How will play 500 elo stronger engines?

[cdani]

I think this gives a nice idea:

http://en.chessbase.com/post/better-tha ... ubicic-1-2

[S.Taylor]

Can someone summarize this?

[Damir]

Hi Daniel

How is development of Andsacs going on? It has been some times since you have released a new version. I have birthday March 30th. Any chance you can release it by that time ?

[Ozymandias]

Can someone summarize this?

Know your engines, give them time, be careful what opening you choose and where the position is headed for.

[cdani]

Can someone summarize this?

Leonardo Ljubicic, the new World Champion of Correspondence Chess, talks about how he works, the weaknesses and strengths of engines, and other related stuff. For me this reflects possible ways of improvement of engines.

Hi Daniel

How is development of Andscacs going on? It has been some times since you have released a new version. I have birthday March 30th. Any chance you can release it by that time ?

Hi
I don't want to release new versions that often to not give that much work to testers Also not always one can reach some minimum advances. Seems not to be the case for the moment as I think I have more than 20 elo more, but anyway I will wait. Maybe March 30th is a good date

[Frank Quisinsky]

Hi Daniel,

you should not thinking on testers. I am sure that all of the testers have a system in testing engines. If you like to release a new version, do it.

BTW:
Next Andscacs test from my site will be start in around 10 days. If you like I can test a newer release version as 0.85. Andscacs have to play vs. 59 opponents, so you can looking 7 days if you like (2.950 games need around 7 days).

Best and ... yes, I like your engine too.
Frank

[Nay Lin Tun]

I am confused with your title. Do you really think that Cor player +engine is 500elo stronger than an engine? There are only few exceptions that an engine can miss due to extensive prune + less knowledge. In reality, these Cor players (let say 2200-2500) will feed 9 out of 10 crappy idea to an engine and in fact it will even weaken the engine. Even if a perfect programmer feed certain chess knowledge that is uncommon in most games, it may even weaken the engine. (eg, putting addition code if *** then ***, unless *** in main search and it will somewhat slower an engine). It might be possible to increase elo of an engine in massive long time control by less pruning and adding more knowledge in search or evaluation loop but it is still questionable . IMHO,those correspondence players who hold the Correspondence GM title, will play poorly against GM (from FIDE) or a pure engine in common OTB time setting, and I really dont think their idea can improve an engine.

[Ozymandias]

I am confused with your title. Do you really think that Cor player +engine is 500elo stronger than an engine?

That's what I'm taking from the title, too.

IMHO,those correspondence players who hold the Correspondence GM title, will play poorly against GM (from FIDE) or a pure engine in common OTB time setting, and I really don't think their idea can improve an engine.

Now it's me who's confused. Are you talking about man vs man and man vs machine matches? Otherwise, you're talking Freestyle chess, where you can actually see top correspondence chess players, playing against standalone engines, at FIDE long time control settings (although not always valid, for ELO higher than 2200).

[cdani]

I am confused with your title. Do you really think that Cor player +engine is 500elo stronger than an engine? There are only few exceptions that an engine can miss due to extensive prune + less knowledge. In reality, these Cor players (let say 2200-2500) will feed 9 out of 10 crappy idea to an engine and in fact it will even weaken the engine. Even if a perfect programmer feed certain chess knowledge that is uncommon in most games, it may even weaken the engine. (eg, putting addition code if *** then ***, unless *** in main search and it will somewhat slower an engine). It might be possible to increase elo of an engine in massive long time control by less pruning and adding more knowledge in search or evaluation loop but it is still questionable . IMHO,those correspondence players who hold the Correspondence GM title, will play poorly against GM (from FIDE) or a pure engine in common OTB time setting, and I really dont think their idea can improve an engine.

I had not the idea of hours-long for one move time control, but for example current CCRL tournament time control with engines that will play 500 elo stronger that current ones.

About adding chess knowledge to engines, I can say that I can add relative often improvements to Andscacs just viewing bad moves of it, even if it is a lot stronger than me. The engines simply cannot understand a position when there is some long term positional idea out of its current reach, or maybe they realize too late.

Another interesting thing is that maybe people that plays correspondence chess can give us an estimation of the level of players that only use engines without adding anything, to compare to the current champion, so we can have an idea of the added value of an human player.

Next Andscacs test from my site will be start in around 10 days. If you like I can test a newer release version as 0.85.

You can test 0.85. I really will wait to publish a new one. Thanks!!

[Ozymandias]

Another interesting thing is that maybe people that plays correspondence chess can give us an estimation of the level of players that only use engines without adding anything, to compare to the current champion, so we can have an idea of the added value of an human player.

Given the nature of corr chess, I don't think it possible, to isolate one factor. Over the course of a game, players will change hardware, software or allocate time on a very random basis, because of real life engagements. The consistency of their play will be even lower, when taking into account multiple games. I can tell you, though, that a player using just one unattended engine, will be easily (in a fraction of the time) overcome by a centaur. How much ELO does that translate into, no one can really say.