Very fast play by LuiseR Einstein forum
2 replies. Last post: 2007-09-05
Reply to this topic Return to forumReturn to forum
You have 0 new messages
You have 0 games on move.
You have 0 invitations to game.
2 replies. Last post: 2007-09-05
Reply to this topic Return to forumIn another thread there was some discussion
on speed of play in “EinStein wurfelt nicht”.
To give you an idea, what really fast play
can mean: Recently an experiment took place
in my office. It was a 60-minutes match in
Ewn. One player was the computer program
mcrandom_c, running on a 1-GHz PC. The
opponent was LuiseR, sitting in front of
the PC with mouse in her hand.
Within the 60 minutes 273 (in words: twohundred
and seventythree) single games were played,
so in the average more than 4.5 games per minute.
The final score was 174-99 in favour of LuiseR.
I was there as referee and taking statistical notes.
Translating the result into the “best of 5” system
of LittleGolem, LuiseR would have won by 52-17.
Currently the score of mcrandom_c on LittleGolem
is 1561. The 75-percent win of LuiseR (in best-of-5
counting) would give her a “blitz rating” of about
1761. (For comparison: I myself clearly miss the 50
percent line when playing against mcrandom_c with
speed of 2-3 games per minute.)
There was some incentive for LuiseR to play strong
and quick in the match: For each single win she got
10 European Cent, for each loss no payment was involved.
Each game had to be played to the very end.
(In total, the 60 minutes brought her 17.40 Euro -
from my private pocket.)
\\\********************
When someone else is interested to participate in a
similar experiment (to be executed in Jena), he or she
should contact me well in advance.
Ingo.
Using a self-written perl/tk script that removes the move confirmation step, and reduces the amount of communication with the LG server to just one transaction per move, I just played 10 games (in parallel) with RoRoRo in 23 minutes. I spend most of the time waiting, and believe that if I can get the network communication to happen without blocking the GUI, then this time can be reduced enormously, and I suspect I can sustain 40-50 parallel games, at the cost of hammering the server a lot more (so I could at least double, and probably triple, my throughput). I won 3 of the games.