You know what they say about statistics ...

petrw1 · 2007-10-05, 18:49

Just for fun (for me anyway) I looked at the last Million range before 10 Million Digits ... 32219000 - 33219013 since it seems to be taking "forever" to complete first time LL tests... I think there are anomalies and curiosities. For example:

4355 - Assignments. 7.3% factoring; 92.7% LL First time tests
95.5% assigned in 2007 - oldest is March 14, 2006
14.6% expired; 54.3% overdue; 58.1% no updates repoerted yet.

Of the 562 (12.9%) assigned between 2006/03/14 and 2007/06/30 almost 78% have updates reported; 37% are overdue and 3.7% are expired.

Of the 3793 (87.1%) assigned since 2007/07/01 21% have updates reported; 56.9% are overdue and 15.4% are expired.

I would have expected the new (re-)assignments to quite systematically take care of these old ranges ... but instead the new assignments are expiring more regularly than the older ones.

And the trend is not gradual as you might expect ... right from the oldest to June 30, 2007 you see screen after screen of mostly updated. The change is immediate after July 1, 2007 to mostly NOT updated. With a default 60 day check-in limit one would expect this change in report rate to be closer to early August.

retina · 2007-10-08, 13:29

I am inclined to blame the ISP's for excessive blocking of non-normal traffic in many places. Sometimes only HTTP and HTTPS traffic can be used through my ISP. My guess is that around the date you give above some new deep packet inspection system was released and installed.

Actually, I don't know how the Prime95 software talks to the server. If it is using HTTP:80 then of course all the above I mention above is invalid as the data would most likely pass though fine. If that is the case then perhaps the reason can be due to the old faithful, concluding too much detail from too small data sets. :p

2007-10-05, 18:49	#1
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 2²×5×7×37 Posts	You know what they say about statistics ... Just for fun (for me anyway) I looked at the last Million range before 10 Million Digits ... 32219000 - 33219013 since it seems to be taking "forever" to complete first time LL tests... I think there are anomalies and curiosities. For example: 4355 - Assignments. 7.3% factoring; 92.7% LL First time tests 95.5% assigned in 2007 - oldest is March 14, 2006 14.6% expired; 54.3% overdue; 58.1% no updates repoerted yet. Of the 562 (12.9%) assigned between 2006/03/14 and 2007/06/30 almost 78% have updates reported; 37% are overdue and 3.7% are expired. Of the 3793 (87.1%) assigned since 2007/07/01 21% have updates reported; 56.9% are overdue and 15.4% are expired. I would have expected the new (re-)assignments to quite systematically take care of these old ranges ... but instead the new assignments are expiring more regularly than the older ones. And the trend is not gradual as you might expect ... right from the oldest to June 30, 2007 you see screen after screen of mostly updated. The change is immediate after July 1, 2007 to mostly NOT updated. With a default 60 day check-in limit one would expect this change in report rate to be closer to early August.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Statistics	mahnouman	Information & Answers	4	2013-02-13 13:31
Statistics and scoring	kar_bon	No Prime Left Behind	85	2008-09-19 02:02
321 Statistics	paulunderwood	3*2^n-1 Search	1	2005-02-25 21:41
Statistics	R.D. Silverman	NFSNET Discussion	1	2004-06-14 18:40
New Statistics available	HiddenWarrior	Data	26	2004-04-13 03:38

2007-10-08, 13:29	#2
retina Undefined "The unspeakable one" Jun 2006 My evil lair 6448₁₀ Posts	I am inclined to blame the ISP's for excessive blocking of non-normal traffic in many places. Sometimes only HTTP and HTTPS traffic can be used through my ISP. My guess is that around the date you give above some new deep packet inspection system was released and installed. Actually, I don't know how the Prime95 software talks to the server. If it is using HTTP:80 then of course all the above I mention above is invalid as the data would most likely pass though fine. If that is the case then perhaps the reason can be due to the old faithful, concluding too much detail from too small data sets. :p