For the old hands, a question: What's the ideal setting for machine vs machine matches?
I realize the ultimate test of maximum strength is 40 moves in 2 hours. However, that's a long time for those of us who want a quick fix.
In my own little experiments, I lean toward 10 seconds per move (allowing each machine to think on the other's time, which allows me to get up for a cooling beverage, etc.) Most machines have that setting, and it seems to yield fair results.
However, I'm open to suggestions, and I've noticed others talking about finding a given machine's so-called "sweet spot" for calculation.
At the moment, I'm messing around with a four-machine tournament with the following contestants:
Designer 2265
Mephisto Milano
Novag Obsidian
GK 2100
This contest would match up four machines by four different programmers (Spracklen, Schroder, Kittinger and Morsch). Their ratings are very similar - within 15 points of each other, according to the Wiki-Elo numbers.
I've run a few games already, and I've been especially surprised by Milano's performance. I'll show a few games on the forum when I build up a bigger list of results. One thing is clear - all these machines play endgames badly.
So again - is 10 seconds per move a reasonable setting, or should I jack up the strength?
Ideal settings (machine vs machine)
Moderators: Harvey Williamson, Steve B, Watchman
Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
Ideal settings (machine vs machine)
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
– H.G. Wells
Hello!
If I remember correctly the minimal time control for such tourneys was 30sec/move. Myself prefer 30 or 60 min per game. Also I have a feeling that H8-computers have more advantage against 6502-computers in speed chess - if a time control becomes closer to classical, than they get in more equal / fair terms of play. For example, GK2100 beats Milano at 30min/game mostly because it has ~2x MaxMIPS advantage and counts 2-3 plyes deeper in tactics.
Countdown controls (X-min/game) also show how good the time managment is used by different computers. To my mind, at 30min/game Milano and Polgar play faster than the toughness of a situation demands!
If I remember correctly the minimal time control for such tourneys was 30sec/move. Myself prefer 30 or 60 min per game. Also I have a feeling that H8-computers have more advantage against 6502-computers in speed chess - if a time control becomes closer to classical, than they get in more equal / fair terms of play. For example, GK2100 beats Milano at 30min/game mostly because it has ~2x MaxMIPS advantage and counts 2-3 plyes deeper in tactics.
Countdown controls (X-min/game) also show how good the time managment is used by different computers. To my mind, at 30min/game Milano and Polgar play faster than the toughness of a situation demands!
-
Steve B
- Site Admin
- Posts: 10146
- Joined: Sun Jul 29, 2007 10:02 am
- Location: New York City USofA
- Contact:
Sounds like an interesting tournament
I would concur with IvenGO
on any of the rating lists a minimum of 30 sec a move is used ...
30 minutes for the entire game is also an acceptable minimum i think
this is called the "Active Rating" as opposed to the "Standard rating" which is 40/2
Personally i prefer 1 Min per move Avg. when i play my games because it does seem to avoid most gross blunders that can ruin an otherwise equal game
i can say that from all of the games i've played it is quite common for a computer to ultimately select a different(better) move then the move it would play if only allowed 10 seconds to ponder
one game at 1 Min/Avg. generally takes about 2-3 Hrs to complete
so for a 30 Sec /Avg.. you are looking at 1/2 that time to finish one game
Old Timer Regards
Steve
I would concur with IvenGO
on any of the rating lists a minimum of 30 sec a move is used ...
30 minutes for the entire game is also an acceptable minimum i think
this is called the "Active Rating" as opposed to the "Standard rating" which is 40/2
Personally i prefer 1 Min per move Avg. when i play my games because it does seem to avoid most gross blunders that can ruin an otherwise equal game
i can say that from all of the games i've played it is quite common for a computer to ultimately select a different(better) move then the move it would play if only allowed 10 seconds to ponder
one game at 1 Min/Avg. generally takes about 2-3 Hrs to complete
so for a 30 Sec /Avg.. you are looking at 1/2 that time to finish one game
Old Timer Regards
Steve