Does Houdini 1.5 adjust playing style to opponent?

[BB+]

As far as I know Houdini doesn't adapt its playing style to the opponent.

Another way of answering this would have been "Depends on what you mean" --- obviously if the opponent plays move X rather than move Y, Houdini will "adapt" by responding to move X rather than move Y. This might even be more than just pedantry, as the original post said: "Houdini seems to play aggressive against some engines and defensive against others", which could be more of a reflection of the opponent than of Houdini. While we're on the topic, judging "style" is always dodgy, and CW famously claimed that it essentially a mirage (the "search gap"), at least for fast programs: http://www.thorstenczub.de/ihatematerialists.html

[AnthonyTheSage]

While we're on the topic, judging "style" is always dodgy, and CW famously claimed that it essentially a mirage (the "search gap"), at least for fast programs

I strongly disagree with this. It's obvious that engines have different styles just like people. Some engines like to attack kings, some like to push passed pawns, some like to control the center. Of course these are all programmed in, but in the end it all results in a style of play for the engine. Obviously as computer chess engines advance and searches become deeper and faster, the newer engines will inevitably wipe the board with the older engines. In this case it will be really hard to see the style of the older engines. But with modern engines obvious patterns or move choice emerge, and this results in style of play for the engine.

[ernest]

I've noticed ... Houdini really shows it's learn abilities. For the most part, Houdini plays different lines as if it learned from past mistakes.

Robert has completely denied that...

[Peter C]

I've noticed ... Houdini really shows it's learn abilities. For the most part, Houdini plays different lines as if it learned from past mistakes.

Robert has completely denied that...

Probably because he didn't write the code that does it.

I haven't observed enough of Houdini's games to tell whether or not it changes it's style, but I doubt it. I know some engines (Junior is supposed to) do, but it seems like it would take more effort than it would be worth.

Peter

[Dr. Ivannik]

I don't want to get in the way of your efficiency...

Like the two earlier replies, your reaction is completely off-topic for this thread.
It's a sad day for a forum when even the grand moderator feels the irresistible urge to make counter-productive, off-topic interventions.

Great, I'll start a new thread then, where we can talk these issues out in a productive, on-topic fashion.

Jeremy

Thank you Jeremy for taking it the fxxk out of here. Mr. Houdart has given us the greatest engine of all time. He does not owe anybody proof of anything. Its a free product use it or lose it. Nor are the the origins (justifiably) relevant for some computer chesslers. Robert Houdart owes you exactly what you paid for his product which is nothing.

Thank you

Dr. Ivannik

[Jeremy Bernstein]

I don't want to get in the way of your efficiency...

Like the two earlier replies, your reaction is completely off-topic for this thread.
It's a sad day for a forum when even the grand moderator feels the irresistible urge to make counter-productive, off-topic interventions.

Great, I'll start a new thread then, where we can talk these issues out in a productive, on-topic fashion.

Jeremy

Thank you Jeremy for taking it the fxxk out of here. Mr. Houdart has given us the greatest engine of all time. He does not owe anybody proof of anything. Its a free product use it or lose it. Nor are the the origins (justifiably) relevant for some computer chesslers. Robert Houdart owes you exactly what you paid for his product which is nothing.

Thank you

Dr. Ivannik

That's right, he owes me nothing. He owes the original programmers of his engine (and the contributors who crafted it into Robbolito), responsible for all but the last 50 ELO, a whole fxxk of a lot, though.

viewtopic.php?f=7&t=992

Jeremy

[Dr. Ivannik]

I don't want to get in the way of your efficiency...

Like the two earlier replies, your reaction is completely off-topic for this thread.
It's a sad day for a forum when even the grand moderator feels the irresistible urge to make counter-productive, off-topic interventions.

Great, I'll start a new thread then, where we can talk these issues out in a productive, on-topic fashion.

Jeremy

Thank you Jeremy for taking it the fxxk out of here. Mr. Houdart has given us the greatest engine of all time. He does not owe anybody proof of anything. Its a free product use it or lose it. Nor are the the origins (justifiably) relevant for some computer chesslers. Robert Houdart owes you exactly what you paid for his product which is nothing.

Thank you

Dr. Ivannik

That's right, he owes me nothing. He owes the original programmers of his engine (and the contributors who crafted it into Robbolito), responsible for all but the last 50 ELO, a whole fxxk of a lot, though.

viewtopic.php?f=7&t=992

Jeremy

Jeremy

If you are not one of the original programmers maybe ya should mind your own business.Its not up to you to determine what he owes original programmers of Robbolito. Your position is worth what you paid for his fine product----fxxking nothing.

Thank you

Dr. Ivannik

[Odeus37]

That's right, he owes me nothing. He owes the original programmers of his engine (and the contributors who crafted it into Robbolito), responsible for all but the last 50 ELO, a whole fxxk of a lot, though.

viewtopic.php?f=7&t=992

Jeremy

Even if what you said was true, which is still to demonstrate... Don't forget that Robbolito is about same Elo than single thread Rybka 3 (not rybka 4) from which IPP and such are "supposedly" derived... And I think many (including Vas) are trying since monthes to add 100 Elo (and not only last 50 like you said) to rybka 3 or to "IPP family", and only one didn't fail...

http://www.inwoba.de/bayeselo.html

Code: Select all

   1 Houdini 1.5a             3008   14   14  2600   82%  2747   23% 
   2 Deep Rybka 4             2953   11   11  3700   77%  2745   30% 
   3 Houdini 1.03a            2949   12   11  3200   79%  2729   30% 
   4 Stockfish 2.01 JA        2919   13   13  2300   74%  2743   33% 
   5 Rybka 3 mp               2901   10   10  4200   77%  2704   31% 
   6 Stockfish 1.9.1 JA       2899   11   11  3000   71%  2748   36% 
   7 Critter 0.90 SSE42       2892   12   12  2800   69%  2757   36% 

http://www.amateurschach.de/swcr/_swcr-all.htm

Code: Select all

01. Houdini 1.5 x64              3.005 22 21 1040 82% 2.750 26%
02. Rybka 4 x64 Exp. 42          2.969 20 19 1240 80% 2.719 25%
--. Houdini 1.03a x64            2.948 22 21  920 79% 2.718 29%
--. Rybka 4 x64                  2.938 17 17 1520 80% 2.699 29% + 47 to 32bit
03. IvanHoe B49jA x64            2.933 19 18 1160 76% 2.739 34%
--. IvanHoe B52aC x64            2.922 21 21  880 77% 2.727 34%
04. Stockfish 1.9.1 JA x64       2.906 18 17 1280 73% 2.728 33% + 13 to 32bit
--. Rybka 3 x64                  2.905 23 22  840 78% 2.687 28% + 46 to 32bit
--. Fire 1.31 x64                2.900 21 20  920 75% 2.720 36%
05. Critter 0.90 x64             2.872 18 17 1200 68% 2.741 37%

[Uly]

And I think many (including Vas) are trying since monthes to add 100 Elo (and not only last 50 like you said) to rybka 3 or to "IPP family", and only one didn't fail...

The "failing" was intentional, as Vas had to take out strength from Rybka 4 to make his Cluster model make more sense (when people get 40 cores on their home, what is going to be their incentive to rent the cluster? Better software).