[QUOTE=retina;412059]This line still appears in the report:[code]• Countdown to testing all exponents below M(57885161) once:
[/code][/QUOTE]

Oh... whoops. :smile:

I mean, no, it's not there, what are you talking about?

[QUOTE=Madpoo;412061]I mean, no, it's not there, what are you talking about?[/QUOTE]You are right. It must have been in my imagination only. Sorry to bother you with my mental problems. Please carry on with whatever you were doing.

[QUOTE=GIMPS milestones]Countdown to first time checking all exponents below 61M: [URL="http://www.mersenne.org/assignments/?exp_lo=58496057&exp_hi=61000000&execm=1&exp1=1&extf=1&exdchk=1"]1,560[/URL] [COLOR=black]([B][COLOR=Red]666[/COLOR] [/B]still unassigned)[/COLOR][/QUOTE]
Yarrr, there must be a prime laying there... :devil:

[QUOTE=retina;412062]You are right. It must have been in my imagination only. Sorry to bother you with my mental problems. Please carry on with whatever you were doing.[/QUOTE]

(That's because I quickly removed the evidence before I hit "submit reply") :devil:

[QUOTE=Madpoo;412066](That's because I quickly removed the evidence before I hit "submit reply") :devil:[/QUOTE]

Whoosh!

[QUOTE=Madpoo;411942]Basically what you'd want to know is: What is the average error rate for first time checks?
...
We can say for sure that the error rate for exponents below 34,661,527 (everything below that has been DC'd) is:
61,047 bad out of a total 746,062 exponents tested at least twice (at least 1,492,124 tests) = ~4.1 % error rate
[/QUOTE]

For what it's worth, I was looking into something else (how well the error code tracks with the eventual error rate).

In short, I found out that if you check in a result and Primenet marks it as "Suspect", there's a HIGH probability it's bad. If the known good/bad where it was originally marked suspect:
31185 bad, 27486 good (and 6450 still marked suspect, pending a triple-check).

Counting the known bad/good, we arrive at an error rate of 31185/(31185+27486) = 53.2%. Ouch.

Conversely, the known error rate for "00000000" (again, only counting the "known bad / known good", ignoring the numerous unknowns) is just 1.8%.

1.8% isn't bad... better than the average 4.1% I mentioned. So if you report an error free run, you're actually doing far better than if you had any other kind of error.

Oddly, if you ever get the error code "01000100" you have an even LOWER chance of it being bad... only 0.8%. Must be because that indicates the program encountered some repeatable error (I think that's what it means?) and the fact that it could be repeated means the system is actually more stable than if it went undetected in the first place?

For all of the "non-suspect" error codes, it breaks down thusly so far:
29008 bad, 1467321 good, for a known error rate of 1.9%

In summary, don't let the 4% error rate scare you too much... if you report a non-suspect result, the odds are better than that, and you should only sweat it if you get a suspect result at which point it's a wee bit worse than a flip of the coin.

That is useful information in those last couple of posts about error rates, could a mod maybe split those into a new thread under the FAQ/information subforum?

[QUOTE=Madpoo;411989]That's definitely weird. Oh well... at least Xebecer still doesn't seem to have checked anything in, which I assumed would be the case, but what odd timing that someone with an expired assignment on that *finally* after months and months checks something in at long last, shortly after I did. Weird.[/QUOTE]

Well, with about 2 days to go before Xebecer's assignment would have been recycled (if there hadn't been any poaching), the exponent [url=http://www.mersenne.org/report_exponent/?exp_lo=57398083&exp_hi=&full=1]M57398083[/url] has now been triple-checked. But the triple check was not done by Xebecer or anyone else who had ever been assigned the exponent (as far as I can tell). If Xebecer is actually working on it, and checks in a result, that will make a quadruple-check.

[QUOTE=Madpoo;412393]...
In summary, don't let the 4% error rate scare you too much... if you report a non-suspect result, the odds are better than that, and you should only sweat it if you get a suspect result at which point it's a wee bit worse than a flip of the coin.[/QUOTE]

I had a thought just now... if we assume a roughly 4% error rate, considering the 422,811 exponents left before we verify M48, that means there are potentially another 16-17K bad exponents hiding in there, waiting for someone to do a *proper* check on it. :smile:

But that's not really all of the story... a lot of those "probably bad" have already been DC'd without a match. There are something like 5000 exponents below 58M that are awaiting a triple-check. Many of those were marked suspect the first time around so they got reassigned pretty quick and resulted in the mismatch.

That still leaves quite a bit of uncharted territory out there. Right now the trick for me is trying to find these bad machines, but at some point we'll have to get more creative to try and find the bad stuff in advance. Otherwise, the DC work will hopefully, eventually get done in that range. Until the very last one is done I'll probably still have my theory that there's a hidden, missing prime in there. I reckon there are still 10-11K possibilities for that.

The latest dark background is terrible.
[url]http://www.mersenne.org/report_milestones/[/url]
:yucky: hard to read the black text further down. Hard to read the green text near the top. :yucky:

[QUOTE=retina;412988]The latest dark background is terrible.
[url]http://www.mersenne.org/report_milestones/[/url]
:yucky: hard to read the black text further down. Hard to read the green text near the top. :yucky:[/QUOTE]

I don't think you see what the rest of us see:[ATTACH]13264[/ATTACH] or it has been fixed in the last hour or so...