Here is something you might want to try.
Take your best Low Ci score in auto.
Place zero length bands.
Put into Low Wiggle Power.
Set to full ci.
Set to Auto wiggle power.
Set to High/Med wiggle power.
Repeat whole process from enabling bands in Low wiggle power until no gains are found.
Now enable these bands in High/Med for short iters and disable for wiggling out until you are tired of playing around.
Not sure how it affects the end game but it certainly gains at least an extra ten points or a lot more, before rebuilding in High/Med.
FOLD.IT puzzle results vary widely because of the random factor. I'd like to explore that idea by solving puzzle 1627 as a group with the following strategy:
1. Agree on a list of recipes that we will use. I propose these but am open to any suggestions:
LOW, ci=.10, ZLB secondary structures, wiggle shake, wiggle shake, wiggle shake. Remove bands.
2. DRemix on AUTO, ci=1.0.
3. Idealize on AUTO.
4. Cut and WIggle on AUTO.
5. COMP on HIGH.
6. BANDFUZE on HIGH.
7. endgame recipes????
Feel free to modify or propose your own. We can decide in a day or two on one set of these. For the purpose of this experiment, I don't think its critical that we get it just right at first.
2. Then, we all begin running the first recipe on as many clients, as many times as we can. After a period (say one day, i.e.) submit your best result and take the best of the team's result and run the next recipe. Repeat until we have gone through the entire set of recipes. Of course, you can revisit any team result.
3. IMPORTANT: When you're saving to the team use the script name that was used in the name. For example, "DRemix"
That's it. Any suggestions? Changes? Comments? Shall we try it?