Topic Rybka Support & Discussion / Rybka Discussion / What chess program has the chances to ever surpass HOUDINI?
![[ro]](/mwf/v12/flags/ro.png "Romania")
I'm thinking about RYBKA ofcourse, but there are some odds for CRITTER, KOMODO and STRELKA. We don't have information about the private engines THINKER and CRYPTIC. I'm a little worried about the fact that HOUDINI 3 was announced to be a BEAST and.... I don't know for sure if DEEP RYBKA 5 could be stronger than him..... Unfortunately, we still have to wait for the 15th version of the Chessbase engines to have some chances to compete at the highest level..... Perhaps only Rybka Cluster will show up in tournaments with a possible clash of the titans against HOUDINI on a huge hardware and Rybka 5 will remain just a dream...... Best regards!
![[za]](/mwf/v12/flags/za.png "South Africa")
Strelka
Most progress in shortest time and stealing ideas from commercial engines
Most progress in shortest time and stealing ideas from commercial engines
If Vas released Rybka 5 now, I'm sure it would surpass Houdini. And I think Komodo MP would as well. But for how long remains to be seen.
so if he releases rybka 5 now,and that will surpass Houdini,do you think there is any reason to dont release rybka now?
i have one to dont do it now:he knows that if releases rybka now,it wont surpass Houdini,and to avoid a fail,he doesnt release it...simply and very clear answer to myself
i have one to dont do it now:he knows that if releases rybka now,it wont surpass Houdini,and to avoid a fail,he doesnt release it...simply and very clear answer to myself
Critter 1.6
(of course that's just a guess)
(of course that's just a guess)
as soon as i will give the values inside Houdini2.0z to everyone,and as soon as Houdart will notice it and implement those values,be sure that Critter will need at least another 13 elo points...and appart from that,add the improvement that Houdini 3 will be over Houdini 2
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
Two points:
- have you/anyone yet proven your tuning values across other engines or is it still only based on a head-to-head with H2c?
- assuming point one is not just based on a head-to-head performance against H2c then what makes you think that H3 will benefit from your changes?
- have you/anyone yet proven your tuning values across other engines or is it still only based on a head-to-head with H2c?
- assuming point one is not just based on a head-to-head performance against H2c then what makes you think that H3 will benefit from your changes?
yes,we proved
i ran a match of 1000 games against Critter 1.4,and it gave a +57 elo points for the tuned version
also im running a match of 1000 games against Stockfish 2.2.2,and after 693 games,gives a +128 elo points for Houdini 2.0z
my friend is running a match against Komodo 4 (both running in 1 core),and after 996 games,+84 elo points for Houdini2.0z
also Keoki ran a match of 1000 games Houdini2.0z vs Houdini 2.0c,and it show an improvement of +12 elo points over Houdini 2.0c
and also keoki is running a tournament between Houdini2.0z,Houdini2.0c,rybka and Critter,and after near 300 games,Houdini 2.0z is hirst,and the second,Houdini2.0c,is + 28 elo points below
second question:
look the perfomances of the tuned version,and look how it overpass all the engines,and even some of them,very very overpassed,so it will be foolish dont implement that changes in Houdini 3,dont you think?
i ran a match of 1000 games against Critter 1.4,and it gave a +57 elo points for the tuned version
also im running a match of 1000 games against Stockfish 2.2.2,and after 693 games,gives a +128 elo points for Houdini 2.0z
my friend is running a match against Komodo 4 (both running in 1 core),and after 996 games,+84 elo points for Houdini2.0z
also Keoki ran a match of 1000 games Houdini2.0z vs Houdini 2.0c,and it show an improvement of +12 elo points over Houdini 2.0c
and also keoki is running a tournament between Houdini2.0z,Houdini2.0c,rybka and Critter,and after near 300 games,Houdini 2.0z is hirst,and the second,Houdini2.0c,is + 28 elo points below
second question:
look the perfomances of the tuned version,and look how it overpass all the engines,and even some of them,very very overpassed,so it will be foolish dont implement that changes in Houdini 3,dont you think?
![[it]](/mwf/v12/flags/it.png "Italy")
Read on Houdini facebook page about the current progress and we are not talking about +12 Elo. Houdini 3 is the number one place and maintain. Komodo Multiprocessor is not available and there is no definite date for it. Komodo is still struggling against Houdini 1.5! Critter will not make a big improvement , but it will be a better analysis tool, said Richard Vida. Only HE can even talk about his new Critterversion. Rybka 5 will never appear, there is only talking about it, so the hope is not lost. stockfish is only a hobby experiment for leisure. Mr. Houdart is the only known wecher a specific date. Houdart is great,cheers!
i never told that Houdini isnt the best engine;in fact,im playing a match to prove Houdini is the best engine;i can only use Houdini,and my opponent cant use Houdini
well,if you think a +12 or +13 elo points is only a bit improvement,you can try to tune the engine yourself,and lets see if you can add another 12 or 13 elo points
well,if you think a +12 or +13 elo points is only a bit improvement,you can try to tune the engine yourself,and lets see if you can add another 12 or 13 elo points
I try again to explain my point of view, but also intended as "constructive" criticism.First Where are the results of the default Houdini 2.0c against the same opponents and the same conditions, for example, openings, or you are using books.
Second: Elo compare other rating lists as a starting point not korrekt, other conditions: 40/4, 5 +3, 40 / 20, ponder off, on, etc.
Third: Stockfish and Komodo are very bad engines in short TC, especially Komodo.
Finally, I have never claimed that I create a tuned version, so people no longer have to buy Houdini3 here. Mr. Bernard, but you said that. I'd rather wait for the official developer version, so I know I get the best engine too.
but should be the tuned version only +12, the people will certainly buy the Houdini 3, sorry
Second: Elo compare other rating lists as a starting point not korrekt, other conditions: 40/4, 5 +3, 40 / 20, ponder off, on, etc.
Third: Stockfish and Komodo are very bad engines in short TC, especially Komodo.
Finally, I have never claimed that I create a tuned version, so people no longer have to buy Houdini3 here. Mr. Bernard, but you said that. I'd rather wait for the official developer version, so I know I get the best engine too.
but should be the tuned version only +12, the people will certainly buy the Houdini 3, sorry
Hi
answers to your questions:
1)results of the Houdini2.0c against the same opponents are in all the lists (ipon,ccrl,etc...)
2)i never compared elo against any list;i just took as reference the ipon list,to know the elo of Houdini2.0c...but if you want,you can take as reference ccrl
3)yes,i claimed and i wont regret of my words;im pretty sure that if properly tuned,i can add 15 elo points more to the tuned version Houdini2.0z...for the people that will think im a fool,i must remember them that no one thought i can tune Houdini2.0c to add 10 elo points,and i added between 12 and 13 wlo points
i never told you aren doing a constructive criticism,i just told that i also think Houdini2 is the better enigne
answers to your questions:
1)results of the Houdini2.0c against the same opponents are in all the lists (ipon,ccrl,etc...)
2)i never compared elo against any list;i just took as reference the ipon list,to know the elo of Houdini2.0c...but if you want,you can take as reference ccrl
3)yes,i claimed and i wont regret of my words;im pretty sure that if properly tuned,i can add 15 elo points more to the tuned version Houdini2.0z...for the people that will think im a fool,i must remember them that no one thought i can tune Houdini2.0c to add 10 elo points,and i added between 12 and 13 wlo points
i never told you aren doing a constructive criticism,i just told that i also think Houdini2 is the better enigne
![[il]](/mwf/v12/flags/il.png "Israel")
I am going to wait for the rating lists before buying houdini3.
Note that I did not buy houdini2 because there was no clear improvement relative to 1.5 based on all rating lists
so I cannot use houdini2z.
Note that I can only see 7 elo improvement by comparing houdini2c 4 cpu to houdini1.5a 4 cpu in the ccrl rating list.
http://computerchess.org.uk/ccrl/4040.live/cgi/compare_engines.cgi?family=Houdini&print=Rating+list&print=Results+table&print=LOS+table&print=Ponder+hit+table&print=Eval+difference+table&print=Comopp+gamenum+table&print=Overlap+table&print=Score+with+common+opponents
If I see a significant improvement with houdini3 relative to houdini1.5 based on rating lists(more than 40 elo in all lists) then probably I am going to buy houdini3.
Note that I did not buy houdini2 because there was no clear improvement relative to 1.5 based on all rating lists
so I cannot use houdini2z.
Note that I can only see 7 elo improvement by comparing houdini2c 4 cpu to houdini1.5a 4 cpu in the ccrl rating list.
http://computerchess.org.uk/ccrl/4040.live/cgi/compare_engines.cgi?family=Houdini&print=Rating+list&print=Results+table&print=LOS+table&print=Ponder+hit+table&print=Eval+difference+table&print=Comopp+gamenum+table&print=Overlap+table&print=Score+with+common+opponents
If I see a significant improvement with houdini3 relative to houdini1.5 based on rating lists(more than 40 elo in all lists) then probably I am going to buy houdini3.
hi
take a look at Sedat rating list;Houdini2.0z is 20 elo points above Houdini2.0c,and the new version that im tuning.'s' version ,is stronger than Houdini2.0z
p.s.i recall IM TUNNING,it is NOT my engine or im claiming anything about it
take a look at Sedat rating list;Houdini2.0z is 20 elo points above Houdini2.0c,and the new version that im tuning.'s' version ,is stronger than Houdini2.0z
p.s.i recall IM TUNNING,it is NOT my engine or im claiming anything about it
Hi Barnard,
I'm not sure the tests you are running constitute the use of the word 'proved', having said that, your results certainly look encouraging. Personally I would look to see if Ingo {IPON fame} would be prepared to run a complete test as most of the engines you mention have a strong similarity - also, the IPON time controls are a little better in my view and suspect you are playing with ponder off whereas the IPON list is based on ponder on which simulates a more realistic scenario in my view.
The second point I made is not that an improvement should be ignored, merely questioning how you know that the new H3 could make use of your changes and even if it could how you know that this would have the same positive effect - make sense?
I'm not sure the tests you are running constitute the use of the word 'proved', having said that, your results certainly look encouraging. Personally I would look to see if Ingo {IPON fame} would be prepared to run a complete test as most of the engines you mention have a strong similarity - also, the IPON time controls are a little better in my view and suspect you are playing with ponder off whereas the IPON list is based on ponder on which simulates a more realistic scenario in my view.
The second point I made is not that an improvement should be ignored, merely questioning how you know that the new H3 could make use of your changes and even if it could how you know that this would have the same positive effect - make sense?
i tested ponder off,because my laptop is obsolete,a core 2 duo computer,not for other reason
well,Houdart will improve Houdini 2 to make Houdini 3,and since he promissed me a free license for Houdini 3 if i can only improve the engine 10 elo points ''playing'' with those values,i really think that Houdart are interested on those values
well,Houdart will improve Houdini 2 to make Houdini 3,and since he promissed me a free license for Houdini 3 if i can only improve the engine 10 elo points ''playing'' with those values,i really think that Houdart are interested on those values
![[se]](/mwf/v12/flags/se.png "Sweden")
Which values have you changed in your experiments?
have you downloaded the HoudiConfig program?because i gave the values i changed for everyone,you must only use the program to apply those values inside Houdini2.0c to convert it in Houdini2.0z
![[tr]](/mwf/v12/flags/tr.png "Turkey")
Hello Razor,
It seems you missed to see SCCT Rating List:
http://www.sedatcanbaz.com/chess/scct-rating/
Note that Houdini 2.0z x64 6 cores has the best performance so far that ever i seen, great job by Salvador !!
Its true that those games are played at Blitz,plus the current number of games are still not enough
And for more serious conclusions we need testings with slow time controls (e.g 90m+30s ;40/40;40/120...)
But however I wonder...
1)Do you know what is the Elo difference between Ponder OFF and Ponder ON ?
Note that in my testings,Ponder ON is around 15-20 Elo stronger than Ponder OFF
2)Do you know what is the Elo difference between i7 980X @4.0Ghz 6 cores compared with slower hardwares ?
3)Do you know what is the Elo difference between 6 core and 1 core ?
Note also that in my testings,I noticed around 130-150 Elo in favor for 6 core
4)Do you know too that any opening is playing a big role for gaining (or affecting) high Elo performance ?!
5) In my opinion,the current 3min+2sec on i7 980X @4.0GHz 6 core is approx. equal in Elo to 30min+2sec on QX9650 3.00GHz 1 core
In other words,
For some people SCCT Rating List can be considered as Blitz,but for others can be considered as Slow Time Control
Btw,i don't expect from you to have same view as mine,but however if you have different opinion than mine:
- Its will be great if you can prove it, but with serious data (if it's possible with games-annotations)
Best,
Sedat
It seems you missed to see SCCT Rating List:
http://www.sedatcanbaz.com/chess/scct-rating/
Note that Houdini 2.0z x64 6 cores has the best performance so far that ever i seen, great job by Salvador !!
Its true that those games are played at Blitz,plus the current number of games are still not enough
And for more serious conclusions we need testings with slow time controls (e.g 90m+30s ;40/40;40/120...)
But however I wonder...
1)Do you know what is the Elo difference between Ponder OFF and Ponder ON ?
Note that in my testings,Ponder ON is around 15-20 Elo stronger than Ponder OFF
2)Do you know what is the Elo difference between i7 980X @4.0Ghz 6 cores compared with slower hardwares ?
3)Do you know what is the Elo difference between 6 core and 1 core ?
Note also that in my testings,I noticed around 130-150 Elo in favor for 6 core
4)Do you know too that any opening is playing a big role for gaining (or affecting) high Elo performance ?!
5) In my opinion,the current 3min+2sec on i7 980X @4.0GHz 6 core is approx. equal in Elo to 30min+2sec on QX9650 3.00GHz 1 core
In other words,
For some people SCCT Rating List can be considered as Blitz,but for others can be considered as Slow Time Control
Btw,i don't expect from you to have same view as mine,but however if you have different opinion than mine:
- Its will be great if you can prove it, but with serious data (if it's possible with games-annotations)
Best,
Sedat
Good points. It would be interesting to match up the playing strength of SCCT with the other ratings lists. When taking into account the factors you've mentioned, I'm guessing your blitz tournaments would be very close to the rapid time controls used by the other lists.
Thanks dear Alan
But however,the current my rating list is Blitz (in my eyes)
And i agree with you that the quality of the SCCT Blitz games are almost equal (close) to rapid time controls used by some other rating lists,which are based on 1 core
Best Regards,
Sedat
But however,the current my rating list is Blitz (in my eyes)
And i agree with you that the quality of the SCCT Blitz games are almost equal (close) to rapid time controls used by some other rating lists,which are based on 1 core
Best Regards,
Sedat
1) there are many variables here so no one answer as far as I'm concerned
2) not precise enough Sedat - you need to specify what you mean by slower hardware?
3) again not precise enough for a specific answer - your variance range should be lower as the amount of move thinking time increases
4) I assume you are asking which openings are good for white or black in terms of a positive outcome - true?
5) I assume you are referring to engine matches as opposed to chess position analysis?
2) not precise enough Sedat - you need to specify what you mean by slower hardware?
3) again not precise enough for a specific answer - your variance range should be lower as the amount of move thinking time increases
4) I assume you are asking which openings are good for white or black in terms of a positive outcome - true?
5) I assume you are referring to engine matches as opposed to chess position analysis?
Actually i was quite clear with my previous posting...
And it seems you dont understand me what i mean exactly,probably due to my English is not so good
But anyway,I will try to explain once more
Houdini 2.0z is already tested in SCCT (great performance by Salvador's settings)
My work is not perfect (still i need to improve many things ...)
But however (with respect to all) my work does not look to others
I love and respect all the 'Good' and 'Top' things,where i can touch or at least i can see...
If you still have difficulties to understand me-SORRY
And it seems you dont understand me what i mean exactly,probably due to my English is not so good
But anyway,I will try to explain once more
Houdini 2.0z is already tested in SCCT (great performance by Salvador's settings)
My work is not perfect (still i need to improve many things ...)
But however (with respect to all) my work does not look to others
I love and respect all the 'Good' and 'Top' things,where i can touch or at least i can see...
If you still have difficulties to understand me-SORRY
Sedat:
When are you going to post Houdini's new settings?
When are you going to post Houdini's new settings?
Here are the settings of Houdini 2.0z:
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=25133
Happy testings,
Sedat
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=25133
Happy testings,
Sedat
![[mx]](/mwf/v12/flags/mx.png "Mexico")
> Critter 1.6
>
> (of course that's just a guess)
Nope, his developer is taking a turn to improve analysis features.
Critter 1.2 and 1.4 are virtually identical in the rating list, no elo improvement.
However, Critter 1.4 is much better for analysis thanks to its Session File feature.
> Critter 1.6
> (of course that's just a guess)
Good guess, at least in the case of chess960
http://computerchess.org.uk/ccrl/404FRC/
http://computerchess.org.uk/ccrl/404FRC/rating_list_pure.html
But it is probably not conclusive, there are not enough opponents for the top engines to get a large number of games.
Wondering how could Critter bring those results to normal chess...
@Magnus Friedmann - You are Vegan number 2
Topic Rybka Support & Discussion / Rybka Discussion / What chess program has the chances to ever surpass HOUDINI?
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill