Live Rapid Chess Engine Tournament (TC=15m+10s)

Posts : 880 Join date : 2020-11-25 Location : USA

Hardware: Threadripper 2950x, RTX 2080TI, 64 GBs Ram.

Tablebases = 7 man top 10.
Book 8 move EPD
GUI = Banksai
Hash = 8 GB.
16 Threads.
TC = 15m+10s
Format is all play all RR.
Default setting for all engines.

Players

1. Stockfish 110922
2. Dragon 3.1
3. Flaming Titz 230822
4. Koivisto 8.0
5. Rebel 15.1a
6. Berserk 9
7. Slowchess Blitz 2.9
8. Rubichess 20220813

Posts : 880 Join date : 2020-11-25 Location : USA

Code:: Result:
----------------------------------------------------------------------------------
  # name games wins draws losses score% elo + -
  1. Stockfish 110922 12 4 8 0 66.7 72 125 118
  2. Flaming Titz 230822 15 2 13 0 56.7 23 104 102
  3. Berserk 9 12 2 9 1 54.2 14 118 117
  4. RubiChess 20220813 14 1 12 1 50.0 2 107 107
  5. Dragon 3.1 12 0 11 1 45.8 -14 112 114
  6. Koivisto 8.0 12 0 11 1 45.8 -24 112 114
  7. SlowChess Blitz 2.9 13 0 11 2 42.3 -29 109 112
  8. Rebel 15.1a 14 0 11 3 39.3 -45 107 111

Cross table:
----------------------------------------------------------------------------------
  # name score% games 1 2 3 4 5 6 7 8
  1. Stockfish 110922 66.7 12 x ==---- =1---- =1---- =1---- =1---- ------ ==----
  2. Flaming Titz 230822 56.7 15 ==---- x ------ ==---- ===--- ==---- 1===-- =1----
  3. Berserk 9 54.2 12 =0---- ------ x ==---- =----- =----- =1=--- ==1---
  4. RubiChess 20220813 50.0 14 =0---- ==---- ==---- x =----- ===--- ==---- 1=----
  5. Dragon 3.1 45.8 12 =0---- ===--- =----- =----- x =----- ===--- =-----
  6. Koivisto 8.0 45.8 12 =0---- ==---- =----- ===--- =----- x ------ ===---
  7. SlowChess Blitz 2.9 42.3 13 ------ 0===-- =0=--- ==---- ===--- ------ x =-----
  8. Rebel 15.1a 39.3 14 ==---- =0---- ==0--- 0=---- =----- ===--- =----- x

Tech:
----------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  # name nodes/m NPS depth/m time/m moves time
  1. Stockfish 110922 345498K 13153084 47.0 26.3 50.8 1335.3
  2. Flaming Titz 230822 224947K 9952418 56.0 22.6 62.9 1420.9
  3. Berserk 9 476396K 22152184 33.9 21.5 60.8 1308.3
  4. RubiChess 20220813 478636K 21178037 46.7 22.6 68.2 1541.7
  5. Dragon 3.1 366284K 17242393 48.4 21.2 72.0 1529.5
  6. Koivisto 8.0 423744K 18337048 38.3 23.1 59.9 1384.6
  7. SlowChess Blitz 2.9 160650K 8173122 31.2 19.7 73.3 1440.9
  8. Rebel 15.1a 165060K 8000944 31.5 20.6 63.1 1302.6
   all --- 315043K 14655626 41.7 22.0 64.0 1409.2

It looks like Stockfish is now aware of the scaling issues. And the latest patches are trying to fix this issue.

Author: FauziAkram
Date: Wed Sep 7 07:38:04 2022 +0200
Timestamp: 1662529084

VLTC tuning

Tuning some parameters that scale well with longer time control:

Failed STC:
https://tests.stockfishchess.org/tests/view/6313424d8202a039920e130a
LLR: -2.94 (-2.94,2.94) <-1.75,0.25>
Total: 42680 W: 11231 L: 11540 D: 19909 Elo -2.52
Ptnml(0-2): 191, 5008, 11232, 4737, 172

Passed LTC:
https://tests.stockfishchess.org/tests/view/6311e2cd874169ca52ae7933
LLR: 2.94 (-2.94,2.94) <0.50,2.50>
Total: 53448 W: 14782 L: 14437 D: 24229 Elo +2.24
Ptnml(0-2): 101, 5214, 15740, 5577, 92

Passed VLTC:
https://tests.stockfishchess.org/tests/view/6312530cfa99a92e3002c927
LLR: 2.95 (-2.94,2.94) <0.50,2.50>
Total: 123336 W: 33465 L: 33007 D: 56864 Elo +1.29
Ptnml(0-2): 38, 11466, 38204, 11920, 40

closes https://github.com/official-stockfish/Stockfish/pull/4154

Bench: 5609606

Posts : 880 Join date : 2020-11-25 Location : USA

Flaming Titz is having a great result in the first test. Now at +10 wins in 38 games.

Code:: Result:
----------------------------------------------------------------------------------
  # name games wins draws losses score% elo + -
  1. Flaming Titz 230822 38 10 28 0 63.2 59 71 69
  2. Stockfish 110922 35 7 28 0 60.0 47 72 71
  3. Dragon 3.1 35 5 29 1 55.7 32 71 70
  4. Berserk 9 33 4 27 2 53.0 15 73 73
  5. SlowChess Blitz 2.9 38 2 33 3 48.7 -5 67 67
  6. RubiChess 20220813 32 1 26 5 43.8 -22 74 75
  7. Koivisto 8.0 35 0 31 4 44.3 -25 69 70
  8. Rebel 15.1a 34 0 20 14 29.4 -102 76 80

Cross table:
----------------------------------------------------------------------------------
  # name score% games 1 2 3 4 5 6 7 8
  1. Flaming Titz 230822 63.2 38 x =====- =====- ==1==- 1===== ==1=1- ==1==1 =1=111
  2. Stockfish 110922 60.0 35 =====- x =1==== =1===- ====1- =1===- =1===- ==11--
  3. Dragon 3.1 55.7 35 =====- =0==== x ====-- ====== ==1=1- ===1=- =1=1--
  4. Berserk 9 53.0 33 ==0==- =0===- ====-- x =1===- ====-- ====-- ==111=
  5. SlowChess Blitz 2.9 48.7 38 0===== ====0- ====== =0===- x =====- ====== =1==1-
  6. RubiChess 20220813 43.8 32 ==0=0- =0===- ==0=0- ====-- =====- x ====-- 1===--
  7. Koivisto 8.0 44.3 35 ==0==0 =0===- ===0=- ====-- ====== ====-- x =====-
  8. Rebel 15.1a 29.4 34 =0=000 ==00-- =0=0-- ==000= =0==0- 0===-- =====- x

Tech:
----------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  # name nodes/m NPS depth/m time/m moves time
  1. Flaming Titz 230822 227549K 10514814 52.2 21.6 68.2 1475.6
  2. Stockfish 110922 327667K 14498617 54.9 22.6 63.8 1441.9
  3. Dragon 3.1 361874K 17731000 46.1 20.4 79.7 1626.3
  4. Berserk 9 463158K 22231889 34.0 20.8 68.1 1417.9
  5. SlowChess Blitz 2.9 173927K 8691947 32.1 20.0 68.6 1373.3
  6. RubiChess 20220813 466378K 20245861 45.0 23.0 62.8 1446.2
  7. Koivisto 8.0 396307K 18488317 38.9 21.4 69.9 1498.6
  8. Rebel 15.1a 164336K 8063918 28.7 20.4 65.4 1333.6
   all --- 310938K 14997754 41.6 21.2 68.4 1451.8

Posts : 1254 Join date : 2020-11-17 Location : France

mwyoung wrote:

Flaming Titz is having a great result in the first test. Now at +10 wins in 38 games.

Code:: Result:
----------------------------------------------------------------------------------
  # name games wins draws losses score% elo + -
  1. Flaming Titz 230822 38 10 28 0 63.2 59 71 69
  2. Stockfish 110922 35 7 28 0 60.0 47 72 71
  3. Dragon 3.1 35 5 29 1 55.7 32 71 70
  4. Berserk 9 33 4 27 2 53.0 15 73 73
  5. SlowChess Blitz 2.9 38 2 33 3 48.7 -5 67 67
  6. RubiChess 20220813 32 1 26 5 43.8 -22 74 75
  7. Koivisto 8.0 35 0 31 4 44.3 -25 69 70
  8. Rebel 15.1a 34 0 20 14 29.4 -102 76 80

Cross table:
----------------------------------------------------------------------------------
  # name score% games 1 2 3 4 5 6 7 8
  1. Flaming Titz 230822 63.2 38 x =====- =====- ==1==- 1===== ==1=1- ==1==1 =1=111
  2. Stockfish 110922 60.0 35 =====- x =1==== =1===- ====1- =1===- =1===- ==11--
  3. Dragon 3.1 55.7 35 =====- =0==== x ====-- ====== ==1=1- ===1=- =1=1--
  4. Berserk 9 53.0 33 ==0==- =0===- ====-- x =1===- ====-- ====-- ==111=
  5. SlowChess Blitz 2.9 48.7 38 0===== ====0- ====== =0===- x =====- ====== =1==1-
  6. RubiChess 20220813 43.8 32 ==0=0- =0===- ==0=0- ====-- =====- x ====-- 1===--
  7. Koivisto 8.0 44.3 35 ==0==0 =0===- ===0=- ====-- ====== ====-- x =====-
  8. Rebel 15.1a 29.4 34 =0=000 ==00-- =0=0-- ==000= =0==0- 0===-- =====- x

Tech:
----------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  # name nodes/m NPS depth/m time/m moves time
  1. Flaming Titz 230822 227549K 10514814 52.2 21.6 68.2 1475.6
  2. Stockfish 110922 327667K 14498617 54.9 22.6 63.8 1441.9
  3. Dragon 3.1 361874K 17731000 46.1 20.4 79.7 1626.3
  4. Berserk 9 463158K 22231889 34.0 20.8 68.1 1417.9
  5. SlowChess Blitz 2.9 173927K 8691947 32.1 20.0 68.6 1373.3
  6. RubiChess 20220813 466378K 20245861 45.0 23.0 62.8 1446.2
  7. Koivisto 8.0 396307K 18488317 38.9 21.4 69.9 1498.6
  8. Rebel 15.1a 164336K 8063918 28.7 20.4 65.4 1333.6
   all --- 310938K 14997754 41.6 21.2 68.4 1451.8

From those node rates, I would guess several threads per engine?

Posts : 880 Join date : 2020-11-25 Location : USA

Chris Whittington wrote:

mwyoung wrote:

Flaming Titz is having a great result in the first test. Now at +10 wins in 38 games.

Code:: Result:
----------------------------------------------------------------------------------
  # name games wins draws losses score% elo + -
  1. Flaming Titz 230822 38 10 28 0 63.2 59 71 69
  2. Stockfish 110922 35 7 28 0 60.0 47 72 71
  3. Dragon 3.1 35 5 29 1 55.7 32 71 70
  4. Berserk 9 33 4 27 2 53.0 15 73 73
  5. SlowChess Blitz 2.9 38 2 33 3 48.7 -5 67 67
  6. RubiChess 20220813 32 1 26 5 43.8 -22 74 75
  7. Koivisto 8.0 35 0 31 4 44.3 -25 69 70
  8. Rebel 15.1a 34 0 20 14 29.4 -102 76 80

Cross table:
----------------------------------------------------------------------------------
  # name score% games 1 2 3 4 5 6 7 8
  1. Flaming Titz 230822 63.2 38 x =====- =====- ==1==- 1===== ==1=1- ==1==1 =1=111
  2. Stockfish 110922 60.0 35 =====- x =1==== =1===- ====1- =1===- =1===- ==11--
  3. Dragon 3.1 55.7 35 =====- =0==== x ====-- ====== ==1=1- ===1=- =1=1--
  4. Berserk 9 53.0 33 ==0==- =0===- ====-- x =1===- ====-- ====-- ==111=
  5. SlowChess Blitz 2.9 48.7 38 0===== ====0- ====== =0===- x =====- ====== =1==1-
  6. RubiChess 20220813 43.8 32 ==0=0- =0===- ==0=0- ====-- =====- x ====-- 1===--
  7. Koivisto 8.0 44.3 35 ==0==0 =0===- ===0=- ====-- ====== ====-- x =====-
  8. Rebel 15.1a 29.4 34 =0=000 ==00-- =0=0-- ==000= =0==0- 0===-- =====- x

Tech:
----------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  # name nodes/m NPS depth/m time/m moves time
  1. Flaming Titz 230822 227549K 10514814 52.2 21.6 68.2 1475.6
  2. Stockfish 110922 327667K 14498617 54.9 22.6 63.8 1441.9
  3. Dragon 3.1 361874K 17731000 46.1 20.4 79.7 1626.3
  4. Berserk 9 463158K 22231889 34.0 20.8 68.1 1417.9
  5. SlowChess Blitz 2.9 173927K 8691947 32.1 20.0 68.6 1373.3
  6. RubiChess 20220813 466378K 20245861 45.0 23.0 62.8 1446.2
  7. Koivisto 8.0 396307K 18488317 38.9 21.4 69.9 1498.6
  8. Rebel 15.1a 164336K 8063918 28.7 20.4 65.4 1333.6
   all --- 310938K 14997754 41.6 21.2 68.4 1451.8

From those node rates, I would guess several threads per engine?

Yes the hardware and setup was in the notes above.

Hardware: Threadripper 2950x, RTX 2080TI, 64 GBs Ram.

Tablebases = 7 man top 10.
Book 8 move EPD
GUI = Banksai
Hash = 8 GB.
16 Threads.
TC = 15m+10s
Format is all play all RR.
Default setting for all engines.

It looks like the Stockfish scaling issue has bitten Stockfish in the ass in this tournament.

Here is Flaming Titz release notes. It said it should have better scaling. And it seems that it does indeed have better scaling then Stockfish.

Flaming Titz 230822

In January 2022, Stockfish developer Sopel released a Fat Titz 2. The engine is based on Cfish and the network is the largest of all NNUE networks, with a size of 180 MB! I carefully studied the gaming style of Fat Titz 2. Also with interesting positions from new Stockfish games.

What's new in Flaming Titz?

I adjusted the scaling, I think Flaming Titz scales better. Tactical power has increased, the engine plays more dynamically, but with the same good positional qualities! The engine gets deep quickly, despite 180 MB NNUE. I tested Flaming Titz for a few hours on PlayChess (with 8 cores and a universal book). The engine plays great, I had a lot of fun seeing how it was easily held against 64 cores.

Download Flaming Titz 230822 AVX2 and BMI2, and source code:

https://filehorst.de/d/etqfFknf

Flaming Titz 230822-sse41-popcnt.zip:

https://filehorst.de/d/eAzAzssa

Posts : 880 Join date : 2020-11-25 Location : USA

Flaming Titz won this tournament with a +11 win result in 42 games. Stockfish and Dragon both had a +7 result.

This gives Flaming Titz a +24 Elo advantage over both Stockfish and Dragon in the first Tournament. The Stage 2 tournament is next.

Code:: Result:
----------------------------------------------------------------------------------
  # name games wins draws losses score% elo + -
  1. Flaming Titz 230822 42 11 31 0 63.1 62 68 66
  2. Stockfish 110922 42 7 35 0 58.3 38 65 64
  3. Dragon 3.1 42 8 33 1 58.3 38 66 65
  4. Berserk 9 42 4 34 4 50.0 2 65 65
  5. SlowChess Blitz 2.9 42 2 37 3 48.8 -4 64 64
  6. Koivisto 8.0 42 1 37 4 46.4 -16 64 64
  7. RubiChess 20220813 42 2 34 6 45.2 -21 65 65
  8. Rebel 15.1a 42 0 25 17 29.8 -99 68 72

Cross table:
----------------------------------------------------------------------------------
  # name score% games 1 2 3 4 5 6 7 8
  1. Flaming Titz 230822 63.1 42 x ====== ====== ==1=== 1===== ==1==1 ==1=11 =1=111
  2. Stockfish 110922 58.3 42 ====== x =1==== =1==== ====1= =1==== =1==== ==11==
  3. Dragon 3.1 58.3 42 ====== =0==== x ====11 ====== ===1== ==1=1= =1=11=
  4. Berserk 9 50.0 42 ==0=== =0==== ====00 x =1==== ====== ====== ==111=
  5. SlowChess Blitz 2.9 48.8 42 0===== ====0= ====== =0==== x ====== ====== =1==1=
  6. Koivisto 8.0 46.4 42 ==0==0 =0==== ===0== ====== ====== x ====== =====1
  7. RubiChess 20220813 45.2 42 ==0=00 =0==== ==0=0= ====== ====== ====== x 1====1
  8. Rebel 15.1a 29.8 42 =0=000 ==00== =0=00= ==000= =0==0= =====0 0====0 x

Tech:
----------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  # name nodes/m NPS depth/m time/m moves time
  1. Flaming Titz 230822 235051K 10861835 53.7 21.6 68.7 1486.0
  2. Stockfish 110922 335232K 14479563 55.3 23.2 60.8 1407.9
  3. Dragon 3.1 361069K 17311313 45.4 20.9 76.0 1585.7
  4. Berserk 9 468954K 21994269 33.8 21.3 66.6 1419.4
  5. SlowChess Blitz 2.9 174150K 8757374 32.3 19.9 69.7 1385.4
  6. Koivisto 8.0 402488K 18349245 38.9 21.9 67.2 1473.8
  7. RubiChess 20220813 454905K 20374741 44.3 22.3 66.6 1487.9
  8. Rebel 15.1a 161154K 8058954 28.9 20.0 68.4 1366.9
   all --- 315453K 15129841 41.4 21.4 68.0 1451.6

» Top Chess Engine Fischer Random Rapid Chess Tournament.
» Rapid Chess Engine Tournament Stage 3, Dragon 3.1 and Flaming Titz Advance.
» Live Top Chess Engine Knock Out Tournament
» Rapid Chess Engine Tournament Stage 2, Stockfish, Dragon 3.1, Flaming Titz, and Berserk 9 Advance.
» Live Top Chess Engine Tournament (3m+2s) (1 Core) (Ponder On)