Fat Titz 1.1 released!

[JohnWoe]

Bigger nets are much better. Topology matters of course.

Some experiments. Minimizing MRE.

Joke networks 2x3x3x3x3x3x3x1 <- Horrible, 10 seconds

Code: Select all

Starting from ( 2x3x3x3x3x3x3x1 ): 2.9956076326750214 jnn.js:312:13
... ( 1 % ): 0.7500052998201416 jnn.js:336:36
... ( 2 % ): 0.7500052998201416 jnn.js:336:36
... ( 3 % ): 0.7500052998201416 jnn.js:336:36
... ( 4 % ): 0.7500052998201416 jnn.js:336:36
... ( 5 % ): 0.7500052998201416 jnn.js:336:36
... ( 6 % ): 0.7500052998201416 jnn.js:336:36
... ( 7 % ): 0.7500052998201416 jnn.js:336:36
... ( 8 % ): 0.7500052998201416 jnn.js:336:36
... ( 9 % ): 0.7500052998201416 jnn.js:336:36
Gradient success jnn.js:329:23
... ( 10 % ): 0.5616695718224758 jnn.js:336:36
... ( 11 % ): 0.5616695718224758 jnn.js:336:36
Gradient success jnn.js:329:23
... ( 12 % ): 0.5003095921038487 jnn.js:336:36
... ( 13 % ): 0.5003095921038487 jnn.js:336:36
... ( 14 % ): 0.5003095921038487 jnn.js:336:36
... ( 15 % ): 0.5003095921038487 jnn.js:336:36
... ( 16 % ): 0.5003095921038487 jnn.js:336:36
... ( 17 % ): 0.5003095921038487 jnn.js:336:36
... ( 18 % ): 0.5003095921038487 jnn.js:336:36
... ( 19 % ): 0.5003095921038487 jnn.js:336:36
... ( 20 % ): 0.5003095921038487 jnn.js:336:36
... ( 21 % ): 0.5003095921038487 jnn.js:336:36
... ( 22 % ): 0.5003095921038487 jnn.js:336:36
... ( 23 % ): 0.5003095921038487 jnn.js:336:36
... ( 24 % ): 0.5003095921038487 jnn.js:336:36
... ( 25 % ): 0.5003095921038487 jnn.js:336:36
... ( 26 % ): 0.5003095921038487 jnn.js:336:36
... ( 27 % ): 0.5003095921038487 jnn.js:336:36
... ( 28 % ): 0.5003095921038487 jnn.js:336:36
... ( 29 % ): 0.5003095921038487 jnn.js:336:36
... ( 30 % ): 0.5003095921038487 jnn.js:336:36
... ( 31 % ): 0.5003095921038487 jnn.js:336:36
... ( 32 % ): 0.5003095921038487 jnn.js:336:36
... ( 33 % ): 0.5003095921038487 jnn.js:336:36
... ( 34 % ): 0.5003095921038487 jnn.js:336:36
... ( 35 % ): 0.5003095921038487 jnn.js:336:36
... ( 36 % ): 0.5003095921038487 jnn.js:336:36
... ( 37 % ): 0.5003095921038487 jnn.js:336:36
... ( 38 % ): 0.5003095921038487 jnn.js:336:36
... ( 39 % ): 0.5003095921038487 jnn.js:336:36
... ( 40 % ): 0.5003095921038487 jnn.js:336:36
... ( 41 % ): 0.5003095921038487 jnn.js:336:36
... ( 42 % ): 0.5003095921038487 jnn.js:336:36
... ( 43 % ): 0.5003095921038487 jnn.js:336:36
... ( 44 % ): 0.5003095921038487 jnn.js:336:36
... ( 45 % ): 0.5003095921038487 jnn.js:336:36
... ( 46 % ): 0.5003095921038487 jnn.js:336:36
Gradient success jnn.js:329:23
Stopping ... 1.979297282609655e-23 jnn.js:340:17
Total: 186872 / Success: 71 / Fail: 186801 / NNPS: 18885 jnn.js:345:13
############### jnn.js:346:13
### SUCCESS ### jnn.js:347:13
############### jnn.js:348:13
NeuralNetwork learn: 9903ms - timer ended

2x20x20x1 <-- Much better. 1 second

Code: Select all

Starting from ( 2x20x20x1 ): 2.9999999999999902 jnn.js:312:13
Gradient success jnn.js:329:23
... ( 1 % ): 0.0015843143628259623 jnn.js:336:36
Gradient success jnn.js:329:23
Stopping ... 5.142671879657763e-20 jnn.js:340:17
Total: 5511 / Success: 8 / Fail: 5503 / NNPS: 4943 jnn.js:345:13
############### jnn.js:346:13
### SUCCESS ### jnn.js:347:13
############### jnn.js:348:13
NeuralNetwork learn: 1123ms - timer ended

2x50x50x1 <- Good but slower. 2 seconds

Code: Select all

Starting from ( 2x50x50x1 ): 4 jnn.js:312:13
Gradient success 10 jnn.js:329:23
Stopping ... 2.7808037459788666e-15 jnn.js:340:17
Total: 2315 / Success: 13 / Fail: 2302 / NNPS: 1048 jnn.js:345:13
############### jnn.js:346:13
### SUCCESS ### jnn.js:347:13
############### jnn.js:348:13
NeuralNetwork learn: 2220ms - timer ended

[Viren]

Thank you Sopel. Nice! In the other thread I only wrote serious postings and only topics related to chess. The thread was closed for no reason. Can't the moderator at least write why he is closing the thread? I feel discriminated against, I don't know what I did wrong.

Indeed. It’s not like this is Gamergate. A little light misogyny never hurt anyone, right?

https://en.m.wikipedia.org/wiki/Titz_(surname)

So people with the surname Titz are misogynists?

Please stop with this nonsense, Mr Titz.

(Before any mods get too trigger-happy, quoting wikipedia: "it is an altered form of the frequent masculine given name Dietrich" so it is perfectly legitimate for me to refer to Dietrich as Mr Titz, and removing it would be an insult to all those with the surname Titz)

[dkappe]

Could you elaborate what "Fat Titz by Stockfish" and "Regular Titz Stockfish" are ?

From the OPs GitHub, Fat Titz 1.0 and 1.1 are a fork of CFish.
Official Stockfish hasn't released it.

Since your legal expertise and insider contacts are unmatched in this forum, can you tell us where did you see Official Stockfish release it ?

Don’t get so hung up on names. Back to the chess:

[pgn]

[Event "?"]
[Site "?"]
[Date "2021.08.28"]
[Round "12"]
[White "Fat Titz by Stockfish 1.0"]
[Black "Small n Perky (Stockfish 13)"]
[Result "1-0"]
[ECO "B33"]
[TimeControl "300+5"]
[Termination "normal"]
[UTCDate "2021.08.29"]
[UTCTime "08:27:02"]
[Variant "Standard"]
[Opening "Sicilian Defense: Lasker-Pelikan Variation"]
[Annotator "https://lichess.org/@/AniChess39"]

1. e4 {+0.36/28 32s} 1... c5 {-0.20/31 21s} 2. Nf3 {+0.44/24 2.8s} 2... Nc6
{-0.28/28 4.0s} 3. d4 {+0.50/25 5.5s} 3... cxd4 {-0.32/29 5.0s} 4. Nxd4
{+0.61/25 5.1s} 4... Nf6 {-0.22/29 6.2s} 5. Nc3 {+0.53/26 7.8s} 5... e5
{-0.09/33 19s} 6. Ndb5 {+0.57/26 6.2s} 6... d6 {-0.13/30 5.8s} 7. Nd5
{+0.53/28 6.9s} 7... Nxd5 {-0.05/30 5.4s} 8. exd5 {+0.55/29 11s} 8... Nb8
{-0.10/28 7.3s} 9. a4 {+0.56/28 8.6s} 9... Be7 {-0.17/29 6.5s} 10. Bd2
{+0.60/26 5.8s} 10... O-O {-0.15/33 24s} 11. Be2 {+0.58/26 7.2s} 11... Bf5
{-0.27/32 19s} 12. O-O {+0.53/27 8.2s} 12... Nd7 {-0.21/28 5.4s} 13. a5
{+0.58/26 6.8s} 13... a6 {-0.28/32 12s} 14. Na3 {+0.47/25 8.2s} 14... Nf6
{-0.05/31 6.1s} 15. Be1 {+0.62/29 25s} 15... Rc8 {0.00/27 7.1s} 16. Nc4
{+0.66/24 5.1s} 16... Be4 {-0.21/31 43s} 17. f3 {+1.03/26 8.1s} 17... Bxd5
{-0.14/27 5.0s} 18. Nb6 {+1.06/27 6.8s} 18... Be6 {-0.27/30 6.7s} 19. Nxc8
{+1.09/28 7.6s} 19... Qxc8 {-0.18/31 6.2s} 20. Bf2 {+1.09/28 8.1s} 20... d5
{-0.34/28 9.6s} 21. c3 {+0.93/26 14s} 21... h6 {-0.46/33 68s} 22. Qd2
{+1.20/24 7.0s} 22... Qc7 {-0.42/28 10s} 23. Kh1 {+1.19/30 18s} 23... Nh5
{-0.38/30 30s} 24. Bd1 {+1.20/28 19s} 24... Bg5 {-0.40/31 15s} 25. Be3
{+1.22/28 5.1s} 25... Bxe3 {-0.44/29 4.3s} 26. Qxe3 {+1.03/29 14s} 26... Nf4
{-0.44/31 6.9s} 27. Re1 {+1.00/27 8.6s} 27... f6 {-0.32/30 15s} 28. Bb3
{+0.87/27 13s} 28... Rd8 {-0.45/33 20s} 29. Ra4 {+1.18/30 22s} 29... Bf5
{-0.46/28 3.0s} 30. Rd1 {+0.97/26 8.2s} 30... Bd7 {-0.46/26 4.3s} 31. Raa1
{+1.17/26 9.7s} 31... Bc6 {-0.42/27 5.9s} 32. Rd2 {+1.14/25 7.8s} 32... Kh8
{-0.48/30 28s} 33. Qb6 {+0.70/28 40s} 33... Qd7 {-0.37/25 3.9s} 34. Qg1
{+0.68/26 20s} 34... Qe8 {-0.42/26 7.0s} 35. Rad1 {+0.81/27 28s} 35... Qe6
{-0.30/26 4.7s} 36. Qb6 {+0.74/26 5.1s} 36... Re8 {-0.33/24 2.6s} 37. Qe3
{+0.58/26 19s} 37... Rd8 {-0.49/26 9.5s} 38. Kg1 {+0.92/28 14s} 38... Rd6
{-0.58/27 14s} 39. Kf2 {+1.38/23 5.9s} 39... Rd7 {-0.71/25 11s} 40. g3
{+1.29/24 7.0s} 40... Ng6 {-0.75/24 4.0s} 41. Bc2 {+1.18/24 9.9s} 41... e4
{-1.03/25 1.6s} 42. fxe4 {+1.66/24 5.9s} 42... Re7 {-1.25/25 2.9s} 43. Rxd5
{+1.78/22 1.8s} 43... Bxd5 {-1.00/25 1.6s} 44. Rxd5 {+1.87/24 4.1s} 44... Ne5
{-1.00/25 3.5s} 45. Kg2 {+1.90/23 3.5s} 45... Kh7 {-1.44/28 19s} 46. Qc5
{+1.74/26 19s} 46... Qg4 {-1.36/25 1.5s} 47. Qf2 {+1.93/21 1.5s} 47... Qe6
{-1.28/23 0.89s} 48. h3 {+1.69/29 12s} 48... Nf7 {-1.24/26 4.4s} 49. Qc5
{+1.87/21 3.5s} 49... Ne5 {-1.16/26 1.6s} 50. Rd8 {+2.29/23 2.9s} 50... Re8
{-1.24/28 3.6s} 51. Qd5 {+2.87/22 1.5s} 51... Qe7 {-1.42/25 2.7s} 52. Rxe8
{+2.87/25 2.1s} 52... Qxe8 {-1.53/25 3.0s} 53. h4 {+3.07/25 3.2s} 53... Qb8
{-1.69/29 3.5s} 54. Bb3 {+3.31/26 4.8s} 54... h5 {-1.27/28 2.6s} 55. Qe6
{+3.34/26 2.7s} 55... Qd8 {-2.04/32 7.1s} 56. Bd5 {+3.64/25 4.6s} 56... Kh6
{-2.12/29 2.3s} 57. b4 {+3.64/25 3.9s} 57... Kh7 {-3.14/33 25s} 58. Kf2
{+4.32/25 5.1s} 58... Kh6 {-3.27/27 3.4s} 59. Ke2 {+4.52/24 3.9s} 59... Qb8
{-3.92/31 10s} 60. Qg8 {+4.88/25 3.5s} 60... Qd6 {-4.06/27 0.90s} 61. Qh8+
{+5.26/25 3.7s} 61... Kg6 {-3.50/26 1.0s} 62. Qe8+ {+5.09/31 10s} 62... Kh6
{-4.48/30 3.9s} 63. Qh8+ {+5.30/24 2.3s} 63... Kg6 {-4.85/25 1.3s} 64. Qc8
{+5.50/26 4.2s} 64... Kh6 {-4.99/30 6.1s} 65. Qxb7 {+5.39/28 7.1s} 65... Ng4
{-4.40/26 1.7s} 66. Kf3 {+5.55/28 5.0s} 66... Nh2+ {-4.77/34 18s} 67. Kg2
{+5.62/26 3.4s} 67... Ng4 {-4.85/26 0.63s} 68. Qb6 {+5.57/26 6.0s} 68... Qe5
{-4.86/28 1.5s} 69. Qd4 {+5.88/24 5.3s} 69... Qe7 {-4.59/28 3.5s} 70. Bc4
{+6.39/27 3.4s} 70... Qb7 {-4.80/29 2.4s} 71. Be6 {+6.74/25 4.6s} 71... Ne5
{-5.75/28 5.3s} 72. Qd2+ {+6.77/27 5.3s} 72... Kh7 {-5.32/26 1.7s} 73. Qd5
{+6.95/26 3.5s} 73... Qc7 {-7.10/32 20s} 74. Bg8+ {+7.09/23 3.7s} 74... Kh6
{-6.36/25 1.1s} 75. Qd2+ {+7.15/24 3.8s} 75... g5 {-6.46/26 2.4s} 76. Be6
{+7.08/26 6.4s} 76... Qc6 {-7.20/30 13s} 77. Bf5 {+7.61/26 6.0s} 77... Ng6
{-6.87/25 2.7s} 78. Kf2 {+8.07/22 3.1s} 78... Qa4 {-7.29/24 2.0s} 79. e5
{+8.21/24 3.3s} 79... Nxe5 {-8.51/25 3.4s} 80. hxg5+ {+8.71/23 4.6s} 80... Kg7
{-8.43/26 2.8s} 81. Qd8 {+8.67/24 5.5s} 81... fxg5 {-8.33/28 14s} 82. Qxg5+
{+8.55/25 4.8s} 82... Kf7 {-8.11/23 1.1s} 83. Qf4 {+8.61/27 4.8s} 83... Ng6
{-8.11/25 2.0s} 84. Qc4+ {+8.83/24 5.4s} 84... Kg7 {-8.94/28 7.1s} 85. Qd4+
{+8.90/23 3.6s} 85... Kf7 {-9.07/25 1.6s} 86. Qd5+ {+8.95/22 3.9s} 86... Kg7
{-9.07/27 1.4s} 87. Qd6 {+9.21/24 4.8s} 87... Nf8 {-8.77/26 4.9s} 88. Bd3
{+9.22/23 4.1s} 88... Qb3 {-9.61/27 16s} 89. Qd4+ {+9.45/21 4.4s} 89... Kf7
{-10.27/24 5.9s} 90. Bxa6 {+9.96/22 3.9s} 90... Qc2+ {-10.00/23 4.7s} 91. Be2
{+9.85/23 5.2s} 91... h4 {-10.66/24 6.8s} 92. gxh4 {+10.04/22 4.6s} 92... Qf5+
{-10.57/21 0.38s} 93. Ke3 {+10.13/23 5.3s} 93... Ne6 {-10.56/24 1.1s} 94. Qe4
{+10.35/22 3.9s} 94... Ng7 {-13.45/23 11s} 95. a6 {+11.26/21 4.4s} 95... Qd7
{-13.45/21 0.61s} 96. Qb7 {+61.56/35 4.3s} 96... Qe7+ {-M24/62 4.3s} 97. Kd2
{+61.87/38 16s} 97... Nf5 {-M30/28 2.4s} 98. Qxe7+ {+66.14/33 3.7s} 98... Kxe7
{-M24/53 1.4s} 99. a7 {+78.11/32 3.5s} 99... Nxh4 {-M22/54 1.6s} 100. a8=Q
{+M55/45 3.4s} 100... Ng6 {-M20/58 2.2s} 101. Qe4+ {+M49/61 3.8s} 101... Kf6
{-M16/63 2.3s} 102. Qd4+ {+M41/67 5.2s} 102... Kf5 {-M14/67 2.3s} 103. Bd3+
{+M21/64 3.7s} 103... Kg5 {-M10/96 2.6s} 104. Qg1+ {+M19/80 4.5s} 104... Kf6
{-M10/98 2.7s} 105. Qxg6+ {+M17/85 4.0s} 105... Ke5 {-M8/245 1.1s} 106. Qh6
{+M7/231 6.3s} 106... Kd5 {-M6/245 0.010s} 107. c4+ {+M5/245 0.017s} 107... Ke5
{-M4/245 0.007s} 108. b5 {+M3/245 0.007s} 108... Kd4 {-M2/245 0.005s} 109. Qd6#
{+M1/245 0.004s, White mates}
1-0
[/pgn]

[dkappe]

https://en.m.wikipedia.org/wiki/Titz_(surname)

So people with the surname Titz are misogynists?

Please stop with this nonsense, Mr Titz.

(Before any mods get too trigger-happy, quoting wikipedia: "it is an altered form of the frequent masculine given name Dietrich" so it is perfectly legitimate for me to refer to Dietrich as Mr Titz, and removing it would be an insult to all those with the surname Titz)

Of course not. I have been assured by the author that the name isn’t obscene but is apparently named in my honor. That’s terrific.

To get back to computer chess, some people have compared its play to leela. Does anyone have concrete examples of this?

[Thomas Lagershausen]

To get back to computer chess, some people have compared its play to leela. Does anyone have concrete examples of this?

[d]2brrbk1/qp1n1ppp/pnp5/4p3/1PR5/PN1P1NPP/4PPB1/B1Q1R1K1 w - - 11 24 bm 24.Re4 ! followed by Rxe5 !

An older Stockfish like Cfish 12 or 13 finds this in 1 sec.

But Leela (RTX 2060S) and Stockfish Sopel (Ryzen 5600X) need more than 10 minutes.

FEN: 2brrbk1/qp1n1ppp/pnp5/4p3/1PR5/PN1P1NPP/4PPB1/B1Q1R1K1 w - - 11 24

FatTitz 1.1 avx2:
Hybrid NNUE evaluation enabled.
39/54 07:54 2.318.402.947 4.888.742 +1,20 Tc4-c2 h7-h6 Tc2-d2 Sb6-a4 Sb3-a5 Da7-b6 Dc1-c2 Db6-b5 Te1-b1 c6-c5 Sa5xb7 Lc8xb7 b4xc5 Db5xc5 Tb1xb7 Dc5xa3 Dc2-a2 Da3xa2 Td2xa2 Sa4-c5 Tb7-b1 e5-e4 d3xe4 Te8xe4 Sf3-d4 Te4-e8 La1-c3 Td8-b8 Tb1-d1 Sd7-e5 Td1-c1 Tb8-d8 e2-e3 Td8-c8 Tc1-d1 g7-g6 Lc3-a5 Se5-c4 La5-b4 Sc5-e6 Lb4xf8 Kg8xf8 Ta2xa6 Se6xd4
40/53+ 09:41 2.834.339.618 4.874.564 +1,28 Tc4-e4
40/53+ 09:56 2.908.096.230 4.875.960 +1,37 Tc4-e4
40/54 11:10 3.260.639.097 4.862.314 +1,14 Tc4-c2 h7-h6 Sb3-a5 Da7-b8 Te1-d1 Db8-d6 e2-e3 Dd6-c7 Tc2-d2 g7-g6 e3-e4 Lf8-g7 d3-d4 e5xd4 Sf3xd4 Sd7-f8 f2-f4 h6-h5 Td2-d3 Sb6-a4 Sa5-c4 c6-c5 Sc4-e3 Dc7-b6 Sd4-c2 Td8xd3 Td1xd3 Lg7xa1 Dc1xa1 c5-c4 Td3-d4 Sf8-e6 Td4xc4 Lc8-d7 Kg1-h2 Ld7-b5 Se3-d5 Db6-f2 Sd5-f6+ Kg8-f8 Sf6xe8 Lb5xc4 Se8-d6 h5-h4 Da1-h8+ Kf8-e7 e4-e5 h4xg3+ Kh2-h1

[Thomas Lagershausen]

Second example for the conversion of Stockfish to Leela:

[d]r1b1nbk1/pp4r1/3p1q2/PN1Pp1pp/4Pp1n/R4P1P/1PQBBNP1/2R3K1 w - - 1 23 bm 23.Qxc8 !

Old Stockfish makes this in a few seconds.

Leela and Stockfish Fat Titz needing some minutes.

FEN: r1b1nbk1/pp4r1/3p1q2/PN1Pp1pp/4Pp1n/R4P1P/1PQBBNP1/2R3K1 w - - 1 23

FatTitz 1.1 avx2:
Hybrid NNUE evaluation enabled.
1/1 00:00 402 402.000 +2,83 Sb5xa7
2/2 00:00 955 477.500 +6,87 Sb5xa7 Ta8xa7
3/3 00:00 1.265 632.500 +6,87 Sb5xa7 Ta8xa7 Dc2xc8
4/4 00:00 3.557 1.778.500 +5,59 Sb5xa7 g5-g4 h3xg4
5/5 00:00 6.877 3.438.500 +1,75 Sb5xa7 Lc8xh3 g2xh3 Ta8xa7 Dc2-c8
6/6 00:00 10.460 5.230.000 +1,88 Sb5xa7 Lc8-d7 Sa7-b5 Ta8-c8
7/7 00:00 16.624 5.541.333 +0,49 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 g5-g4 f3xg4 h5xg4 h3xg4
8/11 00:00 20.498 5.124.500 +1,46 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7
9/11 00:00 60.014 5.455.818 +1,45 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 g5-g4 h3xg4
10/11 00:00 104.796 6.164.470 +1,55 Ta3-c3 Lc8-d7 Sb5-c7 g5-g4 h3xg4 Se8xc7 Tc3xc7 h5xg4 Sf2xg4 Sh4xf3+ g2xf3
11/14 00:00 140.682 6.116.608 +1,60 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 g5-g4 h3xg4 Df6-g5 Ld2-e1 h5xg4 Sf2xg4 Ld7xg4
12/16 00:00 165.946 6.146.148 +1,65 Ta3-c3 g5-g4 h3xg4 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Df6-g5 Ld2-e1 h5xg4 Sf2xg4 Ld7xg4
13/21 00:00 393.106 6.239.777 +1,50 Ta3-c3 g5-g4 h3xg4 h5xg4 Sf2xg4 Lc8xg4 f3xg4 Df6-g6 Tc3-h3 Lf8-e7 Ld2-e1 a7-a6 Le1xh4 Le7xh4 Th3xh4 a6xb5
14/26 00:00 523.787 6.162.200 +1,86 Ta3-c3 g5-g4 f3xg4 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 f4-f3 Le2xf3 Sh4xf3+ g2xf3 Df6xf3 Dc2-c3 h5xg4
15/26 00:00 1.083.419 6.121.011 +2,18 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 g5-g4 f3xg4 f4-f3 Le2xf3 Sh4xf3+ g2xf3 Df6xf3 Dc2-c3 h5xg4 Dc3xf3 g4xf3+ Sf2-g4
16/28 00:00 1.936.230 6.225.819 +2,25 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 g5-g4 h3xg4 Df6-g5 Tc7xb7 Ta8-c8 Tb7xd7 Tc8xc2 Td7xg7+ Kg8xg7 Tc1xc2 h5xg4 Tc2-c7+ Kg7-g6 Sf2xg4 Dg5-d8
17/28 00:00 2.981.007 6.010.094 +2,24 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-b3 Df6-d8 Tc1-c2 Sh4-g6 Ld2-e1 Sg6-h4 Db3-b4 Ld7-e8 Tc7xg7+ Lf8xg7 Db4-a3 Lg7-f6 a5-a6 b7xa6 Da3xa6
18/27 00:00 3.762.342 5.962.507 +2,23 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-b3 Df6-d8 Tc1-c2 Sh4-g6 Db3-a3 Ld7-e8 Tc7xg7+ Kg8xg7 a5-a6 Lf8-e7 a6xb7 Tb8xb7 Ld2-a5 Dd8-b8 b2-b4 Le8-b5 Le2xb5 Tb7xb5 Da3-c1 Tb5-b7
19/30 00:00 4.710.189 5.932.227 +2,35 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-b3 Df6-d8 Tc1-c2 Kg8-h7 Ld2-b4 Sh4-g6 Db3-a3 Ld7-e8 Tc7xg7+ Kh7xg7 a5-a6 Lf8-e7 a6xb7 Tb8xb7 Le2-a6 Tb7-b6
20/35 00:00 5.829.391 5.900.193 +2,24 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Tc1-c2 Kg8-h7 Ld2-e1 Sh4-g6 Dc3-b3 Ld7-e8 Db3-c4 a7-a6 Dc4-c3 Le8-a4 Tc2-c1 La4-e8 Le1-d2 Kh7-g8 Tc7xg7+ Kg8xg7 Dc3-c7+ Lf8-e7
21/30 00:01 8.091.304 5.812.718 +2,17 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-b3 Df6-d8 Tc1-c2 Sh4-g6 Db3-c3 Kg8-h7 Ld2-e1 Sg6-h8 a5-a6 b7xa6 Dc3-a3 Ld7-b5 Le1-a5 Dd8-e8 Le2xb5 a6xb5 La5-b4 Sh8-f7 Tc7xa7 Tb8-c8 Tc2-c7 Tc8xc7
22/32 00:01 11.487.687 5.772.707 +2,12 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-b3 Df6-d8 Tc1-c2 Sh4-g6 Db3-c3 Kg8-h7 Ld2-e1 Sg6-h8 Dc3-b4 Sh8-g6 Db4-a3 Ld7-e8 Tc7xg7+ Kh7xg7 a5-a6 Dd8-d7 Sf2-d3 Lf8-e7 Sd3-b4 Le7-d8 Le1-f2 b7xa6 Sb4-c6
23/34 00:02 16.296.217 5.772.659 +2,35 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Tc1-c2 Kg8-h7 Ld2-e1 Lf8-e7 Dc3-b4 Le7-f8 Db4-b3 Sh4-g6 Db3-a3 Tb8-c8 Tc7xc8 Ld7xc8 a5-a6 b7xa6 Le1-a5 Dd8-e8
24/29 00:03 19.178.251 5.735.122 +2,31 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Tc1-c2 Sh4-g6 Ld2-e1 Sg6-h8 Dc3-a3 Tb8-c8 Tc7xc8 Ld7xc8 a5-a6 b7xa6 Le1-a5 Dd8-e8 La5-b4 De8-d8 Kg1-f1 g5-g4 Lb4-a5 Dd8-e8 f3xg4
25/37- 00:03 22.391.364 5.703.353 +2,23 Ta3-c3 Lc8-d7
25/37+ 00:03 22.755.363 5.701.669 +2,31 Ta3-c3
25/37 00:04 26.133.141 5.694.735 +2,33 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-g6 Tc1-c2 Dg6-e8 Dc3-a3 Tb8-c8 Da3-b3 Tc8-b8 Db3-c3 b7-b5 Dc3-a3 a7-a6 Ld2-b4 De8-d8 Tc7-a7 Tb8-c8 Tc2xc8 Ld7xc8 Ta7-a8 Tg7-c7
26/40- 00:05 30.547.540 5.692.795 +2,25 Ta3-c3 Lc8-d7
26/40 00:05 34.113.507 5.686.532 +2,20 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Tc1-c2 Tg7-f7 Ld2-e1 Sh4-g6 Sf2-d3 Sg6-h4 Dc3-b4 Tf7-g7 Le1-f2 Sh4xg2 Tc7xd7 Dd8xd7 Kg1xg2 Tb8-c8 Tc2-c3 g5-g4 h3xg4 Tc8xc3 Db4xc3 h5xg4 f3xg4 Tg7xg4+ Le2xg4 Dd7xg4+ Kg2-f1
27/37+ 00:07 41.364.033 5.682.653 +2,28 Ta3-c3
27/44 00:08 49.990.874 5.690.480 +2,27 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-b3 Df6-d8 Db3-c3 Lf8-e7 Dc3-b4 Le7-f8 Ld2-e1 Sh4-g6 Db4-c3 Sg6-h8 Tc1-c2 Sh8-f7 a5-a6 b7xa6 Tc7xa7 g5-g4 f3xg4 h5xg4 Le2xg4 Ld7xg4 Sf2xg4 Sf7-g5 Ta7xg7+ Lf8xg7 Dc3-d3 Dd8-b6+ Kg1-h2
28/42+ 00:10 60.815.287 5.669.894 +2,36 Ta3-c3
28/42 00:10 61.640.339 5.667.555 +2,36 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Tc1-c2 Tg7-f7 Ld2-e1 Tf7-g7 Dc3-a3 Tb8-c8 Tc7xc8 Ld7xc8 a5-a6 b7xa6 Le1-a5 Dd8-e8 La5-b4 g5-g4 f3xg4 De8-d8 Le2xa6 Lc8xa6 Da3xa6 Dd8-b8 Lb4-a3 Kg8-h7 Tc2-c8 Db8-b3 Tc8-c3 Db3-a2 Da6-c4
29/35- 00:11 66.694.847 5.652.584 +2,27 Ta3-c3 Lc8-d7
29/38+ 00:12 68.106.861 5.652.021 +2,36 Ta3-c3
29/39 00:12 69.689.408 5.653.395 +2,28 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Tg7-f7 Tc1-c2 Tf7-g7 Dc3-b3 Kg8-h7 Db3-b4 Kh7-g8 Db4-c3 Dd8-e8 Dc3-a3 De8-g6 Tc2-c3 Kg8-h7 Da3-b4 Dg6-e8 Tc3-c2 De8-d8 a5-a6 b7xa6 Db4-a5 g5-g4 h3xg4 Tb8-b3
30/41 00:15 89.040.049 5.635.089 +2,34 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Tg7-f7 Tc1-c2 Tf7-g7 Dc3-b3 Kg8-h7 Db3-b4 Kh7-g8 Db4-c3 Dd8-e8 Dc3-a3 De8-g6 Da3-d3 Ld7-a4 Tc7xg7+ Dg6xg7 Tc2-c1 La4-e8 Dd3-c3 Dg7-d7 Sf2-d3 Lf8-e7 Le1-f2 a7-a6
31/40- 00:17 96.702.436 5.612.445 +2,25 Ta3-c3 Lc8-d7
31/40+ 00:17 97.999.894 5.610.895 +2,34 Ta3-c3
31/40 00:19 109.640.446 5.618.265 +2,34 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Lf8-e7 Dc3-b4 Le7-f8 Tc1-c2 Sh4-g6 Db4-c3 Sg6-h8 Dc3-a3 Tb8-c8 Tc7xc8 Ld7xc8 a5-a6 b7xa6 Le2xa6 Lc8-d7 La6-e2 Dd8-b8 Da3-b4 Db8-d8 Kg1-f1 Lf8-e7 Db4-b7 g5-g4 f3xg4 Le7-h4 g4xh5
32/43 00:24 135.136.785 5.615.957 +2,27 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Lf8-e7 Dc3-b4 Le7-f8 Tc1-c2 Kg8-h7 Db4-c3 Dd8-e8 Dc3-d2 De8-d8 Dd2-b4 Dd8-e8 Db4-c3 Kh7-g8 Dc3-a3 De8-g6 Da3-d3 Dg6-f6 a5-a6 b7xa6 Dd3xa6 g5-g4 h3xg4 h5xg4 Sf2xg4 Ld7xg4 Tc7xg7+ Lf8xg7 f3xg4 f4-f3
33/44+ 00:28 157.045.427 5.607.163 +2,36 Ta3-c3
33/44 00:31 177.533.833 5.614.250 +2,39 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Kg8-h7 Tc1-c2 Lf8-e7 Dc3-b4 Le7-f8 Db4-b3 Dd8-e8 Db3-c3 Kh7-g8 Dc3-a3 De8-g6 Da3-b4 Dg6-e8 Sf2-d3 De8-d8 Db4-c3 Lf8-e7 Le1-f2 Dd8-e8 Lf2xa7 Tb8-a8 Tc7xb7 g5-g4 f3xg4 h5xg4 Le2xg4 Ld7xg4 h3xg4 Tg7xg4 Sd3-e1
34/43- 00:36 202.182.372 5.593.647 +2,31 Ta3-c3 Lc8-d7
34/43 00:37 211.378.633 5.590.991 +2,31 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Kg8-h7 Tc1-c2 Dd8-e8 Dc3-b4 De8-d8 Db4-b3 Kh7-g8 Db3-c3 Lf8-e7 Dc3-b4 Le7-f8 a5-a6 b7xa6 Db4-a3 Tg7-f7 Tc7xa7 Ld7-b5 Ta7xf7 Kg8xf7 Le2xb5 a6xb5 Le1-a5 Dd8-e8 Tc2-c7+ Kf7-g8 Da3-c3 Sh4-g6 Kg1-f1 Tb8-a8 b2-b4 Sg6-e7 Tc7-b7
35/43+ 00:45 253.493.985 5.565.913 +2,39 Ta3-c3
35/48 00:51 283.664.450 5.550.728 +2,35 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Tc1-c2 Lf8-e7 Dc3-b4 Le7-f8 Ld2-e1 Ld7-e8 Db4-c4 Tg7xc7 Dc4xc7 Lf8-e7 Dc7xd8 Le7xd8 Le1-b4 Ld8-e7 Tc2-c7 Le7-d8 Lb4xd6 Ld8xc7 Ld6xc7 Tb8-c8 d5-d6 b7-b5 a5xb6/ep a7xb6 Le2-c4+ Kg8-g7 Sf2-d3 Tc8-a8 Sd3xe5 Ta8-a1+ Kg1-f2 Ta1-c1 Lc7xb6 Tc1-c2+ Kf2-f1 Tc2-d2 Kf1-e1 Td2xb2
36/49+ 01:27 482.452.736 5.486.407 +2,43 Ta3-c3
36/52 01:28 484.047.738 5.485.208 +2,33 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Lf8-e7 Dc3-b4 Le7-f8 Tc1-c2 Ld7-e8 Db4-c3 Tg7xc7 Dc3xc7 Lf8-e7 Dc7xd8 Tb8xd8 Sf2-d3 Kg8-f8 Le1-b4 Sh4-g6 b2-b3 Td8-b8 Tc2-c7 a7-a6 Sd3-b2 Sg6-h8 Sb2-c4 Sh8-f7 Kg1-f1 Le7-d8 Tc7xf7+ Kf8xf7 Sc4xd6+ Kf7-f8 Sd6-c4+ Kf8-g7 Sc4xe5 b7-b6 a5xb6 Tb8xb6 Lb4-a5 Tb6-b8 La5xd8 Tb8xd8
37/54- 01:29 490.077.205 5.485.222 +2,25 Ta3-c3 Lc8-d7
37/54+ 01:31 503.421.917 5.481.510 +2,33 Ta3-c3
37/54 01:46 582.349.474 5.467.762 +2,37 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Lf8-e7 Dc3-b4 Le7-f8 Tc1-c2 Tg7-f7 Db4-c3 Lf8-e7 Dc3-a3 Le7-f8 Da3-b4 Ld7-e8 Db4-c3 Le8-d7 Sf2-d3 g5-g4 f3xg4 f4-f3 Le1xh4 Dd8xh4 Le2xf3 Lf8-h6 Dc3-e1 Dh4xe1+ Sd3xe1 Lh6-g5 Tc7-c3 Lg5-h4 Se1-d3 h5xg4 Lf3xg4 Ld7xg4 h3xg4 Tb8-f8 Tc3-c8
38/49- 01:51 611.601.132 5.464.139 +2,28 Ta3-c3 Lc8-d7
38/49 01:52 613.451.773 5.464.367 +2,31 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Lf8-e7 Dc3-b4 Le7-f8 Tc1-c2 Sh4-g6 Db4-c3 Sg6-h8 a5-a6 b7xa6 Dc3-a5 Sh8-f7 Da5xa6 g5-g4 f3xg4 Sf7-g5 h3-h4 f4-f3 g2xf3 Sg5-h7 Sf2-h1 h5xg4 f3xg4 Sh7-f6 Sh1-f2 Sf6-h7 Sf2-d1 Sh7-f6 Sd1-e3 Sf6xe4 Da6xa7 Tb8-a8 Da7-b6 Ta8-a1
39/51 02:09 706.401.594 5.445.209 +2,36 Ta3-c3 Lc8-d7 Sb5-c7 Se8xc7 Tc3xc7 Ta8-b8 Dc2-c3 Df6-d8 Ld2-e1 Lf8-e7 Dc3-b4 Le7-f8 Tc1-c2 Kg8-h7 Db4-b3 Kh7-g8 Db3-a3 Tb8-c8 Tc7xc8 Ld7xc8 a5-a6 b7xa6 Le1-a5 Dd8-e8 La5-b4 De8-d8 Le2xa6 Lc8xa6 Da3xa6 Dd8-b8 Lb4-a3 Db8-b7 Tc2-c6 Db7xa6 Tc6xa6 Tg7-c7 Ta6-c6 Tc7-b7 Kg1-f1 a7-a5 Sf2-d1 g5-g4 f3xg4 h5xg4 h3xg4 Kg8-g7 La3xd6 Lf8xd6 Tc6xd6
40/53- 02:36 844.248.078 5.398.109 +2,28 Ta3-c3 Lc8-d7
40/53+ 03:40 1.167.714.992 5.304.612 +2,36 Dc2xc8
40/53+ 05:09 1.636.093.450 5.285.854 +3,24 Dc2xc8
40/62 05:26 1.726.266.538 5.280.830 +3,26 Dc2xc8 Ta8xc8 Tc1xc8 g5-g4 h3xg4 h5xg4 f3xg4 Df6-g6 Ta3-h3 a7-a6 Sb5-c3 Tg7-h7 g2-g3 f4xg3 Th3xg3 Se8-g7 g4-g5 Sg7-h5 Le2xh5 Th7xh5 Sf2-g4 Kg8-f7 Kg1-f2 Th5xg5 Sg4xe5+ Tg5xe5 Tg3xg6 Sh4xg6 Tc8-d8 Te5-h5 Kf2-g2 Th5-h7 Sc3-e2 Th7-h4 Se2-g3 Th4-g4 Td8-b8 Lf8-e7 Tb8xb7 Kf7-e8 b2-b4 Le7-h4 Ld2-e1 Lh4xg3 Le1xg3 Tg4xe4 b4-b5 a6xb5 a5-a6 Te4-a4 a6-a7 Sg6-e7 Lg3xd6 Se7xd5

[dkappe]

Thomas,

Thanks for these. I wonder if the “buckets” and a search going over one or more bucket boundaries might not skew the search in some cases.

[Sopel]

Thomas,

Thanks for these. I wonder if the “buckets” and a search going over one or more bucket boundaries might not skew the search in some cases.

This is definitely a possibility we've considered with Stockfish. There is a "virtual layer" that is a term in all of those buckets, and the purpose of it is to ensure that the discrepancies are as small as possible, but there still will be some. Empirical evidence with Stockfish shows that the gains are worth it. We've tried doing a weighted sum over neighbouring buckets during the training to reduce the discrepancies further but this made the nets worse.

[Eduard]

E N,Fat Titz 310821 64 - Victorious,Stockfish 310821 0-1 0.0, Rated game, 5 min 2021
Position from my game today:
[d]3k1r1r/3bqp1n/1p1p1np1/pPpPp2p/K1P1P3/PNN1P2P/4BQPR/5R2 b - - 0 1

After move 33. What great knowledge does the engine have when it plays that?

[supersharp77]

The previous thread has been closed for an unknown reason (perhaps for off-topic unrelated to the engine?). As far as I know this was the first engine to have its thread closed, and I feel discriminated.

Fat Titz 1.1


Note: The new nodes speedup formula with 3 fundamental constants by Additional Pylon aka ChickenLogic was unfortunately lost due to a github force-push malpractice. Stockfish should be thankful for this loss as otherwise the engine would just be unbeatable.[/size]

New binaries available here: https://drive.google.com/drive/folders/ ... sp=sharing. The _bin zip doesn't contain the net, as it has not changed.

Fat Titz 1.0

This is a FREE UCI-compatibile chess engine. It is a fork of cfish.

Thanks to Norman Schidt and Albert Sliver for inspiration on creating my own fork of Stockfish. Also big thanks to Robert Houdart for great optimization ideas, overall making the engine almost 20% faster!

Thx (Herr) Sopel... (VIP Cloner Of Clones Of A Clone)....

Hahahahahaha........LMAO....LooL.....Absolutely Hilarious Mr Sopel....Your Hits Keep On Coming