Programmers Open Letter to ICGA on Rybka/Fruit

[BB+]

Things have quieted down a bit, so what is the latest on the investigation?

Attempting to quantify the evaluation feature overlap of Rybka-Fruit (for both Rybka 1.0 Beta and Rybka 2.3.2a).
I think Bob is poking around to try to see if any of the Rybkas used Crafty-specific bitboard code (beyond just the general rotated bitboard idea).

[kingliveson]

Things have quieted down a bit, so what is the latest on the investigation?

Attempting to quantify the evaluation feature overlap of Rybka-Fruit (for both Rybka 1.0 Beta and Rybka 2.3.2a).
I think Bob is poking around to try to see if any of the Rybkas used Crafty-specific bitboard code (beyond just the general rotated bitboard idea).

Well, I guess anymore digging at this point is just for the history books as there's already enough for ICGA to make a decision.

[Gerold]

Things have quieted down a bit, so what is the latest on the investigation?

Attempting to quantify the evaluation feature overlap of Rybka-Fruit (for both Rybka 1.0 Beta and Rybka 2.3.2a).
I think Bob is poking around to try to see if any of the Rybkas used Crafty-specific bitboard code (beyond just the general rotated bitboard idea).

Well, I guess anymore digging at this point is just for the history books as there's already enough for ICGA to make a decision.

I agree with you.
Why is ICGA taking so long.

[Jeremy Bernstein]

I agree with you.
Why is ICGA taking so long.

Good, fast, cheap -- choose two...

Jeremy

[hyatt]

Things have quieted down a bit, so what is the latest on the investigation?

Attempting to quantify the evaluation feature overlap of Rybka-Fruit (for both Rybka 1.0 Beta and Rybka 2.3.2a).
I think Bob is poking around to try to see if any of the Rybkas used Crafty-specific bitboard code (beyond just the general rotated bitboard idea).

Well, I guess anymore digging at this point is just for the history books as there's already enough for ICGA to make a decision.

I agree with you.
Why is ICGA taking so long.

Why don't you dive in and help? This takes time and effort. With no pay.

[BB+]

Well, I guess anymore digging at this point is just for the history books as there's already enough for ICGA to make a decision.

There seems to be a desire to "forget past history" and consider Rybka 2.3.2a (as being representative of WCCC 2007 -- though even that could end up being an issue) de novo, without regard to Rybka 1.6.1 or Rybka 1.0 Beta (or even Rybka 2.3, or the WCCC 2006 version), so all the info about Rybka 1.0 Beta is half-irrelevant by that measure. The amount of evaluation feature overlap for that version was also disputed, and a quasi-scientific quantification of this is ongoing (rather tedisome, list every single feature, determine how close two engines are). I suspect the "investigatory" phase should wrap up next week, and after any relevant stalling tactics expire, the process should then be (gasp) ready to move on to the next phase.

[Uly]

all the info about Rybka 1.0 Beta is half-irrelevant by that measure.

What's the relevancy of the other half? Just putting Vas as a cloner as a precedent?

[BB+]

There seems to be a desire to "forget past history" and consider Rybka 2.3.2a de novo, so all the info about Rybka 1.0 Beta is half-irrelevant by that measure.

What's the relevancy of the other half? [...]

Well, in the context of why there was such "slowness", what I meant here was that (approximately) half of the work with Rybka 1.0 Beta needed to be re-done for Rybka 2.3.2a (and various in-between versions in some cases), while the other half can be "copy/pasted" as it were. EDIT: to be more specific, my comment that "I have not checked every version of Rybka, but have verified that the evaluation function in Rybka 2.3.2a is substantially the same as in Rybka 1.0 Beta" came under criticism as being too vague, leading to the quantification process.

There is one group of voices which posits that everything before approximately Rybka 2.3.2a lacks relevance because nothing prior won an ICGA event. Others take the stand that Turin 2006 is the first version of interest to the ICGA (and that if guilt is found there, at the very least an apology/explanation is necessary for that version). Then a third group expects the Panel [in this instance or separately] to address Rybka 1.0 Beta and (say) Paderborn 2005. [Persons from all three of these groups seem to exist both inside the Panel and in the Internet at large]. There is nothing particular in the Statutes that either prohibits or mandates any of these investigations. I'm not sure that the Panel should become a "clearinghouse" for cloning investigations beyond just ICGA events, but it seems more suited to undertake such work than other entities.
[1]The purpose of the Panel shall be to:
[a] Investigate and discuss allegations of cloning or creating a derivative of strategy games programs;
[...]
[3h][iii] Recommending to other computer event organizers the exclusion of persons who have been found guilty by the Panel.

[Uly]

[3h][iii] Recommending to other computer event organizers the exclusion of persons who have been found guilty by the Panel.

Oh, I finally get it, since it's the person who is getting excluded, and not the software, now, it's irrelevant if Rybka 4.1 and future versions are clean, since proof that Rybka 1.6.1 is a clone of Crafty would be enough to exclude Vas.

[BB+]

Oh, I finally get it, since it's the person who is getting excluded, and not the software, now, it's irrelevant if Rybka 4.1 and future versions are clean, since proof that Rybka 1.6.1 is a clone of Crafty would be enough to exclude Vas.

That's the argument of some, particularly those who favour an "author-based" approach rather than one based merely on entries. Namely, once an author is found to be "cloning", said author should be excluded until such matters are cleared up, apologies are rendered, etc. The example of ElChinito [and maybe List] comes to mind as a precedent for this. However, given the ex post facto nature of much of the Fruit/Rybka question, there seems to be greater inclination than might be otherwise for an "entry"-based diagnosis [which from my standpoint at least means a lot of extra gruntwork]. I fully expect that the ICGA would not accept a future entry from Rajlich until the Crafty/Fruit episodes are satisfactorily resolved to their mind -- this is my interpretation of the intent of Statute [3h][iii] as quoted above. But as Bob (and others) have pointed out, trying to figure out if/how to re-write the tournament history would be a somewhat different matter.

Good, fast, cheap -- choose two...

Thorough, documented... And if it turns out that either of the Turin/Amsterdam WCCC entries from Rajlich differed from the commercial versions, I expect to get another load of work. Even the question of whether the Paderborn entry (late Dec 2005) was "sufficiently close" to Rybka 1.0 Beta (early Dec 2005) has been raised.