[RC5] Stats suggestion - back calculation to day 1

Blain Nelson blainn13 at earthlink.net
Mon Jan 8 12:40:33 EST 2001


Dan Oetting wrote:
> 
> You don't have to start the calculations from the beginning. You can
> arrive at the same result by starting with todays stats and working
> backwards to day 1. The main advantage is you are adding the higher
> weighted recent values first so the early results are closer to the final
> results and can be displayed or used for rankings after only 3 or 4 half
> lives have been processed. After 10 half lives the accuracy will be
> better than 99.9%.

I think his point was that it was easier to pull the data from the db
for a user all at once than to go through all users day by day, and that
it was about as easy to do all data for a user at once as to do it from
now back to a certain other date.  At least, my sql experience would
indicate that -- it's the hits to the db that slow you down, not
handling the data, even if it's seemingly huge.  There won't be many
individuals with more than a couple thousand dates to handle, and that's
not that huge of a deal to process.  

Your point about reasonable approximations is a good one, but this is a
case where I don't think exact values are going to take much more time
or effort -- it might even be simpler.

Take care,
Blain
-- 
Wanted:  Blain Nelson   aka   blain at blainn.cc   aka   Rahab the Tyrant 
aka  anon-20630 at anon.twwells.com   aka   My Kids' Dad  aka  not-Kristi
Subject is considered armed and legged.  Persecutors will be violated. 
http://www.blainn.cc/      ICQ:19857966       My hat says "It's Irony"
--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list