About the Generation listed in Stats

[rsbriggs]

Howard,

Don't know if this was "as designed" or not, but the generation shown in the Beta stats is a little misleading, if you are thinking about using it to "tweak" the folding algorithm.

I, at least, thought that the generation was the generation in which the best RMS was generated. But what it really shows is the last generation that the given Username on the list has uploaded.

In the first iteration, my best was something like 5.52 A, which was reached in about generation 70 (of 250). In the next iteration I seem to have reached my minimum of 5.19 in generation 44. From my limited experience with the Beta so far, given my current laxness levels (75/84/40), and what happened in the last iteration, this number will hold for the rest of the iteration - I'm not likely to improve this number any. So, in theory, the other 210 generations in this iteration (and the entire rest of the weeks worth of crunching) aren't going to improve that 5.19 number, and are probably just being wasted.

I'm thinking that it might be instructive to see what generation the minimum was actually reached in.... Just my $.02

===bob briggs

[tpdooley]

If you read through all the beta threads, it's been mentioned that the stats were listing the last uploaded generation - not the minimum. (having both the best generation, and the last generation uploaded might be nice to see) If you want to see where the minimum was - you can take a look at the "movie" for each of the top 10 folks, or download the program to create the movie if you don't have a top 10 score. (links and directions on how to use are in one of the Beta threads).
Some of the mimimum scores were hit in the 40-70 range and never got lower; some hit it in the 100-150 range; and some found a lower minimum out at 200-250. It's random. But it would be nice (even though probably impossible to identify when you got there until gen 250) to stop at the minimum for the 250 generations. ;)
And by the way.. you slipped from 5.19 down to 4.84 (sometime before gen 58) and moved me into 2nd place. Congratulations..

[rsbriggs]

Yay! Looks like I've joined the sub-5 club ! :|party|: :elephant:

Guess I missed the earlier caveat about the stats showing the highest generation uploaded - I didn't join the beta until late in Beta-7.

IMHO, it wouldn't be a bad idea if it showed the total number of generations uploaded, too, since someone looking at the numbers would think that I'm only on generation 59, rather than having uploaded closer to 600 generations at this point. :idea:

[Mikus]

Originally posted by tpdooley
Some of the mimimum scores were hit in the 40-70 range and never got lower; some hit it in the 100-150 range; and some found a lower minimum out at 200-250. It's random.
It's well and good to say that the results are random. I suggest the 10-best SHOULD show the generation in which the value shown was calculated. Then everyone could SEE immediately where in the 250-generation cycle minimum values were actually being found.

mikus

[Brian the Fist]

Originally posted by Mikus
It's well and good to say that the results are random. I suggest the 10-best SHOULD show the generation in which the value shown was calculated. Then everyone could SEE immediately where in the 250-generation cycle minimum values were actually being found.

mikus

If you click on View Details, one of the many graphs is RMSD vs. time. From this you can glean what you want to know and much more. Perhaps it is just a matter of semantics - I should change the heading on the stats page from 'generation' to 'movie size' or 'cumulative elapsed time' or something like that

[Mikus]

Originally posted by Brian the Fist
If you click on View Details, one of the many graphs is RMSD vs. time. From this you can glean what you want to know and much more. Perhaps it is just a matter of semantics - I should change the heading on the stats page from 'generation' to 'movie size' or 'cumulative elapsed time' or something like that
I guess my point is that if the lowest calculated RMSD is an achievement significant enough to be displayed in the 10-best table, then "how much more work was done in that 250-generation cycle AFTER the best RMSD was found" would be of interest to DF participants.

The number of generations does not really say much -- if you wantd to convey a sense of the effort put in by that participant, you might show the number of work units contributed.

Yes, looking at the "RMSD to native over time" graph (in Details) is quite interesting. But it has to be done for EACH participant listed in the top-10.

My suggestion is to have the top-10 table list the generation corresponding to the lowest RMSD, rather than the number of generations reported.

mikus

[Brian the Fist]

Thats why there's also a score for 'work done'