22. Nd5 exd5 23. exd5 Ra5 24. Qg4+ Kc7 25. Qxb4 b6 26. Qe4 *

[/pgn]

[Rodolfo Leoni]

Hi Rodolfo!

An interesting game. I'd need some time to make an analysis, I think I'll do this evening.

Good to see your on it again too.

If you haven't done so already, look at the follow- up postings about this Sicilian Whole Hog Variation I made in the meantime too, the single one engine- game here was only for fun.

Especially the .pgn with the other lines after 17...Qb6(!?) here,

http://www.talkchess.com/forum/viewtopi ... 57&t=64510

,I'd really like to have your opinion about.
Have fun!

I only have the time to be impressed by your McBrain-AsmFish game. Scores say W is very strong in that game. At first glance, I don't understand 22. ..Rb7. W reply is as strong as spectacular. I'd start an analysis of 22. ..Qa3... this evening! See later!

Rodolfo

Hi Peter,

it seems you're right about the whole variation. I've not been able to find any satisfactory line for black. 22. ..Qa3 is even worse of 22. ..Rb7. Over 120 CPs advantage for W. The whole variation becomes painful for black. Just think, I managed about 1 hour with Critter PHs, and after a lot of score propagation it thinks 17. ..Na4 is the best move for black... Obviously, white has no problems after 18.Nxa4 bxa4 19.c3.

My congrats.

Rodolfo

P.S. In the 16. ..b4 main line, after 31.Rxd6 Bg5, Cerebellum suggests 32.Bh3+ which is a bit incorrect. Black could continue this way: 32. ..Kc7 33.Rd7+ Kxc6 34.Rxf7 Rb4 35.Rxh7 a5. W has 2 pawns advantage, but B can still try to block pawns. Better is 32.c3 and maybe it could continue with 33. ..Kc7 34.Rd5 f6 35.b4 and black can't do anything against white pawns horde.

[Rodolfo Leoni]

Najdorf Variation. Goteborg Argentine

13...Kg7 side variation is losing ->

SF at end: +1.4

[pgn][Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "51"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Be7 8. Qf3
h6 9. Bh4 g5 10. fxg5 Nfd7 11. Nxe6 fxe6 12. Qh5+ Kf8 13. Bb5 Kg7 14. O-O Ne5
15. Bg3 Ng6 16. gxh6+ Rxh6 17. Rf7+ Kxf7 18. Qxh6 axb5 19. Rf1+ Ke8 20. Qxg6+
Kd7 21. Rf7 b4 22. Nd5 exd5 23. exd5 Ra5 24. Qg4+ Kc7 25. Qxb4 b6 26. Qe4 *

[/pgn]

A Scid ECO B98j, it seems. Looks interesting at first glance. I'll take some time to make analyses. Peter's help would be nice.

[peter]

Hi Rodolfo!

it thinks 17. ..Na4 is the best move for black... Obviously, white has no problems after 18.Nxa4 bxa4 19.c3.

17...Na4 is playable, but after 18.Nxa4 bxa4 20.Kb1! Black is as lost as with all the other 17th moves.

After some Forward- Backward and after 20.Kb1 again:

[d]r1b1kb1r/2q2p2/p2ppP2/7p/p2NP3/P4Q2/1PP4P/1K1R1BR1 b kq - 0 1

Analysis by SF-McBrain 2.6 64 POPCNT:

19...Bd7 20.Rg5 Qb6 21.c3 h4 22.Bc4 0-0-0 23.Qf1 Kb8 24.Bxa6 Bh6 25.Rg4 Rdg8 26.Rxg8+ Rxg8 27.Bb5 e5 28.Bxd7 exd4 29.Rxd4 Be3 30.Rd1 Qb3 31.Qb5+ Qxb5 32.Bxb5 Bf4 33.h3 Kc7 34.Bxa4 Rg6 35.Bb3 Rxf6 36.Rf1 Be5 37.Rxf6 Bxf6 38.Bxf7 Kb6 39.Kc2 Ka5 40.b4+ Ka4 41.Kb2 Kb5 42.Kb3 Ka6 43.Kc2
+- (1.79) Depth: 36/65 00:16:45 20158MN

As for my personal pov 17...Nd7 is the least bad, even if 18.Nxe6 fxe6 19.f7+ doesn't give much hope for Black neither.

Thanks for bringing the position, Rodolfo!

It can be well used for testing engines and their evals, yet I'd like to see in output- lines lose Black after ...b4 and lose after ...h5 too.

16...h5 is clearly better, yet Black will lose with perfect play from White, Black can defend longer and with better chances as for White to blunder and give away the full point.

I'd likd to see any of the winning moves at 17, 18, 19, 20 as for both of the two 16th black moves and evals of >1.80 for ...b4, >1.50 for ...h5