[fran]

Is there a page somewhere that defines the meaning of each column in the stats section, and how they are calculated? Some are obvious. Others are less so.

Thanks!

[reader50]

There is no such definitions page at this time. It's been asked for a few times, but I've yet to create one. If you are curious about a specific column, feel free to ask. I can't really say if an oddball stat is "useful", but at least I'll tell you what it is.

[fran]

Okey Dokey. Here are the stats I would like to know more about what they mean and how they are calculated:

o Rate (how is this different from RAC?)
o RAC (seems to be slightly different from SETI's number. Just a delay or calculated differently?)
o Attend
o Power (I added up the ghz of my machines, and it doesn't come close to 36.4)

Thanks!
Erik

[reader50]

• Rate - We use a 7-day average. RAC seems to use a monthly average, and is supplied only by BOINC projects. At the time I wrote our stats backend, we didn't have any super-long workunit projects, like Climate. If I were to rewrite our backend today, I'd set a toggle variable so that a particular project could use a month-long averaged rate instead of a weekly-averaged rate.
• RAC - Recent Average Credit. This is a straight copy of what the BOINC project supplies. Differences will come from update delays.
• Attend - Attendance. The percentage of days on which this person returns results. ie - a person who dumps once a week will show a 14% attendance rate. It uses the past 100 days to figure it, and it's smart enough to compensate if you haven't been there a full 100 days yet.
• Power - This tries to convert the Rate average into MHz. Since the conversion ratio will vary by CPU, bus speed, and assorted other factors, it will never be quite right for all machines. In practice, the conversion factors have not been updated in awhile, so they don't take optimized clients into account. Also, they haven't been figured separately for each BOINC project. The majority of the conversion errors will tend to give an incorrectly high GHz estimate.

[fran]

Thanks for all the info! Great stuff.

Two more, if that's okay:

o Day Stats: Is this the last 24 hours? Or a fixed 24 hour period? If the latter, when does the day start/end? UTC?

o Week Stats Is this the last 7 days? Or the last full week? If the latter, when does the week start/end?

Thanks again.

Erik

[reader50]

If we used a fixed start of the day, then certain stats could reliably update only once per day. Also, we have users all over the world, in most (if not all) time zones. It was planned from the start that our stats could update at any interval, provided the projects supplied fresh enough data.

Our internal code uses the UNIX timestamp, ie - seconds since Jan 1, 1970. This implies that our current setup could update a maximum of once per second and no more frequently. In practice, our servers and data sources do not approach a 1-second update frequency.

On the sidebar, you can see how old that data is. The time ref is from when we parsed the data from the project servers. The Day Stats are stats accumulated in the 86,400 seconds before that update - ie, one standard day. The Week Stats are those accumulated in the 604,800 seconds before that update - one standard week.

In other words, it's for floating intervals, and will be reliable regardless of what time zone you are in when you check.

The Rate figure is not calculated directly from the Week Stats. In other words, Rate != (Week Stats) / 7. Such a calculation would glitch up for new users until they've been here a week. Instead, it shows Insufficient Data for the first 24 hours. After that, it extrapolates to a weekly rate as needed until the user has been here a full week.

[Knightrider]

Originally Posted by reader50

There is no such definitions page at this time. It's been asked for a few times, but I've yet to create one. If you are curious about a specific column, feel free to ask. I can't really say if an oddball stat is "useful", but at least I'll tell you what it is.

Thanks for these definitions, very helpful in understanding the stats. Perhaps you would be kind enough to post these to the 'Support the team stats' sticky.

K.

[zombie67]

I thought I understood the "Intercept" column, but I clearly don't.

Check me out here: http://teamstats.macnn.com/seti2/stats.php?page=s3

I shoud be in the high 70's at the time I post this.

My Intercept says 2:06. Just one ahead of me is thecranes (slightly slower than me) says 1:21. Just one ahead of hime is Colin (building no credits) also says 1:21. Finally, just one ahead of him is tschneider moving, but much slower than me or thecranes with 3:15.

How do I decode the meaning of these times?

TIA!

[reader50]

thecranes will pass Colin in 1 day and 18 hours, as predicted.
You will pass Colin in 2 days and 3 hours.
You will pass thecranes in 3 days and 2 hours.
You will pass tschneider in 3 days and 14 hours.

In each case, the Intercept column shows the nearest intercept for that member. You can get some more info from your Personal page, by setting the time reference to a future date. Look at the graphs at the bottom, showing nearby members' total stats. It will show projected passings. Also, your Tactical Report at the top will list recent passings, from that future time reference.

The graphs don't show everyone projected ahead, but I don't recall the exact reason for that. Maybe it only projects a few near you, so as to keep stats pages loading at acceptable speeds, maybe it's not projecting people with zero rates (this would be a code glitch). I'll look into that sometime.

Note: intercepts listed on a Personal page are all with respect to yourself. Intercepts on a generic Members page are the closest intercept for each member, if they have any projected intercepts.

[zombie67]

Spot? What's that? Is that new? Or have I missed that column for two weeks now?

TIA

[reader50]

Spot Rate or Spot Stats is since the last stats update, regardless of how little time ago that was. If it was 5 minutes ago, it gives you a more immediate response than waiting for hourly data to become available.

In practice, we have our SETI stats update every 30 minutes, so Spot figures aren't that much more immediate than Hourly data. At one time with Classic Seti (and a smaller team), we updated every 10 minutes.