|
< Earlier Kibitzing · PAGE 137 OF 140 ·
Later Kibitzing> |
Nov-10-18
 | | mistermac: So much work! |
|
| Nov-10-18 | | The Kingfish: A remarkable game that was hard fought on both sides. There are 137 pages of commentary to read here. Can someone explain to me where the mistakes were for both sides? |
|
| Nov-11-18 | | kwid: <The Kingfish:>
Well; my assessment as to the accuracy played by white " nearly flaw less".
Black did make one mistake at move 19 which should have been Rf6 instead of h6 leading to a forced loss.
The correctness of the suggested improvement with 19...Rf6 has been partially confirmed by AylerKupp's engines analysis reaching <24.Qd3>.
click for larger view
This position shows the difficult task black still has to overcome since the opening gave white the opportunity to gain material value, put blacks king into a vulnerable weak position plus restricting blacks Queen side development. The line below demonstrates how black should aim for unwinnable endgame positions . <19. Nb5 Rf6 20. Rb1 Kd8 21. Qd5 Bd7 22. Red1 g6 23. Ne4 Rf5 24. Qd3> from here this line still awaits verification via very deep and wide engines analysis as to its correctness. 24... Rxb5 25. Rxb5 Kc7 26. Qxd6+ Qxd6 27. Nxd6 Rf8 28. Be2 a6 29. Rb2 Rxf4 30. Nxb7 Rf2 31. Nd6 Ne5 32. Ne4 Rf4 33. Ng5 h6 34. Nf3 Nxf3 35. Bxf3 (35. Rf1 Bc6 36. Bxf3 Bb5 37. c4 Bd7 38. Kh2 Ba4 39. g3 Rf5 40. Be2 Rxf1 41. Bxf1 Bd4 42. Rb1 Bc2 43. Rc1 Be4 44. Rd1 Bc5 45. Be2 a5 46. h4 a4 47. Kh3 Be7 48. h5 gxh5 49. Bxh5 Kb6 50. Re1 Bc5 51. Be2 Bb4 52. Rd1 Bd2 53. Rf1 a3 54. Kg4 Kc5 55. Kh5 Bc2 56. Rf8 Bb1 57. Rc8+ Kd4 58. c5 Bxa2 59.
c6 Bd5 60. Ra8 Bxc6 61. Rxa3 Be8+ 62. Kg4 Bb4 63. Rd3+ Ke4 64. Rd8 Bf7 65. Kh3 Ke5 66. Kg2 Bd5+ 67. Kh2 Be7 68. Re8 Kd6 69. Kh3 Bc6 70. Rb8 Kc5 71. Rh8 Bg5 72. Rc8 Kd6 73. Rb8 Kc5 74. Rb1 Kd4 75. Ra1 Be4 76. Ra6 Bd5) 35... Bb5 36. Rb4 Rc4 37. a4 Rxb4 38. cxb4 Bxa4 39. Rc1+ Kd6 40. Be2 Bb5 41. Bxb5 axb5 42. g3 Bd4 43. Kg2 Bb2 44. Rc8 Ba3 45. Rd8+ Ke7 46. Rb8 e2 47. Rb7+ Ke6 48. Rb6+ Kf7 49. Kf2 Bxb4 50. Kxe2 Bc3 51. Rxb5 Bf6 52. Rb7+ Be7 1/2-1/2 As you can see after replaying these moves above that there are many more samples necessary to prove my point for Rf6 instead of h6 as played resulting in a win for white. |
|
| Nov-11-18 | | The Kingfish: <kwid>, that you for that excellent analysis! This game is fantastically complicated to say the least. It seems that the Traxler variation has some problems. I look forward to finding out what the final verdict is. Thanks for your continued work on this fascinating game. |
|
Nov-11-18
 | | AylerKupp: <<The Kingfish> Can someone explain to me where the mistakes were for both sides?> I conducted a 14-part, 3-engine (Houdini 6, Komodo 11/12, and Stockfish 8/9) analysis of each of Team Black's moves starting in Team White vs Team Black, 2017 (kibitz #3727). The computer evaluations were a ratings-weighted average (RWAvg) of the evaluations provided by each engine. I also charted the evaluations for each move. I didn't see a drastic evaluation change until the very end when Black's position was hopeless (and the engines recognized it) but mostly a gradual degradation of Black's position. Here is a summary of the degradation based on the RWAvg of the engines for the principal variation: (a) Position considered equal. Never. At no time was the RWavg evaluation lower than [+0.50], my threshold for considering the position equal. (b) Position considered to be slightly better for White. From move 7 through move 20 and from move 24 through move 26 the position was evaluated between [+0.50] to [0.99], my range for considering the position to be slightly better for White. (c) Position considered to be much better for White. From move 21 thorough move 23, from move 27 through move 36, and on move 39 the position was evaluated between [+1.00] to [+1.99], my range for considering the position to be much better for White. (d) Position considered to be winning for White. From move 35 through move 38 and from move 40 through move 55, the last move, the position was evaluated as greater than [+2.00], my threshold for the position considered to be winning for White. These are charted in a spreadsheet provided in http://www.mediafire.com/file/frcfj.... So I suppose that according to the engines White committed 2 inaccuracies: (1) On move 24.Rb1 since it allowed Black to improve its position from an evaluation of [+1.33] (White standing much better) to [+0.89] (White standing slightly better). (2) On move 38.h6 since it allowed Black to improve its position from an evaluation of [+2.39] (White has a wining advantage) to [+1.98] (White standing much better, but barely). Otherwise I think that it was simply a case of the engines recognizing that White's position was getting better and better as the game went on. |
|
Nov-11-18
| | AylerKupp: <<kwid> Black did make one mistake at move 19 which should have been Rf6 instead of h6 leading to a forced loss.> I haven't seen any evidence that 19...Rf6 is better than 19...h6 or that either or both of them lead to a forced loss by Black. In fact, the engine tournaments I conducted with 3 engines starting from both 19...h6 and 19...Rf6 listed in Team White vs Team Black, 2017 (kibitz #3834) indicated that White scored lower with 19...Rf6 than 19...h6, although I might not have conducted enough games for those results to be statistically significant. And I don't consider that "The correctness of the suggested improvement with 19...Rf6 has been partially confirmed AylerKupp's engines analysis reaching <24.Qd3>". If anything they confirm that your long analysis is pretty much meaningless in confirming any definitive result since according to the engines the best response to 19...Rf6 was 20.Bd5 instead of your suggested 20.Rb1, if only by a small amount. So, if your analysis does not start with 20.Bd5 then, according to the engines, White did not play its best move. It's relatively easy to show a draw by Black if White does not play its best moves. I know that you seem to have an emotional attachment to 19...Rf6 but, as I told you before, unless all the moves or the vast majority of them are forced, your long analysis lines are meaningless in terms of reaching a definitive conclusion. There are just too many reasonable alternate lines for a human or even the fastest computer given today's technology to consider. And that's what's needed to arrive at a definitive conclusion, an in-depth analysis of <all> reasonable alternatives. Although I plan to continue my engine analysis of all the moves you suggested because I promised that I would (even though I've been distracted recently), it's not going to provide anything in terms of definitive conclusions. You can certainly take pride in the fact that the engines so far agree that most of the moves you suggested are best under the circumstances, but not all of them. And that's all. Sorry. |
|
Nov-11-18
| | AylerKupp: <kwid> <Summary of White's choices after 23...Rf5> (part 1 of 2) After 23...Rf5 we reach this position:
click for larger viewHere is a summary of 3 engines' evaluations, sorted in order of descending Ratings-Weighted Average (RWAvg) since it is White's move : a. Moves not reported by the engines as one of their top 5 moves are considered to be null when calculating the averages. White's Houdini 6 Komodo 12.1 Stockfish 9
Move d=27 d=34 d=45 <Avg> <RWAvg>
---------- ---------- ----------- ---------- ---------- -----------
24.Qd3 [+0.87] [+0.70] [+0.83] <[+0.80]> <[+0.80]>
24.Qg8+ [0.00] [0.00] [+0.83] <[+0.28]> <[+0.28]>
24.Qxd6 [-0.63] [-0.74] [-0.70] <[-0.69]> <[-0.69]>
24.Nbxd6 [-6.85] [-6.23] [-8.26] <[-7.11]> <[-7.12]>
24.Qxf5 [-7.24] [------] [-8.04] N/A N/A
24.Re1 [------] [-6.50] [------] N/A N/A
b. Moves not reported by the 3 engines as one of their top 5 moves are considered to be evaluated equal to the evaluation of the 5th best move less the average of the absolute differences between each of the evaluations. This is likely to be a more realistic value than just one centipawn below the evaluation of the 5th best move. White's Houdini 6 Komodo 12.1 Stockfish 9
Move d=27 d=34 d=45 <Avg> <RWAvg> <TrueRank>
---------- ---------- ----------- ---------- ---------- ----------- -----------
24.Qd3 [+0.87] [+0.70] [+0.83] <[+0.80]> <[+0.80]> <1>
24.Qg8+ [0.00] [0.00] [+0.83] <[+0.28]> <[+0.28]> <2>
24.Qxd6 [-0.63] [-0.74] [-0.70] <[-0.69]> <[-0.69]> <2>
24.Qxf5 [-7.24] [-8.30] [-8.04] <[-7.77]> <[-7.12]> <3>
24.Nbxd6 [-6.85] [-6.23] [-8.26] <[-7.11]> <[-7.88]> <4>
24.Re1 [-9.27] [-6.50] [-10.59] <[-8.79]> <[-8.78]> <5> True Rank: 1 = [ 24.Qd3 ]; 2 = [ 24,Qg8+, 24.Qxd6 ]; 3 = [ 24.Qxf5 ]; 4 = [ 24.Nbxd6 ]; 5 = [ 24.Re1 ] I find it interesting that both Houdini 6 and Komodo 12.1 thought that a perpetual check was White's second best attempt, but Stockfish 9 thought that White had more after 24.Qg8+ |
|
Nov-11-18
| | AylerKupp: <kwid> <Summary of White's choices after 23...Rf5> (part 2 of 2) And here is a summary of how the 3 engines ranked their top 5 moves, without regard for the value of the evaluation. White's Houdini 6 Komodo 12.1 Stockfish 9
Move d=27 d=34 d=45 <AvgRank> <TrueRank>
---------- ---------- ----------- ---------- ---------- -----------
24.Qd3 1 1 1 <1.0> <1>
24.Qg8+ 2 2 1 <1.7> <2>
24.Qxd6 3 3 3 <3.0> <3>
24.Nbxd6 4 4 5 <4.3> <4>
24.Qxf5 5 6 4 <5.0> <5>
24.Re1 6 5 6 <5.7> <6> True Rank: 1 = [ 24.Qd3 ]; 2 = [ 24,Qg8+ ]; 3 = [ 24.Qxd6 ]; 4 = [ 24.Nbxd6 ]; 5 = [ 24.Qxf5 ]; 6 = [ 24.Re1 ] |
|
Nov-11-18
| | AylerKupp: <kwid> Here is the current state of the comparison through 24.Ne4 between the moves in your initial analysis in Team White vs Team Black, 2017 (kibitz #3860). Analysis PV=1 PV=2 PV=3 PV=4 PV=5
----------- ------- ------- ------- ------- ------- <20.Rb1> 20.Bd5 20.Qd3 20.f5 <20.Rb1> 20.Qf3 <20...Kd8> <20...Kd8> 20...h6 20...g6 20...h6 20...Bc5 <21.Qd5> <21.Qd5> 21.Qh5 21.Qf3 21.Bd5 21.Nxh7 <21...Bd7> <21...Bd7> 21...a6 21...g6 21...Rf5 21...h6 <22.Red1> <22.Red1> 22.Nf7+ 22.Ne4 22.Qg8+ 22.Nxh7 <22...g6> <22...g6> 22...Rxf4 22...Kc8 22...Na5 22...Rf5 <23.Ne4> <23.Ne4> 23.Nxd6 23.a4 23.Be2 23.a3 <23...Rf5> <23...Rf5> 23...Be6 23...Rxf4 23...Bc5 23...Re6 <24.Qd3> <24.Qd3> 24.Qg8+ 24.Qxd6 24.Nbxd6 24.Qxf5 <24...Rxb5> (TBD) |
|
| Nov-12-18 | | kwid: <AylerKupp:> thank you for all your work. And yes you're right about whites Rb1 choice. My best line with Rf6 ends with an unwinnable endgame position where blacks Bishop pair together with the advanced e3 pawn secures a draw. 
click for larger view19. Nb5 <Rf6> 20. Rb1 Kd8 21.
Qd5 Bd7 22. Red1 g6 23. Ne4 Rf5 24. Qd3 Rxb5 25. Rxb5 Kc7 26. Qxd6+ Qxd6 27.
Nxd6 Rf8 28. Be2 a6 29. Rb2 Rxf4 30. Nxb7 Rf2 31. Nd6 Ne5 32. Ne4 Rf4 33. Ng5
Rf5 34. Nxh7 Rf2 35. Rf1 Be6 36. Rxf2 exf2 37. Rb4 a5 38. Re4 Kd6 39. Ra4 Bf5
40. Nf6 g5 41. Ra3 (41. Ne4+ Kd5 42. Nxg5 Bd3) 41... Ke7 42. Ng4 Nxg4 43. hxg4
Bxg4 44. Bf1 Bd7 45. c4 a4 46. Kh2 Bc5 47. Rd3 Bc6 48. Kh3 Be4 49. Rd2 Ke6 50.
g4 Ke5 51. Kg3 Be3 52. Rd7 Bc5 1/2-1/2 |
|
| Nov-12-18 | | morfishine: <kwid> Thank you for you (and <Aylerkupp>'s) continued input to this game! It was a pleasure to be on your team! I will still not play the Traxler though I enjoyed this game. Its a similar feeling when I learned about and ventured the Schliemann defense, which is another story altogether |
|
| Nov-12-18 | | kwid: <morfishine: <kwid> Thank you>
You're welcome. Now the Schlieman would make a good topic for our next challenge game don't you think? But our provider should allow the use of engines to contribute to the current opening theory which would also generate more interest for such a game. Why not ask around to see if there is enough interest for such an event?
My time may be limited due to effects on my mental and physical capacity beyond my control but I would love to take part in such a battle as long as I can. |
|
Nov-13-18
| | AylerKupp: <kwid> <Summary of Black's choices after 24.Qd3> After 24.Qd3 we reach this position:
click for larger viewHere is a summary of 3 engines' evaluations, sorted in order of ascending Ratings-Weighted Average (RWAvg) since it is Black's move. Another relatively infrequent instance in which the 3 engines agreed on their top 5 moves but, as usual, not on their move rankings : Black's Houdini 6 Komodo 12.1 Stockfish 9
Move d=27 d=29 d=46 <Avg> <RWAvg> <TrueRank>
---------- ---------- ----------- ---------- ---------- ----------- -----------
24...Rxb5 [+0.80] [+0.36] [+0.82] <[+0.66]> <[+0.66]> <1>
24...Rh5 [+0.88] [+0.89] [+1.36] <[+1.04]> <[+1.05]> <1>
24...Ke8 [+1.20] [+1.72] [+1.46] <[+1.46]> <[+1.47]> <2>
24...Rxf4 [+1.15] [+1.68] [+2.05] <[+1.63]> <[+1.64]> <2>
24...d5 [+1.33] [+1.51] [+2.15] <[+1.66]> <[+1.67]> <2> True Rank: 1 = [ 24...Rxb5, 24...Rh5 ]; 2 = [ 24...Ke8, 24...Rxf4, 24...d5 ] And here is a summary of how the 3 engines ranked their top 5 moves, without regard for the value of the evaluation. Black's Houdini 6 Komodo 12.1 Stockfish 9
Move d=27 d=29 d=46 <AvgRank> <TrueRank>
---------- ---------- ----------- ---------- ---------- -----------
24...Rxb5 1 1 1 <1.0> <1>
24...Rh5 2 2 2 <2.0> <2>
24...Rxf4 3 4 4 <3.7> <3>
24...Ke8 4 5 3 <4.0> <3>
24...d5 5 3 5 <4.3> <4> True Rank: 1 = [ 24...Rxb5 ]; 2 = [ 24...Rh5 ]; 3 = [ 24...Rxf4, 24...Ke8 ]; 4 = [ 24...d5 ] |
|
Nov-13-18
| | AylerKupp: <kwid> Here is the current state of the comparison through 24.Ne4 between the moves in your initial analysis in Team White vs Team Black, 2017 (kibitz #3860). Analysis PV=1 PV=2 PV=3 PV=4 PV=5
----------- ------- ------- ------- ------- ------- <20.Rb1> 20.Bd5 20.Qd3 20.f5 <20.Rb1> 20.Qf3 <20...Kd8> <20...Kd8> 20...h6 20...g6 20...h6 20...Bc5 <21.Qd5> <21.Qd5> 21.Qh5 21.Qf3 21.Bd5 21.Nxh7 <21...Bd7> <21...Bd7> 21...a6 21...g6 21...Rf5 21...h6 <22.Red1> <22.Red1> 22.Nf7+ 22.Ne4 22.Qg8+ 22.Nxh7 <22...g6> <22...g6> 22...Rxf4 22...Kc8 22...Na5 22...Rf5 <23.Ne4> <23.Ne4> 23.Nxd6 23.a4 23.Be2 23.a3 <23...Rf5> <23...Rf5> 23...Be6 23...Rxf4 23...Bc5 23...Re6 <24.Qd3> <24.Qd3> 24.Qg8+ 24.Qxd6 24.Nbxd6 24.Qxf5 <24...Rxb5> <24...Rxb5> 24...Rh5 24...Ke8 24...Rxf4 24...d5 <25.Rxb5> (TBD) |
|
Nov-13-18
| | AylerKupp: <kwid> But now I have an issue that needs to be resolved relatively quickly. Your original line posted on Oct-08-18 had the following as your main line: 19...Rf6 20.Rb1 Kd8 21.Qd5 Bd7 22.Red1 g6 23.Ne4 Rf5 24.Qd3 Rxb5 25.Rxb5 Kc7 26.Qxd6+ Qxd6 27.Nxd6 <a6> 28.Rbd5 Na5 29.Be2 Ba4 30.Rb1 Bc6 31.Re5 Rf8 32.Ne4 Rxf4 33.Re7+ Kd8 34.Rxh7 Rxe4 35.Rxb6 Rf4 36.Rb1 Rf2 37.Bf3 Bxf3 38.gxf3 Rxa2 39.Re1 Nc4 40.Rxb7 Rf2 41.Rf7 Nd6 42.Rf4 Nf5 43.Kg1 Rc2 44.h4 Ke7 45.c4 Ke6 46.Ra1 a5 47.Re4+ Kf6 48.Rxa5 Nxh4 49.Rxe3 Rxc4. Your second analysis main line posted on Oct-19-18 had the following as a later main line: 19...Rf6 20.Rb1 Kd8 21.Qd5 Bd7 22.Red1 g6 23.Ne4 Rf5 24.Qd3 Rxb5 25.Rxb5 Kc7 26.Qxd6+ Qxd6 27.Nxd6 <Rf8> 28.Be2 Be6 29.c4 a6 30.Rbb1 Bd4 31.c5 Rb8 32.Bf3 Bxc5 33.Ne4 Bf8 34.Nc3 Re8 35.Kh2 h5 36.g3 Bf5 37.Rb2 Bg7 38.Rc1 Bd4 39.g4 Bd3 40.Be2 Bxe2 41.Rxe2 hxg4 42.hxg4 g5 43.fxg5 Re5 44.g6 Rc5 45.Rec2 Rg5 46.Ne2 Be5+ 47.Kh3 Rxg6 48.Rc4 Kb8 49.Re4 Rh6+ 50.Kg2 Rh2+ 51.Kf1 Bd6 52.Rxe3 Ne5 53.g5 Rh5 54.Rg3 Ng6 55.Rg1 Rh8 56.Nd4 Rf8+ 57.Ke1 Bc7 58.Rg4 Ne5 59.Ne6 Ba5+ 60.Ke2 Re8 61.Ng7 Rg8 62.Re4 Nf7 63.Re8+ Rxe8+ 64.Nxe8 Nxg5 65.Rc5 Bd8 And yesterday you posted a third analysis main line: 19...Rf6 20.Rb1 Kd8 21.Qd5 Bd7 22.Red1 g6 23.Ne4 Rf5 24.Qd3 Rxb5 25.Rxb5 Kc7 26.Qxd6+ Qxd6 27.Nxd6 Rf8 28.Be2 <a6> 29.Rb2 Rxf4 30.Nxb7 Rf2 31.Nd6 Ne5 32.Ne4 Rf4 33.Ng5 Rf5 34.Nxh7 Rf2 35.Rf1 Be6 36.Rxf2 exf2 37.Rb4 a5 38.Re4 Kd6 39.Ra4 Bf5 40.Nf6 g5 41.Ra3 Ke7 42.Ng4 Nxg4 43.hxg4 Bxg4 44.Bf1 Bd7 45.c4 a4 46.Kh2 Bc5 47.Rd3 Bc6 48.Kh3 Be4 49.Rd2 Ke6 50. g4 Ke5 51. Kg3 Be3 52. Rd7 Bc5 I indicated earlier that I was only going to analyze one line so you need to tell me which of the three (or a fourth) you would like me to analyze. Otherwise I will be forced to stop after I analyze Black's best responses to 27.Nxd6, and I have finished analyzing White's best responses to 24...Rxb5 so I don't have too many more responses to analyze. |
|
| Nov-13-18 | | kwid: <AylerKupp:><25.Rxb5> 19. Nb5 <Rf6 20. Rb1 Kd8 21. Qd5 Bd7 22. Red1 g6 23. Ne4 Rf5 24. Qd3 Rxb5 25. Rxb5> Kc7 26. Qxd6+ Qxd6 27. Nxd6 Rf8 28. Be2 a6 29. Rb2 Rxf4 30. Nxb7 Rf2 31. Nd6 Ne5 32. Ne4 Rf4 33. Ng5
h6 34. Nf3 Nxf3 35. Bxf3 Bb5 36. Rb4 Rc4 37. a4 Rxb4 38. cxb4 Bxa4 39. Rc1+ Kd6
40. Be2 Bb5 41. Bxb5 axb5 42. g3 Bd4 43. Kg2 Bb2 44. Rd1+ Ke5 45. Rd8 Kf6 46.
Rf8+ Kg7 47. Re8 Bc3 48. Re7+ Kf6 49. Rxe3 Bxb4 50. Rf3+ Ke7 51. Rb3 Bd2 52.
Rxb5 Bg5 1/2-1/2 |
|
| Nov-13-18 | | kwid: <AylerKupp: <kwid> But now I have an issue that needs to be resolved relatively quickly. Your original line posted on Oct-08-18 had the following as your main line: 19...Rf6> <20.Rb1 Kd8 21.Qd5 Bd7 22.Red1 g6 23.Ne4 Rf5 24.Qd3 Rxb5 25.Rxb5> I was hoping that your findings will show us the " Right Way" confirming 19...Rf6 was the choice move for black to hold white to a draw. |
|
Nov-14-18
| | AylerKupp: <Black holding White to a draw after 19.Nb5> (part 1 of 5)> <<kwid> I was hoping that your findings will show us the " Right Way" confirming 19...Rf6 was the choice move for black to hold white to a draw.> Yes, I know you did. I've told you before that I don't have any confidence that your approach would <conclusively> determine that 19...Rf6 (or any other move) would lead to a draw for Black because there are simply too many reasonable alternative lines to explore for either human or computer. Only if all the moves in a line are practically forced can such an approach succeed and that would be very rare unless they either involve a relatively small number of moves or if at least at the end of a practically forced series of moves you either reach a position with 7 pieces whose evaluation can be conclusively determined by the 7-piece Lomonosov tablebases or reach a relatively simple position that can be analyzed by the FinalGen tablebase generator in a reasonable amount of time. And that's certainly not the case with the positions arising after 19...Rf6. Back on Jun-04-18 starting with Team White vs Team Black, 2017 (kibitz #3727) I posted a summary of a 3-engine analysis of Black's top 2 responses in the actual game to White's moves. The actual summary (not posted) of the analysis evaluating Black's responses to 19.Nb5 on the basis of a Ratings-Weighted Average (RWAvg) was as follows: Move Houdini 6 Komodo 11 Stockfish 8 <Avg> <RWAvg> <TrueRank> ---------- ---------- ----------- ---------- ---------- ----------- ---------- 19...Kd8 [+0.97] [+0.66] [+0.99] <[+0.87]> <[+0.88]> <1> 19...Rf6 [+1.00] [+0.87] [+0.94] <[+0.94]> <[+0.94]> <1> 19...Rxf4 [+1.03] [+0.96] [+1.20] <[+1.06]> <[+1.06]> <1> 19...h6 [+1.37] [+0.59] [+1.28] <[+1.08]> <[+1.10]> <1> 19...g6 [+1.75] [+1.50] [+2.06] <[+1.77]> <[+1.78]> <2> True Rank: 1 = [ 19...Kd8, 19...Rf6; 19...h6, 19...Rxf4 ]; 2 = [ 19...g6 ] So the first four moves were effectively of equal value when you take into account a human analyst's "resolution" of a move/position which I consider to be the equivalent of [ ± 0.50] centipawns. |
|
Nov-14-18
| | AylerKupp: <Black holding White to a draw after 19.Nb5> (part 2 of 5)> If you use the thresholds of:
[-2.00] & Down = <Black has a winning advantage> [-1.99] to [-1.00] = <Black is significantly better> [-0.99] to [-0.50] = <Black is slightly better> [-0.49] to [+0.49] = <Equal> [+0.50] to [+0.99] = <White is slightly better> [+1.00] to [+1.99] = <White is significantly better> [+2.00] & Up = <White has a winning advantage> Then even if you were to IMO optimistically increase a human analyst's "resolution" of the position to be the equivalent of [ ±0.25] centipawns, the <TrueRank> using the engine's evaluations would remain the same from the perspective of a human analyst. Using the evaluation ranges indicated then after 19...Kd8 or 19...Rf6 White would be considered slightly better and after 19...Rxf4 or 19...h6 White would be considered to be significantly better by a human analyst. But that's using a hard and fast threshold for the evaluations and clearly if, say, 19...Rf6 had been evaluated at [+0.99] and 19...Rxf4 had been evaluated at [+1.00] then it would be hard to argue that the [+0.01] centipawn difference (actual difference [+0.12]) was sufficiently better to warrant a different human analyst's evaluation of the position after 19...Rf6 than after 19...Rxf4. |
|
Nov-14-18
| | AylerKupp: <Black holding White to a draw after 19.Nb5> (part 3 of 5)> And if you look at how the engines ranked the moves, without looking at the exact value of each engine's evaluation, this is the result: Move Houdini 6 Komodo 11 Stockfish 8 <AvgRank> <TrueRank>
---------- ---------- ----------- ---------- ---------- ----------- 19...Kd8 1 2 2 <1.7> <1> 19...Rf6 2 3 1 <2.0> <1> 19...h6 4 1 4 <3.0> <2> 9...Rxf4 3 4 3 <3.3> <2> 19...g6 5 5 5 <5.0> <3> True Rank: 1 = [ 19...Kd8, 19...Rf6 ]; 2 = [ 19...h6, 19...Rxf4 ]; 3 = [ 19...g6 ] Not too surprising since each engine's evaluation does have a resolution of [ ±0.01] centipawn. But you can clearly see using this ranking method, this time with a [ ±0.5] threshold for Average Rank (which may or may not be correct, but it's at least consistent with the other method) how the 3 engines differed in their ranking of Black's possible responses to 19.Nb5. In this case each engine considered a different move to be "best"; Houdini 6 considered 19...Kd8 to be "best", Komodo 11 considered 19...h6 to be best, and Stockfish 8 considered 19...Rf6 to be "best". I some those that read this keep this in mind when reading articles attempting to evaluate which players are best by comparing how their moves match with what <a single chess engine> considers the "best" move to be. But that's another issue. And this is what I think is the fundamental problem with the position after 19.Nb5 or, for that matter, after many of White's moves. The positions are simply too complex to probably be able to reach a definitive conclusion with regard to which response by Black would definitely lead to a draw or even which response by Black would give the best chances for a draw. |
|
Nov-14-18
| | AylerKupp: <Black holding White to a draw after 19.Nb5> (part 4 of 5)> My earlier attempt (see Team White vs Team Black, 2017 (kibitz #3834)) of running two 12-round, 3-engine tournaments, one with the position after 19...h6 and one with the position after 19...Rf6 was, I think, "interesting". This is not the way to reach a definitive conclusion as to whether either move allows Black to hold White to a draw but maybe it could provide an answer as to which move gives Black <the best chance> to hold White to a draw, statistically speaking. But, after thinking about it, I have many questions about what I did,. For example: 1. The games pitted one engine against another engine. Should I have instead pitted the same engine against itself? Would that have given a better indication on the correctness of each response? 2. I conducted 12 rounds for each move and don't know whether this is an adequate number of rounds to determine if the results would be statistically significant). I personally think not but I still haven't figured out how to calculate the number of games needed to reach a given confidence level (say, 95%) of the results nor how to determine the confidence level that the results are correct given a specific number of games. If anyone knows how to do this, please let me know. 3. I ran the games at Classic time controls, 40 moves in 2 hours and 20 moves an hour afterwards since my GUI (Arena 3.5) does not provide for increments. That allowed me to run 2 games each evening and (most of the time) I then had the game results the following morning. Is this a reasonable approximation to the 2-days that we were allowed to consider a response? I could double the time allowed for each move by making the time control to 40 moves in 4 hours and 20 moves every 2 hours and run only one game each evening but, given the exponential increase in the time it takes an engine to evaluate each move that would likely increase the search depth to only one play or maybe two. And even increasing the time per move to two days might not significantly increase the search depth because of the same exponential growth. Any thoughts and comments by anyone who bothers to read this post would be appreciated. |
|
Nov-14-18
| | AylerKupp: <Black holding White to a draw after 19.Nb5> (part 5 of 5)> Finally (yeah!) <kwid> how would you like to proceed, if at all? One possibility is that in the 19...Rf6 line you could try coming up with a new analysis after 20.Bd5 instead of 20.Rb1 since not only did an overnight analysis of the position after 19...Rf6 (see Team White vs Team Black, 2017 (kibitz #3906)) indicate that 20.Bd5 was the "best" move (although not by much) but the same 3 engines in the engine tournament (which perhaps makes them biased) chose 20.Bd5 as the move to be played in 11 out of the 12 games starting with 19...Rb6. So 20.Bd5 looks to be at least a reasonable White response to 19...Rf6 and should be investigated as an alternative to 20.Rb1. So, if you're up to it, why don't you give it a try? You still seem to have a strong feeling that 19...Rf6 gives Black at least a reasonable chance for a draw and I will help you try to substantiate the feeling if I can although, frankly, I don't think it's possible. We can perhaps consider ourselves to be a modern equivalent of Don Quixote and Sancho Panza, and I'll even let you decide which part you prefer to play. Although I have to tell you that I've lost a lot of weight recently (> 30 pounds) so I'm not sure that "Panza" (which is Spanish for belly) is still applicable to me. :-) I've ran one more set of analyses that I need to reformat before posting and I would suggest that, after I do that, that I wait to see what analysis you come up with indicating what could be the line with best play by both sides. Which is what chess engines try to do :-). Then, after I see the analysis that you come up with, I'll restart my 3-engine analysis after 20.Bd5 and proceed until the engines find a "better" move (if they do) than the one you suggested. This would be an example of "human/engine forward sliding", perhaps the first of this kind. After all, of the 11 moves you've analyzed in your main line, the engines so far have agreed with the move that you considered best 10 of those times. So I would say that your record is pretty good! But, of course, don't force yourself to do any more than your health (and, of course, your interest) allows you to do. I'm not going anywhere and can wait for you. And if anyone who bothers to read these posts has any other thoughts, including thinking that we're both crazy, please share. |
|
| Nov-14-18 | | kwid: <AylerKupp: <Black holding White to a draw after 19.Nb5> (part 5 of 5)>
Finally (yeah!) <kwid> how would you like to proceed, if at all? One possibility is that in the 19...Rf6 line you could try coming up with a new analysis after 20.Bd5 instead of 20.Rb1> Well,If you look at my last Rb1 line your engines may agree to a draw and thus game over.
As to Bd5 I still have a game analysis which also shows no win for white in sight and thus if your engines agree we should stop searching for a win if black plays 19...Rf6. [Event "Traxler Bxf7+/Bc4/b4"]
[Site "web"]
[Date "2018.11.14"]
[Round "?"]
[White "Theory"]
[Black "Analysis 20.Bd5"]
[Result "1/2-1/2"]
[ECO "C57"]
[Annotator "WinUser"]
[PlyCount "92"]
[EventDate "2018.??.??"]
1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 Bc5 5. Bxf7+ Ke7 6. Bc4 Rf8 7. O-O Qe8
8. b4 Bxb4 9. c3 Ba5 10. d4 d6 11. f4 Bb6 12. Kh1 exd4 13. e5 $1 Ng4 14. exd6+
cxd6 15. h3 Ne3 16. Re1 Kd7 17. Bxe3 dxe3 18. Na3 Qe7 19. Nb5< Rf6 20. Bd5> Kd8
21. Ne4 (21. a4 Rxf4 22. Nf7+ Rxf7 23. Bxf7 Bxh3 24. gxh3 Qe4+ 25. Kh2 Qf4+ 26.
Kg2 Qe4+ 27. Qf3 Qxf3+ 28. Kxf3 Ne5+ 29. Ke4 Nxf7 30. a5 Bc5 31. Rf1 Ke7 32.
Nd4 Re8 33. Kxe3 Kd7+ 34. Kd2 Ne5 35. Kc2 Re7 36. Kb3 g6 37. Rad1 Bxd4 38. Rxd4
Re6 39. Rf8) 21... Rxf4 22. Nexd6 Bg4 23. Nxb7+ (23. Qd3 Ne5 24. Qxh7 a6 25.
Nxb7+ Kd7 26. Nd4 Raf8 27. Rxe3 (27. Qb1 Bxd4 28. cxd4 Rxd4 29. Rxe3 Rxd5 30.
Qe4 Qe6 31. hxg4 Rc8 32. a4 Rc4 33. Qxc4 Nxc4 34. Rxe6 Kxe6 35. Rc1 Ne5 36.
Nc5+ Kd6 37. Nxa6 Ra5 38. Rd1+ Ke7 39. Nb4 Rxa4) 27... Rf1+ 28. Rxf1 Rxf1+ 29.
Kh2 Qf8 30. Qg8 Nf3+ 31. Bxf3 Qxg8 32. Bxg4+ Kc7 33. Re7+ Kb8 34. Nc6+ Ka8 35.
Nd6 Bg1+ 36. Kg3 Bf2+ 37. Kh2 Bg1+ 38. Kg3 Bf2+) 23... Kc8 24. N7d6+ Kd7 25.
Qc2 (25. Bxc6+ Kxc6 26. Qb3 Qe6 27. Nd4+ Bxd4 28. cxd4 Qxb3 29. axb3 Kxd6 30.
hxg4 Re4 31. Ra5 Re7 32. b4 Rb8 33. Re5 Rxb4 34. Rxe7 Kxe7 35. Rxe3+ Kd6 36.
Re5 Ra4 37. Rh5 h6 38. Rc5 g5 39. Kg1 a5 40. Kf2 Ra3 41. Ke2 Ra1 42. Kd3 a4 43.
Ra5 a3 44. Kc4 Ra2 45. Ra6+ Kc7 46. g3 Kb7 47. Rxh6 Rc2+ 48. Kb3 a2 49. Rh7+
Kb6) 25... a6 26. Nd4 Kxd6 27. Bxc6 Rc8 28. Qa4 Bxd4 29. Qb4+ Ke6 30. Bd5+ Kf6
31. Qxe7+ Kxe7 32. cxd4 Bd7 33. Rxe3+ Kd6 34. Bb7 Rc4 35. Rae1 (35. Bxa6 Ra4
36. Be2 Bc6 37. a3 Raxd4 38. Kg1 Rd2 39. Rb1 Rf5 40. Rd1 Rxd1+ 41. Bxd1 Rf4 42.
Kh2 g5 43. Bg4 Ra4 44. Re6+ Kd5 45. Rh6 Kc5 46. Rxh7 Rxa3 47. Rg7 Kd4 48. Rxg5
Ra2 49. Bh5 Rxg2+ 50. Rxg2 Bxg2 51. Kxg2 Ke5 52. h4 Kf6) 35... Ra4 36. a3 Rfxd4
37. Re7 Rxa3 38. Rxg7 a5 39. Rxh7 Rc3 40. Rh6+ Kc7 41. Bf3 a4 42. Ra6 Bxh3 43.
Re7+ Bd7 44. Ra7+ Kd8 45. Rh7 a3 46. Kh2 Re3 1/2-1/2 |
|
| Nov-14-18 | | centralfiles: <AK> I will try to post an answer to you shortly.
It should include an extensive analysis of 20.Bd5 that hopefully will convincingly show to all that black draws. |
|
| Nov-14-18 | | centralfiles: <AK> For what it's worth to you the positions following 27...Rf8<in the 20.Rb1 line> were analyzed here Team White vs Team Black, 2017 (kibitz #3309)
Conclusion: White was unable to reach a winning endgame. As far as "confirming" 19...Rf6 draws. I'm afraid <AK> and <kwid> clearly both mean well but somehow you guys don't seem to be on the same page.
Kwid's idea of chess engine analysis and <AK>s seem to be very far apart (My own personal view is a third one entirely). In my own view i have already close to confirmed the soundness of 19...Rf6 several months ago with the use of engines. see my posts above Team White vs Team Black, 2017 (kibitz #3684) |
|
 |
|
|
< Earlier Kibitzing · PAGE 137 OF 140 ·
Later Kibitzing> |
Kibitzing
WARNING ABOUT COMPUTER USE:
The use of chess playing computers is strictly prohibited during this game. You may not
post computer analysis or evaluations to the Kibitzer's Corner, or
consult chess playing computer software, including endgame tablebases, to help you decide on a move.
Anybody violating this rule will be suspended from the game.
Chess databases (such as chessgames.com) and reference materials are permitted.
Thank you for preserving the integrity of this contest.
