Houdini 1.03 is available

[Roger Brown]

H has possibly overtaken R4 in strength now.

Possibly. But why on earth have I not yet seen a match of 5000 games to prove for any of the Ippolit derivatives that it has overtaken Rybka? Anyone can make strength claims or guesses.

Hello Gabor,

It is a little odd that a member of a testing group which has a policy of not testing specifically those engines (as is their right I might add before I am vilified!) asks that question.



Later.

[Frank Quisinsky]

Hi Gabor,

the reality.
And each one believe what others wrote without a proof for it. All of the Ipp fam. engine are strong in blitz chess, Rybka 4 too ... but with longer time controls the engines lose a bit on points to the others.

Now I think it would be great for the community if Houdini comes with an ELO improvement. The problem the most clones have. No ELO improvements, only strong blitz results but for each new version ... look here 20 ELO points more.

From version Houdini 1.0.0 to 1.0.3 it should now around 60 points if each version is 20 points stronger ... if we believe what the programmer wrote.

Reality is ... 0-10 points with longer time controls if you test the engine vs. the TOP-20 ... think so (but clear, not tested ... speak from the knowledge which such things I have).

So I think if Houdini win ELO points the "cloner's" can celebrate it. A little sensation

Today a friend made his test-suits with the new Houdini 1.0.3. He told me that this version reach a new record in his private test suite. First engine which solved all positions.

That's indeed great!
Let us wait of the first results (eng-eng).

Best
Frank

[perejaslav]

Hi Gabor,

No ELO improvements, only strong blitz results but for each new version ... look here 20 ELO points more.

I don't think so. Here are the latest testing results from the obvious forum:



«ELO improvements» are rather YES than NO

[Frank Quisinsky]

Hi,

that's is interesting ... but not enough, should be around 250 games more for each version.

You are sure that all the versions haven't other LMR / NullMove settings only? If so, all will be "the other way round". That's one of the problems, and the users are thinking ... wow, what for great improvements. But only different settings are changed ... most try it out in blitz and possible that with longer time controls the settings are better in the bad versions in your list.

Nobody can test it ... or the testers need 8-10 Quad Core PCs.

Not sure about it what I wrote here, but so far I believe it.

Best
Frank

[perejaslav]

Hi,

that's is interesting ... but not enough, should be around 250 games more for each version.

Best
Frank

IIRC there are 1005 games for each engine planned. So let's wait for the final result

[Houdini]

Frank,

From version Houdini 1.0.0 to 1.0.3 it should now around 60 points if each version is 20 points stronger ... if we believe what the programmer wrote.

Please start by reading my actual announcements:
- From Houdini 1.0 to 1.01: +15/20 Elo
- From Houdini 1.02 to 1.03: +20 Elo in fast TC, somewhat less in long TC
That's about 30 to 35 Elo between version 1.0 and version 1.03, which I think is a fair assessment.

The debate would actually be much more interesting if you didn't make false claims about "what the programmer wrote"...

Robert

[Frank Quisinsky]

Hope too ... I can't await it.
A life without Houdini must be terribly.
But we have IvanHoe, Firebird ... thanks goodness!

Joking ...

I try since years to do a little bit but nothing is really good. Allways the same problem with myself (see the work I do 4 years for the Arena development). I started my mission in my first year, 43 years later ... all what I have on my harddisk is Houdini and IvanHoe.

Again ... thanks goodness!

But in five years, you can be sure I have 200 Houdini versions, each one is 20 ELO betters as the preview version. On my 32Core PCs I have 16 version for each Houdini ... 2Core, 4Core, 6Core ... all will be config, nothing I will forget.

Edit:
The programmer will read it ... 20 ELO better for each one isn't right, sorry!

I think the Houdini programmer dislike my person.
One more ...

Perhaps I have a chance with ...
Houdini is so great, in my first 1+1 matches with 6-pieces endgame bases the results are clearly better as the results from Firebird with 6-pieces Robber bases.

Best
Frank

[Frank Quisinsky]

Hello Robert,

OK "Programmer" ...
So you get my sorry!

Best
Frank

[Frank Quisinsky]

Hi Robert,

I am joking!
Its in your hand to create a program with will be stronger and stronger with each new version. 50 ELO per year is great.

If I see that you get my sorry for my bad messages in TalkChess about Houdini.

I have nothing against your work but I think talented programmer should not start with ipp sources.

Waste on time!

Its my opinion, must not be right but I wish you good luck with Houdini or other things you do for us. Really, not joking!

Best
Frank

[Matthias Gemuh]

I think the Houdini programmer dislike my person.

Frank

Hi Frank,

Robert is too nice to dislike anybody
A stable Houdini can uproot trees. Try and see

Regards,
Matthias.