| FAH Stats |
Update: Current
05.19.13, 9am CDT |
Time Now: 11:28am
Next Run: 32 min |
|
|
EOC Folding @ Home - Stats News & Updates
Pages (14): « First ... « 4 5 6 7 [8] 9 10 11 12 » ... Last »
| Fixed Team 0 Script07.13.05, 11:29am CST |
| I guess my mind must of been elsewhere the other day. Anyhow, the code changes that I mentioned below have been implemented in the Team 0 processing script. So now *all* teams should have the same processing functionality. |
|
| Slight Hiccup With Noon Update...07.11.05, 12:37pm CST |
I was cleaning up the code in the processing scripts, and I managed to break one of the files which in turn stopped the noon update.
I fixed the glitch and re-ran the update, so everything should be back to normal.
One change I made was when a new user is added. The way it *used* to go was a row was added with their figures all at zero, so points & WUs would go from zero to their current amount. This worked fine for new folders. However, for people who have been folding and get added (like teams working their way up the ranks or Team 0 people), this would cause a HUGE spike in their averages, throwing off many calculations.
To overcome this I devised a simple check. If they have less than 500 Points & 5 WUs, then their starting totals are all zeros, since more than likely they *are* a new folder. If their points are higher than that then basically I subtract 100 Points and 1 WU and insert the "inital" record with those figures. Then the current record gets inserted with their actual numbers. This prevents people from having huge spikes, and also from initially showing up as inactive. Not perfect but should be satisfactory for the time being. |
|
| Few Updates & Code Cleanup...07.10.05, 5:33pm CST |
Okay, so I didn't get around to doing everything the other day that I wanted to, however I'm working on the stats as much as I can spare...
I updated the overlib javascript library and the chartdirector library. Both of these changes should be transparent, however let me know if any of the hover-boxes look wrong, or the graphs don't display properly.
I managed to clean up and optimize a couple of the processing scripts, still have a couple left. I still have a few tricks up my sleeve, hopefully I can shave off a couple more minutes of processing when I'm done.
I don't remember if I mentioned this before or not (I think I just posted it in the forums). I did a test with a script that was multi-threaded, processing two teams at a time (since the server has two CPUs)... The performance improvement just wasn't good enough to justify eating up both CPUs for processing.
Anyhow, once I finish the code cleanup, I've got a couple requests for various modifications (they are in the folding stats section on the forum)... So I guess those will be my next tasks before tackling anything else...
|
|
| More Code Changes! (Yippee)07.07.05, 11:53am CST |
I made some more modifications to the processing scripts. I accidently killed the _ / (No Name) user in team 0, so it will take a week for those numbers to settle back down unfortunatly. (One reason why I hate putting an exception in the code for little instance.)
Anyhow, hopefully the changes should speed up processing time a little more (Assuming nothing goes wrong with the next update.) It's getting so close to the update (8min) that I didn't want to make any more changes. After the run I'll probably make some more and have them implemented by the 3pm update.
Keep your fingers crossed... |
|
| Minor Update...06.30.05, 4:03pm CST |
I re-wrote some of the code that checks each team for duplicate usernames making it MUCH more efficient.
The old code would take each name (as it was read from the txt file) and search through the entire array of names already processed for that team. It was very simple to code, however the downside was that the array being created was not in alphabetical order, thus there is no efficient way to search through it. The more people on the team the longer it took to search, and that time grew almost exponentially. Example being, OCAU with 6.4k people took 16s to dupe check & sort, HOCP with 8.4k took 28s. (s = seconds if you didn’t know.)
The improved code takes the first letter of each name, and drops it into an array that contains names that start with just that letter. (There is another array for all others like numbers or weird characters.) So now instead of having to do a random search through potentially thousands of names, it only has to check through a much smaller fraction. How much more efficient is it? Again with OCAU it shaved the time down to a mere 1s flat, and HOCP down to 1.7s!
Below is a little image I made to help visualize what is going (in a very basic sense). All the names that are grouped together really just point to their row in the master array containing all names for that team. After all names have been checked for duplicates, the master array is sorted (by total points) to determine team-rank, then inserted into the database.
The end results is processing time has been reduced by about 7 minutes. (I only have one update to compare to though.) |
|
| Updated a lot of HTML...06.23.05, 12:18pm CST |
One visitor was kind enough to point out some various HTML mistakes in the code, so I went ahead and corrected them. This pretty much meant touching every template, and several files.
Everything should still look and function the same, but if you happen to notice any link is borked or something is not showing up correctly, please send me an email. |
|
| Team 0 - Top 2,000 Individuals Now Processing!06.22.05, 2:19pm CST |
I completed the script to process *only* the top 2,000 individuals in team 0. I know it shows 1,997, but that is because of some duplicate usernames (stanford filtering out part of a username for various reasons).
It will take a week or so for the numbers to settle on that team, afterwards it should be intersting to see how they stack up. The best part for most people is that it fills in a lot of the gaps for the individual rankings list. |
|
| Forgot to mention...06.22.05, 1:18pm CST |
Something I forgot to mention, I'll probably put it on the FAQ page also, but if anyone wants an old team file for some reason, I keep an archive of them from each update. I usually prune all the files every now and then because they do take up quite a bit of space.
You can get a file by using a URL as follows:
http://folding.extremeoverclocking.com/files/teamid/YYYYMMDDHH0000.gz
So, for example:
http://folding.extremeoverclocking.com/files/11314/20050622120000.gz
If you want the file of all teams, it is in the "summary" directory (instead of a teamid) but it follows the same time/date format.
Enjoy... |
|
| Team 0 Users...06.21.05, 2:08pm CST |
I just did a preliminary load of about 2,000 top users from team 0 (zero). The _ (underscore) user I changed the name so it was a little more visible. This should fill in a lot of the gaps for individual user rankings.
I still have some issues to work out, so until then I'm only loading team 0 data manually, it will probably be tomorrow before it is all done automatically. |
|
| User Summary Page Slightly Changed06.21.05, 11:14am CST |
Someone asked if it was possible to see the date when a person joined. The best I can do is show the date when I first started to record their data. The earliest date that will show up for a person is May 1st, 2004. I had some dates for people going back to 2003, but I was unsure of the accuracy so I just mass-updated them to all start the same date (which is also the earliest date I have to team/user data). I still have the old data, so I can re-update users once I do some further investigation into these older dates if they seem accurate & reasonable. People that have joined since then will show their respective date. Here's an example below:
Also, you can browse through all the old news posts via the page navigation links on the right.
More features on the way... |
|
Pages (14): « First ... « 4 5 6 7 [8] 9 10 11 12 » ... Last » |
|
