White wins: 372'246 (34.7%)
Black wins: 276'998 (25.8%)
Draws: 422'578 (39.4%)
White score: 54.4%
From CCRL website
Computer vs Computer
Sort:
Aug 4, 2019
0
#42
Ziryab wrote:
White wins: 372'246 (34.7%)
Black wins: 276'998 (25.8%)
Draws: 422'578 (39.4%)
White score: 54.4%
From CCRL website
Yes, probably combined 40 min and 4 mins CCRL, but most CCRL games are 4 mins Time control. And the hardware they use is 2 GH AMD , 2 cores cpu. And the engines they test are from 1500 to 3500( average 2500 rating). So the average engines do mistakes here and there like human OTB games between 2000 to 2800 database. ( even among human 2700-2800 database, white advantage will be significantly less than common database)
But in today technology, like 32 or 64 cores hardware with 3500 top 3 engines ( SF, K, H) in 3 hours games. White advantage would be 52-53% at most. ( dont look at tcec statistics, they use unbalanced opening books to get decisive results, to kill draw rate)
Aug 4, 2019
0
#44
drmrboss wrote:
Ziryab wrote:
White wins: 372'246 (34.7%)
Black wins: 276'998 (25.8%)
Draws: 422'578 (39.4%)
White score: 54.4%
From CCRL website
Yes, probably combined 40 min and 4 mins CCRL, but most CCRL games are 4 mins Time control. And the hardware they use is 2 GH AMD , 2 cores cpu. And the engines they test are from 1500 to 3500( average 2500 rating). So the average engines do mistakes here and there like human OTB games between 2000 to 2800 database. ( even among human 2700-2800 database, white advantage will be significantly less than common database)
But in today technology, like 32 or 64 cores hardware with 3500 top 3 engines ( SF, K, H) in 3 hours games. White advantage would be 52-53% at most. ( dont look at tcec statistics, they use unbalanced opening books to get decisive results, to kill draw rate)
Nope. That's 40/40
Black does better at the faster time controls. For instance, 40/4 is
White wins: 806'607 (38.8%)
Black wins: 652'435 (31.4%)
Draws: 617'840 (29.7%)
White score: 53.7%
This is really odd to my knowledge. ( I am surprised)
Aug 9, 2019
0
#46
rocketbobcat wrote:
I have done games comp v comp in the past. black always wins
So far 5, and black won every one
BlunderFish escribió:
EscherehcsE wrote:
BlunderFish wrote:
How did you make two bots play against each other?
My guess is the old trick of two accounts (hopefully not owned by the same person). Account 1 plays the comp as White, Account 2 plays the comp as Black. The two human players just relay the comps' moves, and the result is that the comps end up playing a game between the two comps.
I was hoping there was a more simple way to do it than manually moving the pieces >-<
There is a more easy way. Play against a computer on the level you want. On the match settings, switch the pieces you're using repeatedly to have the CPU play every single turn.
I love how Na3 is called sodium attack!
cool
BlunderFish wrote:
How did you make two bots play against each other?
Just for future reference, under more you can select "Computer Championship"
Jul 18, 2021
0
#52
SnowJaguar0 wrote:
I love how Na3 is called sodium attack!
Shame it's a pretty crappy move - lol.
May 8, 2022
0
#53
AshwathNarayanAS wrote:
You can do computer vs. computer games using chess engine GUIs such as Arena.
I ran Arena tournaments with 64 engines, then played against the loser. Beat it. Second to last also beat. Third from the bottom, the best I could do was draw. This was 15 years ago.
[COMMENT DELETED]
Hey guys ! Sorry for commenting on this after 3 years but regardless, is there An actual way within chess.com to pit bots against each other? For example, Martin vs Mittens, Or Juan vs Pokimane, I Hope you guys get the idea.
Mittens can beat even the strongest bot
White scores about 55% in my computer vs. computer database.
At what time control? 4 min or 40 min from CCRL games?
If you go longer time control like, 3 hours per game, white winning chance will decrease like 52-53%.
In extremely long time control , results may be 51% for white. ( e.g In ICCF, 95% games are draw)
In theory with unlimited time control, results should be 50-50% (assuming balanced opening book)