Latest Update: 03.04.2022
Hello Chess Friends,
I'm pleased to announce that,
Here is another new GUI/Engine competition...!
Meanwhile,
For anyone interested/missed, the previous testing: Results
Well, the main target is simply to see the Elo influence again...
Because, this time a newer Cutechess GUI version is used
If I'm not wrong, the previous Cutechess ver was custom build...
Now any engine's move overhead setting is increased to 200
Plus for these testings: clean fresh updated Windows installed
And I have just great news...for example now,
If we compare, the both GUIs results are more close to each other...!
In other words,
E.g in previous Cutechess GUI testing, Cfish performed 65+ Elo points
But via latest GUI testings: now it seems much more stable for sure...I
As we see, the Elo difference is 38+ Elo ..almost same as Arena GUI
Sure, we can't expect same points, but much closer than older Results
And here are the latest new played results about which I'm referring:
2x E5-2686, Cutechess Qt 5.11.1 GUI, 30sec+0.5sec:
Cfish 031120 performed 38+ Elo better:..and once more in previous GUI testing, The Elo difference (between Cfish and SF) was 65+Elo
1 Cfish 031120 +133/-25/=842 55.40% 554.0/1000
2 Stockfish 12 +25/-133/=842 44.60% 446.0/1000
-----------------------------------------------------------------------------------------------------------Btw, as alternatives, here are another new GUI and Engine Testings
In short, I realized to run more testings..as you know... more better..!)
In this way, I think that we will have more possibility to compare, check...
And once more,
SF PB 130322 engine managed to win...under both GUIs and Machines!!
Well-done and my congratulations to Skynet and Stockfish team as well!
Regarding mostly of rest players, very close Elo points,it seems not so bad
Meanwhile, as I mentioned before,
Via short strong openings...not required thousands of games per player!
E.g 500-1000 games is enough about to determine any engine strength!
And right now a new idea is born I will organize a new opening testings
Where I hope to use critical, weak openings...let's see after 700 games...
What will be the situation...can we expect again similar, close points ?)
1st Test AMD Ryzen7, Arena 3.5.1 GUI, 20s+0.6s, for more Details
# PLAYER : RATING POINTS GAMES (%) 1 SF PB 130322 : 3740.2 367.5 704 52.2% 2 SF Polyglot 110322 : 3738.1 365.5 704 51.9% 3 Swordfish 14.8 : 3735.9 363.0 704 51.6% 4 Kayra 1.3 : 3735.5 361.5 704 51.3% 5 Corchess 280322 : 3734.5 361.0 704 51.3% 6 Stockfish 220322 : 3731.8 358.0 704 50.9% 7 Raubfisch X48d2 : 3729.8 356.0 704 50.6% 8 ShashChess 21.1 : 3728.3 354.0 704 50.3% 9 BrainLearn 16.1 : 3726.5 352.5 704 50.1% 10 Fat Titz 2 130122 : 3723.2 349.5 704 49.6% 11 Cfish 130821 : 3700.0 323.5 704 46.0% 12 Rfish 230122 : 3689.1 312.0 704 44.3%
Note: During the active matches,several Arena 3.5.1 GUIs are played And for each Arena GUI is installed separate folder of chess engines -----------------------------------------------------------------------------------------------------------
2nd Test AMD Ryzen7, Cutechess Qt 5.11.1 GUI, 20sec+0.6sec:
# PLAYER : RATING POINTS GAMES (%) 1 SF PB 130322 : 3742.5 366.0 704 52.0% 2 Stockfish 220322 : 3740.0 363.5 704 51.6% 3 SF Polyglot 110322 : 3739.3 363.0 704 51.6% 4 Kayra 1.3 : 3738.6 361.5 704 51.3% 5 Swordfish 14.8 : 3738.2 361.5 704 51.3% 6 Corchess 280322 : 3734.6 358.0 704 50.9% 7 BrainLearn 16.1 : 3731.8 355.0 704 50.4% 8 ShashChess 21.1 : 3729.1 352.0 704 50.0% 9 Fat Titz 2 130122 : 3727.5 350.0 704 49.7% 10 Raubfisch X48d2 : 3726.2 348.5 704 49.5% 11 Rfish 230122 : 3704.6 325.0 704 46.2% 12 Cfish 130821 : 3700.0 320.0 704 45.5%
Note: During the active matches, only 1 (one) Cutechess GUI is played... That means all of the games are played by same folder of chess engines -----------------------------------------------------------------------------------------------------------
3rd. Test 2x E5-2686, Cutechess Qt 5.11.1 GUI, 30sec+0.6sec:
# PLAYER : RATING POINTS GAMES (%) 1 SF PB 130322 : 3741.3 365.5 704 51.9% 2 SF Polyglot 110322 : 3738.1 362.0 704 51.4% 3 Stockfish 220322 : 3737.7 361.5 704 51.3% 4 Kayra 1.3 : 3736.7 360.5 704 51.2% 5 Raubfisch X48d2 : 3734.9 358.5 704 50.9% 6 BrainLearn 16.1 : 3733.1 356.5 704 50.6% 7 Corchess 280322 : 3732.6 356.0 704 50.6% 8 Swordfish 14.8 : 3730.3 353.5 704 50.2% 9 ShashChess 21.1 : 3726.7 349.5 704 49.6% 10 Fat Titz 2 130122 : 3723.4 346.0 704 49.1% 11 Rfish 230122 : 3712.4 334.0 704 47.4% 12 Cfish 130821 : 3700.0 320.5 704 45.5%
Note: During the active matches, only 1 (one) Cutechess GUI is played... That means all of the games are played by same folder of chess engines -----------------------------------------------------------------------------------------------------------
GAMES More Details, The Ratings are calculated by Ordo, Cfish 130821 fixed to 3700 EloAll chess engines are participated as NNUE (via default Evaifiles)As opening books, Balsa suite and Perfect books are used to play..On 2x E5-2686: 128 MB Hash, where on Ryzen: 64 MB Hash usedOn my Ryzen machine, all matches played via Concurrent option: 8Where on my Xeon 2x E5-2686 machine, the Concurrent option: 32Plus on Xeon, 30sec+0.6sec is used, due to it's approx. 20% slower...
As Final Words,According to my all chess GUI and Engine testings,There is no any perfect GUI or Engine....in short, Each chess GUI has own advantages or disadvantages, Just a few examples, it seems Rfish suffers via Arena 3.5.1And Arena GUI does not support Parallel (Concurrent) matchesAbout Cutechess GUI: no common Hashtable, Threads option etc.So each engine setting is needed to be configurated manually....Also the tours via Cutechess, no chances to be paused/resumed There are a lot of other futures, which can be improved...but inOverall, both GUIs are great and we have to be satisfied with that!
Ok..that's all for now, and thanks for your interest.. )
Greetings,Sedat