The ICGA Process

[BB+]

Another statement, seemingly from an alternate reality, from the Rybka Forum.

[PST] was roughly half of Fruit vs Rybka document that formed a central role in the final verdict.

The PST comparison I produced formed slightly under 5 pages of the RYBKA_FRUIT document, the length in part due to the size of the tables. The document was 16 pages proper, plus 8 or so (depending upon the version) in appendices.

Zach's PST comparison consists of just under 3.5 pages in the PDF format, which ran to 24 pages (closer to 22 when subtracting page breaks).

My parsing of modifiers is unsure, but I think the statement refers to the document forming a central role in the final verdict, which may be true, though PST itself did not. Indeed, the Panel Report only mentions it (page 5) as one item among 4 or 5 other elements of Zach's enumeration.

Furthermore, the by-now-typical expostulation that minor bits here and there (in PST, or more generally for evaluation features) are not "copyrightable" remains in tune [so too on TalkChess], though (yet again) I point out that, either from the standpoint of copyright or originality, this misses the point, namely that: quite similar collections of specific bunches of "minor bits" (and/or specific renditions of ideas) are indeed protectable creative expression. See, as one example, the canonical Romeo and Juliet versus West Side Story "plot-stealing" discussion in Nimmer On Copyright.

[BB+]

I really hate to keep discussing this, but some statements are so erroneous, and get repeated ad infinitum, it seems:

They used an engine that did not even participate in the ICGA to convict an engine that did participate ... assuming that they are the same. That is just so stupid as to not even warrant discussing.

You do realize that ICGA based their disqualification of an engine that participated, by analyzing an engine that did not participate ... I mean that alone is beyond idiotic (sorry but can't think of a better word).

The Panel looked at many Rybka versions, including Rybka 2.3.2a, which was released exactly the same week as the WCCC 2007 in Amsterdam. The Panel listed as one of its "questions for Vas" as to whether this was an acceptable surrogate [and similarly whether something like Rybka 2.1 was a reasonable approximation to the 2006 entry -- later WCCCs were also mentioned, but are less relevant here]. This question, which seems quite a basic informational inquiry, was not answered.

The Panel thus made the evidential assumption (rebuttable, of course, but as just noted -- it wasn't) that the Rybka versions that were investigated (principally R1-R232a) did in fact suffice. It appears that the Board concluded similarly. [On the other hand, it was indeed a Panel issue as to whether "merely" Rybka 1.0 Beta sufficed (where much of the RE work had been done already), and as there was sufficient disagreement, the later Rybka versions were also considered, up to the point where it was thought essentially inarguable that the issue of Rybka versions vis-a-vis WCCC entries had not been adequately addressed, at least to extent that was possible w/o input from VR on the issue].

There is a (wide) variety of other issues/claims that appear in conjunction with the above posts, but they don't relate to the ICGA process per se, so I don't see any reason to explicate them here.

[BB+]

Why does one bother...? Well, at least this isn't as bad as other posts at Rybka Forum (where anyone who made an advisory vote in the Panel has been lumped into a "mob of killers" -- no idea what they think of the Board).

The purpose of the ICGA was to see if Vas used a cloned engine in the tournament.

The purpose of the ICGA Investigative Panel is:

The purpose of the Panel shall be to:
[a] Investigate and discuss allegations of cloning or creating a derivative of strategy games programs;
Report to the ICGA as to the veracity or otherwise of such allegations;
[c] Make recommendations to the ICGA as to what action if any should be taken against those found by the Panel to have been guilty of cloning or creating a derivative;
[d] Publish the findings of the Panel.

There is no mention of an ICGA tournament, though obviously any recommendations made in (c) could vary depending upon where the clone/derivative was entered (if at all).

The "evidence" used to convict Vas was based on Rybka 1.6 and Rybka 1.0beta, two engines that never participated in the ICGA.

The principal evidence (and indeed, requested by Panel members) was with "evaluation overlap" of R232a to Fruit 2.1.

Now we have that they checked 2.3.2a and that since in one post that since 2.3.2a at one point participated in a game due to a malfunction of the main engine (which was clearly explained that it was a different engine) as being proof that Vas is guilty because R2.3.2a participated in the game where the main engine was not participating.

The 2.3.2 version was also checked, which, up to hard-coding some parameters seems to be claimed to be the 2007 WCCC participant for the first games.

The games are running on my computer - remotely controlled by a Laptop in Amsterdam.
All games till yesterday ran on version 2.3.2 x64 mp, a special version with hard coded access to my tablebases and hard coded hash size of 2 GB.
The current game runs on 2.3.2a. Except of hard coded TBs and hash there are no other tweaks.
Settings are all default except Nalimov usage - it is set to normal, as TBs are now on a superfast solid state disk.

Furthermore, I am unable to track down any "malfunction of the main engine", but rather it seems R232 had a few minor bugs that were fixed in R232a. VR himself said: This is a minor update for Rybka 2.3.2. A few small bugs are fixed, the most serious being a problem with zugzwang detection in Rybka 2.3.2. The style is identical to the style of 2.3.2, and the playing strength will be very close to identical (within 2 Elo or so). It's perfectly fine for test groups to merge results for these two versions.

So if Rybka 2.3.2a has a few things in common with Rybka 1.0beta ... thus automatically Rybka 2.3.2a and all future versions of Rybka are guilty.

This was not the principal comparison carried out by the Panel. There was an evaluation comparison R1B to R232a (and intervening versions) to determine how much [if any] had changed, but the basis of the Report relied much more on comparing the R232a evaluation to the Fruit 2.1 evaluation directly.

I think engine programmers should stick with what they do best and that is write engine code. Let lawyers or people that deal with law deal with legalities of what is "guilty". [...] I think it is the same here ... programmers are not lawyers ... and the quicker they understand that the better off everyone is.

The ICGA would seem to be most competent body to decide whether or not someone has contravened their rules via his entries. I am not aware of (say) Formula 1 or the IOC calling in lawyers to handle purely internal disputes.

[Adam Hair]

What sort of benefits do you get for being a member of the ICGA hit squad? Is the renumeration more than adequate? What about the hours?

psst
Hey.
Does membership help with the chicks?

Seriously, I understand the origins of the bad feelings some people have concerning this whole situation. But it is disappointing to see intelligent people abandon logic and reason to the extent that I have seen.

[kingliveson]

Now here is a very interesting quote from Larry Kaufman, someone who worked on Rybka 3 with Vas:

Don has a very high level of concern about ethical issues. He was a strong supporter of Vas and Rybka when the Ippo programs came out and were accused of copying Rybka, even though we were already developing Komodo. My own view of the Fruit/Rybka matter is that there is evidence that early Rybka versions took some shortcuts to save time, and that it was reasonable for ICGA to revoke the first or perhaps the first two titles, but that at least by the time of Rybka 3 (probably earlier), all significant traces of Fruit code were replaced and so I don't think the later titles should have been revoked. I also don't think that the use of Fruit (and/or Crafty) code in early Rybka versions contributed much to Rybka's strength, although the use of Fruit IDEAS (which is allowed) certainly did. In other words Vas could have probably made just as strong a Rybka without doing anything questionable with an extra few weeks of work.

[hyatt]

What sort of benefits do you get for being a member of the ICGA hit squad? Is the renumeration more than adequate? What about the hours?

psst
Hey.
Does membership help with the chicks?

Seriously, I understand the origins of the bad feelings some people have concerning this whole situation. But it is disappointing to see intelligent people abandon logic and reason to the extent that I have seen.

Personally speaking, I've made a small fortune off of this, and also Crafty has gained about +800 elo as well.

[/sarcasm]

[Harvey Williamson]

What sort of benefits do you get for being a member of the ICGA hit squad? Is the renumeration more than adequate? What about the hours?

psst
Hey.
Does membership help with the chicks?

Seriously, I understand the origins of the bad feelings some people have concerning this whole situation. But it is disappointing to see intelligent people abandon logic and reason to the extent that I have seen.

Personally speaking, I've made a small fortune off of this, and also Crafty has gained about +800 elo as well.

[/sarcasm]

I just bought a Yacht and am going to sail around the World with my payment )

[Adam Hair]

Personally speaking, I've made a small fortune off of this, and also Crafty has gained about +800 elo as well.

[/sarcasm]

I just bought a Yacht and am going to sail around the World with my payment )

Well gentlemen, I expected as much. After all, you are Levy's right hand men.

[hyatt]

Actually, _I_ am his "left-hand man." I KNOW what he does with his "right hand".