Topic Rybka Support & Discussion / Rybka Discussion / BETTER way of get a reliable move in a position?
hi, with a strong hardware like an i7 980x what will you used to get the better moves (better line) in a tricky chess positions?
please explain your choice
1) 3 to 5 days with infinity analisis
2) 100k+ games in montecarlo
3) 3 to 5 days with idea
4) 1000+ games 3min+4sec from a tournament with the 6 better engines begining in that position
5) other
thanks for your comments
Manuel
please explain your choice
1) 3 to 5 days with infinity analisis
2) 100k+ games in montecarlo
3) 3 to 5 days with idea
4) 1000+ games 3min+4sec from a tournament with the 6 better engines begining in that position
5) other
thanks for your comments
Manuel
1) or 3) .
3) will give your much more valuable information, but 1) may give the better move, although the time to depth issue could make it worse.
3) will give your much more valuable information, but 1) may give the better move, although the time to depth issue could make it worse.
You can probably get much deeper by using inter active analysis. Also a lot really depends on the position ... in a very complicated tactical position I would probably just leave the engine on 3 mpv and let it analyze for a couple of days. In a position where it seems like there is no progress to be made (blocked position) I would probably run MC overnight. Again interactive will probably take you further as you can quickly eliminate moves that are obviously making no progress.
> I would probably just leave the engine on 3 mpv and let it analyze for a couple of days
...not possible with Rybka 4...
have at least 3 computers and run on IA with top 3 moves,better if you have 4th computer to run with your own ideas.
You have enough computers to do that if you wanted to 

not at the moment,just trying to improve present set-up atm.
You mean using 3 to 4 computer using idea for analisis the better 3 to 4 moves? each computer for 1 move? if yes, i will want to know more about that (i am new to idea software) i allready have 5 computers, and rybka aquarium 4, all my pcs are 64bit and use win 7 64bit ultimated. i have 3 dual core, 1 with 4 core, and 1 with 4 core and 8 treath. in total i think are 18 logical cores and 14 fisical cores.
i have to make a win 7 home network first?
then run idea from rybka aquarium and it will see all 18 cores for use it?
thanks for help,
Manuel
i have to make a win 7 home network first?
then run idea from rybka aquarium and it will see all 18 cores for use it?
thanks for help,
Manuel
sorry,never used IDEA,just use IA.
I think that your games against Crawly are evidence that IDEA isn't necessary!
Very true but not for long ... I hope!
sorry i dont get MC? :(
what softwre you prefer for your interactive analisis?
thanks
Manuel
what softwre you prefer for your interactive analisis?
thanks
Manuel
MC is Monte Carlo analysis and it comes with the chessbase version of Rybka. Basically what it does is play thousands or even hundreds of thousands of very quick games in quick succession and giving an output of how the move does and thus feeding moves (even moves that seem very silly) and checking out how they score. In the end it gives you the output as a tree giving you the percentage of how well a certain move did in self play. I have used it successfully quite a few times to find obscure draws in positions that seemed totally winning and to find breakthroughs in otherwise totally blocked positions.
Next time you come across any positions where MC finds obscure draws or breakthroughs, you should post them. I'm sure they'd result in a lot of interesting discussions. For example, I'd like to test IDeA in one of these positions.
Actually I did that in a post a while back on CCC where a position was put up for discussion that was also on the chessgames forum. Everyone thought that there was no way for black to hold the position but when I left MC overnight I found that there was an obscure bishop move that by some miracle would draw in all lines. This was later confirmed by others, and pretty much convinced me of how useful MC can be sometimes. IDEA is also supposed to be very similar to MC although I have never used it.
> IDEA is also supposed to be very similar to MC
I let IDeA fans answer that!...
I'll take a look at CCC. I probably read it at the time, but don't remember it now.
Also, IDeA is nothing like MC. You should try it!
Also, IDeA is nothing like MC. You should try it!
What settings do you usually use to run MC sessions of about 6 hours? (4 cores)
Thoroughly exploring all available alternatives is the best way to consistently find good moves. How you do it doesn't really matter - interactive infinite analysis, IDeA, whatever -- that's largely personal preference (I prefer IDeA). It doesn't matter so much how you do it, just that it gets done!
do you used more than 1 computer simultaneus with idea?
> do you used more than 1 computer simultaneus with idea?
Sometimes. It depends on the project.
(5) Sell PC on EBAY for £1-2k; pay Magnus £100/hour for advice; get a great plan {and great company talking about chess all afternoon} to execute against; use money left over to over-indulge in drink should plan not work out.
+1 


This plan is flawed. Magnus won't even spit on you for £100/hour.
give me £100 an hour and i will spit on him.
i still believe the best move will be found by a combination of engine + skilled player. i think a strong correspondence GM with a relatively weak engine, say a 32bit rybka 3 will still find the better moves over a 64bit quad core rybka 4 that has no human assistance. if this is no longer the case i might as well give up coresspondence chess right now.
i agree with djbl. the best way to find better moves is 50% hard ware 50% player- you need to get skilled in the basics and understand what makes a move strong or weak better-
once you learn enough from games- start looking at all the moves of your engine matches and trying to find the fall off that is bound to happen to each move in the essence of the basic principles in chess principles.
i found a good way to learn is not to always play your strongest engines against each other because the point where one side capitalizes on the others errors might be to small for the player to fully understand because of the metric level they play on is so high- buy chess master for ps and watch and learn how rybka beats the gm opponents- you learn more from these games then you will from top engine games esp if your elo is is under 2200. the key to good chess is spotting a mis calc and knowing how to capitalize.
which would you learn more from if you were an inexperienced fighter.... two of the best fighters in the world fighting for 1 minute, or watching the best fighter against an ok fighter? you'll learn from both-but if your already one of the best you would benefit from the first, if you are just a good chess player/fighter you will learn more from the latter-
once you learn enough from games- start looking at all the moves of your engine matches and trying to find the fall off that is bound to happen to each move in the essence of the basic principles in chess principles.
i found a good way to learn is not to always play your strongest engines against each other because the point where one side capitalizes on the others errors might be to small for the player to fully understand because of the metric level they play on is so high- buy chess master for ps and watch and learn how rybka beats the gm opponents- you learn more from these games then you will from top engine games esp if your elo is is under 2200. the key to good chess is spotting a mis calc and knowing how to capitalize.
which would you learn more from if you were an inexperienced fighter.... two of the best fighters in the world fighting for 1 minute, or watching the best fighter against an ok fighter? you'll learn from both-but if your already one of the best you would benefit from the first, if you are just a good chess player/fighter you will learn more from the latter-
Topic Rybka Support & Discussion / Rybka Discussion / BETTER way of get a reliable move in a position?
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill