Houdini 1.03 is available

[bob]

Hi Alex,

could you see on how many positions (which are important) Houdini was changed.

Thinking on the thread from Thinker programmer.

Best
Frank

Hello Frank Quisinsky,

Let me see if I understand you clearly:

You are advocating that someone disassemble a closed source engine to provide you with answers.

Soooooo...where does that stop exactly?

Does this need to know only apply to engines which you are certain are derivatives? Or can any commercial program now be ripped apart?

I find that request odd coming from you.

I would prefer open to closed source BUT one is free not to download and use the thing.

Requesting disassembly be done and the results published seems a dangerous road to tread on.

Later.

I think here things are a bit different. Houdini is not the "classic closed source." It is a clearly plagiarized program that has been modified very slightly, which does not exactly meet any existing standard of "new engine." I don't get the idea of copying someone's source, making a few changes, then hiding the modified source and making waves about a "new engine." One of those "things that make you go hmmm." Ethically, _any_ engine based on ip* should be open-source since ip* is open-source. Of course, not everyone is ethical...

Bob,
This is a QUESTION so don't get offended. But can you explain to a non-programmer like myself how someone can come along and make substantial improvements that may put Houdini above Rybka 4 and yet still be blamed for just copying?

Although there may be code from the clones in Houdini, houdini itself must be more than just a straightforward copy because the guy makes 20 elo improvements every new release.

As a punter I'm excited that progress is being made at the very top. It's your phrase 'modified very slightly' that perplexes me; surely he's done quite a lot of modification.

Also he didn't seem to hide the fact that a lot of ideas (code?) came from other programs - but perhaps I'm missing something.

The same way I do this in Crafty. In some cases, we gain significant chunks of Elo just by tuning. Changing a score a little or a lot. Or changing a search control a little or a lot. Sometimes we add code to correct or improve something.

There is no program that is "perfect" yet. And most changes are modest. If you start off near the top, little changes make a difference. If you write your own code, it takes a few years to reach that point, if you are good enough.

[gaard]

My tests reported at open-chess.org confirm all statements by Robert Houdart relating to improvements, though I am not sure if I qualify as "independent". More games will paint a clearer picture, but as of right now I have Houdini 15 Elo over IvanHoe and Rybka 4, with a 3-sigma result in favor of Houdini in heads up matches with R4. My time controls are similar to 3'+1", 40/160", or g/4', when comparing average move time, which in my tests is a constant 4" per move.