Just a warning to computer chess fans here that are buying commercial chess programs, what you buy may be reverse engineered and then sold as copyrighted, independent work by escape artists. Interesting discussion on Talkchess Origins forum.
Quoting a post from Mark Lefler:
The hashes of the files on https://github.com/alureon/houdini (from Fishcooking) match up with the H6 files I have locally. Seems like the DMCA-able code was stripped (yay?) so I expect the repository to be able to stay up. Should be trivial to get it to build. Nobody has an excuse to push reasonless doubts anymore.
So that's that. The code is as open as it can get. All evidence that could be released has been released, and if you are not satisfied with that, there is nothing that I am comfortable doing that can change that. I am incapable of adding anything else to the conversation.
Ovyron and other doubters, go wild. Everyone else who is still arguing over legal issues, you can go home.
I started to write this but it vanished somehow so I will start again.
I looked over the code in the github, and I see a lot of Stockfish there, and also, to my surprise, Komodo. I can give examples. One surprise was a big table that has been in Komodo since I started working on it. The possibility that someone else would come up with the exact same numbers is quite close to 0. That table was created with a formula Don and Larry came up with, but the thief must not have been able to figure it out. They must have want to more finely tune the use of that table, so in the code they just copied the table to a larger array and interpolated numbers in between, using crude straight line interpolation. About a year or so ago I had the same idea, so was able to use the formula to generate new values. The thief did not have the formula so he had to do as best he could.
Other stolen ideas indicate the theft occured on Komodo 10.4 or later. I know because we changed how that code works by the Komodo 11 release.
I know of no Komodo code breaches, so I assume Komodo was reverse engineered. Taking a table is easy and it would not have been that hard to figure out how it was used (but not how it was created). As are the other things I see from Komodo. I do not remember Houdart ever crediting Komodo for his program. He certainly never asked me or Larry if he could use Komodo in his program.
I would be happy to share the evidence with big computer chess tournament like WCCC, should Houdini be entered. I can also share it with the big online tournaments, should they doubt me.
But you have to be logged in first, just like with 'The Edge' in the Rybka Forum here.
> Meh, has Houdart said anything so far?
No. But Houdini 6 being a Stockfish fork (violating the GPL) with a few original ideas has been proven beyond reasonable doubt (so when you compile the code you get Houdini 6 and the code is mostly Stockfish 8.5.) If you're to believe the Komodo team, it also has a lot of code stolen from reverse engineered Komodo.
I've now reached page 28 of the thread, where people are discussing how Houdini divides its node count by 7 and multiplies it by 8, people are discussing what can we do to stop something like this from happening again in the future, and Chris Whittington (Chess System Tal's programmer) being in denial (claiming that no evidence has been presented... (???)), oh dear.
Personally, I'm a big fan of the engine. People talk about licences, and crimes, and credits, and about Houdart being a crook, but the engine itself is fantastic, if the next engine of its kind comes from someone that steals code from many engines and breaks laws and scams people out of their money for something that should be open source, I'd rather live in that world than in one where the engine never got to exist at all.
(I'm Ovyron at Talkchess, heh, I actually made it as the incredulous one)
The ONLY problem was the license method Houdart used. I absolutely HATE even imagine ''buy'' a program to only being able to use it on a limited number of computers! I never did, and it doesn't make any sense at all! Which means that if years later Houdart abandon computer chess, stops responding, or whatever else, we left with a totally unusable program if our machine breaks down. No thanks, no program with such characteristics worth buying, and the SAME goes for any other program! Yes, I'm a bit radical here.
On the other side, I'm really glad that ''Houdidit'' NO longer needs any license method anymore! I want to clarify that I am NOT totally happy with what has happen to Houdart; after all, I deeply appreciate all his hard work (no doubt Houdini had unique ideas too!) and the way he vastly improved Houdini over the years, when he took the big crown from it predecessor Rybka! I wanted to see him longer on the battle for computer chess!
Anything from one or more of these?
# Stockfish, a UCI chess playing engine derived from Glaurung 2.1
# Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
# Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad
# Copyright (C) 2015-2019 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad
It'd be different if it was known that Houdart did this and the source remained closed, so the GPL violation would stand. But now it's open source so at least this part has been amended.
It's hard to sue for "damages" because if this was never known about it'd have made no difference for the authors. Mainly because Stockfish is free, so any sum covering the missing sales wouldn't be an option.
The case with Kodomo is more tricky, Mark Lefter looked at the code and obviously knew things were stolen from Komodo, but we don't have Komodo's sources so we only have his word for it. I believe him but the problem is Houdart didn't steal Komodo source and copied parts of it. He reverse engineered it to see what it was doing, and copied that. In one of the parts Komodo had a function that produced some numbers, Houdart couldn't get the code that produced the function but could see the output, so he just approximated those numbers in the table.
Mark looked at the table and instantly knew it had to come from Komodo's function, but then it's a case of an implementation of an idea witnessed in reverse-engineered code. We don't know how much of Komodo is in Houdini, but the Similarity Report had a bigger similarity between Houdini and Komodo than between Houdini and Stockfish, so it could be a lot. Basically Stockfish's search with Komodo's eval or something like that.
It doesn't seem anybody is doing anything, what is worrying is people that continue to buy Houdini 6 (I still see ads for Houdini on the main section of this forum) without knowing it's now a free open source engine, so the injustice continues...
> From the Stockfish team there's no basis. The licence that was broken allows
GPL, though widely used, is just one type of EULA that copyright holders can use: US/international copyright "exclusive rights" for the authors laws.
No government, local through international treaties, explicitly support GPL, and it can not supersede any laws.
Only a judge can decide the remedy if found guilty of EULA contract breach and/or copyright infringement. Only the copyright holder(s) can assert either.
> without knowing it's now a free open source engine,
It is innocent, lawfully copyright protected, until legally proven otherwise.
To me, someone is guilty at the moment they make the crime regardless of if it can be proven or not. And in this case, it was proven (lest, nothing ever happens if a judge isn't involved).
> So I guess I can kill somebody and if I'm never caught I'm innocent until legally proven otherwise?
Welcome to reality.
Houdini 6 (IPON)
# PLAYER : RATING ERROR (%) D(%) OppAvg CFS(next) POINTS W D L PLAYED
1 Houdini 6 : 3354 11 82.4% 31.1 3060 100 2719.0 2206 1026 68 3300
2 Komodo 11.2.2 : 3315 10 78.9% 34.5 3063 99 2604.0 2035 1138 127 3300
3 Stockfish 8 : 3299 10 77.3% 39.5 3064 100 2551.5 1899 1305 96 3300
4 Shredder 13 : 3120 8 56.3% 51.1 3076 100 1859.0 1016 1686 598 3300
5 Fizbo 1.9 : 3070 8 49.6% 42.6 3079 91 1635.5 933 1405 962 3300
6 Ginkgo 2.0 : 3062 8 48.5% 50.3 3080 83 1599.5 770 1659 871 3300
7 Gull 3 : 3056 8 47.7% 47.6 3080 100 1574.0 789 1570 941 3300
Stockfish 9 (IPON)
PLAYER : RATING ERROR %) D(%) OppAvg CFS(next) POINTS W D L PLAYED
1 Stockfish 9 : 3379 11 83.2% 29.9 3073 100 2744.0 2251 986 63 3300
2 Houdini 6.02 : 3350 10 80.7% 31.8 3074 100 2663.0 2138 1050 112 3300
3 Komodo 11.2.2 : 3322 10 78.1% 34.0 3076 100 2577.0 2016 1122 162 3300
( Stockfish 8 : 3308 8 77.8% 39.0 3064 86 5821.5 4362 2919 199 7480)
4 Shredder 13 : 3125 8 55.3% 50.2 3090 100 1823.5 996 1655 649 3300
5 Fizbo 2 : 3097 8 51.5% 40.9 3091 100 1701.0 1026 1350 924 3300
6 Ginkgo 2.0 : 3069 8 47.8% 49.9 3093 90 1577.0 754 1646 900 3300
7 Gull 3 : 3061 8 46.7% 46.1 3094 96 1542.5 782 1521 997 3300
Stockfish DD ---- Stockfish 9
29 Nov 2013 50 mos. 4 Feb 2018
3235 +197 3432
Komodo TCEC ---- Komodo 11.3.1
4 Dec 2013 51 mos. 6 Mar 2018
3220 +180 3400 *!?
Houdini 4 ---- Houdini 6
25 Nov 2013 46 mos. 15 Sep 2017
3242 +138 3400 *!?
CCRL 40/15 http://ccrl.chessdom.com/ccrl/4040/rating_list_all.html
dates web-search: ' "program version" released ' ; https://blog.stockfishchess.org/archive
Ko 11.3 6mar2018 3400 (+45 from Ko 10.2, days from Sf 8) << Sf 9 4feb2018 3432 (+52 from Sf 8) << Ho 6 15sep2017 3400 (+28 from Ho 5, days from Sf 8) << Ko 11.2 17jul2017 3393 Ko 11.01 23may2017 3384 Ko 10.4 14mar2017 3374 Ko 10.3 20dec2016 3368 Ho 5 7nov2016 3372 < Sf 8 1nov2016 Ko 10.2 29oct2016 < 3379 3355 < Ko 10 23may2016 3347 Ko 9.42 21mar2016 3342 Sf 7 2jan2016 3327 Ko 9.3 27nov2015 3323 Ko 9.2 3326 Ko 9.1 3310 Ko 9 15apr2015 3308 Sf 6 28jan2015 3292 Ko 8 5sep2014 3285 Sf 5 31may2014 3270 Ko 7a 21may2014 3228 Ko TCEC 4dec2013 3220 Sf DD 29nov2013 3235 Ho 4 25nov2013 3242
> similarity report 2019
"In addition, you will find some custom compiled versions of the engine available to all ( allowed under Stockfish's GPL license). "
The engine cannot be made available under the Stockfish GPL because Houdini contains Nalimov code that is Copyright of Nalimov and cannot be released as free to use code under GPL. That is why Stockfish and any other GPL engine cannot use Nalimov code.
Yet the following:
Project ID: 17720419
The version of Stockfish used by Robert Houdart in Houdini"
has several Nalimov folders, the egtb_nalimov.cpp has lines 60-70:
/* -------------------------------------------------------------------- */
/* Probe chess endgame database ("tablebase") */
/* Copyright (c) 1998--2001 Eugene Nalimov */
/* The code listed below should not be used in any product (software or */
/* hardware, commercial or not, and so on) without written permission */
/* from the author. */
/* -------------------------------------------------------------------- */ ...
> It may be a big blunder by "MikeB" who has posted this on Github quote,
I am not entirely sure what you meant, but if it is what I was looking at yesterday that is gone today:
Now there is this on TalkChess http://www.talkchess.com/forum3/viewtopic.php?f=2&t=73639 :
Joined: Thu Mar 09, 2006
Location: Pen Argyl, Pennsylvania
Re: Houdini for Android
Post by MikeB » Thu Apr 16, 2020
daniel71 wrote: ↑
Thu Apr 16, 2020
Looks like all the files have been pulled, too bad I didn't download anything in time...
There was an element of risk, and the quickest and easiest way for me to mitigate my risk was to take it down.
It will not be going back up and I will not be responding to any requests.
If there is a breach of copyright there is due legal process for the copyright owner(s) that is better than people self appointing themselves as judge and jury.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill