TCEC stage 3 , New Houdini starts with a bang

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
bob
Posts: 20550
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: TCEC stage 3 , New Houdini starts with a bang

Post by bob » Thu Jul 28, 2016 12:54 am

syzygy wrote:
Raptor wrote:I suppose I follow your arguement now, all you are saying is that the rules should be clear and unbiased. Which is to say, if program A was misconfigured, and it was noted that it was misconfigured, either replay the game irrespective of the result or don't replay THAT game no matter what.

Your arguement being if we do not do that, the mis-configured engine has an opportunity that _all_ other programs are not offered (the 2nd shot at the game).

That sounds fair to me. And I would advocate having that rule clearly there, so that there is no confusion from here on.

PS: I know I stated exactly what you did, but I just wanted to make sure we are on the same page.
Exactly.

The only problem with such a rule is that I'm not sure how to phrase it. As was also already discussed, what to do in case the mistake is detected only much later?

In the end, this should be a very rare thing to happen. There are surely other things that could go wrong and are not covered by the rules and that we now could not think of even if we tried. But Murphy's law guarantees they will happen at some point.

In such cases it might be better to continue with the other games for the moment and take one or two days to consult with the participants and decide which solution is fairest to both sides (and to the other participants). Or one could let the TD decide and set up an appeals committee.

Of course this is all just a very insignificant quibble. But so is chess in general ;-)
There are an endless variety of possible glitches. I once played on a machine that slowly forgot how to correctly subtract two integer values (This was a Cray C90, so not a cheap box). How do we know that one of the machines used didn't do something equally bizarre, but only when an engine used some specific oddball instruction (such as popcnt) that many others do not?

I think there should be some statute of limitations (such as 24 hours) where such a mistake can be caught and corrected, otherwise the game stands whatever was done wrong.

User avatar
Ozymandias
Posts: 1126
Joined: Sun Oct 25, 2009 12:30 am

Re: TCEC stage 3 , New Houdini starts with a bang

Post by Ozymandias » Thu Jul 28, 2016 5:34 am

Milos wrote:I assume that ppl in general are never fully honest or saints, and most of the time behave just in their own interest.
I also firmly believe that ppl believing otherwise are either naive or simply dishonest and hypocritical.
Honest opinion, not popular to say out loud.

Post Reply