Gull 3 vs Stockfish 14053109 modern

Questions and comments related to CCRL testing study
Post Reply
bezzy3004
Posts: 3
Joined: Wed Jan 21, 2015 5:23 pm
Sign-up code: 10159

Gull 3 vs Stockfish 14053109 modern

Post by bezzy3004 »

Hi all,
I run a chess engine against human players (Authorised) on a chess server site, Stockfish 14053109 x64 modern is proving to be unbeatable almost, with a record of W214 D9 L12.
I was curious as to if any other engines could replace Stockfish as my main engine so set about testing the top 10 available engines apart from Komodo 8 which i dont own. I tested them at 1 min blitz as that is what i use to play at upto ten mins time control, I was very suprised to see Gull 3 at the top and Stockfish at position 6 Gull is clearly very comfortable playing rapid games. All engines using 4 cores. Apologies if the table is out of alignment i cant seem to get it right.

Rank Engine Rating Score % Gu Ho Fi Ko Cr St Eq Bo De Bl S-B
01 Gull 3 x64 12.0/18 66.6 · · 0= 0= =1 1= 10 11 1= =1 11 99.00
02 Houdini_4_Prox64 11.0/18 61.1 1= · · =1 =0 01 0= == 11 1= 1= 93.75
03 Fire_4_x64 10.0/18 55.5 1= =0 · · 10 0= 1= == 01 =1 =1 86.50
04 Komodo 6-64bit 10.0/18 55.5 =0 =1 01 · · 00 01 1= 01 11 1= 84.00
05 Critter_1.6a_64bit 9.5/18 52.7 0= 10 1= 11 · · 10 =0 0= == 1= 84.50
06 Stockfish_14053109_x64 9.0/18 50.0 01 1= 0= 10 01 · · =1 10 =0 10 82.75
07 Equinox 3.20-x64 9.0/18 50.0 00 == == 0= =1 =0 · · 1= =1 =1 74.00
08 Bouquet 1.8 x64 3157 7.0/18 38.8 0= 00 10 10 1= 01 0= · · == =0 63.25
09 Deep Rybka 4.1 x64 6.5/18 36.1 =0 0= =0 00 == =1 =0 == · · 10 57.00
10 BlackMamba_MP_x64 6.0/18 33.3 00 0= =0 0= 0= 01 =0 =1 01 · · 50.75

90 games played
Tournament start: 2015.01.21, 13:43:19
Level: Blitz 1/0
Hardware: Intel(R) Core(TM) i5-3570K CPU @ 4.1GHz with 16.0 GB Memory
Operating system: Windows 7 Home Premium Home Edition Service Pack 1 (Build 7601) 64 bit
Conditions: Hash: 256MB, Tablebases: On, 4 man TB, Ponder Off

So that is why i was curious as to Gulls performance against Stockfish 14053190 x64 Modern under CCRL testing conditions which translate to 40/2 repeating on my hardware. Here is the result of 30 games.

Ccrl Engine Tournament.

Rank Engine Score %
1 Stockfish_14053109_x64_modern 16.5/30 55.0
2 Gull 3 x64 13.5/30 45.0


30 games played
Tournament start: 2015.01.25, 12:35:25
Level: Tournament 40/2
Hardware: Intel(R) Core(TM) i5-3570K CPU @ 4.1GHz with 16.0 GB Memory
Operating system: Windows 7 Home Premium Home Edition Service Pack 1 (Build 7601) 64 bit
Conditions: Hash: 256MB, Tablebases: On, 4 man TB, Ponder Off

Stockfish wins by 3 points, has anybody any ideas why Gull was so good at 1min blitz and stockfish so bad im pretty confident its nothing to do with individual engine parameters as all where checked and set to unified settings?
User avatar
Graham Banks
Posts: 27017
Joined: Sun Dec 18, 2005 5:47 pm
Sign-up code: 0
Location: Auckland, NZ

Re: Gull 3 vs Stockfish 14053109 modern

Post by Graham Banks »

If you ran another couple of 30 game matches, you might come up with completely different results.
bezzy3004
Posts: 3
Joined: Wed Jan 21, 2015 5:23 pm
Sign-up code: 10159

Re: Gull 3 vs Stockfish 14053109 modern

Post by bezzy3004 »

Graham Banks wrote:If you ran another couple of 30 game matches, you might come up with completely different results.
What is the reason for that Graham? I would of thought 1 engine is superior to the other and thats it, forgive me for my ignorance on the subject.
User avatar
Graham Banks
Posts: 27017
Joined: Sun Dec 18, 2005 5:47 pm
Sign-up code: 0
Location: Auckland, NZ

Re: Gull 3 vs Stockfish 14053109 modern

Post by Graham Banks »

bezzy3004 wrote:
Graham Banks wrote:If you ran another couple of 30 game matches, you might come up with completely different results.
What is the reason for that Graham? I would of thought 1 engine is superior to the other and thats it, forgive me for my ignorance on the subject.
There will always be margins of error, even after several hundred games.
However, usually after 150 games or so have been played, an engine's rating rarely changes by much.
This is the reason that an engine must have over 200 games before it becomes established in our lists.
Post Reply