SWCR-64: Houdini 1.5 x64 started in around 15 hours ...

[Frank Quisinsky]

Hi Robert,

if I compare statistics it seems that your engine have comparable results to the exp. Rybka settings.

I know you don't like it to read that and the following things I will write you:

Rybka Exp. Settings 42 and I think 61 too are around 25 ELO stronger with the time control I used. I will test setting 61 after the Houdini 1.5 x64 test.

It's easy to see that in the latest three months the IvanHoe's are 10 ELO stronger (my list, Ahmed versions, PeterPan versions).

= 35 ELO

Rybka Exp. Settings are more aggressive, better results vs. weaker and not better results vs. stronger engines.

Same Houdini 1.5 has ... with the conditions I used for testing!

Houdini 1.03a is 10 ELO stronger as Rybka 4 Standard in my list + 10 ELO more (IvanHoe +) = 20 ELO more as Rybka 4 with Exp. Settings.

10 ELO + 10 ELO from the newest IvanHoes + 25 ELO from the Rybka Exp Settings = 45 ELO + ... Houdini 1.5 to Houdini 1.03a

Same ponder reactions as the latest IvanHoes.

This one only if I look in results and statisicts. But all can be random!

---

All this is more easy to see with longer time controls because IPP family is stronger with shorter time controls.

---

This one can be in the brains by testers or others persons with interest because my databases are open and each one can make such statistics.

So it's possible, that you used one of the latest IvanHoe compiles and add Rybka Exp. settings. After this one you added your own code again, your own settings, Gaviota support other things you do around your work.

...

Positiv ...
All this can be random, nobody can say such things without to know your sources.

Perhaps you will give a little statement about it?

Fact is: Houdini is on 1 and you improved Houdini a lot.

Hope you understand it without to think ... I am searching only bad points!!!

This isn't in my interest!

And to your answers of my "impressions". That is very interesting. Have thanks. Let us wait now of the final results.

Have a nice xMas!
Frank, without to have only negativ things in my brain. My interest is to inform about ratings with the work I do here with SWCR testing.

Best
Frank

[Roger Brown]

So it's possible, that you used one of the latest IvanHoe compiles and add Rybka Exp. settings. After this one you added your own code again, your own settings, Gaviota support other things you do around your work.

Hello Frank,

What would be the most likely outcome of combining parts in this way - analogous to building a Frankenstein monster?

An engine that plays stronger than anything out there or a weaker thing?

I am curious as to exactly how a stronger engine would result from the process as described by you. Sounds like it would be easier to write something stronger....

Positiv ...
All this can be random, nobody can say such things without to know your sources.

Perhaps you will give a little statement about it?

I hope that when he is ready that he releases the source as he said he would but given your hypothesis, would you grant interviews or release source at this time?

You were willing to test it and look at the hostile response (pm's and emails) you received. I cannot imagine the author receiving any less of a hostile response.

All of this over a chess program.

Later.

[Houdini]

Frank,

Like I showed in my previous post, your analysis doesn't match my knowledge as developer of the engine. It saddens me to see you speculating freely about issues about which you have little to no real information.

Best regards,
Robert

[Frank Quisinsky]

Hi Robert,

I can read sources perhaps to 15%, not more. What I have are statistics to all of the top engines, created by the databases of SWCR. You are the programmer and can give information about programming, I cann't do this.

What I can is to try to find out more with games and statistics. Not more not less and all others can do such things with the free available SWCR database too, not a secret.

Yes, all what I wrote is a speculation not more not less. And this have nothing to do with your person or your work you do. In the first time you release your engine we have some problems in discuss. Not all what I here wrote are right, sorry for it.

Honest:
I changed my opinion to IPP family engines and to much other things in computer chess. For me important alone are the engine I got. If strong sources are free available, why not programmers used this sources and started directly with strong sources. I have nothing again it.

Again, what I wrote are speculations and such things I got from different users as comments to different engines in my list.

I am sure you are a good programmer after all what I can see to Houdini. Better endgame statistics as Rybka, your SMP support, GaviotaBases and your latest improvements.

No doubt about it from my point of view.

But what I wrote here is in the brain by much others. And after my statistics it could indeed possible.

Look in my latest message, I wrote ... "it can be random".

I think that so strong engines like Rybka / Houdini could produced comparable statistics without clonening.

Again ...
No negative things in my brain with the things I wrote.

Why:
It's abolutely clear, that I don't have the knowledge and proofs to write ... x is a clone from y. I can say ... it could be possible that x and y after statistics I created.

Best
Frank

[beram]

Frank,

You mustn't jump to quickly to speculative conclusions. Please just stuck to providing us with SWCR results (>2000 games per engine) and leave it by that.
Robert Houdart is better chessplayer too.

Have a nice Christmas

[Frank Quisinsky]

Hi Roger,

if you are working in computer chess are you got many mails, each day. Same in the time I work with Martin in Arena or on my older webpages Frank's Chess Page, today WBEC Ridderkerk.

What I wrote here comes not from myself. Such things and much other things users of chess programs have in his brain.

For my SWCR not very important and its quit clear that the programmer of Houdini don't like it to read such things.

At the moment I am not at home and follow the Houdini results on my webpage. Houdini 1.5 produced sensational results.

At the moment 3.012 ELO.
You have to calculate around -7 because it's a Shredder GUI calculation. Bayesian will give around -7 ELO = 3007 ELO.

Best
Frank

[Frank Quisinsky]

Hi Bram,

I don't know how strong is Robert Houdert. I am playing in a chess club without to play official tournaments. For around 20 years my level are around 2.000 ELO. Today I am playing with players in my chess club which have around 2.200 ELO. My chess club SG Trier is playing in the first German leage. My performance in around 30-40%. But this one is for me not very important because my chess interest go to other things as to play tournament games in a chess club. Don't have the time to make both, testing chess programs for programmers (beta test), SWCR and other things or the day must have 36 hours

Very interesting which results Houdini 1.5 has in SWCR. Possible that Houdini 1.5 is the first engine with more as 3.000 ELO.

And if Robert Houdert read my message:
What I wrote here comes not from myself, OK I wrote it.

I will give you a reason why I am not believe what I got by others per mail.

One of the persons which compiled IvanHoe versions try to add the Exp. 42 settings. It was IvanHoe T0.7. IvanHoe T.07 plays clearly weaker as the preveiw version T0.5.4.1 without this settings.

Now, if persons are thinking that Houdini 1.5 used the Exp. Settings from Rybka is this nonsens in my opinion.

Best
Frank

[beram]

Hi Bram,

Very interesting which results Houdini 1.5 has in SWCR. Possible that Houdini 1.5 is the first engine with more as 3.000 ELO.

And if Robert Houdert read my message:
What I wrote here comes not from myself, OK I wrote it.

I will give you a reason why I am not believe what I got by others per mail.

One of the persons which compiled IvanHoe versions try to add the Exp. 42 settings. It was IvanHoe T0.7. IvanHoe T.07 plays clearly weaker as the preveiw version T0.5.4.1 without this settings.

Now, if persons are thinking that Houdini 1.5 used the Exp. Settings from Rybka is this nonsens in my opinion.

Best
Frank

well Frank,
it seems Houdini will be the first one, I just copied this from your SWCR page:

Code: Select all

Rank Name                          Elo    +    -  games score oppo. draws 
   1 Houdini 1.5 x64              3014   25   25   779   83%  2742    NEW +  66 to 1.03a *
   2 Rybka 4 x64 Exp. 42          2963   23   23   873   80%  2721   25% NEW +  24
   - Houdini 1.03a x64            2948   22   21   920   79%  2718   29% 
   - Rybka 4 x64                  2939   17   17  1520   80%  2700   29% 

grts Bram

[Houdini]

Very nice results against the top 3 after 800 games:

Code: Select all

Houdini 1.5 x64 - Rybka 4 x64 Exp. 42 (2963)      21.5 - 11.5     65.15%     Perf=3071
Houdini 1.5 x64 - IvanHoe B49jA x64 (2931)        20.0 - 13.0     60.61%     Perf=3005
Houdini 1.5 x64 - Stockfish 1.9.1 JA x64 (2905)   23.5 -  8.5     73.44%     Perf=3081

So much for the claim "better results vs. weaker and not better results vs. stronger eng".

Robert

[Frank Quisinsky]

Hi Bram,

rating goes higher and higher.
After 800 games now 3.018 ... around -7 for the Shredder calculation = 3.011.

= 2.948 has Houdini 1.03a = + 63 ELO !!
Again around 10 ELO more as in IPON.

Logical comment could be ...
A little bit stronger with longer time controls as in blitz.

Also a differenz to the IvanHoe's ... this engines are a little bit stronger in blitz.

Best
Frank