On 02/11/06, Gavin Kistner <gavin.kistner / anark.com> wrote:
> The second line is the result of recording, each time the fuzzy time
> changed, how many minutes ahead or behind the actual time the change
> occurred. Ideally, you should see the same percentages spread along the
> whole range.

I see now why the quiz mentioned that we might need to assume a
regular observation period of e.g. 7 seconds - to allow us to even out
the distribution of times.  However, to do this would also require
that we know what the total duration of the observation was.
Otherwise surely the only solution to ensure roughly average counts of
-5 to +5 would be to start by wandering slowly from -5 mins to +5 mins
then back (a bit more quickly) to -5 mins again... and repeat?

Then after doing that a few times you might think you can become a bit
more random and wiggle the wandering a bit... but who's to say that
the observer isn't going to start the statistics from this point on?
Then you're going to look a bit silly with all that extra wiggling and
not averaging out the times.

So better stick to just wandering back and forth, back and forth like
clockwork... is this random? :)

When I tried increasing the RUNS_PER_SET value in Gavin's
FuzzyTimeTester from 1 to about 15, I noticed some results take on a
more gaussian distribution with a slight weighting to the right
because of the "can't go back in time" condition.

I'm starting to think that this may be the only truely random result
given some of the conditions of the quiz... but I hope I can be proved
wrong!

Now looking forward to the conclusion and letting my brain relax at last :)

-- 
Marcel