Stats - Why a 0?

[cjohnsto]

Here is a section of my stats page for seiving. Why does the last entry have a 0 for PRP's saved? Already factored?

Factors next to enter (double check) 'active window' (@n=1881538)

p (T) k n Score Factor found Score changed Score was Score could be Reqd bias PRP saved
631.138T 10223 2718089 6.311 Sat 22-Jan-2005 8742.845 (1)
672.947T 21181 3504404 6.729 Tue 01-Mar-2005 15495.664 (1)
711.054T 24737 4468063 7.111 Fri 04-Mar-2005 26615.993 (1)
672.455T 24737 4620271 6.725 Sun 13-Feb-2005 26915.315 (1)
631.448T 4847 5177487 6.314 Tue 01-Feb-2005 31737.827 (1)
711.043T 21181 5419628 7.110 Fri 04-Mar-2005 39159.396 (1)
631.486T 22699 5432518 6.315 Wed 02-Feb-2005 34943.550 (1)
631.312T 4847 5996703 6.313 Fri 28-Jan-2005 42566.772 (1)
631.192T 4847 7020783 6.312 Mon 24-Jan-2005 58335.602 (1)
711.194T 10223 7730825 7.112 Wed 09-Mar-2005 26565.623 0

[MikeH]

Here is a section of my stats page for seiving. Why does the last entry have a 0 for PRP's saved? Already factored?

Already 2 PRP results returned. Might sound mad, but it's true. Basically this test was issued for PRP, that user's client failed to report progress for 90 days(?), so the test was re-issued, but both came back with results. Or it could be one of those that was first tested by the duff version of the client.

My opinion as to why a higher than expected number of tests of n well above the DC level have 2 PRPs is due to the little box on the client that reads "Transmit intermediate blocks". Unfortunately the consequences of unchecking this are not spelt out.

[vjs]

Mike,

Correct me if I'm wrong here.... It's also prossible that your dat it not totally upto date and someone has already sent in a factor at lower p. Basically the lowest p factor will be the one that scores.

[Nuri]

No, that's not the case. Just check Mike's factors excluded ... file to see it's the only factor for that k/n pair.

[MikeH]

No, that's not the case. Just check Mike's factors excluded ... file to see it's the only factor for that k/n pair.

Correct. If it was a duplicate it would score (a little), but it wouldn't be displayed on these pages.

Basically the lowest p factor will be the one that scores.

It's the first factor to be found for a k/n pair that scores not the lowest. Originally it was the lowest p that scored, but that changed in July 2003.

[vjs]

Very insteresting and very good.

Joe and I were holding back on processing our low p/high-n factexcl.txt partly b/c we were worried about removing peoples scores with lower p-value factors.

Looks like I'll go ahead and snail mail him the factexcl.txt's.