HOUDINI is to be polite another derivative.
Moderators: hgm, Rebel, chrisw
-
- Posts: 508
- Joined: Fri Jun 04, 2010 7:23 am
Re: HOUDINI is to be polite another derivative.
If the evaluation and search is improved then that implies there is a great chance that there is NO EXACT CODE that matches Rybka 3.
-
- Posts: 305
- Joined: Sat Jan 02, 2010 11:31 pm
- Location: bristol,uk
Re: HOUDINI is to be polite another derivative.
perhaps you would like to play me in a corr game.no engine is ever right on every move but i would be prepared to use only Rybka as my base engine if you did same with Houdini.we often play corr games on Hiarcs forum where this happens and it is a matter of trust.i welcome your reply.
this was in reply to Damir Desevac
this was in reply to Damir Desevac
-
- Posts: 2041
- Joined: Wed Mar 08, 2006 8:30 pm
Re: HOUDINI is to be polite another derivative.
Would that really prove something?NATIONAL12 wrote:perhaps you would like to play me in a corr game.
Corr game has a lot to do with experience, skill and a clever opening book.
Besides, not everybody owns a 12-core monster!...
-
- Posts: 10296
- Joined: Thu Mar 09, 2006 12:37 am
- Location: Tel-Aviv Israel
Re: HOUDINI is to be polite another derivative.
I am afraid that in correspondence games you are going to get 90% drawsNATIONAL12 wrote:perhaps you would like to play me in a corr game.no engine is ever right on every move but i would be prepared to use only Rybka as my base engine if you did same with Houdini.we often play corr games on Hiarcs forum where this happens and it is a matter of trust.i welcome your reply.
this was in reply to Damir Desevac
because engines are almost perfect in correspondence games so you will need many games to get a conclusion.
My experience is that
often a move that is not the best move is not enough to lose the game but only to give the opponent an advantage that is not enough to win.
-
- Posts: 10296
- Joined: Thu Mar 09, 2006 12:37 am
- Location: Tel-Aviv Israel
Re: HOUDINI is to be polite another derivative.
I think that the rules in correspondence games to test should include no opening book and games from fixed positions.ernest wrote:Would that really prove something?NATIONAL12 wrote:perhaps you would like to play me in a corr game.
Corr game has a lot to do with experience, skill and a clever opening book.
Besides, not everybody owns a 12-core monster!...
I do not have Rybka4 but people who have Rybka4 do not need an opponent and they can simply play a correspondence game against themselves from the noomen positions or from different positions that they choose.
It may be good if they post the positions that they choose for the match and the time control in infinite mode that they use before the beginning of the game so they cannot discard quietly results that they do not like.
-
- Posts: 44
- Joined: Wed Jan 27, 2010 2:58 pm
Re: HOUDINI is to be polite another derivative.
And then, there is that little item that 800+ games are needed to establish a reliable result.
-
- Posts: 1766
- Joined: Wed Jun 03, 2009 12:14 am
Re: HOUDINI is to be polite another derivative.
ingo bauer ran 1900...and uniformly taken you will easily find another independent 800+ that confirm his results.
-
- Posts: 44
- Joined: Wed Jan 27, 2010 2:58 pm
Re: HOUDINI is to be polite another derivative.
The 800+ games was meant to refer to correspondence games.
-
- Posts: 10121
- Joined: Thu Mar 09, 2006 12:57 am
- Location: van buren,missouri
Re: HOUDINI is to be polite another derivative.
NOt much differences. They still use say R4 moves.John_F wrote:The 800+ games was meant to refer to correspondence games.
Then they call themselves GM's hahahaha.
Best,
Gerold.
-
- Posts: 5106
- Joined: Tue Apr 29, 2008 4:27 pm
Re: HOUDINI is to be polite another derivative.
Where did you post the proof? I would like to take a look at it. It's my understanding that it is a much improved Robbolitto. Is that what you are seeing?CThinker wrote:I disassembled Houdini, posted the proof, and there are people who still refuse to believe that it is nothing but an Ippolit.gaard wrote:Calling something that trumps its predecessor from which you believe it is a clone of, by ~100 Elo, "just another fast clone" is anything but polite. Better would be "Houdini is to be insulting another derivative."
Houdini 1.5 has been out, what, 72 hours? Yet, you have already determined that it "fails miserably" at correspondence time controls, for example, 48 hours/move? Better to wait till you reach a sample size of 1 first before you draw such far reaching conclusions.
Don't we have a separate forum for these _Engine Origins_ discussions?
There are people who want their engine, even if it is a product of dishonesty.