Surprise...

[Don]

I just looked at this more carefully, I thought it was a hand picked example but it's right after playing "e4" in the opening. Not only do the nodes match exactly, but so does the PV. That is pretty damning evidence even by itself.

Do you have that version of Houdini somewhere that I can get a copy? I want to check this myself.

By the way, when I built the similarity tester I found that 2 of the closest matching programs was Robolitto and Houdini. The match percentage was higher than even 2 different releases of most programs.

Forgive my confusion, but I do not understand this thread at all. That Houdini is an IPPO is very old news, so why is this being discussed here as if it were some sort of revelation?

I think it's because now Houdini is trying to go commercial with a GPL program without releasing the source code or even admitting its origins.

Don,

What if you are wrong?

Houdart has come out and said (finally) that his program contains no licensed or GPL code at all.

And unlike others he has at least acknowledged where his inspiration came from.

I have been actually waiting for Houdart to make such a blanket statement. And because he has in such a public forum I am tending to believe him.

Will you not give him the benefit of the doubt?

Of course I give him the benefit of the doubt - but there is no doubt.

I tried several positions and did one ply searches with both Robolitto and the Houdini build 2010-05-15 and the moves match very time so far (out of 7 positions I have tried near the opening) and the node counts either match perfectly, or are very close. That's almost impossible to see happen so consistently.

e.g.:

Code: Select all

Houdini w32 1_CPU
build 2010-05-15 
by Robert Houdart
position startpos moves g1f3 g8f6 d2d4 d7d5
go depth 1
info depth 1 seldepth 2 score cp -5  time 1 nodes 30 nps 30000 pv f3g5
info depth 1 seldepth 2 score cp 6  time 1 nodes 31 nps 31000 pv f3e5
info depth 1 seldepth 4 score cp 8  time 1 nodes 38 nps 38000 pv h2h3
info depth 1 seldepth 5 score cp 16  time 1 nodes 55 nps 55000 pv d1d3
info depth 1 seldepth 5 score cp 28  time 1 nodes 64 nps 64000 pv b1c3
info depth 1 seldepth 5 score cp 28  time 2 nodes 66 nps 33000 pv b1c3
bestmove b1c3 ponder NULL
fixme:console:CONSOLE_DefaultHandler Terminating process 21 on event 0
drd@i7:~/Desktop/robbolito$ ./RobboLito.exe 
RobboLito 0.09 w32
Aug  4 2010 19:40:41
basic windows version by kranium, improved by sentinel
position startpos moves g1f3 g8f6 d2d4 d7d5
go depth 1
info depth 1 seldepth 2 score cp -9  time 0 nodes 30 nps 0 pv f3g5
info depth 1 seldepth 2 score cp 11  time 0 nodes 31 nps 0 pv f3e5
info depth 1 seldepth 4 score cp 13  time 0 nodes 38 nps 0 pv h2h3
info depth 1 seldepth 5 score cp 26  time 1 nodes 55 nps 55000 pv d1d3
info depth 1 seldepth 5 score cp 46  time 1 nodes 64 nps 64000 pv b1c3
info depth 1 seldepth 5 score cp 46  time 1 nodes 66 nps 66000 pv b1c3
bestmove b1c3 ponder NULL

Here is an e-pawn opening which is more advanced and should excercise the evaluation a lot differently:

Code: Select all

rd@i7:~/Desktop/robbolito$ ./RobboLito.exe 
RobboLito 0.09 w32
Aug  4 2010 19:40:41
basic windows version by kranium, improved by sentinel
position startpos moves e2e4 e7e5 g1f3 b8c6 f1b5 a7a6
go depth 1
info depth 1 seldepth 4 score cp 41  time 1 nodes 37 nps 37000 pv b5c6 d7c6 f3e5
info depth 1 seldepth 5 score cp 41  time 1 nodes 71 nps 71000 pv b5c6 d7c6 f3e5
bestmove b5c6 ponder d7c6
fixme:console:CONSOLE_DefaultHandler Terminating process 8 on event 0
drd@i7:~/Desktop/robbolito$ ./Houdini_w32_1CPU.exe 
fixme:heap:HeapSetInformation (nil) 1 (nil) 0
Houdini w32 1_CPU
build 2010-05-15 
by Robert Houdart
position startpos moves e2e4 e7e5 g1f3 b8c6 f1b5 a7a6
go depth 1
info depth 1 seldepth 4 score cp 26  time 0 nodes 37 nps 0 pv b5c6 d7c6 f3e5
info depth 1 seldepth 5 score cp 26  time 0 nodes 71 nps 0 pv b5c6 d7c6 f3e5
bestmove b5c6 ponder d7c6

I did not fish for these positions, these were the first 2 that I tried after the e4 position (where the node counts vary only slightly even after 8 ply.)

Don

Code: Select all

RobboLito eval/Houdini eval/Predicted/Diff

-21  -13  -13	0
 -9	-5	-5	0
 -8	-5	-5	0
 -3	-1	-1	0
  1	 0	 0	0
  2	 1	 1	0
  6	 3	 3	0
  7	 4	 4	0
 11	 6	 6	0
 13	 8	 8	0
 19	11	11	0
 21	13	13	0
 26	16	16	0
 29	18	18	0
 41	26	25	1
 46	28	28	0
 58	36	36	0

H = int(R - (3*R / 8))

That seems to get you within 1 ELO. It's off by 1 point for 41.

[geots]

I think it's because now Houdini is trying to go commercial with a GPL program without releasing the source code or even admitting its origins.

Funny, for you the problem only arises because Houdini is now commercial?

Let me state one more time, for the record, that Houdini does NOT contain any licensed code, GPL or other, except for the Gaviota probing code which is copyright M. Ballicora.

Let me state one more time, for the record, that I have acknowledged FROM THE START the sources of inspiration for Houdini, unlike most other engine developers (including yourself) who freely use all kind of sources (including Ippolit, Stockfish, etc, apparently now you even use Critter sources) without giving the slightest acknowledgement whatsoever.

Robert

Robert, the sad fact is that if you had to struggle for 6 years to get your program in the "Top 20", no one would even be caring to look at where it came from. Larry Kaufman, who I think has hit the nail on the head with the Rybka/Fruit issue, admits Robbolito has been able to use things that Vas used, but on the other end- using it caused a loss in another aspect of the program for Vas. He said Robbolito has figured out how to make the big improvement withou losing anything on the other end. And he says that is what he just cannot figure out- tho he has tried. But in a legal way, it may be that you have figured it out.

But that is off the point of this thread a bit, which is- expect trouble. Some have even stated their suspicions are because you could develop and improve Houdini so fast. Not willing to give you the credit for just plainly being better than most programmers, if you assume engine strength is a main deciding factor.

Larry mentioned that proving Robbolito, Ivanhoe, Fire, etc. were illegal could be almost impossible. But he left out the fact that some will say it will be easy. It is sad that when you give the chess community a wonderful program- you have to pay the price of people doubting your morals and ethics. And question your program.

After the ICGA debacle with Rybka/Fruit, and then seeing how they are going to treat you in the same manner, I have had enough. I don't think you are the type of person who would cheat and lie to reach your goal. In fact, your honesty and down-to-earth attitude is extremely impressive to me. I imagine the only thing you are guilty of is being able to do something almost everyone else cannot. And for that you will be hounded.

I had had enough when I saw this coming. I immediately got IvanHoe, Fire, Robblito and a few more and loaded and now use them. To those who would be critical of me- I say "tough shit". And the same "tough shit" for anyone who questions my reasons or morals. I have had enough of this belly-aching, whining bunch of people. I will use them plus Rybka and not give a dime to the programs from the "old school" programmers, whose main answer to strength increase is faster hardware.

Again, thank you for your wonderful program. I will be here if you need me for anything.

Best,

George

PS: I have quit completely reading any responses to my threads, simply because I don't care what anyone thinks- so it is a complete waste of my time. They can answer to the blowing wind. Of course I WILL READ any response from you.

[michiguel]

Since when was Ippolit GPL ?? I thought it was released as Public Domain code with absolutely zero restrictions.

GPL requires any derivative work to be released FREE of charge with Source code too. However you are allowed to SELL it or value added services with it, but you still must make the free version available.

Red Hat Linux is a good example. It's open source but it's a commercial distribution which means you can buy it. If you buy it you get "support" and I nice printed handbook or manual or something. However, you can also go right the red hat site and download the distribution and burn your own CD if you want to. If you purchase they will send you the CD with documentation.

What this means for example is that you can sell Robbolitto if you want - however you must acknowledge that it's a derivative work and provide the source code and you must make a free version available too.

Wasn't the original version of Robbolito public domain?

Miguel

[bob]

Since when was Ippolit GPL ?? I thought it was released as Public Domain code with absolutely zero restrictions.

GPL requires any derivative work to be released FREE of charge with Source code too. However you are allowed to SELL it or value added services with it, but you still must make the free version available.

Red Hat Linux is a good example. It's open source but it's a commercial distribution which means you can buy it. If you buy it you get "support" and I nice printed handbook or manual or something. However, you can also go right the red hat site and download the distribution and burn your own CD if you want to. If you purchase they will send you the CD with documentation.

What this means for example is that you can sell Robbolitto if you want - however you must acknowledge that it's a derivative work and provide the source code and you must make a free version available too.

Wasn't the original version of Robbolito public domain?

Miguel

I thought ippolit was public domain, and the rest were just derived works, some kept private, some released as GPL, etc...

[Guerrero]

Wasn't the original version of Robbolito public domain?

Miguel

I thought ippolit was public domain, and the rest were just derived works, some kept private, some released as GPL, etc...

Originals Robbolito (0.84, 0.85, 0.85c, 0.85d, 0.85f1a) are public domain.
Versions 0.85g3 and 0.09 are forks of version 0.85d with license GPL.

[kinderchocolate]

Hi,

I don't think Houdini is illegal and anybody who is complaining that it's illegal is an idiot. First, what can you do? Take Robert to the court? You can't hurt his sales by whinging here. You have to make something better than Houdini. Second, only the copyright holder of a GPL license can complain, this is stated clearly. Do you know exactly who published GPL Robblitto and does he care? If he doesn't, why should you?

[wims]

Since when was Ippolit GPL ?? I thought it was released as Public Domain code with absolutely zero restrictions.

GPL requires any derivative work to be released FREE of charge with Source code too. However you are allowed to SELL it or value added services with it, but you still must make the free version available.

Red Hat Linux is a good example. It's open source but it's a commercial distribution which means you can buy it. If you buy it you get "support" and I nice printed handbook or manual or something. However, you can also go right the red hat site and download the distribution and burn your own CD if you want to. If you purchase they will send you the CD with documentation.

What this means for example is that you can sell Robbolitto if you want - however you must acknowledge that it's a derivative work and provide the source code and you must make a free version available too.

That is not how the GPL license works. There is no requirement to make a free version available (at least not free as in free of charge). The only requirement is that the source code must be made available free of charge to a license holder. The programmer is allowed to sell his program and is only required to make the source available to the persons who have bought a license to run the program. The programmer doesnt have to create a free-of-charge version, nor does he have to make the source code available for free to anyone that asks.

[M ANSARI]

Since when was Ippolit GPL ?? I thought it was released as Public Domain code with absolutely zero restrictions.

GPL requires any derivative work to be released FREE of charge with Source code too. However you are allowed to SELL it or value added services with it, but you still must make the free version available.

Red Hat Linux is a good example. It's open source but it's a commercial distribution which means you can buy it. If you buy it you get "support" and I nice printed handbook or manual or something. However, you can also go right the red hat site and download the distribution and burn your own CD if you want to. If you purchase they will send you the CD with documentation.

What this means for example is that you can sell Robbolitto if you want - however you must acknowledge that it's a derivative work and provide the source code and you must make a free version available too.

Wasn't the original version of Robbolito public domain?

Miguel

I thought ippolit was public domain, and the rest were just derived works, some kept private, some released as GPL, etc...

Exactly what I thought too. But can you take public domain code and modify it a little and then stick a GPL license to it? I think Robbolito is just a small modification of Ippolit, so if Robbolito is useful to Houdini then Robert could just as easily go back to the code base of Ippolit. Since Vas has not taken legal action against Ippolit, that code base is and will stay as public domain. Since it is public domain, everyone and anyone can use it without worry of breaking any license.

Sometimes I think we need an IT law firm to wade through this forum .

[rodolfoleoni]

As a chess user, my interest is to see engines which grow stronger and stronger. And my simple questions are? What if Vasik didn't built that old Rybka 1.0? What if he didn't work on it to get a V3 and V4 versions? What if Ippo team didn't disassemble something (if they really did)? What if nobody could look at their code?

Any of the top engine wouldn't exist, I guess, including Stockfish, Komodo, Critter, Houdini. And, as many programmers look at how Stockfish works, many others would be weaker.

I'm not a crusader. I only think science should mean progress, not regression.

[rbarreira]

But can you take public domain code and modify it a little and then stick a GPL license to it?

The definition of "public domain" is that you can do whatever the hell you want, so yes, I believe you can.

It's rather pointless to do so though, since it will be almost impossible for you to prove that anyone showing up with a derivative of your program did in fact copy you and not the earlier public domain work.

There's an issue of public domain not being a valid concept in all countries, but that is a whole other can of worms...