Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka Elo per depth ?
- - By Hetman (*****) Date 2008-11-16 16:39
Hello,

have anyone idea what are Rybka Elos strength at play depths:

10:
11:
12:
13:
14:

Thanks for the answers ?
Rgds
Hetman
Parent - - By turbojuice1122 (Gold) [us] Date 2008-11-16 18:05
It depends very, very strongly on the opening.  I don't think that there is a meaningful answer for this.  Even if you asked it for each specific opening, it would still be quite tough, as the English Attack in the Sicilian is going to be different than my 2...b6 variation.
Parent - - By Hetman (*****) Date 2008-11-16 19:12
Elo rating is expressing medium strength so it is not dependent on opening. :-)
Parent - - By turbojuice1122 (Gold) [us] Date 2008-11-16 23:58
I assume that by "Rybka Elo strength", you're referring to a playing level such that if Rybka was to play those moves in a game at tournament time controls against grandmasters (or, for that matter, against other computers), it would have such and such Elo.  I am saying that this depends on the opening: in some openings, there tend to be many forced moves so that one piles on the depth much more quickly, and if one fails to get passed a certain depth, the engine will potentially lose due to the horizon effect.  In other openings, this isn't so much the case.
Parent - - By Uri Blass (*****) [il] Date 2008-11-17 06:57
I think that it is very easy to test it.
give rybka to play at fixed depth in CCRL or CEGT and see the rating of it.

Rating in games against humans can be dependent on the opening book but it is correct not only for fixed depth.
I think that we have no money for thousands of games against humans so the interesting question is about CCRL or CEGT rating against  computer programs when the opponent get fixed time control with fixed hardware when rybka get a fixed depth.

Uri
Parent - - By stvs (***) [gr] Date 2008-11-17 13:06
what about positional only understanding at 1 ply related with human elo? its 2000 as vas said? or 2300  as larry said?
Parent - By Uri Blass (*****) [il] Date 2008-11-17 13:54
You cannot give numbers here.

humans use search and rybka 1 ply can be outsearched by humans at long time control when the opposite may happen at 1 minute per game
so the level of rybka at 1 ply is going to prove nothing.

Uri
Parent - - By Hetman (*****) Date 2008-11-17 19:36
lets go from opposite side, Rybka strength at 20ply is about 3000 Elo, could we make any conclusions ?
Parent - - By Uly (Gold) [mx] Date 2008-11-17 20:45

> lets go from opposite side, Rybka strength at 20ply is about 3000 Elo, could we make any conclusions ?


Hm, I think Rybka at 15ply is already 3000 elo, so she should gain something from the extra 5plies.

The problem it's this is pure speculation.
Parent - By Dennis Breder (***) Date 2008-11-18 15:33
I think Rybka at 15ply is already 3000 elo
I doubt that. If some pieces are exchanged, rybka reaches 15ply much quicker so the strengh of rybka should go down during the game. Maybe rybka plays like 3000 elo in the early middlegame but then the strengh should start decreasing with fixed depth. In the endgame rybka should be much weaker at 15ply compared to a normal tournament game.
Parent - - By Hetman (*****) Date 2008-11-20 17:13
3000 ? GM Milov match is showing sth else ;-)
Parent - - By saxon (**) [hr] Date 2008-11-20 18:52
Milov handicap match shows almost nothing regarding these matters.Maybe it shows something along the lines of "handicap rating" (if there's such thing),but very little then.Milov played too small number of games and mixed (various) handicaps vs Rybka. 
Parent - By Hetman (*****) Date 2008-11-21 20:25
Non handicap games Rybka scores 75 %
Parent - - By Roland Rösler (****) [de] Date 2008-11-21 00:11
Nobody knows what´s a ply in Rybka! So why this question? What you see in your GUI is only bullshit!
Parent - - By Uly (Gold) [mx] Date 2008-11-21 01:22

> Nobody knows what´s a ply in Rybka!


Vas does :)
Parent - By Roland Rösler (****) [de] Date 2008-11-21 02:14
Not really! :-)
Parent - - By oddg (**) [no] Date 2008-11-21 08:18
I'm not sure what you want but using fixed ply depth to make a sparring partner is a bad idea. The difference in strength of the engine will be too big in different part of the game. This will lead you to adopt a playingstyle that is not suitable in a normal game.
Much better would be to use time or cpu power if this is an option.
Parent - - By Uly (Gold) [mx] Date 2008-11-21 09:07

> cpu power


How does that work?
Parent - - By oddg (**) [no] Date 2008-11-21 11:00
It looks like some engines use this to lower their strength.
If you want to try this for any engine you could use inbetween (or wb2uci), both have a switch to set the cpu-power.

What it do is to use suspend/resume on the thread to set the cpu usage.
Parent - - By keoki010 (Silver) [us] Date 2008-11-21 15:01
Isn't that just about the same as setting the depth to 1 or the time to x seconds  or less?
Parent - - By Uly (Gold) [mx] Date 2008-11-21 21:06

> Isn't that just about the same as setting the depth to 1 or the time to x seconds  or less?


Yes, but Rybka will take longer to move instead of moving instantly, so the human gets more time to think. I wasn't aware this could be used to handicap the engine.
Parent - By keoki010 (Silver) [us] Date 2008-11-22 14:58
Well, I knew if you gave the engines less time you could weaken it.  Some other engines use that as their strength limits.  I didn't think about using it in Aquarium though. 
Live and learn!
Parent - By Quapsel (****) [de] Date 2008-11-21 15:27

> have anyone idea what are Rybka Elos strength at play depths:
> 10:
> 11:
> 12:
> 13:
> 14:


I think, this depends on the time controls, which are choosen for that game.
If doing 'Blitz' with 5 Minutes per game (or 2 minutes per game) this will result in a rather (very) strong play.
If playing a tournament with 3 minutes per move it will be much weaker (in relation to the strength of the opponents).

Quap
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka Elo per depth ?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill