Fuzz doesn't work with short maximum interval

I find that the maximum interval is applied after fuzz_bounds. So if the original interval is longer than maximum interval, the fuzz intervals will be longer than maximum interval, too. The current implementation will clamp the fuzz lntervals with maximum interval, so they will be equal to maximum interval. Should we apply maximum interval to the original interval before fuzz?

I think that intervals shouldn’t exceed max. interval. So the following inequality:
(interval + fuzz) <= max. interval
should always hold.

What about the (interval - fuzz)? For example, the original interval is 900 day and the (lower, upper) is (850, 950). If the max. interval is 365, the result from current method is (365, 365). I think (347, 365) would be better.

I agree that (347, 365) would be better. So I suppose we need some sort of asymmetric clamping.

I think the implementation is very simple. Just add an extra line to clamp the interval before fuzz_bounds. @dae, what do you think of?

I agree with @Expertium that the interval should never exceed the maximum, and agree with @L.M.Sherlock’s proposed solution, which keeps things simple. It’s a bit of a corner case already, and having half the normal fuzz is fine I think.

2 Likes

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.