Thread: CRT offsets
View Single Post
Old 2021-04-06, 06:33   #16
CraigLo
 
Mar 2021

53 Posts
Default

Quote:
Originally Posted by henryzz View Post
As far as I understand it optimising a range of x should provide better odds of finding a gap of size y with y < x. We just need to work out the correct ratio of x and y.

I think that is correct. For a target gap size of G, all gaps >= G are going to contain some candidates < -G/2 or > G/2. You can't optimize the result while ignoring all of these.

Probably the best way to calculate this would be to scale the remaining candidates by the probability that they contribute to a gap of size G. Candidates near 0 would have a probability near 1. Candidates near +/- G/2 would have a probability near 0.5. Candidates near +/- G would have a probablility near 0. So run from -G to G and sum the probabilities for all remaining candidates and then minimize this.

I'm guessing you could get similar performance by doing what you suggested. Minimize the remaining candidates for some range > G which needs to be determined.
CraigLo is offline   Reply With Quote