MATCH sanity

Discussion of chess software programming and technical issues.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Guenther
Posts: 2826
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: MATCH sanity

Post by Guenther » Fri May 05, 2017 6:22 pm

Rebel wrote:
Ferdy wrote: Based from this,

Code: Select all

103. Kf1 {0.00/33 0.29s, Black disconnects} 1-0
It is the engine that disconnected in the first place.
Doubt that because it can't :wink: But I think the -draw option will solve the problem.
I wouldn't do that. You'll hide problems. It seems ProDeo '2.4 future' crashed...
Current foe list count : [84 - still rising]
http://rwbc-chess.de/chronology.htm

User avatar
hgm
Posts: 23230
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: MATCH sanity

Post by hgm » Fri May 05, 2017 9:17 pm

I don't see how you can guarantee that. Any program can crash, even if it is just by bugs in the libraries you link to.

User avatar
Rebel
Posts: 4431
Joined: Thu Aug 18, 2011 10:04 am

Re: MATCH sanity

Post by Rebel » Sun May 14, 2017 7:13 am

Guenther wrote:I wouldn't do that. You'll hide problems. It seems ProDeo '2.4 future' crashed...
hgm wrote:I don't see how you can guarantee that. Any program can crash, even if it is just by bugs in the libraries you link to.
You both got me think twice and indeed after a long search (a week :x) I got the deeply hidden bastard in it's full gory glory making the engine crash in extreme circumstances.

Glory, because when I fixed the bug I expected at least a small improvement. The opposite was true, a regression of aprox. 10 elo. The world upside down, bugs that provide elo!

Now that I have insight what the bug is actually doing I further exploited it and got an extra 6 elo :lol:

So now I am on the crossroad fixing the bug losing 10 elo (no more crashes) or tune the bug and gain 6 elo with the risk of an exceptional crash.

Isn't CC life wonderful?


:lol:

User avatar
hgm
Posts: 23230
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: MATCH sanity

Post by hgm » Sun May 14, 2017 7:32 am

I don't know the details, but it seems to me it should always be possible to detect the conditions that would lead to a crash (like out-of-bound array accesses, or infinite recursion), and take appropriate action in that case. While leaving the unintended behavior that seems to produce Elo just do its thing when it would not produce a crash.

User avatar
Rebel
Posts: 4431
Joined: Thu Aug 18, 2011 10:04 am

Re: MATCH sanity

Post by Rebel » Sun May 14, 2017 8:15 am

In a nutshell, what happens is that the bug removes legal moves from the move list. I noticed these moves in general are extremely bad and thus play no role and it's where the elo comes from. new kind of pruning :wink:

Naturally sooner or later Murphy's Law shows up and things go wrong especially in the late endgame.

User avatar
hgm
Posts: 23230
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: MATCH sanity

Post by hgm » Sun May 14, 2017 9:02 am

Removing moves can cause poor play in unfavorable situations, but it should not be able to cause crashing.

User avatar
Rebel
Posts: 4431
Joined: Thu Aug 18, 2011 10:04 am

Re: MATCH sanity

Post by Rebel » Sun May 14, 2017 9:44 am

hgm wrote:Removing moves can cause poor play in unfavorable situations, but it should not be able to cause crashing.
You are lacking imagination.

Not every program setup the same.

User avatar
hgm
Posts: 23230
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: MATCH sanity

Post by hgm » Sun May 14, 2017 12:15 pm

Well, I say 'should not' and not 'cannot'. I imagine there is something horrebly wrong with the design if pruning moves makes it crash. Pruning moves is basically just playing a game with somewhat different rules that would make the pruned moves not pseudo-legal..

User avatar
Rebel
Posts: 4431
Joined: Thu Aug 18, 2011 10:04 am

Re: MATCH sanity

Post by Rebel » Sun May 14, 2017 12:45 pm

I understood what you were saying the first time.

It's a typical human fallacy to reject | deny | judge things people can not imagine without having experienced those things themselves.

Heck, on the issue at hand I might have even agreed with you one week ago although my choice of words (with regard to the above) would have been more careful.

Assumptions (judging etc.) without testing is not very scientific, you of all forgot?

User avatar
hgm
Posts: 23230
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: MATCH sanity

Post by hgm » Sun May 14, 2017 7:54 pm

Not really. When someone says something that is inconsistent, no testing is needed at all to know it is wrong. That is logic, i.e. mathematics. Mathematics is not an empirical science. One doesn't prove mathematical theorems by testing them.

Post Reply