Komodo 4 on long time control

[Houdini]

As I said before, and as you have acknowledged before, you hardly know anything about Houdini or Houdini's scaling. See your "we are not interested enough in Houdini to study this in any detail" above.

Stop presenting any of your stupid facts about Houdini 2 or Houdini 1.5, I don't make any stupid comments about Komodo either.

Robert

[gerold]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

That will be a big jump.
I have Houdini 1.5 about 85-90 elo above Komodo 2. I don't have the latest Komodo.

[Frank Quisinsky]

Hi Maurizio,

that would be great but Komodo 4 x64 must be 60 ELO stronger as Houdini 2.0c x64. I added the results from the ended round robin and games on my webpage, and TalkChess tournament / matches selection.

60 ELO is a lot for the Komodo team.

Best
Frank

[Frank Quisinsky]

Hi there,

Code: Select all

SWCR is played at 40/10 with Ponder ON. This is not miles away from the 40/40 with ponder off.

And on faster hardware as CCRL!
40 in 10 with ponder = on on my hardware is around 40 in 16 CCRL. CCRL have the highest time control but many things I missed in this list. But of course a very good list.

Houdini 2.0c x64 is after statistics I made in all game phases better as 1.5. Only in tactic middlegame a little bit more lost games as the preview version 1.5. Endgame strength is perhaps a very little bit better. Stability and time mangement is perfect. No crashes, no time loses, all runs 1a.

All in all ... a rating list can not give an absolutly clear ELO for a number 1 or the latest places. Furthermore, not many other opponents with an ELO higher as 2.900 are available. Not enough really good oponents for Houdini 2.0c so far.

In my opinion 30 ELO more for Houdini 2.0c x64 should be right with the time control I used.

Don wrote that Komodo is stronger with longer time controls. That's it logical for engines with positional strengths. But Komodo have some tactic problems and I can't believe that the distance with longer time controls between Houdini 2.0c x64 and Komodo 3.0 x64 will be smaler. For a new place one Komodo need more as 60 ELO.

Code: Select all

   1    1 Houdini 2.0c x64               3022   21   20  1200   82%  2751   24% Update, + 24
   -    2 Houdini 1.5 x64                2998   14   14  2320   78%  2771   29% 
   -    3 Houdini 1.5 w32                2983   19   19  1202   79%  2755   29% 
   3    6 Komodo 3.0 x64                 2962   17   17  1480   77%  2751   32% 
   -   11 Houdini 1.03a x64              2943   21   21  1000   80%  2710   30% 
   -   19 Komodo 2.03 JA x64             2926   17   17  1480   73%  2749   31% 
   -   20 Robbolito 0.09 x64             2918   17   17  1480   74%  2729   33% 

By the way:
+104 to Robbolito 0.09 x64

Best
Frank

[Don]

As I said before, and as you have acknowledged before, you hardly know anything about Houdini or Houdini's scaling. See your "we are not interested enough in Houdini to study this in any detail" above.

Meaning that we do not intend to waste our time running Houdini at long time controls just to find out how good or bad it scales beyond the time controls we already understand. We spend time testing Komodo, not Houdini.

I never said we were not interested, I said, "not interested enough." We are interested in data on any chess program, even the illegitimate ones, that are this close to Komodo in strength.

Stop presenting any of your stupid facts about Houdini 2 or Houdini 1.5, I don't make any stupid comments about Komodo either.

Of course you get annoyed in a big hurry whenever someone reminds the forum about the origins of Houdini - and that is one of the "stupid facts" that you want people to ignore.

Robert

[Frank Quisinsky]

Hi Don,

all engines runs in SWCR with Q9550 processors. The 600 games on i7 I delete from the database.

With other words, no SSE for all tested SWCR engines.

I am sure that Komodo benefit from SSE more as others, and you wrote it can be 15 ELO. I think Houdini benefit too from SSE, perhaps 5 ELO, I don't know.

So the different is perhaps 50 ELO from Komodo 3 x64 to Houdini 2.0c x64. Possible, I don't know that.

All what I have to say ...
I like the work you and Larry do and Komodo is playing a total other chess as all other TOP-7 SWCR engines. Never I saw an engine which try to used each very little advantage in detail so aggressive Komodo do that in early and late endgames. I think in the most of the endgames Komodo is really the number 1.

Good luck with your development of Komodo.
I hope that Komodo, Critter or Stockfish will be the next number 1. Because I don't like Houdini and Rybka.

IvanHoe and Fire are fully OK for me too

But I have to test all engines and fact is, that the strongest engine is Houdini by Robert Houdart. If you are looking in SWCR Houdini games it's a massaker what this engine do in middlegames with the others ... Don a massaker. In tactical statistics in the early middlegame Houdini 2.0c x64 is now on the same level as Stockfish 2.1.1 PHQ. I say it, I think in all game phases Houdini is clear - a bit better. Unfortunately, only to see in statistics ... to high is the level and ...

My list can not give a clear ELO for a number 1. I believe if more engines are available with 2.900 ELO Houdini can go to 3.030 or 3.040 ... bayesian and elostat have problems to calculate a number 1 or the latest places in a list.

Weeknesses can be search in tactical middlegames only (Houdini 2.0c). But here you will be found not many things too. I mean have a look on the Houdini lost games.

Best
Frank

[Houdini]

Of course you get annoyed in a big hurry whenever someone reminds the forum about the origins of Houdini - and that is one of the "stupid facts" that you want people to ignore.

No, you may have noticed that I only intervene when you make *technical* comments about Houdini that are outside your field of knowledge or experience.
I wonder whether you even own a copy of Houdini 2?

Your repeated off-topic rants about Houdini's origin - in about every discussion we have - just demonstrate your level of frustration about an engine that has taken the #1 spot you think you deserve, I actually find them quite amusing.

Robert

[Frank Quisinsky]

And why I think thats Houdini ELO will go higher if more strong engines are available.

1. 24% remis quote
2. In the most important game phase, the early endgame, this engines plays very very strong.
3. In the late middlegame I think Houdini is 75-100 ELO stronger as the other TOP-7 engines (OK, Komodo is here also very strong).

It will be not easy for you, Richard or the Stockfish Team to be the next number 1.

3.024 ELO in SWCR ... I think it could be higher, perpaps 3.040 with more strong opponents. Can see it in the statistics I made.

Best
Frank

[Don]

Hi Don,

all engines runs in SWCR with Q9550 processors. The 600 games on i7 I delete from the database.

With other words, no SSE for all tested SWCR engines.

I am sure that Komodo benefit from SSE more as others, and you wrote it can be 15 ELO. I think Houdini benefit too from SSE, perhaps 5 ELO, I don't know.

It's pretty substantial for Komodo, nobody is really sure why but that is the case. That explains why your list disagree's with others with respect to Komodo.

So your test probably fairly reflects the difference on hardware that does not support SSE. I know that there is also a large discrepency between 32 bit and 64 bit and it's clear that we have strongly taken the path of making Komodo run well on modern hardware, much more so that the Ipoo based programs and most others. Stockfish is a good example of a program that runs well on any hardware. So this is one downside to Komodo which we know about.

So the different is perhaps 50 ELO from Komodo 3 x64 to Houdini 2.0c x64. Possible, I don't know that.

It isn't really going to make any difference to identify the exact difference between the two programs, the only thing that makes it interesting is that the other programs are closing the gap on the Ippo clones such as Houdini and have passed most of them already. Two years ago this would not have even been an interesting discussion as it was "no contest."

All what I have to say ...
I like the work you and Larry do and Komodo is playing a total other chess as all other TOP-7 SWCR engines. Never I saw an engine which try to used each very little advantage in detail so aggressive Komodo do that in early and late endgames. I think in the most of the endgames Komodo is really the number 1.

Good luck with your development of Komodo.
I hope that Komodo, Critter or Stockfish will be the next number 1. Because I don't like Houdini and Rybka.

IvanHoe and Fire are fully OK for me too

But I have to test all engines and fact is, that the strongest engine is Houdini by Robert Houdart.

I don't have any problem with that, in fact I don't want Komodo to get number 1 by default because the strongest program was not tested. It really tics me off that we are held hostage by a program just because it happens to be strong - we are forced to deal with it. Not a single person I know is fighting for the "Richard Hall", the guy who copied Mark Leflers program. Nobody is saying that he was treated unfairly by being booted out of tournaments for what HE did. It's kind of scary to think that if he had had somehow tricked Houdart out of the Houdini code everyone would be fighting for his cause and cheering him on.

That's why it's important to us that Komodo or for that matter Critter or Stockfish or some other legitimate program ends up on the rating lists on top. I cannot do anything about Critter or Stockfish but I can do something about Komodo. I strongly oppose the dishonest methods and lies of Houdart and that is regardless of any rivalry between Komodo and Houdini.

If you are looking in SWCR Houdini games it's a massaker what this engine do in middlegames with the others ... Don a massaker. In tactical statistics in the early middlegame Houdini 2.0c x64 is now on the same level as Stockfish 2.1.1 PHQ. I say it, I think in all game phases Houdini is clear - a bit better. Unfortunately, only to see in statistics ... to high is the level.

Weeknesses can be search in tactical middlegames only (Houdini 2.0c). But here you will be found not many things too. I mean have a look on the Houdini lost games.

Best
Frank

[gerold]

Hi Maurizio,

that would be great but Komodo 4 x64 must be 60 ELO stronger as Houdini 2.0c x64. I added the results from the ended round robin and games on my webpage, and TalkChess tournament / matches selection.

60 ELO is a lot for the Komodo team.

Best
Frank

Thanks for your work Frank. I have downloaded Komodo 3. Thanks to their team. Now i have match going Komodo 3. Vs. Houdini. I like Komodo style. Seems to be a little weak in end game. To early to tell.
1k games will show some interesting games.

Best,
Gerold.

P.S. Using the Perfect 12 book and Blitz 5/6 TC. for both Engines.