What does "fail low at the root" mean?

Code, algorithms, languages, construction...

What does "fail low at the root" mean?

Postby watersky33 » Wed Dec 24, 2014 1:16 pm

I often see quotes such as "fail low at the root" is problematic, but what does that mean really? Why would I get into a situation that none of the root moves is good? I mean, one of the root moves must be the actual move. Thanks for some nice explanation.
watersky33
 
Posts: 1
Joined: Fri Jun 14, 2013 7:36 pm

Re: What does "fail low at the root" mean?

Postby hyatt » Wed Dec 24, 2014 7:03 pm

Most programs use an aspiration window at the root. If you complete iteration N with a score of something like +0.16, then when you start the next iteration, the search is more efficient if you start with alpha = 0.0, beta = 0.32 or something similar. Now your search can prune branches with scores < 0.0 or > 0.32, which speeds the search up. But on occasion that window is not wide enough and the score returns 0.0, which is not the real score, but the original lower bound you gave the search. You know the score is at least dropping to 0.0 and it might be dropping further, so you have to reduce the alpha bound and search again. Getting this 0.0 lower bound on the search is called "failing low at the root".

It is, in general, a bad thing to happen, because it means the score is going to be less than you expected.
hyatt
 
Posts: 1242
Joined: Thu Jun 10, 2010 2:13 am
Location: University of Alabama at Birmingham


Return to Programming and Technical Discussions

Who is online

Users browsing this forum: No registered users and 1 guest

cron