Proper settings for accurate ELO play

Get your specific HIARCS/Junior support questions answered here as well as up-to-the-minute news!

Moderators: Watchman, Mark Uniacke, mrudolf

Post Reply
vsolimine
Member
Posts: 2
Joined: Sun Feb 03, 2008 10:44 pm

Proper settings for accurate ELO play

Post by vsolimine »

I have just downloaded and begun using Hiarcs 11.2 SP. It is as impressive as it is purported to be.

My question is…what are the correct/minimum settings to insure that Hiarcs is playing at the requested ELO strength.

I am using the Arena 1.99beta4 GUI and there are so many engine settings to play with, not to mention playing time per game as well.

Thank you in advance for any assistance.
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Post by Harvey Williamson »

On its default settings it will play at maximum strength. If you want to reduce the strength then tick the uci limit strength parameter and then set the elo you want Hiarcs to play at.
vsolimine
Member
Posts: 2
Joined: Sun Feb 03, 2008 10:44 pm

Post by vsolimine »

Thanks for the reply Harvey. So if I set it to 1800 ELO with the defaults set it should play at 1800. But what about the Ponder option? Should that be on or off for accurate 1800 play? Does Ponder=on make it play stronger than the selected ELO (1800) or at that level?

Also, what about time limits for the game? What is the minimum amount of time that Hiarcs needs to play at the selected (1800) strength? 30 minutes or 1 hour or more?
User avatar
Dylan Sharp
Senior Member
Posts: 2431
Joined: Fri Aug 10, 2007 12:07 am

Post by Dylan Sharp »

Harvey Williamson wrote:On its default settings it will play at maximum strength. If you want to reduce the strength then tick the uci limit strength parameter and then set the elo you want Hiarcs to play at.
How accurate is this? Will Hiarcs play at these 1800 ELO the same at an old 1Ghz computer than in the latest over-clocked Quad?

That's hard to grasp.
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Post by Harvey Williamson »

vsolimine wrote:Thanks for the reply Harvey. So if I set it to 1800 ELO with the defaults set it should play at 1800. But what about the Ponder option? Should that be on or off for accurate 1800 play? Does Ponder=on make it play stronger than the selected ELO (1800) or at that level?

Also, what about time limits for the game? What is the minimum amount of time that Hiarcs needs to play at the selected (1800) strength? 30 minutes or 1 hour or more?
You must remember to enable uci limit strength in the engine parameters as well as setting the elo.

I have always left ponder on when trying this but why not experiment with it on and off!?
Henrik Dinesen
Member
Posts: 18
Joined: Tue Jul 31, 2007 4:06 pm
Location: Denmark

Post by Henrik Dinesen »

How accurate is this? Will Hiarcs play at these 1800 ELO the same at an old 1Ghz computer than in the latest over-clocked Quad?
I agree - how to define accurately - since it have to be given the hardware too. I know that the limit strength parameter does nothing but attempt to paly at a gicen strenght, but surely hardware (and time) must make some big differences for the accuracy when talking ELO-estimation.

Is the setting calibrated to a specific cpu-speed/type for the estimate to work?
User avatar
turbojuice1122
Senior Member
Posts: 2315
Joined: Thu Aug 23, 2007 9:11 pm

Post by turbojuice1122 »

Just thinking about it now, I think that as long as the strength of the hardware is such that the program's maximum strength is considerably stronger than the maximum UCI limit strength, it would probably play at about the same UCI limit strength for a particular on all such hardware. This would be because it would still make the same number of "mistakes" of certain severity per number of moves or whatever.
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Post by Harvey Williamson »

turbojuice1122 wrote:Just thinking about it now, I think that as long as the strength of the hardware is such that the program's maximum strength is considerably stronger than the maximum UCI limit strength, it would probably play at about the same UCI limit strength for a particular on all such hardware. This would be because it would still make the same number of "mistakes" of certain severity per number of moves or whatever.
A simpler way and to get sensible chess rather than stupid moves played is restrict to a number of nodes searched or depth - this can be applied on any hardware and should give similar results.
User avatar
turbojuice1122
Senior Member
Posts: 2315
Joined: Thu Aug 23, 2007 9:11 pm

Post by turbojuice1122 »

Harvey Williamson wrote:
A simpler way and to get sensible chess rather than stupid moves played is restrict to a number of nodes searched or depth - this can be applied on any hardware and should give similar results.
I think you would still get stupid moves here, possibly just as much as before: the horizon effect would become glaringly obvious on forced lines, and there would be some stupid moves that lead into forced losing lines.
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Post by Harvey Williamson »

turbojuice1122 wrote:
Harvey Williamson wrote:
A simpler way and to get sensible chess rather than stupid moves played is restrict to a number of nodes searched or depth - this can be applied on any hardware and should give similar results.
I think you would still get stupid moves here, possibly just as much as before: the horizon effect would become glaringly obvious on forced lines, and there would be some stupid moves that lead into forced losing lines.
Your right but the computer would not be deliberately playing stupid moves and the elo set would be consistent on any hardware.
Post Reply