Page 1 of 1
Random mover win rate
Posted: Wed Jul 24, 2019 7:56 am
by xr_a_y
I add a random mover level to Minic and i'd like to "validate" it.
What is the expected draw/win rate of the random mover playing against itself ?
Re: Random mover win rate
Posted: Wed Jul 24, 2019 9:52 am
by Sven
I would expect a 50% score against itself (i.e., two identical random movers facing each other), with sufficient number of games.
Re: Random mover win rate
Posted: Wed Jul 24, 2019 10:06 am
by Guenther
xr_a_y wrote: ↑Wed Jul 24, 2019 7:56 am
I add a random mover level to Minic and i'd like to "validate" it.
What is the expected draw/win rate of the random mover playing against itself ?
I expect a draw rate of may be 95% between to rnd movers as they seldomly can enforce wins, no matter how much material difference
inbetween a game.
Re: Random mover win rate
Posted: Wed Jul 24, 2019 10:10 am
by hgm
15.3% wins, 84.7% draws. See
http://www.talkchess.com/forum3/viewtopic.php?t=40193 for a subdivision of the draws. (This assumes 3-fold-rep and 50-move draws will be claimed at the first opportunity.)
Re: Random mover win rate
Posted: Wed Jul 24, 2019 11:13 am
by JohnWoe
By Sapeli. Quickly 10_000 games. "-aiw 0" is a random mover.
Code: Select all
...
~~~ Results ~~~
[ Score: 740 - 717 - 8543 [0.50] 10000 ]
Win rate: (740+717)/10000 ~14.57%
Draw rate: 8543 / 10000 ~ 85.43%
Edit: These numbers are for random chess960 positions...
Re: Random mover win rate
Posted: Wed Jul 24, 2019 1:32 pm
by AlvaroBegue
After searching the web for about 3 minutes:
http://wismuth.com/chess/random-games.html
Enjoy!
Álvaro.
Re: Random mover win rate
Posted: Wed Jul 24, 2019 2:40 pm
by xr_a_y
Yeah I was a bit lazy to ask that here ... sorry guys ... Minic is near this with 86% draw, i'll check deeper tonight.
Thank you all
Re: Random mover win rate
Posted: Wed Jul 24, 2019 2:51 pm
by hgm
Note that the win rate is slightly larger here than in the Edwards study (both at least a billion games, so the difference is significant): 15.46% vs 15.3%. This is because it uses 6-fold-rep and 75-move rules for declaring draw. Some of the games that would have been draw after 3-fold-rep or 50-move will still result in a checkmate in the mean time.
Re: Random mover win rate
Posted: Wed Jul 24, 2019 3:23 pm
by xr_a_y
More details for Minic gives
Code: Select all
709 0-1 {Black mates}
644 1-0 {White mates}
1524 1/2-1/2 {Draw by 3-fold repetition}
1503 1/2-1/2 {Draw by fifty moves rule}
4689 1/2-1/2 {Draw by insufficient mating material}
923 1/2-1/2 {Draw by stalemate}
9992 games (there was 8 disconnection of program ... I'll check for the bug later ...)
White win 6.44%
Black win 7.09%
3fold 15.25%
50move 15.04%
insifficient 46.92%
stalemate 9.23%
I'll have to investigate
Re: Random mover win rate
Posted: Thu Jul 25, 2019 11:06 am
by xr_a_y
Ok last figures after a bug fix
Code: Select all
1540 0-1 {Black mates} -> 7.62%
4 0-1 {White's connection stalls}
6 1-0 {Black's connection stalls}
1498 1-0 {White mates} -> 7.41%
497 1/2-1/2 {Draw by 3-fold repetition} -> 2.46%
4499 1/2-1/2 {Draw by fifty moves rule} -> 22.27%
10883 1/2-1/2 {Draw by insufficient mating material} -> 53.87%
1272 1/2-1/2 {Draw by stalemate} -> 6.29%
total 20199
This is near expected ...
Still 10 games went wrong