Houdini 1.03a The New Nr 1 !!!

[James Constance]

Clearly R4 different from R3 in many ways.

Now how different is ippolit/houdini from R3?

Clearly ippolit/houdini is different from R3 in many ways. Not many.

Now how different is R3 from R4? More different than Ippolit is from R3.

If ippolit was so different from R3 - then we would not have ANY controversy. It had to be shown that ippolit was a legal derivative .. I think the paper by BB maybe shows it to be a legal derivative not to mention the considerable evidence on this forum to show that its evaluation is very close to R3 --- see Larry Kaufman's and Dan's posts.

...not to mention Graham Banks might never add these engines to his rating list!

Sorry I've missed these posts, as I sometimes spend time away from the forum - do you have a link? Is there a scientific way of measuring how similar one engine is to another and has this been done in the case of Rybka and Ippolit? My own feeling from using the engines is that Houdini probably differs in eval more than R4 from R3.

[Matthias Gemuh]

If ippolit was so different from R3 - then we would not have ANY controversy.

...not to mention Graham Banks might never add these engines to his rating list!

Interesting proof.

[mwyoung]

I have played long time control games with Houdini and Houdini is keeping its advantage over Rybka 4. Yes, some IPPO programs fall off badly at longer time controls. I have found this with Ivanhoe. But Houdini is the strongest program I have tested. And long time controls have not changed this so far.

Here is my current match to 50 games with Houdini Vs Rybka 4 at 40 moves in 2 hours. Played on a Quad 2.4 Ghz. 13 games of 50 played.




1 Houdini 1.03a w32 4_CPU 2500 ½½½1½½½½1½½½½ 7.5/13
2 Deep Rybka 4 w32 2500 ½½½0½½½½0½½½½ 5.5/13

[Milos]

1. All Ippo family engines are worse than Rybka at long time controls.

Wishful thinking .

2. The above behaviour is probably due to the obfuscation code present in Rybka, which disadvantages her at fast time controls.

There is no obfuscation code in Rybka. That's a provable fact. There are bunch of proofs in disassembled code. Stop making up things.
If you are referring to wrong depth and nps reporting as obfuscation, these things are trivial and literally have zero slowing effect.
Why do you keep repeating false claims about things that you obviously do not understand?

[S.Taylor]

The best Rybka 4 on the CCRL 40/4 is only +45 Elos compared to the best Rybka 3, so it didn't take much to overtake it. As Ippo is only better at fast controls, I guess its gain is merely due to removing obfuscation code. In the case of Ippo, the gain was not spectacular. In the case of Robbo/Fire the removal of obfuscation code may have been accompanied by some bug fixing and optimizing, so a further gain was obtained. No original ideas were necessary (although I don't claim there weren't any), only some programming skills were needed.
In the case of Houdini the base may have been Robbolito or Fire, so it is not unlikely that there wasn't any progress at all as far as the introduction of new ideas are concerned.

Vincent's quote is the most appropriate comment to this:

Both programmers wrote a 3000+ elo engine and obviously are not doing the testing themselves. They got: "a bunch of idiots" for that.

Houdini and Ippo derivatives programmers don't have such a luxury.
And despite "a bunch of idiots" doing testing for one of them, he can't even tune TM properly. And without external support, even with strong cluster can't improve his program for more than 45elos in 2 years. His real skills were clear to some ppl when he released the first and totally mediocre 1600 elo version of his engine. After that came fruit, and than external help financed from who knows whom with a huge amount of money to do the tuning for him (since he didn't have, and still doesn't have a clue about engine parameter tuning)...

Well said...I'd like to add that Vasik has hit the wall regarding further improvements and I personally think that the so called Rybka cloud project is just a hoax to mislead people that he has a much stronger version of Rybka that can be rented online....the truth is that he couldn't improve Rybka 4 more than 40 Elo and by that still full of nasty bugs....
Dr.D

If he hit the wall and could not improve much more, but tried his best, then I would think better about him than i had.
Obviously, he should then have made it the free upgrade.
But i don't know what his problems actually were.

IF he could not improve it further, but made Rybka 4 after one year (of Rybka 3), as it is now, then i would say that it passes, as an upgrade, just that he is at the end of his abilities.

Do you say your firebird personality is a whole level higher than R4 and all the competition? (and would win with clear margin in the Martin Thoresen matches, in which Rybka 4 has been shining in all its splendor?)

[S.Taylor]

I have played long time control games with Houdini and Houdini is keeping its advantage over Rybka 4. Yes, some IPPO programs fall off badly at longer time controls. I have found this with Ivanhoe. But Houdini is the strongest program I have tested. And long time controls have not changed this so far.

Here is my current match to 50 games with Houdini Vs Rybka 4 at 40 moves in 2 hours. Played on a Quad 2.4 Ghz. 13 games of 50 played.




1 Houdini 1.03a w32 4_CPU 2500 ½½½1½½½½1½½½½ 7.5/13
2 Deep Rybka 4 w32 2500 ½½½0½½½½0½½½½ 5.5/13

Mark, how is it that Houdini is clearly weaker than Rybka 4 in Thoresens matches?

[Dann Corbit]

I have played long time control games with Houdini and Houdini is keeping its advantage over Rybka 4. Yes, some IPPO programs fall off badly at longer time controls. I have found this with Ivanhoe. But Houdini is the strongest program I have tested. And long time controls have not changed this so far.

Here is my current match to 50 games with Houdini Vs Rybka 4 at 40 moves in 2 hours. Played on a Quad 2.4 Ghz. 13 games of 50 played.




1 Houdini 1.03a w32 4_CPU 2500 ½½½1½½½½1½½½½ 7.5/13
2 Deep Rybka 4 w32 2500 ½½½0½½½½0½½½½ 5.5/13

Mark, how is it that Houdini is clearly weaker than Rybka 4 in Thoresens matches?

Because everyone runs 30 games and then makes a decision instead of 300 like they ought to.

Now, running 30 games is a great idea. But then don't imagine that you have decided who is strongest yet.

[michiguel]

Clearly R4 different from R3 in many ways.

Now how different is ippolit/houdini from R3?

Clearly ippolit/houdini is different from R3 in many ways. Not many.

Now how different is R3 from R4? More different than Ippolit is from R3.

If ippolit was so different from R3 - then we would not have ANY controversy. It had to be shown that ippolit was a legal derivative .. I think the paper by BB maybe shows it to be a legal derivative not to mention the considerable evidence on this forum to show that its evaluation is very close to R3 --- see Larry Kaufman's and Dan's posts.

...not to mention Graham Banks might never add these engines to his rating list!

Sorry I've missed these posts, as I sometimes spend time away from the forum - do you have a link? Is there a scientific way of measuring how similar one engine is to another and has this been done in the case of Rybka and Ippolit?

Yes, it has been done, in terms of "move selection".

Just an example of a long discussion
http://www.talkchess.com/forum/viewtopi ... 02&t=32112

Miguel

My own feeling from using the engines is that Houdini probably differs in eval more than R4 from R3.

[Martin Thoresen]

Now, running 30 games is a great idea. But then don't imagine that you have decided who is strongest yet.

Dann, this is presicely what I am trying to explain to mr. Taylor.

My matches at tournament time control are not in any way intended to give a final answer to whether engine A is stronger than engine B.

To reach such conclusions at the very slow time control I use, a large number of computers would have to be used in order to get the game numbers high enough within a humane time-frame.

Even with 1500 games the error margins are quite high.

My matches and upcoming tournaments are intended to provide the viewers with some quality chess entertainment, nothing else.


Best Regards,
Martin

[Dann Corbit]

Now, running 30 games is a great idea. But then don't imagine that you have decided who is strongest yet.

Dann, this is presicely what I am trying to explain to mr. Taylor.

My matches at tournament time control are not in any way intended to give a final answer to whether engine A is stronger than engine B.

To reach such conclusions at the very slow time control I use, a large number of computers would have to be used in order to get the game numbers high enough within a humane time-frame.

Even with 1500 games the error margins are quite high.

My matches and upcoming tournaments are intended to provide the viewers with some quality chess entertainment, nothing else.


Best Regards,
Martin

I would like to add that I greatly appreciate your efforts.
I never examine the games at 40/4 because the quality is low.
Games at slow time control are the ones that contain the brilliancies and are the most enjoyable to watch move by move.
If we want to mine computer generated chess games for great ideas, these ideas will mostly come from the ponderous time control games.

I guess that more great ideas will come from 10 correspondence paced chess games between high level opponents than from 10,000 lightning games.

IMO-YMMV.

P.S.
Did I ever mention that I am not a blitz or lightning fan?
Yes, now that I think of it, I have.