P-1 factoring anyone? - Page 30

petrw1 · 2010-03-12, 22:39

Quote:

Originally Posted by wblipp

Almost exactly 5% probability of 5 or more successes in 30 trials. Lucky, but not incredible.

William

Now 6 in 31 .... (this one in phase 1)

and another crash with a "problem addressing memory" error just after reporting.

but still registered on the DB.

sdbardwick · 2010-03-12, 22:50

Yeah, I gave up on P-1 testing because of all of the crashes after finding factors. Too much babysitting.

petrw1 · 2010-03-12, 23:10

Quote:

Originally Posted by sdbardwick

Yeah, I gave up on P-1 testing because of all of the crashes after finding factors. Too much babysitting.

AHHH so it's not just me and my PC...there have been others...

Uncwilly · 2010-03-12, 23:41

Quote:

Originally Posted by petrw1

I know I have had others suggesting P-1 completions in closer to 1 day but I have never been able to acheive that ... am I doing something wrong?

I have never seen the times that many claimed. I think that the memory timing may be a big issue.

petrw1 · 2010-03-12, 23:53

Quote:

Originally Posted by Uncwilly

I have never seen the times that many claimed. I think that the memory timing may be a big issue.

Considering the points awarded and the time spent it works out pretty close for me:

4.4 ponits over 2 days for a E6550 = 2.2 per day per core
Pretty much exactly what it gets for DC or LL.

cheesehead · 2010-03-13, 00:05

Quote:

Originally Posted by petrw1

Do I simply have a statistical anomaly

Quote:

Originally Posted by Mini-Geek

I'd call it an anomaly. (as long as no pedantic people want to consider the word as being overly-specific, but a quick dictionary check suggests it'd be good enough) An unlikely event, for sure, but not so impossibly unlikely as to be a sure indicator of a problem.

The Strong Law of Small Numbers may be relevant here, too.

http://mathworld.wolfram.com/StrongL...llNumbers.html

Mini-Geek · 2010-03-13, 01:27

Quote:

Originally Posted by cheesehead

The Strong Law of Small Numbers may be relevant here, too.

http://mathworld.wolfram.com/StrongL...llNumbers.html

I don't see how.

The Strong Law of Small Numbers is talking about small numbers (or sequences of them) appearing in more places than you might expect because there's so few of them. I don't see how that applies...unless you meant some sort of related "law" regarding the statistics of a small sample size and how that can produce "unexpected" results at a relatively high rate.

petrw1 · 2010-03-13, 01:36

For what stats are worth....

I've had 10 PC's that have done 12 P-1 tests in the last 365 days (the lowest is 12 tests). The success rates are:
(I understand that the smaller the sample size the greater the chance of being away from the expected value)

2/25 = 8.0%
1/25 = 4.0%
3/49 = 6.1%
1/27 = 3.7%
4/58 = 6.9%
2/12 = 16.7%
8/88 = 9.1% (this is the one that is 6/31 recently)
4/60 = 6.7%
1/34 = 2.9%
4/61 = 6.6%
===========
30/439 = 6.8%
34/471 = 7.2% LIFETIME

Guess I'm within the statistical norms...

davieddy · 2010-03-13, 07:34

Quote:

Originally Posted by wblipp

Almost exactly 5% probability of 5 or more successes in 30 trials. Lucky, but not incredible.

William

From "almost exactly" and knowing you, I would guess you
used the binomial distribution with n=30 and p=0.066.

To get a "feel" for the likelihood of the actual number of
successes when the expected number is 2, I resort to
the Poisson distribution whenever it is approximately
applicable:

Probability of 0,1,2,3 or 4 successes is

(1 + 2 + 2 + 4/3 + 2/3) / e² = 0.95

David

Noting also that the standard deviation is sqrt(2).

petrw1 · 2010-03-19, 02:28

My P-1 assignments jumped from 51.8M to 53.0M recently.
At the same time about 30,000 TF Available exponents appeared in the 51-52M ranges.

I'm going to take a guess that the 51-52M range was released to do the final bit of TF.

I'm going to further guess that these assignments will then be assigned the old (v4 way) letting the tester do P-1 and then LL.

Finally, I'm going to guess this is to give the P-1 workers a better chance to stay ahead of the LL testers.

Anyone know the real facts?

cheesehead · 2010-03-19, 04:24

Quote:

Originally Posted by petrw1

Finally, I'm going to guess this is to give the P-1 workers a better chance to stay ahead of the LL testers.

Or, to put it another way, to keep the P-1 workers from holding up LL assignments.

2010-03-13, 01:36	#327
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 2²·3·17·23 Posts	My P-1 Success rates For what stats are worth.... I've had 10 PC's that have done 12 P-1 tests in the last 365 days (the lowest is 12 tests). The success rates are: (I understand that the smaller the sample size the greater the chance of being away from the expected value) 2/25 = 8.0% 1/25 = 4.0% 3/49 = 6.1% 1/27 = 3.7% 4/58 = 6.9% 2/12 = 16.7% 8/88 = 9.1% (this is the one that is 6/31 recently) 4/60 = 6.7% 1/34 = 2.9% 4/61 = 6.6% =========== 30/439 = 6.8% 34/471 = 7.2% LIFETIME Guess I'm within the statistical norms...

2010-03-19, 02:28	#329
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 2²×3×17×23 Posts	A gap in the P-1 assignments.... My P-1 assignments jumped from 51.8M to 53.0M recently. At the same time about 30,000 TF Available exponents appeared in the 51-52M ranges. I'm going to take a guess that the 51-52M range was released to do the final bit of TF. I'm going to further guess that these assignments will then be assigned the old (v4 way) letting the tester do P-1 and then LL. Finally, I'm going to guess this is to give the P-1 workers a better chance to stay ahead of the LL testers. Anyone know the real facts?

2010-03-12, 22:50	#321
sdbardwick Aug 2002 North San Diego County 685₁₀ Posts	Yeah, I gave up on P-1 testing because of all of the crashes after finding factors. Too much babysitting.