Page 17 of 17
Re: Thoughts on Fruit=Rybka EVAL
Posted: Sat Sep 03, 2011 9:28 pm
by hyatt
If you actually believe that, more power to you. But you believe something that (a) is contrary to any good book on software engineering and (b) has NOT been claimed as a defense by Vas, publicly.
It is funny to watch a few of you try this "imaginary defense" strategy where you postulate explanation after explanation, with absolutely no supporting evidence, and then wait for someone to shoot the last one down, before coming out with a new one...
Re: Thoughts on Fruit=Rybka EVAL
Posted: Sun Sep 04, 2011 8:47 pm
by Uly
hyatt wrote: with absolutely no supporting evidence
Are you calling me a liar when depicting the incidents of breaking time management and having no previous TM source code to just copy there?
Re: Thoughts on Fruit=Rybka EVAL
Posted: Mon Sep 05, 2011 12:12 am
by hyatt
I thought I was pretty clear. The idea that ALL old versions are gone, just because one deletes code, is bogus. Nobody would incrementally update a program and then toss the old version. You never know when you might want to go back after discovering a serious bug you had not noticed.
Software just isn't normally developed in that ad hoc way, it wastes LOTS of time.
How could I call you a liar when I am absolutely certain that you have no idea what you are talking about here. Did Vas directly tell you that little saga? Or is it just one of those "could have happened" type hypotheses??
Re: Thoughts on Fruit=Rybka EVAL
Posted: Mon Sep 05, 2011 3:27 am
by Uly
hyatt wrote:Did Vas directly tell you that little saga?
I was there, I
lived the saga.
hyatt wrote:Software just isn't normally developed in that ad hoc way, it wastes LOTS of time.
Agreed, Vas wasted a lot of time testing a badly functioning time management code that in the end was suboptimal and probably cost Rybka the #1 place against Houdini, as the new time management depends on the system (what is optimal in some computer isn't in another), time control style (dead match/incremental/repeating), the time control itself (what works at blitz doesn't at slower time control) and even, opponent (what works against one opponent doesn't work against another) or conditions (no matter what values you use for incremental time controls, Rybka will play worse than in repeating time controls).
But it happened. Vas could have saved all that if he had working time management code from any older version, but he didn't have any.