Strategic Double Clicking

[cuBerBruce]

I matched Bruno's residue for M58475341.

[Madpoo]

Originally Posted by cuBerBruce

I matched Bruno's residue for M58475341.

Here are the current stats for UBR47K's machine that he was concerned about:

Good	Bad	Susp	Unk	Fact	Solo	Mis
31	5	0	20	0	14	6

6 more mismatches in the bunch, so the concerns about this system probably mean those 6 results were bad but the double-check got it right.

In general it's unusual for 2 runs of the same exponent to *both* be bad. I mean, it happens, but if we say that 1% of non-suspect runs end up bad, then the odds of 2 non-suspect runs being bad is 0.01%

Which reminds me, I meant to play with my queries a little more (no jokes, I realize how that sounds) and see what happens if I do something with the data to assume mismatches are bad.

I realize that could be erroneous for people like us who are *deliberately* looking for mismatches so I'll just keep that in mind, and then I have to consider which of the two runs seems bad, so I'll probably just do something to compare the two systems and see which one is more likely to be wrong.

I've already scoured the mismatches that exist now, looking for any where neither machine seemed to be bad (and also looked for the ones where *both* looked bad). What's left are the ones where one is sketchy, one is generally okay.

Okay, enough digressing.

[cuBerBruce]

UBR47K did a self-double-check of M58491773, poaching my assignment. Since it is a self-double-check, I think I'll let my half-done assignment continue to provide an independent double-check.

[Madpoo]

Originally Posted by cuBerBruce

UBR47K did a self-double-check of M58491773, poaching my assignment. Since it is a self-double-check, I think I'll let my half-done assignment continue to provide an independent double-check.

Oh yeah, he sure did. If you could finish up yours that'd be great since I'm still doing independent triple-checks as needed (much to LaurV's dismay) Or if you wanted to free up that machine you could send me your save file and I can finish it up for you.

[cuBerBruce]

Originally Posted by Madpoo

Oh yeah, he sure did. If you could finish up yours that'd be great since I'm still doing independent triple-checks as needed (much to LaurV's dismay) Or if you wanted to free up that machine you could send me your save file and I can finish it up for you.

I can have it finished in less than 48 hours. I'll just let it finish.

[Madpoo]

In case anyone wants more of these fun assignments...

I thought it might be fun to sort these by the bad/good ratio for the machine that ran it... I picked stuff with a 3:1 bad to good or higher. So it's sorted by that ratio (descending), then bad (descending), suspect (descending), good (ascending) and finally exponent (ascending). That should bubble the most likely to be wrong to the top of the list... in theory.

exponent	Ratio	Bad	Good	Unk	Sus	Solo	Mis	worktodo
38130749	4.00	4	1	1	2	1	2	DoubleCheck=38130749,71,1
43919461	4.00	4	0	12	1	11	2	DoubleCheck=43919461,72,1
43919527	4.00	4	0	12	1	11	2	DoubleCheck=43919527,72,1
43919671	4.00	4	0	12	1	11	2	DoubleCheck=43919671,72,1
44771917	4.00	4	0	12	1	11	2	DoubleCheck=44771917,72,1
44946037	4.00	4	0	12	1	11	2	DoubleCheck=44946037,72,1
45535901	4.00	4	0	12	1	11	2	DoubleCheck=45535901,72,1
46300987	4.00	4	0	12	1	11	2	DoubleCheck=46300987,72,1
47425219	4.00	4	0	12	1	11	2	DoubleCheck=47425219,72,1
47568511	4.00	4	0	12	1	11	2	DoubleCheck=47568511,72,1
37595837	3.00	6	2	3	2	2	3	DoubleCheck=37595837,71,1
42036389	3.00	3	0	3	1	3	1	DoubleCheck=42036389,72,1
42407327	3.00	3	0	6	4	6	4	DoubleCheck=42407327,72,1
42719503	3.00	3	0	6	4	6	4	DoubleCheck=42719503,72,1
43389751	3.00	3	0	3	1	3	1	DoubleCheck=43389751,72,1
43612379	3.00	3	0	6	4	6	4	DoubleCheck=43612379,72,1
43927669	3.00	3	0	6	4	6	4	DoubleCheck=43927669,72,1
44486327	3.00	3	1	3	0	2	1	DoubleCheck=44486327,72,1
45381379	3.00	3	1	3	0	2	1	DoubleCheck=45381379,72,1
45663269	3.00	3	0	2	1	2	1	DoubleCheck=45663269,72,1
45663271	3.00	3	0	2	1	2	1	DoubleCheck=45663271,72,1
51064841	3.00	3	1	1	1	1	1	DoubleCheck=51064841,73,1

[sdbardwick]

Originally Posted by Madpoo

Oh, what the heck, if anyone wants to try some of these out, here's a group where the cpu in question had a 5:1 bad/good ratio:

exponent	Bad	Good	Unk	Sus	Solo	Mis	worktodo
36099269	11	2	8	2	6	4	DoubleCheck=36099269,71,1
36527723	5	1	6	1	5	2	DoubleCheck=36527723,71,1
36574903	10	2	7	0	5	2	DoubleCheck=36574903,71,1
36631909	11	2	8	2	6	4	DoubleCheck=36631909,71,1
36717617	11	2	8	2	6	4	DoubleCheck=36717617,71,1

Finished the above.
M36574903 needs a TC; others matched.

[Madpoo]

Originally Posted by sdbardwick

Finished the above.
M36574903 needs a TC; others matched.

Hmm... I would have hoped for more mismatches, especially that one that had just 11 bad, 2 good.

I looked into that one and all of it's results came in through Primenet version 4 which doesn't have the same detail about the date the result came in. When that happens I just lump all of their results into an arbitrary "2007" year. So it might be trickier to nail down a bad period in that machine's history by a certain date.

Narrowing by the app version turned out to be a help for those old results, but if they used the same app version for several years then it didn't do much.

This particular user, I think, submitted results from 2004 up to 2008 so the bad results could have been a particular stretch of time but it's hard to say.

I'm afraid this may make me look at updating the dates for all of the LL results using the old version 4 logs we recently added to the "history" section of the report. Matching up the raw logs with the LL result stuff could be tricky but may be worth it for things like this.

[endless mike]

Originally Posted by Madpoo

In case anyone wants more of these fun assignments...

I thought it might be fun to sort these by the bad/good ratio for the machine that ran it... I picked stuff with a 3:1 bad to good or higher. So it's sorted by that ratio (descending), then bad (descending), suspect (descending), good (ascending) and finally exponent (ascending). That should bubble the most likely to be wrong to the top of the list... in theory.

exponent	Ratio	Bad	Good	Unk	Sus	Solo	Mis	worktodo
38130749	4.00	4	1	1	2	1	2	DoubleCheck=38130749,71,1
43919461	4.00	4	0	12	1	11	2	DoubleCheck=43919461,72,1
43919527	4.00	4	0	12	1	11	2	DoubleCheck=43919527,72,1
43919671	4.00	4	0	12	1	11	2	DoubleCheck=43919671,72,1
44771917	4.00	4	0	12	1	11	2	DoubleCheck=44771917,72,1
44946037	4.00	4	0	12	1	11	2	DoubleCheck=44946037,72,1
45535901	4.00	4	0	12	1	11	2	DoubleCheck=45535901,72,1
46300987	4.00	4	0	12	1	11	2	DoubleCheck=46300987,72,1
47425219	4.00	4	0	12	1	11	2	DoubleCheck=47425219,72,1
47568511	4.00	4	0	12	1	11	2	DoubleCheck=47568511,72,1
37595837	3.00	6	2	3	2	2	3	DoubleCheck=37595837,71,1
42407327	3.00	3	0	6	4	6	4	DoubleCheck=42407327,72,1
42719503	3.00	3	0	6	4	6	4	DoubleCheck=42719503,72,1
43389751	3.00	3	0	3	1	3	1	DoubleCheck=43389751,72,1
43612379	3.00	3	0	6	4	6	4	DoubleCheck=43612379,72,1
43927669	3.00	3	0	6	4	6	4	DoubleCheck=43927669,72,1

Taking these 16

[cuBerBruce]

Originally Posted by cuBerBruce

I can have it finished in less than 48 hours. I'll just let it finish.

Triple check completed. (M58491773)

[cuBerBruce]

Originally Posted by Madpoo

In case anyone wants more of these fun assignments...

exponent	Ratio	Bad	Good	Unk	Sus	Solo	Mis	worktodo
44486327	3.00	3	1	3	0	2	1	DoubleCheck=44486327,72,1

I took this one.