Can I please get one of your signatures.
Account number: 717121

Because then all the fun of updating your globe will be gone after only 2 years run time. You for instance would be stuck with the same globe for the last 23 years.

My thinking would be that if you retain the daily export for 30 days and take the newest and one nearest the 30 day mark, take run time difference and divide by days between those 2 exports, you'd have a fair mean. For the starters you could accept to just use that formula or test join date against last export date to see if less than 30 days, and use that as an avarager... molto complicato ;?

As I commented in the ''Going under...'' and "Building 600 TFL machine" threads, [we] may have to review if to stick to the old 100 Cobblestones per GFL for performance indicating, or go the new BOINC 200 Cobblestones/GFL. For the moments the credits have though not at all doubled [to what SETI gives per second, but including their GPU credits] per the new server 7 credit system. We used to run 1.59 TFL per year and after few days on new system it's 1.806/year, so holding off for a little to maybe when we get GPU contributions running at WCG. Anticipating, but not overly, that this will lift credit substantially [not having been allowed to sniff the tech kitchen at all how the credits will be approached for that part. Time will though remain time as I understand 1 day GPU is 1 day runtime as clocked by the CPU elapsed, meaning the term Runtime has to remain and is not clean CPU time or GPU time [GPU's do not understand the time concept, which is why Elapsed it taken]

To everyone else... plz leave this TFL/GFL/Runtime exchange to SNURK and Me.... wanting to make sure we coordinate any sig performance numbers with what's shown in the supplemental charts. SNURK, if you want to break this out to separate thread from this one or take it off-line to mail, please do.

thx

--//--

edit: added a expansion to the possible formula for starting crunchers.

What marvey11 said.
Indeed an all time average is readily available, but I think that value is of little interest. Only a recent (say 30 days) average would be interesting to add to the signature in my opinion.

SNURK, there might even be an alternative to keeping historical data for individual users. You could calculate the recent average runtime in the same way that the RAC is calculated. In that case you need store only the current value for the recent average runtime and the timestamp when this value was last calculated.

I already do this in my own suite of statistics scripts and it works quite well. For the implementation of the algorithm I used the article in the Unofficial BOINC Wiki for the Recent Average Credit and adapted the formulae to use runtime seconds instead of credits. It's even possible to simplify the formulae if you know in advance that you're only updating the data exactly once a day (which you're currently doing anyway).

There's one drawback, however: if your own data gets somehow corrupted, there's no way to recalculate the current value if you do not store historical data. You would have to start from scratch in that case.