[FaD] Are points such a bad thing?

[RustyGT]

They seem to be working for me atm.

Aussie FaD Stats

Top of the list!

Not sure who is most pleased.
Me, who built this mess.

Or the Adminatrix [Donna] who keeps it running while I'm away!

cheers Rusty.

Edit: We are both pleased!

Hey, I make #7 in that list!!

Pity I ain't quite so high when you do it by mollies.

Tell Donna she can always PM/email me if something breaks and she can't get ya.

Cheers
Mick

[RustyGT]

Mick, I was always a stickler for jobs/molecules in the stats.
[and probably still am]

But you need to take the kudos when it's offered.
The points system is a metric of the project.

Not every day you see No.1 beside your name.

Gave me a thrill!

Also noted, I got past '3 years' processor time recently.

R.

[PresterJohn]

i know when the points system came out, there was an outcry among many of the long-time FAD crunchers who wanted to stay with the old 'molecules completed' standings. i really didn't care either way because the primary goal of FAD is to help find good drug candidates.

the real measure of 'meaningful' contribution to the project remains, imo, hit counts.

this is especially true in light of the fact that many people now have knowingly stopped crunching the 'tougher' molecules in favor of the Methodology work units. so as a result, their 'mollies completed' total shoots up and (even their points go up though, not at such an exponential rate). more and more now, i am starting to discount 'molecules completed' as a 'proper' yardstick.

i like to look at a users hits count/moleclues completed ratio. to me, that is a more indicative figure of whether they contribute to FAD.

user1 100 jobs completed-180 hits ==ratio of 1.8
user2 4 jobs completed-210 hits ==ratio of 52.5

clearly, user1 is running a lot of methodology WU's. user2 is not.

[tony_j15]

IMO, they should have never started counting by points, as you can cheat on it. I prefer mollies and hit count. BTW, I HATE methedology WU. They are extremely slow, get no hits(or very few). Last one took 4 days to get done. Usually when they comne across I delete them, but last time it seemed that was the only thing I could d/l.

[PresterJohn]

Originally posted by tony_j15
IMO, they should have never started counting by points, as you can cheat on it. I prefer mollies and hit count. BTW, I HATE methedology WU. They are extremely slow, get no hits(or very few). Last one took 4 days to get done. Usually when they comne across I delete them, but last time it seemed that was the only thing I could d/l.

the current crop of methodology WUs finish very quickly.

they typically take 10-20% less time to complete than the other WUs. there isn't a way to deselect them and i will admit that after crunching thru a slew of 1p9t's that averaged 4-6 days to finish on p3-1ghz machines, i took a little break and downloaded some methodology units since a number of people on our team and others did so also.

i was primarily a 'Cancer only' cruncher here on FAD but i've switched to 'respiratory diseases' as my primarily project what with the recent SARS emergency. also, my feeling is that the cancer mollies are the more popular units which get taken on by the 'converted' UD crunchers and that respiratory and MS and others get a little less support.

i have a quad pentium 200mhz server crunching 3 methodology wu's (along w/1 UD process) just because a 200mhz cpu seems ideally suited for those WU's while i can use my higher end systems to crunch the tough WU's.

as for cheating on points, i didn't know how or why people would be so desperate to climb the rankings that way. AFAIK, the cheat is only on win9x systems (none of which i run anyway)...

[sharder8]

Originally posted by PresterJohn
i like to look at a users hits count/moleclues completed ratio. to me, that is a more indicative figure of whether they contribute to FAD.

user1 100 jobs completed-180 hits ==ratio of 1.8
user2 4 jobs completed-210 hits ==ratio of 52.5

clearly, user1 is running a lot of methodology WU's. user2 is not.

Not true!!

I'm not, and have not been running methodology jobs, but rather HIV, MS, and SARS, and all 3 seem to be taking excessively long times. The 1f45-q2's (MS) seem to be some of the slowest and many are coming up with 0 - 10 hits after 8-14 days.

I like Points, Mollies, Time, and Hits, but I'm in it for the cures!

Harder

[PresterJohn]

>but rather HIV, MS, and SARS, and all 3 seem to be taking excessively long times. The 1f45-q2's (MS) seem to be some of the slowest and many are coming up with 0 - 10 hits after 8-14 days

>user1 100 jobs completed-180 hits ==ratio of 1.8
>clearly, user1 is running a lot of methodology WU's. user2 is

sharder8,

the keyword in your post is 'long time' in relation to the above ratio. let's use some numbers to better illustrate...

user1 100 jobs completed-180 hits ==ratio of 1.8
user3 3 jobs-8 hits ==ratio of 2.66

if the daily stat page consistently show a pattern for user1 (over the period of 3-4 several days) which match the above, ie. high # of jobs completed w/very few hits...odds are very likely that they are running large counts of methodlogy WU's. this is especially true if this is a user who before averaged say '450K' of completed mollies and who is now consistently putting out over 2mil per week.

>The 1f45-q2's (MS) seem to be some of the slowest and many are coming up with 0 - 10 hits after 8-14 days

the example you quote is not a good example to use to argue your case because if it takes 8-14 days for each WU, then that person's stats are going to look nothing like the examples i am referring to.

my point to this whole thread is that while climbing the ranks is fun and evokes some good spirited competition (whether it's between fellow teammates or between competing teams)...it can quickly degenerate into a contest of who finishes the most wu's and *this* can have a negative impact on the FAD project for 2 reasons: (1) participants will single out only methodology WU's to run, and (2) long-time FAD crunchers will leave the project because of this abuse.

it is for these reasons why the points system has some definite 'pluses' over the standard "WU's completed" stat of which, many, seem too fond of and which is sometimes, over-emphasized (imo).

[edit]
another case in point...until about 2-3 weeks ago, the "WU's completed" stats between ThinkTank and TechIMO were roughly the same. then there was a period of several weeks where ThinkTank's output outmatched us by a factor of 3, yet their weekly *point* count was still less than ours even though they were finishing 3 times MORE wu's than we were. the only way that could be is if some of their members were concentrating on metho WU's and that was indeed the case.

[DPA]

i think that points should be distributed soley on the # of hits [per saw "bonus" points] and the true CPU time.

perhaps a CPU usage logging system needs to be integrated into the program so that there is no speculation of true "work" between 9x, NT, 2K, XP systems.

but this is more than likely too much to ask as this program is fronted towards the general idea of getting as much work done for as little cost as possible.

ive done many jobs [so to speak] but yet my points remain very low, people with less completed jobs and less hits than me will have more points than me, this is what i do not understand.

perhaps the ranking system needs to be a combonation of molecules+hits/cpu time=points? i don't know, im sure others have thought of this, and already have the best median point system set up.

regardless, im doing this because it is fun and competitive, as well as helpful to mankind in general. i suppose the ranking system is a good way to keep me interested.

[Idigguns]

Yeah, I always look at the hits/jobs ratio. It tells you a lot! I'm proud to say, mine is one of the highest.... 51.50! I guess that's because I only do Cancer. I'm thinking about trying something else though. I have a Celery 666 that takes FOREVER to do these 1e7u's. Any suggestions for something that will move a bit faster but still contribute to finding cures?