Forget my previous post:
>This I find interesting. I assume the initial 1.4 billion iterations
>estimate was based on
>the probability of finding 23 1's in a random set of 108 bits. Since this
>probability
>cannot change how was the client able to significantly cheat the odds to
>find distinguished
>points 20% faster than expected? Is the recent change in the way the
>client starts the
>next point after a distinguished point is found responsible for the
>average iterations
>returning to the original estimate?
I finally figured out what you were saying all along. When a client starts
out with a new saved.state it will find the first point in an average of
1.4 billion iterations. But this point will only have been iterated an
average of 1.4 billion/PARAL or about 44 million times. These first points
will significantly pull down the overall average of the returned points.
The missing iterations are still in the saved.state file and will bring the
average back up in the long run.
-- Dan Oetting <oetting@ghtmail.cr.usgs.gov>
This archive was generated by hypermail 2b29 : Wed Mar 08 2000 - 01:49:14 MET