Still possible

[carldaman]

Hi Carl.
Thanks for the answer.

I am weak, but not such a weak player after all. My best FIDE rating was a bit over 2100 (but only because I played too few FIDE rated games), my Bulgarian rating a bit over 2200 (there is basically no distinction between Bulgarian and FIDE rating, as with USCF ratings, with Bulgarian only the time control was without increment, say 1 hour per the entire game; once they did not have digital clocks att all tournaments ), but those ratings are based just on a couple of years' record, very few tournaments, playing after work hours, etc.
My last 2 complete tournaments (for Bulgarian rating) end 2005/beginning 2006 exhibited a performance rating of some 2400 points, with wins against a couple of GMs, IMs, etc. If you are able to do this once or twice, you should be able to do it on a regular basis, but the fact is I stopped playing competitive chess after that. In the meanwhile, I have trained quite seriously (but only privately), first with Rybka, and then especially with the latest generation of chess engines, meaning I could have added a couple of 'unregistered' elos in the process.

But that is completely unimportant for me, as I do not compete and do not have the wish to do so in the future. My only point was to suggest that I am not so weak after all.

But of course, you got it right on many fronts. I think a strong GM with the necessary preparation (Kasparov, the 2004 Bilbao guys, Kramnik, etc., definitely almost did not have a clue how they should play against engines, and that was one of the reasons for the bad results) could be able to play much more successfully with engines. But in order to prepare, you need a couple of years of specific engine experience, for which most GMs do not have time or will to do.

I think the stronger the engine, the more difficult it is to play against, no matter how it plays. Non-standard entities might be sometimes more dangerous to humans, but they also allow a significant amount of opportunities, so basically it should be easier to win against those. Spark comes to my mind: sometimes it plays fantastic, but at other times, looking at its pawn structure, you see the engine does not have a clue about pawns, much easier to win against it in general as against any top-rated engine.

When I speak of weakness, it is only in relative terms, by comparison to top humans or strong chess engines. I've always thought you were a strong player from a strictly human perspective, and you would not be weak at all in an absolute sense, just the contrary. Statistically, you don't have to be 2700 rated to be stronger than most players on earth. I'm of candidate master strength myself, but I don't excel at blitz and can't play well against engines. My style is more suitable against human players, as I tend to make fairly frequent exploitable mistakes when playing engines, where human players would struggle to find refutations. Even strong humans will punish you much less than a tactical monster like SF or H3.

I think you did very well exposing weak points in Stockfish's play. No way such a 3000+ strong player should allow massive simplification into a basically drawn game, if it was truly that strong at all phases of the game. Again, it's good and commendable that you are playing these games, so every one can see with their own eyes that SF is far from perfect, despite substantial recent Elo gains.

If I were a programmer, I'd just love to have one or two strong players on my team to serve as 'style tester', for lack of a better term. Call it 'quality-over-quantity' control. This is what many developers neglect. To them, it's all about running thousands of hyper-bullet game (yes, that's very useful, don't get me wrong) and little else.

I'm thinking that it's probably easier to find players with good anti-computer skills *below* 2700 FIDE. First of all, there are a lot more players below 2700, and secondly, a natural anti-machine style doesn't fare well in human play above 2700 anyway, so it is more likely to find such players in a lower bracket. I get the feeling that it would be too much of an adjustment for the strongest humans to begin playing this way. This is why, in my view, Kasparov, Kramnik, Anand, Adams, etc., never managed to play convincingly in their respective matches against computer opponents.

Another factor was the book the engines used. I believe future man vs machine matches, if we ever see any, ought to have a short 4-move book, and wide enough to provide variety, but thus forcing the engine to play most of the opening on its own. As I wrote before, it's best for the sake of progress that the developers deal with the weaknesses and not hide them.

We've reached a point where a proverbial step backwards may have to be taken, before two steps forward can be made. Deep Fritz 10 was a very promising engine, but Fritz 11 and later versions changed completely, as it became stronger tactically, while throwing away all that was great about it, its ability to navigate semi-closed positions and flank attacks. This is an example of how not to do things. Where is Fritz now? Nowhere near the top. It would've been much wiser to stick with the old approach. Something like that is needed now - to revisit that same paradigm, needlessly discarded.


Re: Spark -- yes, an exciting attacking engine, with serious positional eval issues. Chiron also attacks well, but allows weak doubled pawns all over the place.

Cheers,
CL

[carldaman]

I find it significant that in the game you drew, SF played 8. c5 ?! creating a partially locked up position, with one open file for both sides, and easy exchanges to follow. The Exchange French is also a lousy variation for White, in general. In the game you lost, there was no locking of the position, and the play was a lot more dynamic, so even an advantage for White could be lost after 1-2 bad moves.

I do not have much time to look at the specific positions, but Black's control of the e-file and White's light-square weakness look to balance White's Q-side advanced pawn majority. The question is if White can do better (win or draw) with optimal play, as I'm tempted to think so, but at what point? So many questions, so little time...

[Lyudmil Tsvetkov]

Hi Carl.
Thanks for the answer.

I am weak, but not such a weak player after all. My best FIDE rating was a bit over 2100 (but only because I played too few FIDE rated games), my Bulgarian rating a bit over 2200 (there is basically no distinction between Bulgarian and FIDE rating, as with USCF ratings, with Bulgarian only the time control was without increment, say 1 hour per the entire game; once they did not have digital clocks att all tournaments ), but those ratings are based just on a couple of years' record, very few tournaments, playing after work hours, etc.
My last 2 complete tournaments (for Bulgarian rating) end 2005/beginning 2006 exhibited a performance rating of some 2400 points, with wins against a couple of GMs, IMs, etc. If you are able to do this once or twice, you should be able to do it on a regular basis, but the fact is I stopped playing competitive chess after that. In the meanwhile, I have trained quite seriously (but only privately), first with Rybka, and then especially with the latest generation of chess engines, meaning I could have added a couple of 'unregistered' elos in the process.

But that is completely unimportant for me, as I do not compete and do not have the wish to do so in the future. My only point was to suggest that I am not so weak after all.

But of course, you got it right on many fronts. I think a strong GM with the necessary preparation (Kasparov, the 2004 Bilbao guys, Kramnik, etc., definitely almost did not have a clue how they should play against engines, and that was one of the reasons for the bad results) could be able to play much more successfully with engines. But in order to prepare, you need a couple of years of specific engine experience, for which most GMs do not have time or will to do.

I think the stronger the engine, the more difficult it is to play against, no matter how it plays. Non-standard entities might be sometimes more dangerous to humans, but they also allow a significant amount of opportunities, so basically it should be easier to win against those. Spark comes to my mind: sometimes it plays fantastic, but at other times, looking at its pawn structure, you see the engine does not have a clue about pawns, much easier to win against it in general as against any top-rated engine.

When I speak of weakness, it is only in relative terms, by comparison to top humans or strong chess engines. I've always thought you were a strong player from a strictly human perspective, and you would not be weak at all in an absolute sense, just the contrary. Statistically, you don't have to be 2700 rated to be stronger than most players on earth. I'm of candidate master strength myself, but I don't excel at blitz and can't play well against engines. My style is more suitable against human players, as I tend to make fairly frequent exploitable mistakes when playing engines, where human players would struggle to find refutations. Even strong humans will punish you much less than a tactical monster like SF or H3.

I think you did very well exposing weak points in Stockfish's play. No way such a 3000+ strong player should allow massive simplification into a basically drawn game, if it was truly that strong at all phases of the game. Again, it's good and commendable that you are playing these games, so every one can see with their own eyes that SF is far from perfect, despite substantial recent Elo gains.

If I were a programmer, I'd just love to have one or two strong players on my team to serve as 'style tester', for lack of a better term. Call it 'quality-over-quantity' control. This is what many developers neglect. To them, it's all about running thousands of hyper-bullet game (yes, that's very useful, don't get me wrong) and little else.

I'm thinking that it's probably easier to find players with good anti-computer skills *below* 2700 FIDE. First of all, there are a lot more players below 2700, and secondly, a natural anti-machine style doesn't fare well in human play above 2700 anyway, so it is more likely to find such players in a lower bracket. I get the feeling that it would be too much of an adjustment for the strongest humans to begin playing this way. This is why, in my view, Kasparov, Kramnik, Anand, Adams, etc., never managed to play convincingly in their respective matches against computer opponents.

Another factor was the book the engines used. I believe future man vs machine matches, if we ever see any, ought to have a short 4-move book, and wide enough to provide variety, but thus forcing the engine to play most of the opening on its own. As I wrote before, it's best for the sake of progress that the developers deal with the weaknesses and not hide them.

We've reached a point where a proverbial step backwards may have to be taken, before two steps forward can be made. Deep Fritz 10 was a very promising engine, but Fritz 11 and later versions changed completely, as it became stronger tactically, while throwing away all that was great about it, its ability to navigate semi-closed positions and flank attacks. This is an example of how not to do things. Where is Fritz now? Nowhere near the top. It would've been much wiser to stick with the old approach. Something like that is needed now - to revisit that same paradigm, needlessly discarded.


Re: Spark -- yes, an exciting attacking engine, with serious positional eval issues. Chiron also attacks well, but allows weak doubled pawns all over the place.

Cheers,
CL

Hi Carl.
Thanks for your conceptions.

My case is a bit different story: for some unknown reason I have always played much much better when enjoying quiet surroundings. My performance when you have a lot of quiet/concentration, and when the level of quiet/concentration drops, is very much different. I would play very well, when able to concentrate deep, and very bad, when unable to. For this reason I am not well suited for competitive tournament chess, when conditions between tournaments in terms of noise differ significantly.

I remember that when I was still a relatively weak player (2000 elo), I would play blitz games for fun against some master level players. As I usually was younger in this case, I would draw a match under normal noise conditions (for Bulgaria that might mean playing in a cafe, for example, as in the last 25 years or so a lot of the existing town chess clubs closed their doors, or in the park), with normal level of noise, just a few kibitzers, etc.; then, when the level of noise started rising, for example when you have a throng of kibitzers around eagerly suggesting your next move , I would start playing very bad, losing most of the games, with my opponent usually unaffected; but, when we transferred to a more quiet location to continue the match, a private room, even better during the night, I would suddenly start winning every single game.

It was like that once, and nothing has changed till now. Therefore, I am not a good competitive player, but that does not mean that I am objectively weak.

Style testing' seems like an interesting enterprise to me

Concerning books, I think GMs might well avoid all computer opening preparation by simply picking up an inexisting opening line very early on (let's say, by playing seemingly a very slightly inferior move that promises unchartered waters; in this case, you need no opening preparation at all). The real challenge would be to remain unruffled during the course of the entire game, which is easy for computers, that do not have affections, but difficult or fully impossible for humans who are influenced by a variety of non-chess situations (someone distracting your attention, intrusive thoughts unrelated to the event, getting nervous under certain conditions, etc.). I think engines still have less knowledge than humans, but you need to make just a single bigger mistake and it is over (like Kramnik missing a mate in 1 against Fritz).

[Lyudmil Tsvetkov]

I find it significant that in the game you drew, SF played 8. c5 ?! creating a partially locked up position, with one open file for both sides, and easy exchanges to follow. The Exchange French is also a lousy variation for White, in general. In the game you lost, there was no locking of the position, and the play was a lot more dynamic, so even an advantage for White could be lost after 1-2 bad moves.

I do not have much time to look at the specific positions, but Black's control of the e-file and White's light-square weakness look to balance White's Q-side advanced pawn majority. The question is if White can do better (win or draw) with optimal play, as I'm tempted to think so, but at what point? So many questions, so little time...

Yes, time is important, the more time you have, the better you play.
Btw., referring to your previous post, Fritz is still all around. If I were asked what do I associate computer chess primarily with, I would say: 'Fritz!', and only after that add: 'Houdini, Stockfish and Komodo'.

Concerning the lost game (the interesting one), purely untuitively I think d5 was a mistake on the part of black, while b5 for white is an interesting continuation, offering a piece for a pawn with a tremendous positional compensation (I am almost sure, provided that the sacrifice is correct, that no engine would see this, even with more time to think).

I played another game in the same line in analysis mode (meaning giving myself a lot more time, as well as a few take-backs, if necessary ), and, although white played a somewhat more inferior variation of the line, the game ended with a draw with certain white pressure, probably suggesting that in the actual game white might have an advantage already after d5.

Here the game in analysis mode:

[pgn][PlyCount "100"]
[Event "Blitz 2m+2s"]
[Site "Sofia"]
[Date "2013.09.13"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish 4 64 SSE4.2"]
[Result "1/2-1/2"]
[ECO "E11"]
[TimeControl "120+2"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "50"]
[MLFlags "010100"]

{1024MB, Dell XPS 4Cores} 1. Nf3 {[%emt 0:00:00]} 1... Nf6 {[%emt 0:00:05]} 2.
c4 {[%emt 0:00:02]} 2... e6 {[%emt 0:00:08]} 3. d4 {[%emt 0:00:02]} 3... Bb4+
{[%emt 0:00:09]} 4. Bd2 {[%emt 0:00:02]} 4... Bxd2+ {[%emt 0:00:10]} 5. Nbxd2
{[%emt 0:00:01]} 5... O-O {[%emt 0:00:10]} 6. e3 {[%emt 0:00:02]} 6... b6
{[%emt 0:00:03]} 7. Be2 {[%emt 0:00: 03]} 7... c5 {[%emt 0:00:04]} 8. O-O
{[%emt 0:00:03]} 8... Nc6 {[%emt 0:00:01]} 9. Rc1 {[%emt 0:00:04]} 9... cxd4
{[%emt 0:00:01]} 10. exd4 {[%emt 0:00:02]} 10... Bb7 {[%emt 0: 00:04]} 11. Qc2
{[%emt 0:00:23]} 11... d5 {[%emt 0:00:05]} 12. c5 {[%emt 0:00:03]} 12... bxc5
{[%emt 0:00:04]} 13. dxc5 {[%emt 0:00:07]} 13... e5 {[%emt 0:00:10]} 14. Bb5
{[%emt 0:00:10]} 14... Qc7 {[%emt 0:00:00]} 15. Nb3 {[%emt 0:00:22]} 15... a6
{[%emt 0:00: 05]} 16. Ba4 {[%emt 0:00:06]} 16... e4 {[%emt 0:00:03]} 17. Bxc6
{[%emt 0:00:00]} 17... Bxc6 {[%emt 0:00:04]} 18. Nfd4 {[%emt 0:00:06]} 18... Ng4
{[%emt 0:00:03]} 19. g3 {[%emt 0:00:02]} 19... Ne5 {[%emt 0:00:04]} 20. Nxc6
{[%emt 0:00:38]} 20... Nxc6 {[%emt 0: 00:08]} 21. Qc3 {[%emt 0:00:03]} 21... a5
{[%emt 0:00:04]} 22. Nd4 {[%emt 0:00:08]} 22... Nb4 {[%emt 0:00:07]} 23. Ra1
{[%emt 0:00:00]} 23... Nd3 {[%emt 0:00:06]} 24. c6 {[%emt 0:00:03]} 24... Rfc8
{[%emt 0:00:01]} 25. b3 {[%emt 0:00:22]} 25... Qe5 {[%emt 0:00: 08]} 26. a3
{[%emt 0:00:17]} 26... Ra6 {[%emt 0:00:05]} 27. b4 {[%emt 0:00:21]} 27... Raxc6
{[%emt 0:00:04]} 28. Qxc6 {[%emt 0:00:03]} 28... Rxc6 {[%emt 0:00:09]} 29. Nxc6
{[%emt 0:00:01]} 29... Qc7 {[%emt 0:00:19]} 30. b5 {[%emt 0:00:19]} 30... Kf8
{[%emt 0:00: 10]} 31. a4 {[%emt 0:00:00]} 31... Qb6 {[%emt 0:00:03]} 32. Rab1
{[%emt 0:00:21]} 32... Ke8 {[%emt 0:00:03]} 33. Nxa5 {[%emt 0:00:35]} 33... Nc5
{[%emt 0:00:03]} 34. Nc6 {[%emt 0:01:49]} 34... Nxa4 {[%emt 0:00:01]} 35. Rfc1
{[%emt 0:00:06]} 35... f6 {[%emt 0:00:00]} 36. Rb4 {[%emt 0:00:58]} 36... Nc5
{[%emt 0:00:01]} 37. Rxc5 {[%emt 0:00:17]} 37... Qxc5 {[%emt 0:00:00]} 38. b6
{[%emt 0:00:03]} 38... Qc1+ {[%emt 0:00:00]} 39. Kg2 {[%emt 0: 00:04]} 39... Qa3
{[%emt 0:00:00]} 40. h4 {[%emt 0:00:00]} 40... Qf3+ {[%emt 0:00:03]} 41. Kg1
{[%emt 0:00:03]} 41... e3 {[%emt 0:00:00]} 42. b7 {[%emt 0:00:15]} 42... Qxf2+
{[%emt 0:00:00]} 43. Kh1 {[%emt 0:00:08]} 43... Qe1+ {[%emt 0:00:00]} 44. Kg2
{[%emt 0:00: 03]} 44... Qe2+ {[%emt 0:00:00]} 45. Kg1 {[%emt 0:00:05]} 45...
Qd1+ {[%emt 0:00:04]} 46. Kg2 {[%emt 0:00:07]} 46... Qc2+ {[%emt 0:00:00]} 47.
Kg1 {[%emt 0:00:49]} 47... Qc1+ {[%emt 0:00:00]} 48. Kg2 {[%emt 0:00:02]} 48...
Qc2+ {[%emt 0:00:05]} 49. Kg1 {[%emt 0:00:02]} 49... Qc1+ {[%emt 0:00:08]} 50.
Kg2 {[%emt 0:00:02]} 50... Qc2+ {[%emt 0:00:02]} 1/2-1/2
[/pgn]

[Lyudmil Tsvetkov]

[d]2rq1rk1/pb2nppp/5n2/1PPp4/8/5N2/P3BPPP/1QRR2K1 w - - 0 19

Without much analysis, I think 19.Nd4 is the natural and strong move. I played 19.Qb4, very much inferior and unnecessary, I considered Nd4 but dropped it because of 19...Qa5; but on Qa5 white has 20.c6, then Qb2, Rc3 (or Rd3), followed by Ra3, etc.

If black does not play 19...Qa5, white simply continues pushing all its queen-side pawns. The d5 pawn does not count as it is blocked, so more or less white has 2 far advanced connected passers for a piece, with much better piece coordination because of the restricting quality of the connected passers, i.e., according to me (might be absolutely wrong, would be happy for someone to unmask me), white has a clear edge here after 19.Nd4

[Lyudmil Tsvetkov]

I find it significant that in the game you drew, SF played 8. c5 ?! creating a partially locked up position, with one open file for both sides, and easy exchanges to follow. The Exchange French is also a lousy variation for White, in general. In the game you lost, there was no locking of the position, and the play was a lot more dynamic, so even an advantage for White could be lost after 1-2 bad moves.

I do not have much time to look at the specific positions, but Black's control of the e-file and White's light-square weakness look to balance White's Q-side advanced pawn majority. The question is if White can do better (win or draw) with optimal play, as I'm tempted to think so, but at what point? So many questions, so little time...

Yes, time is important, the more time you have, the better you play.
Btw., referring to your previous post, Fritz is still all around. If I were asked what do I associate computer chess primarily with, I would say: 'Fritz!', and only after that add: 'Houdini, Stockfish and Komodo'.

Concerning the lost game (the interesting one), purely untuitively I think d5 was a mistake on the part of black, while b5 for white is an interesting continuation, offering a piece for a pawn with a tremendous positional compensation (I am almost sure, provided that the sacrifice is correct, that no engine would see this, even with more time to think).

I played another game in the same line in analysis mode (meaning giving myself a lot more time, as well as a few take-backs, if necessary ), and, although white played a somewhat more inferior variation of the line, the game ended with a draw with certain white pressure, probably suggesting that in the actual game white might have an advantage already after d5.

Here the game in analysis mode:

[pgn][PlyCount "100"]
[Event "Blitz 2m+2s"]
[Site "Sofia"]
[Date "2013.09.13"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish 4 64 SSE4.2"]
[Result "1/2-1/2"]
[ECO "E11"]
[TimeControl "120+2"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "50"]
[MLFlags "010100"]

{1024MB, Dell XPS 4Cores} 1. Nf3 {[%emt 0:00:00]} 1... Nf6 {[%emt 0:00:05]} 2.
c4 {[%emt 0:00:02]} 2... e6 {[%emt 0:00:08]} 3. d4 {[%emt 0:00:02]} 3... Bb4+
{[%emt 0:00:09]} 4. Bd2 {[%emt 0:00:02]} 4... Bxd2+ {[%emt 0:00:10]} 5. Nbxd2
{[%emt 0:00:01]} 5... O-O {[%emt 0:00:10]} 6. e3 {[%emt 0:00:02]} 6... b6
{[%emt 0:00:03]} 7. Be2 {[%emt 0:00: 03]} 7... c5 {[%emt 0:00:04]} 8. O-O
{[%emt 0:00:03]} 8... Nc6 {[%emt 0:00:01]} 9. Rc1 {[%emt 0:00:04]} 9... cxd4
{[%emt 0:00:01]} 10. exd4 {[%emt 0:00:02]} 10... Bb7 {[%emt 0: 00:04]} 11. Qc2
{[%emt 0:00:23]} 11... d5 {[%emt 0:00:05]} 12. c5 {[%emt 0:00:03]} 12... bxc5
{[%emt 0:00:04]} 13. dxc5 {[%emt 0:00:07]} 13... e5 {[%emt 0:00:10]} 14. Bb5
{[%emt 0:00:10]} 14... Qc7 {[%emt 0:00:00]} 15. Nb3 {[%emt 0:00:22]} 15... a6
{[%emt 0:00: 05]} 16. Ba4 {[%emt 0:00:06]} 16... e4 {[%emt 0:00:03]} 17. Bxc6
{[%emt 0:00:00]} 17... Bxc6 {[%emt 0:00:04]} 18. Nfd4 {[%emt 0:00:06]} 18... Ng4
{[%emt 0:00:03]} 19. g3 {[%emt 0:00:02]} 19... Ne5 {[%emt 0:00:04]} 20. Nxc6
{[%emt 0:00:38]} 20... Nxc6 {[%emt 0: 00:08]} 21. Qc3 {[%emt 0:00:03]} 21... a5
{[%emt 0:00:04]} 22. Nd4 {[%emt 0:00:08]} 22... Nb4 {[%emt 0:00:07]} 23. Ra1
{[%emt 0:00:00]} 23... Nd3 {[%emt 0:00:06]} 24. c6 {[%emt 0:00:03]} 24... Rfc8
{[%emt 0:00:01]} 25. b3 {[%emt 0:00:22]} 25... Qe5 {[%emt 0:00: 08]} 26. a3
{[%emt 0:00:17]} 26... Ra6 {[%emt 0:00:05]} 27. b4 {[%emt 0:00:21]} 27... Raxc6
{[%emt 0:00:04]} 28. Qxc6 {[%emt 0:00:03]} 28... Rxc6 {[%emt 0:00:09]} 29. Nxc6
{[%emt 0:00:01]} 29... Qc7 {[%emt 0:00:19]} 30. b5 {[%emt 0:00:19]} 30... Kf8
{[%emt 0:00: 10]} 31. a4 {[%emt 0:00:00]} 31... Qb6 {[%emt 0:00:03]} 32. Rab1
{[%emt 0:00:21]} 32... Ke8 {[%emt 0:00:03]} 33. Nxa5 {[%emt 0:00:35]} 33... Nc5
{[%emt 0:00:03]} 34. Nc6 {[%emt 0:01:49]} 34... Nxa4 {[%emt 0:00:01]} 35. Rfc1
{[%emt 0:00:06]} 35... f6 {[%emt 0:00:00]} 36. Rb4 {[%emt 0:00:58]} 36... Nc5
{[%emt 0:00:01]} 37. Rxc5 {[%emt 0:00:17]} 37... Qxc5 {[%emt 0:00:00]} 38. b6
{[%emt 0:00:03]} 38... Qc1+ {[%emt 0:00:00]} 39. Kg2 {[%emt 0: 00:04]} 39... Qa3
{[%emt 0:00:00]} 40. h4 {[%emt 0:00:00]} 40... Qf3+ {[%emt 0:00:03]} 41. Kg1
{[%emt 0:00:03]} 41... e3 {[%emt 0:00:00]} 42. b7 {[%emt 0:00:15]} 42... Qxf2+
{[%emt 0:00:00]} 43. Kh1 {[%emt 0:00:08]} 43... Qe1+ {[%emt 0:00:00]} 44. Kg2
{[%emt 0:00: 03]} 44... Qe2+ {[%emt 0:00:00]} 45. Kg1 {[%emt 0:00:05]} 45...
Qd1+ {[%emt 0:00:04]} 46. Kg2 {[%emt 0:00:07]} 46... Qc2+ {[%emt 0:00:00]} 47.
Kg1 {[%emt 0:00:49]} 47... Qc1+ {[%emt 0:00:00]} 48. Kg2 {[%emt 0:00:02]} 48...
Qc2+ {[%emt 0:00:05]} 49. Kg1 {[%emt 0:00:02]} 49... Qc1+ {[%emt 0:00:08]} 50.
Kg2 {[%emt 0:00:02]} 50... Qc2+ {[%emt 0:00:02]} 1/2-1/2
[/pgn]

A similar position comes to mind.

[d]3q2k1/5b1p/6p1/3p1p2/3NpP2/1P2P3/4N1PP/1R4K1 w - - 0 1

For example, how many engines would recognize white has a probably decisive advantage here?

At first glance, black might be a pawn and a half better, but that is not so.

White might be much better because:

- the d5 pawn does not count, firmly blocked by the white d4 knight, that can not be expelled/attacked by an enemy piece
- the b white passer is very dangerous because, for once, the black king is very far from it (but that is quite a different story from being an edge/extreme/outside passer); and secondly, because having a passer with pieces of lower power against the queen is a big plus, deserving a special bonus in many cases, as the less powerful pieces will be able to efficiently support the advance of the passer, alternatively controlling squares all along the way to promotion, while the queen being unable to thwart this plan/block the advance of the passer.

Really interesting how many engines see white as almost winning here.

PS. But of course, this has nothing to do with the real game, Stockfish with more time is obviously a monster.

[Lyudmil Tsvetkov]

There are all types of crazy people: some run thousands and millions of games between engines, believing they measure something, while others play endless sparring games themselves against engines. Which one are crazier I am not very certain, but I obviously belong to the second category.

Believe it or not, on the 3rd day of actively sparring against Stockfish, I finally managed to win a game! I allowed myself more time as I wanted to score 'first blood', but that is not really important to me right now: I know that I can do clearly better even with less time when I am in a better form.

Below the game:

[pgn][PlyCount "79"]
[Event "Blitz 2m+2s"]
[Site "Sofia"]
[Date "2013.09.14"]
[White "Tsvetkov, Lyudmil"]
[Black "Stockfish 4 64 SSE4.2"]
[Result "1-0"]
[ECO "D05"]
[TimeControl "120+2"]
[Annotator "Tsvetkov,Lyudmil"]
[MLNrOfMoves "39"]
[MLFlags "100100"]

{1024MB, Dell XPS 4Cores} 1. e3 {[%emt 0:00:00]} 1... e6 {[%emt 0:00:04]} 2. d4
{[%emt 0:00:02]} 2... c5 {[%emt 0:00:08]} 3. Nf3 {[%emt 0:00:02]} 3... Nf6
{[%emt 0:00:06]} 4. Bd3 {[%emt 0:00:02]} 4... d5 {[%emt 0:00:04]} 5. c3
{[%emt 0:00:02]} 5... Nc6 {[%emt 0:00:05]} 6. O-O {[%emt 0:00:03]} 6... Be7
{[%emt 0:00:05]} 7. Nbd2 {[%emt 0: 00:03]} 7... O-O {[%emt 0:00:08]} 8. Ne5
{[%emt 0:00:02]} 8... Qc7 {[%emt 0:00:06]} 9. f4 {[%emt 0:00:02]} 9... Bd7
{[%emt 0:00:06]} 10. Rf3 {[%emt 0:00:05]} 10... a6 {[%emt 0:00: 08]} 11. Rh3
{[%emt 0:00:04]} 11... g6 {[%emt 0:00:08]} 12. g4 {[%emt 0:00:08]} 12... Rac8
{[%emt 0:00:08]} 13. g5 {[%emt 0:00:02]} 13... Nxe5 {[%emt 0:00:06]} 14. fxe5
{[%emt 0:00:05]} 14... Ne8 {[%emt 0:00:05]} 15. Nf3 {[%emt 0:00:28]} 15... Ng7
{[%emt 0:00:00]} 16. Qe1 {[%emt 0:00:12]} 16... Rfd8 {[%emt 0:00:00]} 17. Rh6
{[%emt 0:00:06]} 17... Nh5 {[%emt 0:00:06]} 18. Qh4 {[%emt 0:00:18]} 18... Bb5
{[%emt 0:00:08]} 19. Bc2 {[%emt 0: 00:22]} 19... Be8 {[%emt 0:00:03]} 20. Rxh7
{[%emt 0:01:19]} 20... Kxh7 {[%emt 0:00:01]} 21. Qxh5+ {[%emt 0:00:01]} 21...
Kg8 {[%emt 0:00:04]} 22. Qh4 {[%emt 0:00:12]} 22... Qb6 {[%emt 0:00:04]} 23. h3
{[%emt 0:00:28]} 23... Qb5 {[%emt 0:00:11]} 24. Kf2 {[%emt 0: 00:08]} 24... Qa5
{[%emt 0:00:00]} 25. Bd2 {[%emt 0:01:14]} 25... Bc6 {[%emt 0:00:00]} 26. Nh2
{[%emt 0:00:07]} 26... Rd7 {[%emt 0:00:00]} 27. Ng4 {[%emt 0:00:04]} 27... Qd8
{[%emt 0:00:01]} 28. Nf6+ {[%emt 0:00:00]} 28... Bxf6 {[%emt 0:00:03]} 29. gxf6
{[%emt 0:00: 02]} 29... Rcc7 {[%emt 0:00:16]} 30. Rg1 {[%emt 0:00:00]} 30... Qe8
{[%emt 0:00:09]} 31. Rg5 {[%emt 0:00:25]} 31... Qf8 {[%emt 0:00:04]} 32. Rh5
{[%emt 0:00:07]} 32... Qg7 {[%emt 0:00:00]} 33. fxg7 {[%emt 0:00:03]} 33... gxh5
{[%emt 0:00:04]} 34. Qf6 {[%emt 0:00: 09]} 34... Rc8 {[%emt 0:00:00]} 35. Qh6
{[%emt 0:00:02]} 35... f5 {[%emt 0:00:02]} 36. exf6 {[%emt 0:00:01]} 36... Kf7
{[%emt 0:00:00]} 37. Qg6+ {[%emt 0:00:07]} 37... Kg8 {[%emt 0: 00:00]} 38. Qh7+
{[%emt 0:00:01]} 38... Kf7 {[%emt 0:00:00]} 39. g8Q+ {[%emt 0:00: 01]} 39...
Kxf6 {[%emt 0:00:00]} 40. Qgg6# {[%emt 0:00:05]} 1-0
[/pgn]

Some diagrams:

[d]2r1nrk1/1pqbbp1p/p3p1p1/2ppP1P1/3P4/2PBP2R/PP1N3P/R1BQ2K1 w - - 0 15

I do not know how someone might believe that this f7 black pawn (backward-fated part of the king shelter) is not a tremendous liability. I am certain it loses the game. It is true I missed many wins in similar lines, but am fully convinced it was mainly because I played bad, and not because the position setup would not favour white strongly (and being strongly in disfavour of the side with the backward-fated pawn). It is also evident that Stockfish 4, in distinction to previous versions, and also even to Houdini 3, defends in an incomparably more brilliant way, really amazing, but I stick to my opinion this setup favours white by a wide margin.

[d]2rrb1k1/1pq1bp1R/p3p1p1/2ppP1Pn/3P3Q/2P1PN2/PPB4P/R1B3K1 b - - 0 20

Tricking the trickster, it is not very often that one is able to sacrifice against Stockfish to good avail.

[d]2rq2k1/1p1rbp2/p1b1p1p1/2ppP1P1/3P2NQ/2P1P2P/PPBB1K2/R7 w - - 0 28

Well, it is obvious that f7 loses the game: the shelter is completely inflexible.

[d]5qk1/1prr1p2/p1b1pPp1/2ppP2R/3P3Q/2P1P2P/PPBB1K2/8 b - - 0 32

This is already very easy to see.

[d]2r5/1p1r3Q/p1b1pkQ1/2pp3p/3P4/2P1P2P/PPBB1K2/8 b - - 0 40

I like it

PS. Please, do not think I am assuming something or being presumptuous; I must have a topic for a message, and this time it is winning against Stockfish, a rare affair. Later I will post some fantastic Stockfish games.

And, of course, once again, enormous thanks to the big Stockfish team, the engine has brought joy to so many hearts, and it is really a very cute engine. Actually I think, the stronger the engine, the cuter it is

[carldaman]

Very well done, Lyudmil. Almost incredible that in 2013 a top engine can still be defeated by the good ole Stonewall Attack (an underrated opening!). Back in the old days, 15-20 years ago, this was a routine strategy that pretty much guaranteed a win for the human player, but now?!

[shrapnel]

I haven't heard much discussion about SF's contempt. Houdini, on the other hand, is a different engine with its contempt changed.

As for testers of engine vs. engine, this issue is a bit of a dilemma, however. Do we change Houdini's default when playing equal engines?! It seems that H3 with C=0, or C=2 is a different entity. H3 C=0 does well vs latest SF, but H3 C=1 (default) now struggles vs SF. Which is the real Houdini now ?!

HA ! THIS is what I've been saying for quite some time, but some Stockfish fans seem to avoid this point for reasons unknown !

[mcostalba]

Believe it or not, on the 3rd day of actively sparring against Stockfish, I finally managed to win a game!

Congratulations Lyudmil !!

A real win of a real game ! Really impressive. I'd suggest to repost this game on the main forum, I think many people would be interested.