Thank you, Ed.
I still use both Pro Deo and Benjamin when analyzing, so it is meaningful to me, although depth has a big influence on the Speedy Rating List.
I rather miss an "anti-GM" option in many engines, and I was more interested in the ERL, but it was a project nobody was interested in. Lately, most engines have an evaluation tuned automatically, it gives good results in eng-eng matches, but it is ugly chess.
The Speedy Rating List
Moderators: hgm, Rebel, chrisw
-
- Posts: 365
- Joined: Mon May 14, 2007 8:20 pm
- Full name: Boban Stanojević
-
- Posts: 1871
- Joined: Sat Nov 25, 2017 2:28 pm
- Location: France
Re: The Speedy Rating List
Thanks for considering Minic in here !
Can you please give a shot to Minic 2.32 (unofficial release that can be found here https://github.com/tryingsomestuff/Mini ... ter/Minic2)
I suspect it might be stronger on this type of test ...
Can you please give a shot to Minic 2.32 (unofficial release that can be found here https://github.com/tryingsomestuff/Mini ... ter/Minic2)
I suspect it might be stronger on this type of test ...
-
- Posts: 6995
- Joined: Thu Aug 18, 2011 12:04 pm
Re: The Speedy Rating List
True that, search rules.
Same cause, search.matejst wrote: ↑Thu May 28, 2020 12:49 pm I rather miss an "anti-GM" option in many engines, and I was more interested in the ERL, but it was a project nobody was interested in. Lately, most engines have an evaluation tuned automatically, it gives good results in eng-eng matches, but it is ugly chess.
I have seen it happen to my hand-tuned evaluation because back then there were no eng-eng facilities, nowadays search with all its reductions and pruning (while adding hundreds of elo points) damages what you carefully have programmed and tuned, the evaluation function and with it the playing style. As a positional oriented chess player I hate it, as computer chess programmer I have no choice
BTW, there is a replacement for the anti-GM option, actually 3, all of them function better than the original anti-GM.
[ATTACKING = 100]
[CHAOS = 100]
[CONFUSION = 100]
They all 3 try to complicate the position.
The activation is a bit non-standard, 100 means not active, but any other value than 100 it becomes alive. Forgot why I did it this strange way.
90% of coding is debugging, the other 10% is writing bugs.
-
- Posts: 6995
- Joined: Thu Aug 18, 2011 12:04 pm
Re: The Speedy Rating List
Oki doki, but which executable to pick?xr_a_y wrote: ↑Thu May 28, 2020 4:21 pm Thanks for considering Minic in here !
Can you please give a shot to Minic 2.32 (unofficial release that can be found here https://github.com/tryingsomestuff/Mini ... ter/Minic2)
I suspect it might be stronger on this type of test ...
Code: Select all
minic_2.32_mingw_x64_nehalem.exe minic 2.32 May 28, 2020
minic_2.32_mingw_x64_skylake.exe minic 2.32 May 28, 2020
minic_2.32_mingw_x64_x86-64.exe minic 2.32 May 28, 2020
90% of coding is debugging, the other 10% is writing bugs.
-
- Posts: 1871
- Joined: Sat Nov 25, 2017 2:28 pm
- Location: France
Re: The Speedy Rating List
In this order, best to worst, depending on your hardware :Rebel wrote: ↑Thu May 28, 2020 4:34 pmOki doki, but which executable to pick?xr_a_y wrote: ↑Thu May 28, 2020 4:21 pm Thanks for considering Minic in here !
Can you please give a shot to Minic 2.32 (unofficial release that can be found here https://github.com/tryingsomestuff/Mini ... ter/Minic2)
I suspect it might be stronger on this type of test ...
Code: Select all
minic_2.32_mingw_x64_nehalem.exe minic 2.32 May 28, 2020 minic_2.32_mingw_x64_skylake.exe minic 2.32 May 28, 2020 minic_2.32_mingw_x64_x86-64.exe minic 2.32 May 28, 2020
skylake == avx2/bmi2
nehalem == sse4.2
x86-64 == just popcnt
-
- Posts: 5960
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
Re: The Speedy Rating List
The range of the ratings shown is roughly half that of the range on the 1' + 0.6" FastGM list, which should be the closest comparison (although it doesn't have all of these engines). I suppose the scaling was matched to some rating list, but what rating list has such a narrow range of ratings from top to bottom? Other than this scaling detail, it does look surprisingly accurate.Rebel wrote: ↑Thu May 28, 2020 11:25 am Added Winter 0.8, ProDeo 2.2 and Benjamin, provisional (text) result as I adding more:Code: Select all
EPD : epd\lc1.epd Time : 1000ms Max Time Hash Engine Points Used Time Found Pos Elo Score Score ms Mb Cpu Errors 1 Stockfish 11 324287 11:30:00.8 23191 40000 3242 400000 81.07% 1000 128 1 0 2 Komodo 14 317110 11:09:21.4 22398 40000 3171 400000 79.28% 1000 128 1 0 3 Houdini 6.03 314596 11:27:54.8 22218 40000 3146 400000 78.65% 1000 128 1 0 4 rofChade 2.3 309642 11:24:20.2 21587 40000 3096 400000 77.41% 1000 128 1 41 5 Ethereal 12 309088 11:30:20.9 21840 40000 3090 400000 77.27% 1000 128 1 0 6 Laser 1.7 308178 11:29:30.5 21320 40000 3081 400000 77.04% 1000 128 1 0 7 Schooner 2.2 306808 10:22:01.9 21335 40000 3068 400000 76.70% 1000 128 1 0 8 Xiphos 0.6 306796 10:03:55.9 21333 40000 3068 400000 76.70% 1000 128 1 0 9 RubiChess-1.7.2 305900 11:26:03.9 21215 40000 3059 400000 76.48% 1000 128 1 0 10 Booot 6.4 305082 12:56:19.6 21267 40000 3050 400000 76.27% 1000 128 1 124 11 Defenchess 2.2 303754 10:12:00.7 21208 40000 3037 400000 75.94% 1000 128 1 41 12 Andscacs 0.95 301560 12:10:30.8 20745 40000 3015 400000 75.39% 1005 128 1 41 13 Shredder 13 300619 11:28:47.0 20946 40000 3006 400000 75.15% 1000 128 1 0 14 Arasan 22 298663 11:27:32.7 20460 40000 2986 400000 74.67% 1000 128 1 0 15 ice 4.0 297995 11:49:51.1 20767 40000 2980 400000 74.50% 1000 128 1 0 16 Fire 7.1 296996 11:26:35.7 20356 40000 2970 400000 74.25% 1000 128 1 0 17 Vajolet 2.8 296269 11:25:51.6 20327 40000 2962 400000 74.07% 1000 128 1 0 18 Demolito 200514 295686 10:53:58.5 20429 40000 2956 400000 73.92% 1000 128 1 0 19 Wasp 3.75 293908 11:44:35.3 20203 40000 2939 400000 73.48% 1000 128 1 0 20 Gogobello 2.2 293824 11:44:24.1 20296 40000 2938 400000 73.46% 1000 128 1 41 21 igel 2.40 293056 11:28:38.5 20261 40000 2930 400000 73.26% 1000 128 1 41 22 Rodent 4 291647 11:45:03.3 19774 40000 2916 400000 72.91% 1000 128 1 0 23 Winter 0.8 291596 10:02:03.0 20057 40000 2916 400000 72.90% 1000 128 1 41 24 Monolith 2 290807 11:31:27.5 19972 40000 2908 400000 72.70% 1000 128 1 0 25 Deuterium 2019 290578 11:31:42.7 19819 40000 2905 400000 72.64% 1000 128 1 0 26 Amoeba 3.1 289712 09:40:26.2 19682 40000 2897 400000 72.43% 1000 128 1 0 27 FabChess 1.15 289296 11:50:40.2 19849 40000 2892 400000 72.32% 1000 128 1 41 28 Marvin 3.60 288732 10:48:52.8 19848 40000 2887 400000 72.18% 1000 128 1 41 29 Minic 2.25 287730 10:54:48.6 19780 40000 2877 400000 71.93% 1000 128 1 0 30 Counter 3.5 285137 11:25:20.1 19329 40000 2851 400000 71.28% 1000 128 1 0 31 Asymptote 0.8 284305 12:19:43.9 19245 40000 2843 400000 71.08% 1000 128 1 41 32 Weiss 0.10 281037 11:06:09.2 18927 40000 2810 400000 70.26% 1000 128 1 0 33 Fruit 2.3 275618 11:30:55.1 18270 40000 2756 400000 68.90% 1000 128 1 0 34 ProDeo 2.2 273793 11:29:46.3 18263 40000 2738 400000 68.45% 1000 128 1 0 35 Benjamin 271801 11:29:55.8 17985 40000 2718 400000 67.95% 1000 128 1 0 36 Fruit 2.1 270830 11:29:45.7 17982 40000 2708 400000 67.71% 1000 128 1 0 37 Ruffian 2 265592 11:13:01.7 17740 40000 2656 400000 66.40% 1000 128 1 0
Komodo rules!
-
- Posts: 6995
- Joined: Thu Aug 18, 2011 12:04 pm
Re: The Speedy Rating List
Thanks Larry for your kind words. Regarding the elo calculation, it's quite simple: Komodo 14 scores 79.28% multiply this with 40 and you get the elo rating, 3.171. I could easily increase the value to 41, 42, 43 till it matches the known rating lists. But that step has to be taken when the 4000ms list is more voluminous because 4000ms is comparable with for instance CCRL 40/2. I don't recall an elo list based on 1 second per move.lkaufman wrote: ↑Thu May 28, 2020 6:40 pmThe range of the ratings shown is roughly half that of the range on the 1' + 0.6" FastGM list, which should be the closest comparison (although it doesn't have all of these engines). I suppose the scaling was matched to some rating list, but what rating list has such a narrow range of ratings from top to bottom? Other than this scaling detail, it does look surprisingly accurate.Rebel wrote: ↑Thu May 28, 2020 11:25 am Added Winter 0.8, ProDeo 2.2 and Benjamin, provisional (text) result as I adding more:Code: Select all
EPD : epd\lc1.epd Time : 1000ms Max Time Hash Engine Points Used Time Found Pos Elo Score Score ms Mb Cpu Errors 1 Stockfish 11 324287 11:30:00.8 23191 40000 3242 400000 81.07% 1000 128 1 0 2 Komodo 14 317110 11:09:21.4 22398 40000 3171 400000 79.28% 1000 128 1 0 3 Houdini 6.03 314596 11:27:54.8 22218 40000 3146 400000 78.65% 1000 128 1 0 4 rofChade 2.3 309642 11:24:20.2 21587 40000 3096 400000 77.41% 1000 128 1 41 5 Ethereal 12 309088 11:30:20.9 21840 40000 3090 400000 77.27% 1000 128 1 0 6 Laser 1.7 308178 11:29:30.5 21320 40000 3081 400000 77.04% 1000 128 1 0 7 Schooner 2.2 306808 10:22:01.9 21335 40000 3068 400000 76.70% 1000 128 1 0 8 Xiphos 0.6 306796 10:03:55.9 21333 40000 3068 400000 76.70% 1000 128 1 0 9 RubiChess-1.7.2 305900 11:26:03.9 21215 40000 3059 400000 76.48% 1000 128 1 0 10 Booot 6.4 305082 12:56:19.6 21267 40000 3050 400000 76.27% 1000 128 1 124 11 Defenchess 2.2 303754 10:12:00.7 21208 40000 3037 400000 75.94% 1000 128 1 41 12 Andscacs 0.95 301560 12:10:30.8 20745 40000 3015 400000 75.39% 1005 128 1 41 13 Shredder 13 300619 11:28:47.0 20946 40000 3006 400000 75.15% 1000 128 1 0 14 Arasan 22 298663 11:27:32.7 20460 40000 2986 400000 74.67% 1000 128 1 0 15 ice 4.0 297995 11:49:51.1 20767 40000 2980 400000 74.50% 1000 128 1 0 16 Fire 7.1 296996 11:26:35.7 20356 40000 2970 400000 74.25% 1000 128 1 0 17 Vajolet 2.8 296269 11:25:51.6 20327 40000 2962 400000 74.07% 1000 128 1 0 18 Demolito 200514 295686 10:53:58.5 20429 40000 2956 400000 73.92% 1000 128 1 0 19 Wasp 3.75 293908 11:44:35.3 20203 40000 2939 400000 73.48% 1000 128 1 0 20 Gogobello 2.2 293824 11:44:24.1 20296 40000 2938 400000 73.46% 1000 128 1 41 21 igel 2.40 293056 11:28:38.5 20261 40000 2930 400000 73.26% 1000 128 1 41 22 Rodent 4 291647 11:45:03.3 19774 40000 2916 400000 72.91% 1000 128 1 0 23 Winter 0.8 291596 10:02:03.0 20057 40000 2916 400000 72.90% 1000 128 1 41 24 Monolith 2 290807 11:31:27.5 19972 40000 2908 400000 72.70% 1000 128 1 0 25 Deuterium 2019 290578 11:31:42.7 19819 40000 2905 400000 72.64% 1000 128 1 0 26 Amoeba 3.1 289712 09:40:26.2 19682 40000 2897 400000 72.43% 1000 128 1 0 27 FabChess 1.15 289296 11:50:40.2 19849 40000 2892 400000 72.32% 1000 128 1 41 28 Marvin 3.60 288732 10:48:52.8 19848 40000 2887 400000 72.18% 1000 128 1 41 29 Minic 2.25 287730 10:54:48.6 19780 40000 2877 400000 71.93% 1000 128 1 0 30 Counter 3.5 285137 11:25:20.1 19329 40000 2851 400000 71.28% 1000 128 1 0 31 Asymptote 0.8 284305 12:19:43.9 19245 40000 2843 400000 71.08% 1000 128 1 41 32 Weiss 0.10 281037 11:06:09.2 18927 40000 2810 400000 70.26% 1000 128 1 0 33 Fruit 2.3 275618 11:30:55.1 18270 40000 2756 400000 68.90% 1000 128 1 0 34 ProDeo 2.2 273793 11:29:46.3 18263 40000 2738 400000 68.45% 1000 128 1 0 35 Benjamin 271801 11:29:55.8 17985 40000 2718 400000 67.95% 1000 128 1 0 36 Fruit 2.1 270830 11:29:45.7 17982 40000 2708 400000 67.71% 1000 128 1 0 37 Ruffian 2 265592 11:13:01.7 17740 40000 2656 400000 66.40% 1000 128 1 0
90% of coding is debugging, the other 10% is writing bugs.
-
- Posts: 5960
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
Re: The Speedy Rating List
The FastGM 1' + 0.6" list comes the closest of major lists, it's about 2 seconds per move. Multiplying by 80 rather than 40 and subtracting a suitable constant, somewhere around 2950 I think, should produce a fairly good match. Of course a regression would produce a closer match, this is just a quick estimate.Rebel wrote: ↑Thu May 28, 2020 7:27 pmThanks Larry for your kind words. Regarding the elo calculation, it's quite simple: Komodo 14 scores 79.28% multiply this with 40 and you get the elo rating, 3.171. I could easily increase the value to 41, 42, 43 till it matches the known rating lists. But that step has to be taken when the 4000ms list is more voluminous because 4000ms is comparable with for instance CCRL 40/2. I don't recall an elo list based on 1 second per move.lkaufman wrote: ↑Thu May 28, 2020 6:40 pmThe range of the ratings shown is roughly half that of the range on the 1' + 0.6" FastGM list, which should be the closest comparison (although it doesn't have all of these engines). I suppose the scaling was matched to some rating list, but what rating list has such a narrow range of ratings from top to bottom? Other than this scaling detail, it does look surprisingly accurate.Rebel wrote: ↑Thu May 28, 2020 11:25 am Added Winter 0.8, ProDeo 2.2 and Benjamin, provisional (text) result as I adding more:Code: Select all
EPD : epd\lc1.epd Time : 1000ms Max Time Hash Engine Points Used Time Found Pos Elo Score Score ms Mb Cpu Errors 1 Stockfish 11 324287 11:30:00.8 23191 40000 3242 400000 81.07% 1000 128 1 0 2 Komodo 14 317110 11:09:21.4 22398 40000 3171 400000 79.28% 1000 128 1 0 3 Houdini 6.03 314596 11:27:54.8 22218 40000 3146 400000 78.65% 1000 128 1 0 4 rofChade 2.3 309642 11:24:20.2 21587 40000 3096 400000 77.41% 1000 128 1 41 5 Ethereal 12 309088 11:30:20.9 21840 40000 3090 400000 77.27% 1000 128 1 0 6 Laser 1.7 308178 11:29:30.5 21320 40000 3081 400000 77.04% 1000 128 1 0 7 Schooner 2.2 306808 10:22:01.9 21335 40000 3068 400000 76.70% 1000 128 1 0 8 Xiphos 0.6 306796 10:03:55.9 21333 40000 3068 400000 76.70% 1000 128 1 0 9 RubiChess-1.7.2 305900 11:26:03.9 21215 40000 3059 400000 76.48% 1000 128 1 0 10 Booot 6.4 305082 12:56:19.6 21267 40000 3050 400000 76.27% 1000 128 1 124 11 Defenchess 2.2 303754 10:12:00.7 21208 40000 3037 400000 75.94% 1000 128 1 41 12 Andscacs 0.95 301560 12:10:30.8 20745 40000 3015 400000 75.39% 1005 128 1 41 13 Shredder 13 300619 11:28:47.0 20946 40000 3006 400000 75.15% 1000 128 1 0 14 Arasan 22 298663 11:27:32.7 20460 40000 2986 400000 74.67% 1000 128 1 0 15 ice 4.0 297995 11:49:51.1 20767 40000 2980 400000 74.50% 1000 128 1 0 16 Fire 7.1 296996 11:26:35.7 20356 40000 2970 400000 74.25% 1000 128 1 0 17 Vajolet 2.8 296269 11:25:51.6 20327 40000 2962 400000 74.07% 1000 128 1 0 18 Demolito 200514 295686 10:53:58.5 20429 40000 2956 400000 73.92% 1000 128 1 0 19 Wasp 3.75 293908 11:44:35.3 20203 40000 2939 400000 73.48% 1000 128 1 0 20 Gogobello 2.2 293824 11:44:24.1 20296 40000 2938 400000 73.46% 1000 128 1 41 21 igel 2.40 293056 11:28:38.5 20261 40000 2930 400000 73.26% 1000 128 1 41 22 Rodent 4 291647 11:45:03.3 19774 40000 2916 400000 72.91% 1000 128 1 0 23 Winter 0.8 291596 10:02:03.0 20057 40000 2916 400000 72.90% 1000 128 1 41 24 Monolith 2 290807 11:31:27.5 19972 40000 2908 400000 72.70% 1000 128 1 0 25 Deuterium 2019 290578 11:31:42.7 19819 40000 2905 400000 72.64% 1000 128 1 0 26 Amoeba 3.1 289712 09:40:26.2 19682 40000 2897 400000 72.43% 1000 128 1 0 27 FabChess 1.15 289296 11:50:40.2 19849 40000 2892 400000 72.32% 1000 128 1 41 28 Marvin 3.60 288732 10:48:52.8 19848 40000 2887 400000 72.18% 1000 128 1 41 29 Minic 2.25 287730 10:54:48.6 19780 40000 2877 400000 71.93% 1000 128 1 0 30 Counter 3.5 285137 11:25:20.1 19329 40000 2851 400000 71.28% 1000 128 1 0 31 Asymptote 0.8 284305 12:19:43.9 19245 40000 2843 400000 71.08% 1000 128 1 41 32 Weiss 0.10 281037 11:06:09.2 18927 40000 2810 400000 70.26% 1000 128 1 0 33 Fruit 2.3 275618 11:30:55.1 18270 40000 2756 400000 68.90% 1000 128 1 0 34 ProDeo 2.2 273793 11:29:46.3 18263 40000 2738 400000 68.45% 1000 128 1 0 35 Benjamin 271801 11:29:55.8 17985 40000 2718 400000 67.95% 1000 128 1 0 36 Fruit 2.1 270830 11:29:45.7 17982 40000 2708 400000 67.71% 1000 128 1 0 37 Ruffian 2 265592 11:13:01.7 17740 40000 2656 400000 66.40% 1000 128 1 0
Komodo rules!
-
- Posts: 6995
- Joined: Thu Aug 18, 2011 12:04 pm
Re: The Speedy Rating List
xr_a_y wrote: ↑Thu May 28, 2020 4:46 pmIn this order, best to worst, depending on your hardware :Rebel wrote: ↑Thu May 28, 2020 4:34 pmOki doki, but which executable to pick?xr_a_y wrote: ↑Thu May 28, 2020 4:21 pm Thanks for considering Minic in here !
Can you please give a shot to Minic 2.32 (unofficial release that can be found here https://github.com/tryingsomestuff/Mini ... ter/Minic2)
I suspect it might be stronger on this type of test ...
Code: Select all
minic_2.32_mingw_x64_nehalem.exe minic 2.32 May 28, 2020 minic_2.32_mingw_x64_skylake.exe minic 2.32 May 28, 2020 minic_2.32_mingw_x64_x86-64.exe minic 2.32 May 28, 2020
skylake == avx2/bmi2
nehalem == sse4.2
x86-64 == just popcnt
Code: Select all
Max Time Hash
Engine Points Used Time Found Pos Elo Score Score ms Mb Cpu Errors
29 Minic 2.32 289394 11:01:03.5 19880 40000 2894 400000 72.35% 1000 128 1 0
32 Minic 2.25 287730 10:54:48.6 19780 40000 2877 400000 71.93% 1000 128 1 0
90% of coding is debugging, the other 10% is writing bugs.
-
- Posts: 144
- Joined: Sun Oct 14, 2018 8:21 pm
- Full name: JSmith
Re: The Speedy Rating List
Wow, this is really cool. Do you think you could talk about the methodology behind Lc1.epd? Also, would you be able to test Fire 7.1 again? I'm surprised to see it so low. It seems like a major outlier compared to what rating lists show.