Return on Use Less Precise Wiggle

Case number:845813-996208
Topic:Game: Tools
Opened by:bertro
Opened on:Wednesday, October 30, 2013 - 20:51
Last modified:Wednesday, November 27, 2013 - 17:10

Hi Foldit dev,

Notwithstanding this feedback:

I would like to present some of the reasons to keep the "Less Precise Wiggle" (LPW for short):

A) Normal wiggle is too slow for casual players or players with less powerful computers who cannot play for any extended time. This limits these players to hand-folding and short recipes or a relatively few cycle of longer ones. With LPW, it is possible for them to enjoy foldit.

B) Some types of recipe do really well with LPW mainly because they are types that win points by trying MANY possible solutions (I think of GAB and Quake, for example). With normal wiggle, these recipes cannot achieve many cycles to try out different configurations but with LPW they can and make progress possible.

I know I use LPW in some occasion because it is more suited even if, I think, I'm a power user and can let a recipe run for long period of time.

Thanks for taking the time to consider.

(Wed, 10/30/2013 - 20:51  |  11 comments)

Susume's picture
User offline. Last seen 24 min 3 sec ago. Offline
Joined: 10/02/2011

I agree that regular wiggle is too slow for running scripts. Where I now run a script that rebuilds the entire protein one area at a time, I would only be able to rebuild perhaps a third of the protein. Where I now have time for 1000 sets of rubber bands in GAB, I would maybe have time for 300. The more precise wiggle may yield a few more points when you run it once, but by making scripts do only a third of the work in the available time, it will end up leaving lots of points on the table because of possible changes that never get tried.

spmm's picture
User offline. Last seen 2 weeks 3 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

Agree with both of the above posts and mention that imo precise wiggle can lock in a bad pose early, regardless of scripts, even if just wiggling, it can be difficult to recover from that bad position.

Also if you need to reset structure once you start using precise wiggle you may not be able to do a small 'wiggle sidechains' to reset the high score with the structure.

Joined: 09/21/2011
Groups: Void Crushers

One more reason:
When using local wiggle on a open cutpoint the precize one sucks. The not precize one really helps to remove bad backbones.

bertro's picture
User offline. Last seen 21 weeks 6 days ago. Offline
Joined: 05/02/2011
Groups: Beta Folders

Also, I have noticed that regular wiggle will sometimes get stuck for a VERY long time on iteration #2 (maybe a bug but probably only very long because of minute changes in score keeping it from quitting iteration 2). It is most noticeable when I leave the computer for a time and come back and the recipe is still at the same place it was before, wiggling at it.2. Hitting space bar does always let it continue, so not frozen.

Joined: 04/19/2009

As time passes, I find that I am only using PW for refinement puzzles, or perhaps late game, for three reasons:

First, my one year old top of the line Mac desktop cranks up its fans when using PW (or mutating, and the combo of the two is painfully loud). If you don't have the latest and greatest PC, then you can't reasonably use PW.

Second, I would rather watch paint dry than use many scripts with PW... I did a test a while back with a client running a script in PW and another client running the same script with the same start in less precise. The PW client got through maybe half a dozen iterations, while the less precise client did over 2 dozen iterations (and scored much better!). So when I do a refinement puzzle in PW, I know that it will take the entire time the puzzle is live to go through one path with scripts, there will be no time to explore other avenues.

Which brings me to the third reason - especially in a denovo puzzle, I want the protein as loose as possible to have the opportunity for an unusual rebuild to work. I also want to be able to have the luxury of backtracking to an earlier save, and try something different with it. This is only possible for me with less precise.

In the denovo just expired, 803, I used less precise wiggle throughout the entire solution. My solution ranked solo #1. If precise wiggle was so much better (as we have been led to believe) then can a dev explain this result?

frood66's picture
User offline. Last seen 3 hours 31 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch

I have to agree with all that auntdeen says - then add the fact that I run on a 3 yr old macbook.

This only exacerbates auntdeens observations and reasons.

For those players with less than 'top of the range' kit, foldit is becoming less accessible. These days I fall back on my group to play out any solution that I produce. I fear it is only a matter of time b4 other players find themselves in the same situation - effectively blocked from mid and end game play. The long term effect is, I feel sure, obvious.

Joined: 09/24/2012
Groups: Go Science

After reading all these comments, I'm convince LPW should stay for long time (I even use old computers precisely to explore new puzzles).

I then would like to bump on the former proposal to add a LUA command for less precise wiggle on/off:

It would be very usefull in some recipes to begin with LPW (exploring new venues) and end with normal wiggle (from the best avenue).

bertro's picture
User offline. Last seen 21 weeks 6 days ago. Offline
Joined: 05/02/2011
Groups: Beta Folders


Joined: 11/15/2007
Groups: None

Precise wiggle is better at getting to the absolute top of the nearest peak in the score function. But until the end of a search, the extra effort can actually be detrimental as it makes it easier to get stuck at local maxima in the score function. In our automated Rosetta protocols, we find it best to use less precise minimization (wiggle) until near the end of a search trajectory. I'd expect that optimal strategies in FoldIt would similarly start wtih less extensive wiggle searches at the beginning and make them more intensive towards the end.

bertro's picture
User offline. Last seen 21 weeks 6 days ago. Offline
Joined: 05/02/2011
Groups: Beta Folders

Thanks David,

One more reason to keep less precise wiggle in the client AND add the LUA functions.

Joined: 09/24/2012
Groups: Go Science

See test here:
Almost 20% time saved with less precise wiggle on a very big protein:


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