Stockfish Natural TB loses heavily to Stockfish master

[Lyudmil Tsvetkov]

I guess probing Lomonosov, whether you use DTZ or DTM, would be an even better way to go.

[Laskos]

On 5-men regular Wins, it seems to perform a bit better than the first version.

Wins:
1000 games
Suite: Regular 5-men Wins
TC: 0.25s per move
Score of SF Master vs SF Final2 NTB: 500 - 496 - 4 [0.502] 1000
ELO difference: 1.39 +/- 21.49
Finished match

4 failures out of 500 regular 5-men Wins.

One of the goals of Natural, to have shorter Mates, is completely defeated.

After glancing a bit over these PGNs, and on this one with easy regular 5-men Wins at the root, I observed with naked eye that usually the path to Mate of Stockfish Final2 NTB is longer than the path to Mate of Stockfish master. I took the PGN database of this match (500 Wins of master, 496 Wins of Natural) and computed some stats on the lengths of the games to Mate:

Stockfish master:
Mean Mate: 17.6 moves
Median Mate: 14 moves

Stockfish Final2 Natural TB (the newest):
Mean Mate: 32.7 moves
Median Mate: 26 moves

Almost twice longer Mates with SF Final2 NTB! Pretty remarkable, and defeats even "naturalness". It maybe doesn't play "weird moves" so often, but it plays pretty dumb moves often, which lengthen the Mate to 50-80 moves on these easy regular 5-men Wins at the root. SF master very rarely does that. And this can lead SF Natural TB even to fail to convert.

Here are length of the games histograms for Master and Final2 NITB:

[Joerg Oster]

On 5-men regular Wins, it seems to perform a bit better than the first version.

Wins:
1000 games
Suite: Regular 5-men Wins
TC: 0.25s per move
Score of SF Master vs SF Final2 NTB: 500 - 496 - 4 [0.502] 1000
ELO difference: 1.39 +/- 21.49
Finished match

4 failures out of 500 regular 5-men Wins.

One of the goals of Natural, to have shorter Mates, is completely defeated.

After glancing a bit over these PGNs, and on this one with regular 5-men Wins at the root, I observed with naked eye that usually the path to Mate of Stockfish Final2 NTB is longer than the path to Mate of Stockfish master. I took the PGN database of this match and computed some stats on the lengths of the games to Mate:

Stockfish master:
Mean Mate: 17.6 moves
Median Mate: 14 moves

Stockfish Final2 Natural TB (the newest):
Mean Mate: 32.7 moves
Median Mate: 26 moves

Really?

Or is it only that SF NTB starts displaying mate scores much earlier than master,
and with a higher distance for that reason, of course!?

[Laskos]

On 5-men regular Wins, it seems to perform a bit better than the first version.

Wins:
1000 games
Suite: Regular 5-men Wins
TC: 0.25s per move
Score of SF Master vs SF Final2 NTB: 500 - 496 - 4 [0.502] 1000
ELO difference: 1.39 +/- 21.49
Finished match

4 failures out of 500 regular 5-men Wins.

One of the goals of Natural, to have shorter Mates, is completely defeated.

After glancing a bit over these PGNs, and on this one with regular 5-men Wins at the root, I observed with naked eye that usually the path to Mate of Stockfish Final2 NTB is longer than the path to Mate of Stockfish master. I took the PGN database of this match and computed some stats on the lengths of the games to Mate:

Stockfish master:
Mean Mate: 17.6 moves
Median Mate: 14 moves

Stockfish Final2 Natural TB (the newest):
Mean Mate: 32.7 moves
Median Mate: 26 moves

Really?

Or is it only that SF NTB starts displaying mate scores much earlier than master,
and with a higher distance for that reason, of course!?

No, the games are played to the point Mate is on the board.

[Guenther]

On 5-men regular Wins, it seems to perform a bit better than the first version.

Wins:
1000 games
Suite: Regular 5-men Wins
TC: 0.25s per move
Score of SF Master vs SF Final2 NTB: 500 - 496 - 4 [0.502] 1000
ELO difference: 1.39 +/- 21.49
Finished match

4 failures out of 500 regular 5-men Wins.

One of the goals of Natural, to have shorter Mates, is completely defeated.

After glancing a bit over these PGNs, and on this one with regular 5-men Wins at the root, I observed with naked eye that usually the path to Mate of Stockfish Final2 NTB is longer than the path to Mate of Stockfish master. I took the PGN database of this match and computed some stats on the lengths of the games to Mate:

Stockfish master:
Mean Mate: 17.6 moves
Median Mate: 14 moves

Stockfish Final2 Natural TB (the newest):
Mean Mate: 32.7 moves
Median Mate: 26 moves

Really?

Or is it only that SF NTB starts displaying mate scores much earlier than master,
and with a higher distance for that reason, of course!?

No, the games are played to the point Mate is on the board.

But this doesn't answer the question of Joerg?
(of course you play until mate otherwise the stats would be meaningless,
but the question is when mate scores appear first.)

Can you show one or two examples?

[mcostalba]

Kai, thanks for keeping testing, this is useful. I have already copied your hard position and doing some tests locally.

But in this case your test is flawed (as already pointed out): if for instance SF NTB is programmed to only show mate in 1 it would easily win your statistic....

[Laskos]

On 5-men regular Wins, it seems to perform a bit better than the first version.

Wins:
1000 games
Suite: Regular 5-men Wins
TC: 0.25s per move
Score of SF Master vs SF Final2 NTB: 500 - 496 - 4 [0.502] 1000
ELO difference: 1.39 +/- 21.49
Finished match

4 failures out of 500 regular 5-men Wins.

One of the goals of Natural, to have shorter Mates, is completely defeated.

After glancing a bit over these PGNs, and on this one with regular 5-men Wins at the root, I observed with naked eye that usually the path to Mate of Stockfish Final2 NTB is longer than the path to Mate of Stockfish master. I took the PGN database of this match and computed some stats on the lengths of the games to Mate:

Stockfish master:
Mean Mate: 17.6 moves
Median Mate: 14 moves

Stockfish Final2 Natural TB (the newest):
Mean Mate: 32.7 moves
Median Mate: 26 moves

Really?

Or is it only that SF NTB starts displaying mate scores much earlier than master,
and with a higher distance for that reason, of course!?

No, the games are played to the point Mate is on the board.

But this doesn't answer the question of Joerg?
(of course you play until mate otherwise the stats would be meaningless,
but the question is when mate scores appear first.)

Can you show one or two examples?

I cannot make statistic about that, and several examples are pretty meaningless. I will show several games of SFNTB in won 5-men positions, if you want full 1000 games PGN, I can probably upload it. But is that so important, that a probably incorrect M19 appears first or not? TB score +132.39 is also Mate (in master, not necessarily in Natural, it is sometimes a Draw there).

Code: Select all

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "78"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "137"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "1k6/3K4/8/8/R7/8/3P1r2/8 w - - 0 1"]

1. Rb4+ {+10.80/22 0.25s} 1. ... Ka7 {-132.79/20 0.25s} 2. d4 {+5.67/41 
0.25s} 2. ... Rf7+ {-132.79/21 0.25s} 3. Kc6 {+58.20/20 0.25s} 3. ... Rf4 
{-132.79/20 0.25s} 4. Kc5 {+21.60/24 0.25s} 4. ... Rf2 {-132.79/22 0.25s} 
5. d5 {+132.66/20 0.25s} 5. ... Rc2+ {-132.79/21 0.25s} 6. Kd6 {+121.79/21
0.25s} 6. ... Ka6 {-132.79/23 0.25s} 7. Rb1 {+132.66/21 0.25s} 7. ... Rc4 
{-132.79/20 0.25s} 8. Rb8 {+121.62/21 0.25s} 8. ... Rc2 {-132.79/23 0.25s}
9. Ke6 {+132.60/19 0.25s} 9. ... Rd2 {-132.79/24 0.25s} 10. d6 {+132.71/21
0.26s} 10. ... Re2+ {-132.79/18 0.25s} 11. Kd7 {+132.67/20 0.25s} 11. ... 
Ka7 {-132.79/19 0.25s} 12. Rb1 {+132.69/19 0.25s} 12. ... Rc2 {-132.79/19 
0.25s} 13. Kd8 {+132.62/18 0.25s} 13. ... Ka6 {-132.79/21 0.25s} 14. Kd7 {
+59.15/17 0.25s} 14. ... Ka5 {-132.79/22 0.25s} 15. Ke7 {+132.70/17 0.26s}
15. ... Rd2 {-132.79/25 0.25s} 16. Kd7 {+132.68/20 0.25s} 16. ... Ka4 {
-132.79/23 0.25s} 17. Rb8 {+125.47/18 0.25s} 17. ... Rd1 {-132.79/26 0.25s
} 18. Rf8 {+132.68/19 0.25s} 18. ... Kb3 {-132.79/16 0.25s} 19. Rf3+ {
+132.60/18 0.25s} 19. ... Kb4 {-132.79/16 0.25s} 20. Kc7 {+123.32/20 0.25s
} 20. ... Rc1+ {-132.79/15 0.25s} 21. Kb7 {+127.45/24 0.25s} 21. ... Rd1 {
-132.79/16 0.25s} 22. Kc7 {+53.29/20 0.25s} 22. ... Rc1+ {-132.79/16 0.25s
} 23. Kb7 {+12.92/20 0.25s} 23. ... Rd1 {-132.79/18 0.25s} 24. Kc6 {
+132.70/19 0.25s} 24. ... Ka4 {-132.79/16 0.25s} 25. Kc7 {+132.74/18 0.25s
} 25. ... Rc1+ {-132.79/17 0.25s} 26. Kd8 {+132.72/20 0.25s} 26. ... Rc2 {
-132.79/19 0.25s} 27. Rf4+ {+132.64/18 0.25s} 27. ... Kb3 {-132.79/17 
0.25s} 28. d7 {+132.73/18 0.25s} 28. ... Kc3 {-132.79/21 0.25s} 29. Rf3+ {
+64.24/17 0.25s} 29. ... Kd4 {-132.79/22 0.25s} 30. Ke7 {+58.15/17 0.25s} 
30. ... Re2+ {-132.79/25 0.25s} 31. Kf7 {+132.75/20 0.25s} 31. ... Ke4 {
-132.79/28 0.25s} 32. Rf4+ {+132.75/19 0.26s} 32. ... Kxf4 {-132.79/20 
0.25s} 33. d8=Q {+5.88/35 0.25s} 33. ... Ke4 {-132.79/22 0.25s} 34. Ke6 {
+5.92/22 0.25s} 34. ... Kf4+ {-132.79/25 0.25s} 35. Kf6 {+5.97/22 0.25s} 
35. ... Ke4 {-132.79/22 0.25s} 36. Qe7+ {+5.97/22 0.25s} 36. ... Kd3 {
-132.79/19 0.25s} 37. Qd6+ {+5.97/20 0.25s} 37. ... Kc3 {-132.79/22 0.25s}
38. Qc5+ {+6.01/23 0.25s} 38. ... Kd3 {-132.79/21 0.25s} 39. Qb5+ {
+6.01/24 0.26s} 39. ... Kd2 {-132.79/20 0.25s} 40. Kf5 {+6.01/25 0.25s} 
40. ... Re3 {-132.79/20 0.25s} 41. Kf4 {+6.01/25 0.25s} 41. ... Rh3 {
-132.79/20 0.25s} 42. Qb2+ {+6.05/21 0.26s} 42. ... Kd1 {-132.79/19 0.25s}
43. Qf2 {+6.05/23 0.25s} 43. ... Ra3 {-132.79/20 0.25s} 44. Ke4 {+6.05/20 
0.25s} 44. ... Kc1 {-132.79/20 0.25s} 45. Qc5+ {+6.09/23 0.25s} 45. ... 
Kb2 {-132.79/22 0.25s} 46. Qb4+ {+7.40/23 0.25s} 46. ... Rb3 {-132.79/25 
0.25s} 47. Qd2+ {+52.57/20 0.25s} 47. ... Ka3 {-132.79/26 0.25s} 48. Qa5+ 
{+52.89/23 0.25s} 48. ... Kb2 {-132.79/29 0.25s} 49. Kd4 {+132.50/23 0.25s
} 49. ... Kc2 {-132.79/26 0.25s} 50. Qa2+ {+132.60/25 0.25s} 50. ... Rb2 {
-132.79/28 0.25s} 51. Qc4+ {+132.60/25 0.25s} 51. ... Kd1 {-132.79/26 
0.25s} 52. Qf1+ {+132.61/25 0.25s} 52. ... Kc2 {-298.84/25 0.25s} 53. Kc4 
{+132.63/26 0.25s} 53. ... Kd2 {-M42/30 0.25s} 54. Qf2+ {+132.65/25 0.25s}
54. ... Kc1 {-M46/31 0.25s} 55. Qe1+ {+132.65/23 0.25s} 55. ... Kc2 {
-M40/34 0.25s} 56. Qa1 {+132.65/23 0.25s} 56. ... Rb8 {-M26/27 0.25s} 57. 
Qc3+ {+132.66/20 0.25s} 57. ... Kd1 {-M24/33 0.25s} 58. Qf3+ {+132.75/22 
0.25s} 58. ... Kc2 {-M22/35 0.25s} 59. Qg2+ {+M53/24 0.25s} 59. ... Kc1 {
-M20/36 0.25s} 60. Qg1+ {+M23/28 0.25s} 60. ... Kd2 {-M18/38 0.25s} 61. 
Qh2+ {+M19/30 0.25s} 61. ... Ke3 {-M16/39 0.25s} 62. Qg3+ {+M17/32 0.25s} 
62. ... Ke4 {-M16/31 0.25s} 63. Qxb8 {+M15/34 0.25s} 63. ... Kf5 {-M14/34 
0.25s} 64. Kd5 {+M11/32 0.25s} 64. ... Kg5 {-M10/40 0.25s} 65. Ke5 {+M9/38
0.25s} 65. ... Kg4 {-M8/58 0.25s} 66. Qb3 {+M7/64 0.25s} 66. ... Kh5 {
-M6/101 0.25s} 67. Qg8 {+M5/127 0.085s} 67. ... Kh6 {-M4/127 0.18s} 68. 
Kf4 {+M3/127 0.005s} 68. ... Kh5 {-M2/127 0.005s} 69. Qg5# {+M1/127 
0.002s, White mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "80"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "87"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "8/8/8/R7/1P6/6r1/5k2/1K6 w - - 0 1"]

1. Ra2+ {+4.53/25 0.25s} 1. ... Ke1 {-132.79/18 0.25s} 2. Rb2 {+130.61/29 
0.25s} 2. ... Rg6 {-132.79/20 0.25s} 3. b5 {+132.73/26 0.25s} 3. ... Rb6 {
-132.79/19 0.25s} 4. Kc2 {+10.65/21 0.25s} 4. ... Ke2 {-132.79/19 0.25s} 
5. Kc3+ {+120.40/18 0.25s} 5. ... Kd1 {-132.79/19 0.25s} 6. Rb1+ {
+132.75/20 0.25s} 6. ... Ke2 {-132.79/19 0.25s} 7. Kd4 {+132.69/20 0.25s} 
7. ... Kf3 {-132.79/18 0.25s} 8. Kc5 {+132.71/18 0.25s} 8. ... Rg6 {
-132.79/16 0.25s} 9. Rb3+ {+132.70/18 0.25s} 9. ... Ke4 {-132.79/16 0.25s}
10. Rb4+ {+132.67/21 0.26s} 10. ... Kd3 {-132.79/17 0.25s} 11. b6 {
+115.23/33 0.25s} 11. ... Rg5+ {-132.79/17 0.25s} 12. Kd6 {+132.69/20 
0.25s} 12. ... Kc3 {-132.79/18 0.26s} 13. b7 {+7.58/27 0.25s} 13. ... Kxb4
{-132.79/20 0.25s} 14. b8=Q+ {+5.92/33 0.25s} 14. ... Kc3 {-132.79/19 
0.25s} 15. Qf8 {+5.97/17 0.25s} 15. ... Rg3 {-132.79/19 0.25s} 16. Kc5 {
+6.01/17 0.25s} 16. ... Re3 {-132.79/18 0.25s} 17. Qf4 {+5.92/18 0.25s} 
17. ... Kd3 {-132.79/22 0.25s} 18. Qc4+ {+6.01/19 0.25s} 18. ... Kd2 {
-132.79/21 0.25s} 19. Kd4 {+6.01/22 0.25s} 19. ... Rg3 {-132.79/17 0.25s} 
20. Qa2+ {+6.01/20 0.25s} 20. ... Ke1 {-132.79/18 0.25s} 21. Qh2 {+6.09/23
0.25s} 21. ... Ra3 {-132.79/21 0.25s} 22. Qd6 {+6.09/23 0.25s} 22. ... Rf3
{-132.79/21 0.25s} 23. Qe5+ {+6.24/21 0.25s} 23. ... Kf1 {-132.79/23 0.25s
} 24. Ke4 {+52.81/21 0.25s} 24. ... Kg2 {-132.79/25 0.25s} 25. Qg5+ {
+132.63/23 0.25s} 25. ... Rg3 {-132.79/27 0.25s} 26. Qd2+ {+132.63/22 
0.25s} 26. ... Kh3 {-132.79/26 0.25s} 27. Qd1 {+132.63/26 0.25s} 27. ... 
Rg4+ {-132.79/26 0.25s} 28. Kf3 {+132.69/23 0.25s} 28. ... Rg3+ {-M28/28 
0.25s} 29. Kf4 {+132.65/25 0.25s} 29. ... Kh2 {-M26/32 0.25s} 30. Qe1 {
+132.67/22 0.25s} 30. ... Rg8 {-M28/28 0.25s} 31. Qf2+ {+M57/21 0.25s} 
31. ... Kh3 {-M28/32 0.25s} 32. Qf3+ {+M29/31 0.25s} 32. ... Kh2 {-M26/35 
0.25s} 33. Qh5+ {+M27/31 0.25s} 33. ... Kg1 {-M24/37 0.25s} 34. Qd5 {
+M21/31 0.25s} 34. ... Rg2 {-M22/38 0.25s} 35. Kf3 {+M19/37 0.25s} 35. ...
 Kh2 {-M18/44 0.25s} 36. Qh5+ {+M17/42 0.25s} 36. ... Kg1 {-M16/50 0.25s} 
37. Qh4 {+M15/45 0.25s} 37. ... Rg8 {-M14/51 0.25s} 38. Qe1+ {+M13/48 
0.25s} 38. ... Kh2 {-M12/58 0.25s} 39. Qe5+ {+M11/53 0.25s} 39. ... Kg1 {
-M10/62 0.25s} 40. Qa1+ {+M9/63 0.25s} 40. ... Kh2 {-M8/66 0.25s} 41. Qa2+
{+M7/70 0.25s} 41. ... Kg1 {-M6/71 0.25s} 42. Qxg8+ {+M5/127 0.078s} 
42. ... Kf1 {-M4/127 0.003s} 43. Qg2+ {+M3/127 0.008s} 43. ... Ke1 {
-M2/127 0.002s} 44. Qe2# {+M1/127 0.002s, White mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "82"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "19"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "8/8/2P5/2K5/8/6k1/3P4/3B4 w - - 0 1"]

1. d4 {+132.73/14 0.25s} 1. ... Kf2 {-M16/27 0.25s} 2. c7 {+132.71/15 
0.26s} 2. ... Ke1 {-M14/31 0.25s} 3. d5 {+M19/16 0.25s} 3. ... Kxd1 {
-M14/28 0.25s} 4. c8=Q {+M13/24 0.25s} 4. ... Ke2 {-M12/33 0.25s} 5. Qg4+ 
{+M11/26 0.25s} 5. ... Kd2 {-M10/41 0.25s} 6. d6 {+M9/32 0.25s} 6. ... Kc1
{-M8/56 0.25s} 7. d7 {+M7/46 0.25s} 7. ... Kb1 {-M6/127 0.20s} 8. d8=Q {
+M5/110 0.25s} 8. ... Ka1 {-M4/127 0.009s} 9. Qd2 {+M3/127 0.013s} 9. ... 
Kb1 {-M2/127 0.001s} 10. Qgd1# {+M1/127 0.004s, White mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "86"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "41"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "4R3/1K6/2P4r/8/8/8/8/k7 w - - 0 1"]

1. Re1+ {+121.54/19 0.25s} 1. ... Kb2 {-132.79/20 0.25s} 2. Re2+ {
+127.57/20 0.25s} 2. ... Kc3 {-132.79/21 0.25s} 3. c7 {+132.72/16 0.25s} 
3. ... Rh7 {-132.79/23 0.25s} 4. Kb8 {+47.80/24 0.25s} 4. ... Rh8+ {
-M42/25 0.25s} 5. c8=Q+ {+47.80/34 0.25s} 5. ... Rxc8+ {-M38/35 0.25s} 6. 
Kxc8 {+127.82/45 0.25s} 6. ... Kd3 {-M36/37 0.25s} 7. Re6 {+M33/31 0.25s} 
7. ... Kc4 {-M30/38 0.25s} 8. Kc7 {+M29/33 0.25s} 8. ... Kd5 {-M28/38 
0.25s} 9. Re1 {+M25/35 0.25s} 9. ... Kc4 {-M24/38 0.25s} 10. Kc6 {+M23/36 
0.25s} 10. ... Kd3 {-M22/40 0.25s} 11. Kc5 {+M21/38 0.25s} 11. ... Kd2 {
-M20/42 0.25s} 12. Re4 {+M19/40 0.25s} 12. ... Kd3 {-M18/44 0.25s} 13. Re6
{+M17/42 0.25s} 13. ... Kc2 {-M16/46 0.25s} 14. Kc4 {+M15/42 0.25s} 
14. ... Kd2 {-M14/51 0.25s} 15. Re8 {+M13/47 0.25s} 15. ... Kc2 {-M12/57 
0.25s} 16. Rd8 {+M11/52 0.25s} 16. ... Kb2 {-M10/69 0.25s} 17. Rd2+ {
+M9/65 0.25s} 17. ... Kc1 {-M8/78 0.25s} 18. Kc3 {+M7/126 0.25s} 18. ... 
Kb1 {-M6/115 0.25s} 19. Kb3 {+M5/127 0.011s} 19. ... Kc1 {-M4/83 0.25s} 
20. Rd4 {+M3/127 0.003s} 20. ... Kb1 {-M2/127 0.004s} 21. Rd1# {+M1/127 
0.002s, White mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "88"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "25"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "8/8/7K/8/4k1P1/4P3/5B2/8 w - - 0 1"]

1. g5 {+132.68/20 0.26s} 1. ... Kf5 {-M26/21 0.25s} 2. Be1 {+132.76/18 
0.25s} 2. ... Ke4 {-M26/26 0.25s} 3. g6 {+132.77/18 0.25s} 3. ... Kxe3 {
-M22/26 0.25s} 4. Bb4 {+56.87/17 0.25s} 4. ... Kd4 {-M22/29 0.25s} 5. g7 {
+132.72/16 0.25s} 5. ... Kc4 {-M18/29 0.25s} 6. Bf8 {+M19/22 0.25s} 6. ...
 Kc3 {-M14/31 0.25s} 7. g8=Q {+M13/25 0.25s} 7. ... Kd3 {-M12/33 0.25s} 8.
 Bc5 {+M11/26 0.25s} 8. ... Kc3 {-M10/37 0.25s} 9. Qd5 {+M9/29 0.25s} 
9. ... Kb2 {-M8/53 0.25s} 10. Qc4 {+M7/40 0.25s} 10. ... Ka1 {-M6/127 
0.25s} 11. Bd4+ {+M5/127 0.10s} 11. ... Kb1 {-M4/127 0.004s} 12. Qb3+ {
+M3/127 0.006s} 12. ... Kc1 {-M2/127 0.002s} 13. Be3# {+M1/127 0.003s, 
White mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "90"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "21"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "8/8/3B3P/6K1/8/3P4/1k6/8 w - - 0 1"]

1. d4 {+132.71/20 0.25s} 1. ... Kc3 {-M20/24 0.25s} 2. Be5 {+132.72/17 
0.25s} 2. ... Kb2 {-M18/28 0.25s} 3. d5+ {+M17/20 0.25s} 3. ... Kb3 {
-M16/27 0.25s} 4. d6 {+M15/23 0.25s} 4. ... Kb4 {-M14/29 0.25s} 5. d7 {
+M13/23 0.25s} 5. ... Kc4 {-M12/29 0.25s} 6. h7 {+M11/23 0.25s} 6. ... Kb3
{-M10/31 0.25s} 7. d8=Q {+M9/24 0.25s} 7. ... Kc2 {-M8/40 0.25s} 8. h8=Q {
+M7/31 0.25s} 8. ... Kb3 {-M6/93 0.25s} 9. Qd3+ {+M5/77 0.25s} 9. ... Kb4 
{-M4/127 0.015s} 10. Qb8+ {+M3/127 0.017s} 10. ... Kc5 {-M2/127 0.001s} 
11. Qbb5# {+M1/127 0.004s, White mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "92"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "25"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "8/8/8/8/P7/3k3P/1K5B/8 w - - 0 1"]

1. a5 {+132.65/16 0.25s} 1. ... Ke4 {-M18/24 0.25s} 2. a6 {+132.78/16 
0.25s} 2. ... Kf3 {-M16/26 0.25s} 3. Ka1 {+132.71/20 0.25s} 3. ... Kg2 {
-M22/25 0.25s} 4. a7 {+132.78/14 0.25s} 4. ... Kxh2 {-M18/27 0.25s} 5. h4 
{+M17/24 0.25s} 5. ... Kg1 {-M16/31 0.25s} 6. h5 {+M15/26 0.25s} 6. ... 
Kg2 {-M14/33 0.25s} 7. h6 {+M13/28 0.25s} 7. ... Kf3 {-M12/37 0.25s} 8. h7
{+M11/30 0.25s} 8. ... Ke2 {-M10/40 0.25s} 9. a8=Q {+M9/32 0.25s} 9. ... 
Kd3 {-M8/54 0.25s} 10. h8=Q {+M7/42 0.25s} 10. ... Ke3 {-M6/80 0.25s} 11. 
Qc3+ {+M5/127 0.22s} 11. ... Kf4 {-M4/127 0.047s} 12. Qaf3+ {+M3/127 
0.009s} 12. ... Kg5 {-M2/127 0.003s} 13. Qcf6# {+M1/127 0.003s, White 
mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "94"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "17"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "8/8/8/1P6/3K2k1/1P6/5B2/8 w - - 0 1"]

1. Be3 {+132.72/15 0.25s} 1. ... Kf5 {-M16/29 0.25s} 2. b6 {+M17/22 0.25s}
2. ... Kg6 {-M14/32 0.25s} 3. b7 {+M13/25 0.25s} 3. ... Kf5 {-M12/34 0.25s
} 4. b8=Q {+M11/26 0.25s} 4. ... Ke6 {-M10/38 0.25s} 5. Qf8 {+M9/28 0.25s}
5. ... Kd7 {-M8/57 0.25s} 6. Qf7+ {+M7/38 0.25s} 6. ... Kc6 {-M6/86 0.25s}
7. Kc4 {+M5/127 0.14s} 7. ... Kd6 {-M4/127 0.060s} 8. Bf4+ {+M3/127 0.005s
} 8. ... Kc6 {-M2/127 0.001s} 9. Qc7# {+M1/127 0.003s, White mates} 1-0

[Event "?"]
[Site "?"]
[Date "2017.09.06"]
[Round "96"]
[White "SFNTB"]
[Black "SFM"]
[Result "1-0"]
[PlyCount "37"]
[SetUp "1"]
[TimeControl "0.25/move"]
[FEN "8/8/8/3B2B1/8/6rk/8/5K2 w - - 0 1"]

1. Be6+ {+49.34/30 0.25s} 1. ... Kh2 {-M40/35 0.25s} 2. Bf4 {+M41/34 0.25s
} 2. ... Kh1 {-M36/39 0.25s} 3. Bd5+ {+M35/36 0.25s} 3. ... Kh2 {-M34/41 
0.25s} 4. Kf2 {+M33/37 0.25s} 4. ... Kh3 {-M32/41 0.25s} 5. Bxg3 {+M31/37 
0.25s} 5. ... Kg4 {-M28/40 0.25s} 6. Bd6 {+M29/35 0.25s} 6. ... Kg5 {
-M30/35 0.25s} 7. Kg3 {+M25/34 0.25s} 7. ... Kf5 {-M26/36 0.25s} 8. Kh4 {
+M23/34 0.25s} 8. ... Kg6 {-M24/38 0.25s} 9. Kg4 {+M21/35 0.25s} 9. ... 
Kf6 {-M20/40 0.25s} 10. Kf4 {+M19/36 0.25s} 10. ... Kg6 {-M18/39 0.25s} 
11. Be7 {+M17/35 0.25s} 11. ... Kh5 {-M16/41 0.25s} 12. Bf7+ {+M15/35 
0.25s} 12. ... Kh6 {-M14/44 0.25s} 13. Bf8+ {+M13/36 0.25s} 13. ... Kh7 {
-M12/49 0.25s} 14. Kf5 {+M11/41 0.25s} 14. ... Kh8 {-M10/59 0.25s} 15. Kf6
{+M9/52 0.25s} 15. ... Kh7 {-M8/74 0.25s} 16. Bb3 {+M7/127 0.25s} 16. ... 
Kh8 {-M6/127 0.018s} 17. Kf7 {+M5/127 0.008s} 17. ... Kh7 {-M4/127 0.002s}
18. Bc2+ {+M3/127 0.004s} 18. ... Kh8 {-M2/127 0.002s} 19. Bg7# {+M1/127 
0.002s, White mates} 1-0

[Guenther]

But this doesn't answer the question of Joerg?
(of course you play until mate otherwise the stats would be meaningless,
but the question is when mate scores appear first.)

Can you show one or two examples?

I cannot make statistic about that, and several examples are pretty meaningless. I will show several games of SFNTB in won 5-men positions, if you want full 1000 games PGN, I can probably upload it. But is that so important, that a probably incorrect M19 appears first or not? TB score +132.39 is also Mate (in master, not necessarily in Natural, it is sometimes a Draw there).

...

For me it is/was not very important. I just asked the logical question, which immediately arose from your 'answer' to Joerg.

And from a first look at your posted examples it really seems SFNTB needs longer and starts later to show mate scores.

Guenther

[Laskos]

Kai, thanks for keeping testing, this is useful. I have already copied your hard position and doing some tests locally.

But in this case your test is flawed (as already pointed out): if for instance SF NTB is programmed to only show mate in 1 it would easily win your statistic....

I don't quite understand the questions in these last posts. Could you make Natural actually Mate the master faster than vice-versa? I don't care too much about a showing M27 earlier, often not the shortest one. TB score of +132 in master (not always in Natural) is also M with unknown value, so what's this showing off of Natural amounts for? That it is showing a line (often not the shortest) to actually Mate in 27? Master will show too a line to Mate, maybe a bit later, and never misses its TB score of +132.

Or you are saying that if Natural finds early a M40, while really it's M20, it will stick to M40 line?

[Joerg Oster]

Thanks, Kai.

So SF NTB not only produces longer lasting games than Master,
but also resolves mate later on a regular basis.
This is quite unexpected.

Another ugly thing is the drop of the score from one move to the other:

Code: Select all

32. Rf4+ {+132.75/19 0.26s} 32. ... Kxf4 {-132.79/20
0.25s} 33. d8=Q {+5.88/35 0.25s} 

(This is from the first game you posted.)