Server feature request

[Mikus]

I would like to compare what kind of results I am getting vs. what kind of results other participants are getting. Unfortunately, ALL the statistics being shown are for cumulative structures submitted. Meaning that if a given team member is shown as having submitted 20,000,000 structures, I do not know how many of those were submitted for the current protein vs. how many are "credits" carried forward from previous activity.

I think an option to ask the server to show statistics from the current protein only would allow a much better understanding of how other participants are currently faring.

mikus

[Louis]

I agree. Another thing I would like to see would be how many results I've returned in the past 24 hours (or some other convenient period). That way I can make sure my work computer is actually running, and not aborted like this past weekend.:swear:

[wirthi]

Louis, check out http://www.statsman.org/; there you find what you are looking for.

Mikus, I'm on your side - such a feature would be great.

[Louis]

Excellent...eventually I'll figure out all the cool stuff. Now if I can just break into the top 1000. Muchas thanks-a-lot!

[MAD-ness]

I am not sure if it would be worthwhile for Howard to add protein-specific stats but you might try lobbying one of the stats gurus (perhaps reader50 at MacNN or Dyyryath at Free-DC) to do so.

You could do a crude kind of thing and just capture people's production at the time of each change-over.

They wouldn't have past data, but a decent scripter could track this in the future.

[wirthi]

Hmm .. and how could Dyy (and all the others) differ between uploads of the new and the old protein? During the 48 hours after the protein change, uploads could come from both the old and the new protein, and I see no chance how 3rd party stats could differ.

[Brian the Fist]

The stats for the current protein are stored separately in our database already, just not displayed, so this would be easy. But before I consider doing it, you'll have to propose exactly HOW it will work - i.e. user will click a button, or go to a separate page, or whatever. Be very specific about your idea as I think the stats is pretty cluttered as is and doesnt need any more info added to it.

[Scotttheking]

nm

[Brian the Roman]

Brian;
I like this idea a lot.
You could do this by creating a new column in almost all of your current stats tables that showed '# of structures on current protein'. This is VERY interesting from an ROI perspective.

Specifically;
1) in the table under the termometer you could make col 1 much narrower, either remove or narrow col 2, make 'view struct ' button col narrower and add a new column.
2) In the table underneath the above, make the rank col narrower and add the same col.
3) In the table underneath that and in the team tables, there seems to be plenty of room to simply add the column with no other changes required.

[Mikus]

Originally posted by Brian the Fist
The stats for the current protein are stored separately in our database already, just not displayed, so this would be easy. But before I consider doing it, you'll have to propose exactly HOW it will work - i.e. user will click a button, or go to a separate page, or whatever. Be very specific about your idea as I think the stats is pretty cluttered as is and doesnt need any more info added to it.

To avoid clutter, the simplest thing would be to have TWO pages in the present high-flyers format -- one that shows structure counts from the beginning of the project, the other that shows in the very same places structure counts from the beginning of the current protein. There would be two links on the logon page, one labeld "logon and show project stats" and the other labelled "logon and show cumulative stats". In addition, each of the two kinds of user-statistics pages would have a link to the other (so someone would not have to logon a second time to see the other kind of numbers).

[I'm not sure about the team statistics -- simplest would be if the user clicks from a "user stats this project" page, one would see team (and team member) structure counts in the very same places from the beginning of the current protein. And if clicked from a "user stats cumulative" page, one would see team (and team member) counts from the beginning of the project.]


If you did not MIND clutter, the thing to do would be to have TWO numbers on the page (one vertically below the other) *every* place where there now appears a "structure count" value. Then you could keep the present pages, just make each a little longer.


Please note that I asked for a way to see STRUCTURE COUNT values since the beginning of the current protein (rather than since the beginning of the project). I did *not* ask for changes to any other kind of values being shown on the stats pages.


One definite recommendation I would like to make -- put the "participants" graphics on a separate page of their own (which requires clicking on a link somewhere, to view). These graphics probably change very slowly (so don't have to be looked at each time), while it seems to me that these graphics require extra transmission resources (and maybe rendering resources) whenever the present "user statistics" page is being shown.

mikus

[reader50]

Interesting idea. I'll look into it later tonight (and likely tomorrow night too). wirthi is correct, late uploaders will skew (their own) results a little bit. But CASP season is over, and the average protein lasts much longer than a week. The typical error should be less than 5 percent, unless people were deliberately turning in units late.

[wirthi]

Originally posted by reader50
Interesting idea. I'll look into it later tonight (and likely tomorrow night too). wirthi is correct, late uploaders will skew (their own) results a little bit. But CASP season is over, and the average protein lasts much longer than a week. The typical error should be less than 5 percent, unless people were deliberately turning in units late.
And even if they do, they don't screw up science-related material but only their own stats. And the stats of their team, but with little effect (unless it's a one man team or the person buffers a huge ammount of workunits)

[reader50]

Ok, I hacked something up quickly. It shows in the All Members current lists only, and is not yet totaling for the team. Bugs would not surprise me, but it's after 1AM and I have to work tomorrow, so...


Originally posted by MAD-ness
...
They wouldn't have past data, but a decent scripter could track this in the future. Actually, there is a way to synthesize the past data, but it would take a bunch more code hacking. Probably not worth it before the next protein change gives a new baseline anyway. Unless you guys want to see totals for dead proteins? I'm not set up to show that kind of data just now, it would require fresh code.