Scripting in NewChapter

Started by AsDawnBreaks

AsDawnBreaks Lv 1

I've decided to just create a feedback about this, as NewChapter is going live in 4 days, and these are major issues.

See my comments here: http://fold.it/portal/node/996724

What would be the ETA to add either the matrix caching or wiggle score gain? This is critical to new scripts.

Also, what will be done about existing scripts? They're virtually useless unless wiggle is re-done as by-gain. There are currently 850 public scripts, and I'm sure many group shared ones, and it would be very difficult to convert them all (and useless until we get a matrix/wiggle till point gain). Will we have to do this?

These are pretty important topics, as scripts are now heavily used, and leaving them in a bugged/broken state would either mean that efficiency would go down or everyone reverts to hand folding-reducing the ability of players either way.

Until a wiggle fix, current scripts could do something like this this (using semi-pseudocode):

score = getscore()
wiggleRun = True
wiggleIter = 2
Quicksave(3)

while wiggleRun = true
wiggle(wiggleIterations)
newscore = getscore()

if (newscore > score + [Some number])
then
wiggleRun = True
wiggleIter = wiggleIter + [Some other number]
Quickload(3)
end
end

This would work, however resuming back to the last state would mean the possibility, and high probability of wiggle taking many times as long as it would otherwise (depending on your settings even 50x+). Obviously it could be changed by changing [Some Number],[Some other number], and the default for wiggleIter, but who knows what the best results would be for that…. That may even change depending on puzzle size, aas, conditions/filters, current state, etc.

Thanks to the devs for your time. :)

AsDawnBreaks Lv 1

It should probably be announced that in order for scripts not to break (unless wiggle is changed and the script writer takes advantage of that) they will have to be on low.

Seth Cooper Staff Lv 1

We'll be looking in to what should be done with the wiggle scripting API next, hopefully having something next week.

Susume Lv 1

Tested: How many wiggle iterations does it take newchapter to reach a stable score?

Started from a stable score in both TR663 and Staph Aureus ED puzzles (each about 150 segs). Generated ~300 different rebuilds on each puzzle as test poses (610 poses in all), shook, then wiggled for various numbers of iterations on each of the poses. Results attached.

Conclusions:
If a script is to wiggle all the way to a stable score, on medium, it should do more than 20 iterations.
If a script is to wiggle all the way to a stable score, on low, it should do 10 or more iterations.

Open question: Is it really necessary to wiggle all the way to a stable score in order to pick the best rebuild poses? I hope to have some data on that soon.

Vredeman Lv 1

Here is an update on wiggle iterations.
In a test aimed at finding an equivalent of the old 2 wiggle iters, and with repeated results on puzzle 847, I have found that, in a simple script, using wiggle, mutate and wiggle again I have found the following.
Starting score 10040.280
Low Wiggle Power, 25 iters 11288,457; 14 iters 11297.892 once and 11288,457 repeatedly; 12 iters 11288,457; 11 iters 11288,457; 10 iters 11422,342 11297.892 11377,609
Possible conclusion (10 iters always wins so I put my money there)
High Wiggle power, too much detail but the results show that you can put your money on 14 iters.
To top it all…a 10 iter wiggle on low, mutate and a 14 iter wiggle on high is magical.
These tests wer done using only global GUI operations.
Good luck scripters.
I recommend scripts have a dialogue box asking if the players are using high or low wiggle power and perhaps an extra option for a slider.
I trust this info is useful and makes for happier folding for all.
porky