12 replies [Last post]
Joined: 09/24/2012
Groups: Go Science

Would it be possible to have more time for design puzzles? Say 10 days instead of 7 for puzzles > 55 residues?

Is it my technique? but personally, I've no time to "finish" my refining and get the higher possible score from my designs. If I start too early with disabling slow filters, I feel I loose good "science" opportunities. And with the new ABEGO filter, it's still slower. And I've no time to do some more remix.

spvincent's picture
User offline. Last seen 3 days 2 hours ago. Offline
Joined: 12/07/2007
Groups: Contenders
I imagine everyone has their

I imagine everyone has their own preference here based on their playing style and the amount of computational resources they have available. For me the time allocated is both too long and too short: too long in the sense that a first design typically grinds to a halt with a few days left: too short in that there is then not enough time to really work on a second design.

gitwut's picture
User offline. Last seen 20 weeks 1 day ago. Offline
Joined: 05/18/2012
Groups: Contenders
I agree

I agree, Bruno, but I seriously doubt it has anything to do with your technique. I feel like the last two puzzles we have been given were poorly considered.

The ED puzzle (1249) has a very poor density map and that leaves us trying to work a somewhat more difficult (lengthy/complex) de-novo than most folks are used to working with. On top of that, working it as a de-novo means you also have to contend with the gravity/tug effect that is counter-productive.

Puzzle 1248 is a PITA because of the "ideal loops" filter. Two days left and still a -200 point penalty. I've wasted far too much time trying to fix something that should be handled by the engine, and not by a filter. I understand that new things must be tried, but the filter needs more time in devprev to work out the kinks and allow scripts to catch up.

The result is that players (particularly newer ones) can become frustrated, particularly with the unstable client issues on top of everything else.

Joined: 06/06/2013
Groups: Gargleblasters
re:Gitwut comments & encouraging participation

Like Bruno, I really don't have the time & capacity to try multiple designs. So I never feel like I've given the design puzzles my best effort.
Gitwut's comment mirrors a concern I share. It is hard to see how we encourage new players to play when even experienced players are having difficulty. I'd love to see Foldit figure out how to get more new players to play. The number playing each puzzle seems to be dwindling, and I think more players might well benefit the science and game of Foldit. I think this is more than a summer vacation lull.
How do we find balance?

spvincent's picture
User offline. Last seen 3 days 2 hours ago. Offline
Joined: 12/07/2007
Groups: Contenders
I agree with Skippy: the

I agree with Skippy: the learning curve for new players has become ever steeper and must be a real deterrent to playing. I think what's needed, among other things, is a tutorial that will bridge the gap between the existing tutorials and getting a reasonable solution on the science puzzles.

Joined: 04/24/2014
Groups: None
Tutorials

It's an interesting observation you've made here. I've boosted this thread to the team in hopes they'll have some time to share some thoughts (if not directly in-thread, than via email which I can sum).

jeff101's picture
User offline. Last seen 1 hour 44 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Encouraging participation:

I think rmsd dot plots as discussed in:
https://fold.it/portal/node/988112
https://fold.it/portal/node/2000880 and
https://fold.it/portal/node/1998460
would help players see better how they are doing
and show them the range of solutions they explored
on a particular puzzle. This might encourage more
participation, because it would show players that
their contribution to Foldit isn't just their
top-scoring solution on a particular puzzle.

jeff101's picture
User offline. Last seen 1 hour 44 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Calculatus Eliminatus

Part of our job as Foldit players is to find many undesirable solutions and show
that their Foldit scores are lower than the most desirable solutions' scores.
If this happens, then Foldit must be scoring things correctly.
If it doesn't happen, the Foldit Team needs to adjust Foldit's scoring
appropriately, with new filters, changes to Rosetta scoring, etc.

This reminds me of a scene from "The Cat in the Hat"
from 8:41-10:12 in https://www.youtube.com/watch?v=4jK6l1WJKUU
"The way to find a missing something is to find out where it's not".

It also reminds me of https://fold.it/portal/node/2001059

jeff101's picture
User offline. Last seen 1 hour 44 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Is the number of Foldit players dwindling?

I've been taking Foldit pretty seriously the past 2 months,
yet my solos have ranked from 4th-58th place during that time.
This means that on various puzzles, 3-57 players have found
better solutions than me. Also, my global soloist rank right
now is 118, meaning that lately 117 players have been finding
more or better solos than me. If my solutions are any good,
it seems like Foldit must be getting plenty of good solutions
on each puzzle. If the number of Foldit players was really
dwindling, I would expect to have better rankings than I do.

Joined: 06/06/2013
Groups: Gargleblasters
Jeff re ranking

Jeff -- my individual puzzle scores also vary greatly. However, I think the rankings are not a good measure of either your abilities or contributions, and take them with a good bit of caution. Due to time constrains and machine size, I can't play every puzzle (caretaking responsibilities that are unpredictable may mean I have to abandon). Those able to play every puzzle get more points and are ranked higher. Also, those with powerful machines can use scripts to fine tune a fold. You or I or many others might have a better concept, but haven't tweaked our solutions.
So please know that I for one do not rank your abilities based on your posted rank. And thanks for the Rama map links
:)
Skippy

jeff101's picture
User offline. Last seen 1 hour 44 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Rankings

I've noticed that some folks treat the de novo puzzles as design puzzles,
changing the secondary structures from what was given, and using things
that may score better but are not so close to what is already known about
the structure. I'm hoping that my solutions are just below this layer of
solutions and are among the highest-scoring ones that make sense with the
known constraints.

jeff101's picture
User offline. Last seen 1 hour 44 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Library of Contact Maps:

http://foldit.wikia.com/wiki/Distance_Maps and http://memorize.com/distance-and-contact-maps show Distance and Contact Maps along-side their respective 3D images for many different proteins. Images are classified as in the SCOPe database (http://scop.berkeley.edu), and http://memorize.com/distance-and-contact-maps can shuffle the images in multiple-choice and matching modes to help you learn the patterns in them.

Skippysk8s - Are these the "Rama Map" links you mentioned above?

Joined: 09/24/2012
Groups: Go Science
Player's Hand fold time / puzzle solution

Just curious how much total "hand fold" time the other players need for several types of puzzles (the total time you need in front of your computer for pure hand interventions and checking of short lasting tools and recipes - not included the time to check and change long lasting recipes).

I would say for myself that I basically "hand fold" only the first hours within the 3 first days, the 4 latest days being mostly covered by long lasting recipes.

Here is my estimated mean HF time per puzzle solution: (I may try several solutions):

1) Revisiting puzzles without bridge: 10 min
2) Revisiting puzzles with expected bridges: 20 min
3) Refining, second round etc: 30 min
4) Peptides, hot spots etc (small designs): 45 min
5) Free design: 1h
6) Symmetric: 1h30
7) Design, ideal loops ("no rebuild"): 8h
8) De Novo free style: 2h
9) Contact: 8h
10) Electron Density refining, already in the cloud: 1h
11) Electron Density refining, not yet in the cloud: 5h
12) Electron Density, short and easy, de novo: 10h
13) Electron Density, long and complex, de novo: 20h

Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons