Foldit Wiki
Register
Advertisement
Local Wiggle Strategy

Local Wiggle Strategy (LWS) is an endgame method used to fix the pose and get some points. It's also useful on start game just after a hand Fold session, for example on a de novo or a design puzzle. It's advantage is that it doesn't break the protein apart, keeping the beautifull 3D design you just managed. While using the graph to watch the progress, it's possible to make many small gains by applying local wiggles to small sections (2 to 5 pieces at a time, mostly) of an "all-loop" puzzle. After that, just keep any gain, and overwrite any losses. Using "control-B" to capture a best score can often find momentary jumps in score that would otherwise be lost.

Use locks to "lock off" sections that I use "local wiggle" on, watch the graph to see if it worked, then move the locks a little at a time, repeatedly applying local wiggles to "squeeze/massage" all the points out of different combinations of pieces.

Procedure[]

Switch the structure all loops/tubes[]

Do this by pressing 2 to go into structure mode. Look for both ends of the structure: one will probably already be a loop. Drag from that end to the other, and all the pieces in between will become loops.

This is done because local-wiggle can now be applied to any subsection we choose. The artificial barriers put up by the transition between helix/sheet/loop no longer apply. I find the loops less cluttering on the screen, but you may find sections that "collapse" when local-wiggled. They will twist and turn while your score plummets to zero. This does not happen if the sections are turned into helix or sheet structures, but you can also move the locks slightly and the collapse will be avoided.

Clear the Graph[]

Press the "Clear" button on the graph so it will be sensitive to any move. Remember that the first move will not show any movement at all as the graph contains only two values.

Pick a section to start from[]

I'll often start from one end of the structure. Count 4 or 5 sections in from the end and lock it. Local-wiggle this small portion of the structure and watch the graph. If it jumps up, wait a couple of seconds to give it time to reach its maximum, then stop and do it again so the graph will now have the opportunity to show some movement. If the score dropped, however, you should undo and overwrite it with another move instead. Try moving the lock in or out by one spot, or putting another lock at the very end of the structure and wiggling just the intervening 3 pieces.

Move the Locks[]

Now that the graph is telling you what is happening, start adding more locks to the structure, or moving the ones you have already put down. You will get results with as little as two free pieces between locks and, of course, there's no upper limit: you can "local-wiggle" the whole structure with no locks if you want. If you're getting good points in an area, move just one end by one space, either in or out. Continue moving the "window" that you're working on: make it larger, smaller, cut it in half and hit each half etc. There are no rules about what to do, only that one or two attempts will probably not exhaust the area: you can repeatedly hit the same overlapping sections in different combinations and still get more points.

More tips[]

Where to look for[]

Most structures are green with some brown/red areas: these are the areas which are under higher tension/stress and can be relaxed. I will usually lock off a section that is mostly green with one or two brown pieces. I usually start with the brown at the center of the section, but it sometimes works better with the brown near one end. But I have also had success in working on areas that were already completely green.

Rinse and Repeat[]

This method was described as "tedious but effective" and it is: you can move the locks to any combination of overlapping areas, one after the other, finding that some combinations will raise your score and other will drop it. You need to try lots of different combinations in each area of the puzzle to extract all the points that are there.

Occasional Stops[]

Every now and then you can remove all the locks and apply a global wiggle and/or a nudge/wiggle and/or a shake. The repeated local wiggles seem to build up the potential for these large moves to unlock a bunch more points, though this diminishes as you reach the end of the puzzle. If you don't do this, the local wiggles are less effective, but after it's done, they seem to be MORE effective. It's as if you have reset something in the structure and you can apply local-wiggle to areas you though were tapped out and get more points again.

Automation[]

As you can see, this method requires little human intuition. Fortunately, with the development of recipes and Lua scripts, the task of local wiggling can now be entirely automated. Many recipes/scripts for local wiggling are publically available and can be obtained via the "recipes" section of the Foldit website; when searching for them, possible keywords to use include "walk," "local," and "wiggle."

Notes[]

Local Wiggle Strategy is a cousin to Global Lock/Wiggle Strategy and both can be used for similar purposes. You may also want to read Brian's notes on overall strategy to see where this fits in.

It has recently been confirmed that Local Wiggle Strategy is apparently not an artifact of the scoring algorithm or of some bug in the game itself:

There is some question whether the score increases from slicing and dicing are real or are an artifact of the method of rendering used by Foldit. My distillation of Michael's explanation - local wiggle might be allowing segments to assume positions that are energetically more favorable but are physically "illegal" (the amino acid bonds don't line up correctly). If this were true, you would expect the score to decrease again when you global wiggle after lws, as the problem corrects itself. In my experience, the score almost always goes up after full lws/global wiggle, and this argues for the score to be real. They're going to do some testing on this soon; Seth told me to keep slicing until we hear otherwise.
Advertisement