Page 1 of 2
CERL Games
Posted: Fri May 19, 2023 10:17 pm
by FritzUser
The Chess Engine Rating List is located at:
https://chessnerd.net/cerl.txt and is really just an excuse to use all the engines in my UCI folder. In turn, running them at hyper-fast speeds serves to cut down on draws, and gives enough results that it's possible to do the list at all, even with just two computers. Having this many games per tournament is meant to reduce the error rate.
Code: Select all
Rank Name TC Elo +/- Games Wins Losses Draws Points Score Draw
1 Berserk 11.1 1+0.1 238 5 12000 7536 408 4056 9564.0 79.7% 33.8%
2 Clover 4.1 1+0.1 60 5 12000 4336 2276 5388 7030.0 58.6% 44.9%
3 Caissa 1.8 1+0.1 -85 5 12000 1993 4886 5121 4553.5 37.9% 42.7%
4 Velvet 5.2.0 1+0.1 -202 5 12000 974 7269 3757 2852.5 23.8% 31.3%
24000 of 24000 games finished.
Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz
Speed: 4.58 GHz
Cores: 8
Logical processors: 16
Installed RAM: 64.0 GB
CuteChess GUI 3.5.0
Hash size: 16
Threads: 1
Concurrency: 10
Ponder: off
ECO 12ply opening suite
Syzygy 3-4-5 TB adjudication
No stalls, disconnects, timeouts, or illegal moves
All 24,000 games:
https://chessnerd.net/tour/cerl-0046.pgn
Re: CERL Games
Posted: Fri May 19, 2023 10:54 pm
by janus
:greetings-clappingorange:
Love it .. thank you ..
Re: CERL Games
Posted: Sat May 20, 2023 8:29 am
by massimilianogoi
I am taking this direction too, with
this latest tournament especially. The question is: aren't the engines prone to make mistakes with such a short time? I've never used those chess programs, but Stockfish makes jumps from 0.00 to kinda -1.26 sometimes with that short time, single-cored especially.
FritzUser wrote: ↑Fri May 19, 2023 10:17 pm
The Chess Engine Rating List is located at:
https://chessnerd.net/cerl.txt and is really just an excuse to use all the engines in my UCI folder. In turn, running them at hyper-fast speeds serves to cut down on draws, and gives enough results that it's possible to do the list at all, even with just two computers. Having this many games per tournament is meant to reduce the error rate.
Code: Select all
Rank Name TC Elo +/- Games Wins Losses Draws Points Score Draw
1 Berserk 11.1 1+0.1 238 5 12000 7536 408 4056 9564.0 79.7% 33.8%
2 Clover 4.1 1+0.1 60 5 12000 4336 2276 5388 7030.0 58.6% 44.9%
3 Caissa 1.8 1+0.1 -85 5 12000 1993 4886 5121 4553.5 37.9% 42.7%
4 Velvet 5.2.0 1+0.1 -202 5 12000 974 7269 3757 2852.5 23.8% 31.3%
24000 of 24000 games finished.
Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz
Speed: 4.58 GHz
Cores: 8
Logical processors: 16
Installed RAM: 64.0 GB
CuteChess GUI 3.5.0
Hash size: 16
Threads: 1
Concurrency: 10
Ponder: off
ECO 12ply opening suite
Syzygy 3-4-5 TB adjudication
No stalls, disconnects, timeouts, or illegal moves
All 24,000 games:
https://chessnerd.net/tour/cerl-0046.pgn
Re: CERL Games
Posted: Sat May 20, 2023 2:34 pm
by FritzUser
massimilianogoi wrote: ↑Sat May 20, 2023 8:29 am
I am taking this direction too, with
this latest tournament especially. The question is: aren't the engines prone to make mistakes with such a short time? I've never used those chess programs, but Stockfish makes jumps from 0.00 to kinda -1.26 sometimes with that short time, single-cored especially.
Yes. To be honest, this is the weak link in the chain. My hope is that it lowers the playing field in a consistent way, but that might be wishful thinking. I think to do it right, with proper time controls, it would take hundreds of people.
Re: CERL Games
Posted: Sat May 20, 2023 6:28 pm
by massimilianogoi
I'm not understanding... Are you telling that you wish the engines turn weaker?
FritzUser wrote: ↑Sat May 20, 2023 2:34 pm
...My hope is that it lowers the playing field in a consistent way, but that might be wishful thinking. I think to do it right, with proper time controls, it would take hundreds of people.
Re: CERL Games
Posted: Sat May 20, 2023 6:47 pm
by FritzUser
I think that there is an ideal TC (perhaps one hour per game) that allows the engines to think as much as they need to, and nobody can really test at that speed, because it's too slow. So everything is a compromise, in my opinion. You would need hundreds of people to contribute test results if you wanted a rating list based on hour-long games.
What I was meant was, I accept that testing at 1s+0.1s is going to introduce mistakes, but it still works as long as the drop in performance is the same for every engine. So my hope is that if the playing field has to be made lesser, it can at least be so consistently.
Re: CERL Games
Posted: Wed May 24, 2023 9:42 pm
by FritzUser
Code: Select all
Rank Name TC Elo +/- Games Wins Losses Draws Points Score Draw
1 Igel 3.4.0 1+0.1 216 7 8000 5042 628 2330 6207.0 77.6% 29.1%
2 Igel 3.3.0 1+0.1 165 7 8000 4468 926 2606 5771.0 72.1% 32.6%
3 Igel 3.2.0 1+0.1 106 6 8000 3833 1467 2700 5183.0 64.8% 33.8%
4 Igel 3.1.0 1+0.1 67 6 8000 3419 1906 2675 4756.5 59.5% 33.4%
5 Igel 3.0.5 1+0.1 65 6 8000 3384 1901 2715 4741.5 59.3% 33.9%
6 Igel 3.0.0 1+0.1 39 6 8000 3165 2268 2567 4448.5 55.6% 32.1%
7 Seer 2.1.0 1+0.1 -58 7 8000 2605 3927 1468 3339.0 41.7% 18.4%
8 Rybka 4.1 1+0.1 -266 9 8000 1157 6309 534 1424.0 17.8% 6.7%
9 Renegade 0.10.0 1+0.1 -713 25 8000 53 7794 153 129.5 1.6% 1.9%
36000 of 36000 games finished.
Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz
Cores: 8
Logical processors: 16
Installed RAM: 64.0 GB
CuteChess GUI 3.5.0
Hash size: 128
Threads: 1
Concurrency: 8
Ponder: off
ECO Opening Suite to 8 ply
Syzygy 3-4-5 TB adjudication
All 6000 games:
https://chessnerd.net/tour/cerl-0053.pgn
Re: CERL Games
Posted: Sun Jun 25, 2023 3:08 pm
by FritzUser
Code: Select all
Rank Name TC Elo +/- Games Wins Losses Draws Points Score Draw Time Illegal Discon Stall WrClaim
1 SugaR AI SE 1+0.1 278 15 1900 1330 68 502 1581.0 83.2% 26.4% 0 0 0 0 0
2 Stockfish 15.1 1+0.1 275 16 1900 1340 87 473 1576.5 83.0% 24.9% 0 0 0 0 0
3 BrainLearn 25 1+0.1 273 15 1900 1320 75 505 1572.5 82.8% 26.6% 0 0 0 0 0
4 ShashChess 33 1+0.1 271 15 1900 1325 85 490 1570.0 82.6% 25.8% 0 0 0 0 0
5 Dragon 3.2 1+0.1 227 15 1900 1243 153 504 1495.0 78.7% 26.5% 0 0 0 0 0
6 Berserk 11.1 1+0.1 194 15 1900 1191 228 481 1431.5 75.3% 25.3% 0 0 0 0 0
7 Clover 5.0 1+0.1 146 14 1900 1086 331 483 1327.5 69.9% 25.4% 0 0 0 0 0
8 Seer 2.6 1+0.1 106 14 1900 999 437 464 1231.0 64.8% 24.4% 0 0 0 0 0
9 Caissa 1.9 1+0.1 66 14 1900 909 551 440 1129.0 59.4% 23.2% 0 0 0 0 0
10 Viridithas 10.0.0 1+0.1 59 14 1900 909 592 399 1108.5 58.3% 21.0% 0 0 0 0 0
11 Velvet 5.2.1 1+0.1 13 14 1900 812 739 349 986.5 51.9% 18.4% 0 0 0 0 0
12 Komodo 14.1 1+0.1 4 14 1900 799 779 322 960.0 50.5% 16.9% 0 0 0 0 0
13 Svart 5 1+0.1 -88 15 1900 606 1075 219 715.5 37.7% 11.5% 0 0 0 0 0
14 Monolith 2.01 1+0.1 -170 17 1900 439 1300 161 519.5 27.3% 8.5% 0 0 0 0 0
15 Marvin 5.1.0 1+0.1 -182 17 1900 403 1318 179 492.5 25.9% 9.4% 0 0 0 0 0
16 4ku 3.1 1+0.1 -242 18 1900 301 1445 154 378.0 19.9% 8.1% 0 0 0 0 0
17 Willow 2.9 1+0.1 -252 19 1900 282 1461 157 360.5 19.0% 8.3% 0 0 0 0 0
18 Polaris 1.6.1 1+0.1 -261 19 1900 271 1479 150 346.0 18.2% 7.9% 0 0 0 0 0
19 Renegade 0.11.0 1+0.1 -385 24 1900 137 1663 100 187.0 9.8% 5.3% 0 0 0 0 0
20 Abbess2022B 1+0.1 -706 55 1900 18 1854 28 32.0 1.7% 1.5% 0 0 0 0 0
19000 of 19000 games finished.
Intel(R) Core(TM) i5-10400 CPU @ 2.90GHz
Cores: 6
Logical processors: 12
Installed RAM: 12.0 GB
CuteChess GUI 3.5.0
TC: 1+0.1
Hash: 16 MB
Threads: 1
Concurrent: 3
Ponder off
Syzygy 3-4-5 TB adjudication
Named Openings suite, randomized, to whichever ply it needs
All 19,000 games:
https://www.mediafire.com/file_premium/ ... 1.pgn/file
Re: CERL Games
Posted: Wed Jun 28, 2023 3:36 am
by FritzUser
Here is the latest tournament. The results table is pretty big for the code block, but it was a nice bunch of numbers. The only two problems were that Avalanche played an illegal move in one game, and Altair timed out eleven times. I'm not sure how long it took, but as you can see, with 1s+0.1s it's possible to run a big tournament with a proper number of games (to reduce the margin of error) and not have to wait a week.I wish I could take all the CERL games from back when I was also using this TC and plug them into the list, but when I tried that, Berserk 11, I think, ended up at the top with something like 3900. :-)
Code: Select all
Rank Name TC Elo +/- Games Wins Losses Draws Points Score Draw Time Illegal Discon Stall WrClaim
1 Stockfish 15 1+0.1 296 14 2400 1752 91 557 2030.5 84.6% 23.2% 0 0 0 0 0
2 BrainLearn 25 1+0.1 296 14 2400 1743 82 575 2030.5 84.6% 24.0% 0 0 0 0 0
3 BrainLearn 24 1+0.1 293 14 2400 1734 83 583 2025.5 84.4% 24.3% 0 0 0 0 0
4 ShashChess 32 1+0.1 279 14 2400 1713 116 571 1998.5 83.3% 23.8% 0 0 0 0 0
5 Dragon 3.1 1+0.1 234 13 2400 1601 190 609 1905.5 79.4% 25.4% 0 0 0 0 0
6 Berserk 11 1+0.1 221 13 2400 1572 222 606 1875.0 78.1% 25.3% 0 0 0 0 0
7 Koivisto 9.0 1+0.1 185 13 2400 1488 321 591 1783.5 74.3% 24.6% 0 0 0 0 0
8 Clover 5.0 1+0.1 158 13 2400 1400 378 622 1711.0 71.3% 25.9% 0 0 0 0 0
9 Seer 2.6 1+0.1 129 12 2400 1293 440 667 1626.5 67.8% 27.8% 0 0 0 0 0
10 Caissa 1.9 1+0.1 81 12 2400 1209 658 533 1475.5 61.5% 22.2% 0 0 0 0 0
11 Caissa 1.8 1+0.1 65 12 2400 1133 688 579 1422.5 59.3% 24.1% 0 0 0 0 0
12 Fire 9 1+0.1 46 12 2400 1105 791 504 1357.0 56.5% 21.0% 0 0 0 0 0
13 Viridithas 9.1.0 1+0.1 40 12 2400 1067 795 538 1336.0 55.7% 22.4% 0 0 0 0 0
14 Velvet 5.2.1 1+0.1 21 13 2400 1059 915 426 1272.0 53.0% 17.8% 0 0 0 0 0
15 Komodo 13 1+0.1 1 13 2400 1010 1005 385 1202.5 50.1% 16.0% 0 0 0 0 0
16 Svart 5 1+0.1 -89 13 2400 748 1351 301 898.5 37.4% 12.5% 0 0 0 0 0
17 Drofa 4.0.0 1+0.1 -130 14 2400 634 1493 273 770.5 32.1% 11.4% 0 0 0 0 0
18 Avalanche 1.4.0 1+0.1 -180 15 2400 493 1638 269 627.5 26.1% 11.2% 0 1 0 0 0
19 Monolith 2.01 1+0.1 -184 15 2400 501 1665 234 618.0 25.8% 9.8% 0 0 0 0 0
20 Altair 3.0.0 1+0.1 -255 16 2400 342 1843 215 449.5 18.7% 9.0% 11 0 0 0 0
21 Willow 2.9 1+0.1 -273 17 2400 311 1885 204 413.0 17.2% 8.5% 0 0 0 0 0
22 4ku 3.1 1+0.1 -283 17 2400 287 1901 212 393.0 16.4% 8.8% 0 0 0 0 0
23 Inanis 1.2.0 1+0.1 -301 18 2400 260 1939 201 360.5 15.0% 8.4% 0 0 0 0 0
24 4ku 2.0 1+0.1 -377 21 2400 168 2075 157 246.5 10.3% 6.5% 0 0 0 0 0
25 Akimbo 0.3.0 1+0.1 -446 24 2400 110 2168 122 171.0 7.1% 5.1% 0 0 0 0 0
30000 of 30000 games finished.
Intel(R) Core(TM) i5-10400 CPU @ 2.90GHz
Cores: 6
Logical processors: 12
Installed RAM: 12.0 GB
CuteChess GUI 3.5.0
TC: 1+0.1 (50 ms margin)
Hash: 16 MB
Threads: 1
Concurrent: 3
Ponder off
Syzygy 3-4-5 TB adjudication
Draw adjudication after move 80: 30 moves within 1 cp
No resign adjudication
Named Openings Suite to whichever ply it needs
All 30,000 games
https://www.mediafire.com/file_premium/ ... 1.pgn/file
Re: CERL Games
Posted: Thu Jun 29, 2023 5:27 am
by FritzUser
Here's the latest tournament, though it may not be of any value. I wanted to create a small database for the
Bongcloud opening, so I did. Not sure how many doubles are in here. I probably won't use it for the CERL, because some of these engines inexplicably play this opening rather well.
Code: Select all
Rank Name TC Elo +/- Games Wins Losses Draws Points Score Draw Time Illegal Discon Stall WrClaim
1 Berserk 10 1+0.1 220 20 1400 998 213 189 1092.5 78.0% 13.5% 0 0 0 0 0
2 BrainLearn 24 1+0.1 214 19 1400 971 204 225 1083.5 77.4% 16.1% 0 0 0 0 0
3 Koivisto 9.0 1+0.1 212 19 1400 980 217 203 1081.5 77.3% 14.5% 0 0 0 0 0
4 Stockfish 14 1+0.1 163 18 1400 905 292 203 1006.5 71.9% 14.5% 0 0 0 0 0
5 Seer 2.6 1+0.1 147 18 1400 878 320 202 979.0 69.9% 14.4% 0 0 0 0 0
6 Caissa 1.8 1+0.1 45 16 1400 654 472 274 791.0 56.5% 19.6% 0 0 0 0 0
7 Viridithas 9.1.0 1+0.1 16 16 1400 604 538 258 733.0 52.4% 18.4% 0 0 0 0 0
8 Clover 3.3.1 1+0.1 15 17 1400 618 556 226 731.0 52.2% 16.1% 0 0 0 0 0
9 Velvet 5.2.1 1+0.1 8 17 1400 605 574 221 715.5 51.1% 15.8% 0 0 0 0 0
10 Fire 9 1+0.1 2 16 1400 578 570 252 704.0 50.3% 18.0% 0 0 0 0 0
11 Komodo 13 1+0.1 -15 17 1400 599 658 143 670.5 47.9% 10.2% 0 0 0 0 0
12 Svart 5 1+0.1 -131 18 1400 351 856 193 447.5 32.0% 13.8% 0 0 0 0 0
13 Monolith 2.01 1+0.1 -251 22 1400 207 1073 120 267.0 19.1% 8.6% 0 0 0 0 0
14 4ku 3.1 1+0.1 -384 28 1400 97 1220 83 138.5 9.9% 5.9% 0 0 0 0 0
15 Akimbo 0.3.0 1+0.1 -543 43 1400 43 1325 32 59.0 4.2% 2.3% 0 0 0 0 0
10500 of 10500 games finished.
Intel(R) Core(TM) i5-10400 CPU @ 2.90GHz
Cores: 6
Logical processors: 12
Installed RAM: 12.0 GB
CuteChess GUI 3.5.0
TC: 1+0.1 (50 ms margin)
Hash: 16 MB
Threads: 1
Concurrent: 3
Ponder off
Syzygy 3-4-5 TB adjudication
Draw adjudication after move 80: 30 moves within 1 cp
No resign adjudication
All 10,500 games:
https://chessnerd.net/tour/bongcloud.pgn