I think 19th and most of 20th scores were still Ok, then it became totally crazy. I would apply #1 starting 21st or remove totally from 21st to 23rd to make it simpler.

We believe we have enough information (thanks to Bok @ Free-DC) to reverse the credit anomaly which happened between the 19th and 23rd of October.

What happened? For these days users were continuously awarded credit at a rate proportional to how much they'd cumulatively done since the 19th. I can go into further details if anyone is interested.

How could it be remedied:

Solution #1: revert all credit accrued between Friday the 19th and Tuesday the 23rd in the user totals. This would have no effect on accumulated Gridcoin, but would on DHE totals. This would also remove all credit genuinely earned during this time period.

Solution #2: increase credit allocation rate by (a) having a sprint involving DHE and/or (b) increasing the DHE credit rate. This would have the effect of future credit 'drowning out' the anomaly.

Option #3: do nothing.

How do people feel about these? We could put this to a vote.

Kind regards,
Michael

Option #4: re-calculate the results for those three days

If credit during those days is known and there's a ratio of normal credit vs high credit on those days, there could be a ratio applied. If normal credit is 1 and those days PCs were getting 20 then divide by 20 for those days instead of zeroing everyone.

The problem there is that the glitch didn't affect every host, and didn't affect every host evenly; not all hosts that were affected were affected for the whole period. In practice that means that trying to reverse engineer the results would have to be done on a per-user basis and would have poor accuracy. Huge workload + poor results = lots of annoyed people.

Option 1 seems the most practicable option, unless there's enough data in the audit logs to "replay" the timeline with corrected credits. If the project didn't anticipate an issue like this though, the odds of having that data are low.

If credit during those days is known and there's a ratio of normal credit vs high credit on those days, there could be a ratio applied. If normal credit is 1 and those days PCs were getting 20 then divide by 20 for those days instead of zeroing everyone.

The problem there is that the glitch didn't affect every host, and didn't affect every host evenly; not all hosts that were affected were affected for the whole period. In practice that means that trying to reverse engineer the results would have to be done on a per-user basis and would have poor accuracy. Huge workload + poor results = lots of annoyed people.

Option 1 seems the most practicable option, unless there's enough data in the audit logs to "replay" the timeline with corrected credits. If the project didn't anticipate an issue like this though, the odds of having that data are low.

Option 1 affects everyone too and even worse then the proposed option 4. You're contradicting your own argument.

It was said to be Windows only so editing stats for those hosts only makes option 4 even better. The admin has already adjusted credit in the past. When I looked at mikey's host stats they pretty much had the same scale. Each day the credit grew. There will be no way for the admin to know if other things limit a CPU/host on a certain day more than other hosts. Option 4 is the most fair for everyone.

Option #3: do nothing.

As an early cruncher (seti since 1.1.2000) we did/do it for no gains because we knew and still know some projects give/gave more credits/hr than others.
Even if you get rewarded (gridcoin) for most it's still a no-win situation and you crunch just for mankind (my case).
Cheers and enjoy the weekend.
BR/Jarmo

No matter what option is used somebody will be unhappy.

So just pick one (throw a dice if you can not decide) but don't drag this out endless like you are doing with the badges.

OK thanks everyone for the feedback and mmonnin for your chart.

Here's the current plan:

1. Look at average credit per user for the three days prior to the issue.
2. Replace credit during the anomaly period by the average rate calculated above.

This would be done on the user and team totals (yes BOINC stores them all separately) and host data would remain incorrect. This would fix user and team rankings. RAC would not be affected and would settle as normal.

This would be done in the next few days.

Kind regards,
Michael

Done

Michael DHEP wrote:

Done

Nicely done! I only lost a fraction of credit (something like 4.5M to 4.2M) and nothing looks broken.

---

Michael DHEP wrote:

Done

Nicely done! I only lost a fraction of credit (something like 4.5M to 4.2M) and nothing looks broken.

Many thanks to Bok @ Free-DC for providing the life saving stats backup. 👏👏

Hello folks,
I' m here just since the 18th of October.
One question: are the former posts describing and discussing the Circumstances that led to the phenomene loosing 1.4 million points in 1 hour?
greets,
vonboedefeldt

Hi, I've just lost 7.5M in last update.

but still the figures for the hosts are incorrect, the total has been corrected but not the amount for the individuals hosts that are still showing previous scores.

should be possible that the project updates itself e.g. on daily basis or every twelve hours, because I have to update all computers to know the progress??
otherwise it remains at per last update ;-)

---

Over Sailor's grave doesn't grow grass

CRAP!!! Now I'm under the badge level for 5 million. Going to have to crunch like mad to get it back! :)

---

If credit during those days is known and there's a ratio of normal credit vs high credit on those days, there could be a ratio applied. If normal credit is 1 and those days PCs were getting 20 then divide by 20 for those days instead of zeroing everyone.

Free-DC has host credit per day if its allowed to be exported. An example is mikey since he was one of the top users with export and multiple PCs
https://stats3.free-dc.org/stats.php?page=user&proj=dhe&name=1018

Host 994 has info before, during and after the credit spike.
https://stats3.free-dc.org/stats.php?page=host&proj=dhe&hostid=994



Some things to note:
Credit looked to be higher before on some of his PCs before the spike than after the spike.
The 3 high days were not the same credit per day like many other days. Maybe ave out those 3 days.

WOO HOO I'm famous!!! Yes some of my pc's did get GREAT credits during those few days, I have no idea why and it doesn't matter to me what they do about it as I'm not here for the credits. A few million credits and $10 can get me a cup of coffee at Starbucks, but without the $10 I will remain thirsty!!

That being said there could be a 4th option but it could be a bit more difficult to implement, figure out which pc's got the credit advantage and which ones didn't, then give it to those that didn't.

Hi, I've just lost 7.5M in last update.

but still the figures for the hosts are incorrect, the total has been corrected but not the amount for the individuals hosts that are still showing previous scores.

should be possible that the project updates itself e.g. on daily basis or every twelve hours, because I have to update all computers to know the progress??
otherwise it remains at per last update ;-)

I lost 50 million credits today!! -50,291,636

CRAP!!! Now I'm under the badge level for 5 million. Going to have to crunch like mad to get it back! :)

Hey I lost -50,291,636 TODAY!!!

---

CRAP!!! Now I'm under the badge level for 5 million. Going to have to crunch like mad to get it back! :)

Hey I lost -50,291,636 TODAY!!!

hey there mikey :) Good seeing you here :)

Not sure what I lost, maybe only a few hundred thousand. But I hadn't set up all my cpus yet until after it happened. So no complaints from me.

Everyone lost what was due. Nothing to be concerned about. Fair decision by Michael. Balance has been restored.

Thought, wow! what a project! I've never seen CPU credits pile up like that before! Like 17M in three days. I never bothered to check the message boards but did free up as many CPU cores as possible and set them all to 100%. Shocking to see the correction but right thing to do.

---

According to free-dc stats I only lost 119,000 since the last update. no major loss. Time to get my cpu's going again.

Already told but hosts credits should be updated too.

how could be to have hosts with:
- 3,8xx,xxx.0
- 1.8xx,xxx.0
- 1.4xx,xxx.0
- 1.1xx,xxx.0
and total of
- 1.7xx,xxx.0 when this is the real credits and the hosts ones are still adding to the old (incorrect) ones

---

Over Sailor's grave doesn't grow grass