Quote:
Originally posted by Dyyryath
[prokaryote: A 'mega-stat' system will require additional columns (or tables). Before we try to determine what those are and how they should be organized, we'll need to determine how the mega stat will be generated.
I've seen a variety of ways to do this. My favorite is probably using a 'CPU power' style unit for each project. For example, we choose a processor to use as our baseline metric (let's say a P3-1000) and determine how it performs in each project. For Folding@Home, we may determine that the P3-1000 does 'z' points per hour. This becomes the official 'standard' for the Folding@Home megastat. To determine a user's megastat for that project, you simply divide their total points by 'z' to acquire the number of CPU hours they've contributed. The scores for each project they participate in can simply be added together for a total megastat.
This is not the most accurate way to measure contributions, especially on projects like Distributed Folding where the amount of work generated by a processor in a set amount of time can change (i.e. protein changeovers). It is also less than ideal due to the differences in certain processor architectures and how they perform with different projects. However, it is extremely easy for end users to grasp, and I think that is important.
In fact, that's probably my one 'guiding' principle with stats. If you make something that is ultra cool, but difficult to use or understand, you've gone too far. I've seen some unbelievably thorough stats systems over the last couple of years that the creators were justifiably proud of...except that they were a mess to actually use. You don't want to make something that a user can't figure out pretty quickly. Having to spend time 'learning' to use a complicated stats site means that it's either too complex or has a poor user interface.
Now, that *doesn't* mean that I'm married to the idea of using a 'CPU power' style metric. It's just my first stab at the idea. If we can come up with a more elegant solution that can still be easily explained to end users, I'd love to hear it...
Okay, we're both in agreement that a system based upon CPU power is chok full of issues.