That is actually really cool. Good noticing.
Hmm - wondering why bots treat very similar puzzles differently
are very similar and 2 out of 3 bots failed for both
but a different bot succeeded in each case - why ?
is it just random ?
if it’s random and we want to test this more accurately is it possible to ‘fix the random seed’ that the bots start with - so there is the option when testing similar shapes to have them all start on the same series of random bases/changes.
if it’s not random - what is the underlying difference between these two puzzles ?
Though the two designs are seemingly similar, there is difference on the two puzzles. Tilted-baby has adjacent strings, angry puppy has three nucleotide between each arms. Also there is difference in number of nucleotides.
Fair point Eli, but having seen that I made the same change that the labs are undergoing when compared to the player projects
if you take an old ‘lab’ puzzle and then remove the hook by submitting the same shape as a ‘not lab’ puzzle
this one with the hook
this one without the hook
infoRNA bot failed the first but not second one
As long as the hook is there, it ads complexity to the to the puzzle, for the algorithm having to solve it.
But this comparison between old lab and new lab, I’m really looking forward to. How similar does the results for those twosame design look in regular lab and player projects? Great idea making this experiment.
I’m not sure if people know, but you can use INFO-RNA directly at:
Easy to paste in the secondary structure like ((…(…).)) and let it run.
It has several options for its optimization, which make a difference in its “solution”.
Which options does eteRNA use? I don’t know.
I tried it on Buard’s “An actual easy puzzle” a bunch of times, and it got it right about half the time. It almost always got the top part of the puzzle correct, but often butchered the bottom part.
You might enjoy tinkering with your favorite puzzles.
I just tried this on Edward Lane’s Plimsole line.
I copied it’s dot bracket structure (secondary structure) into the puzzle maker, threw the same into the link to Info-Rna and submitted it. Then I copied out the “designed sequence” and posted the letter sequence into the puzzle in the puzzle maker. And InfoRNA stabilized the puzzle in one hit. One would still have to fill out the requirements, as Starry rightfully pointed out.
I guess this loophole, you Ruch found, is exactly why, we are not allowed dumping letter sequences directly into a puzzle.
I think another thing worth exploring is that the bots seem to do worse on puzzles with many unbonded nucleotides (the exception usually being InfoBot). It may be the case that they have no algorithm for handling those separately, which would offer a partial explanation as to why human players tend to do better.
I’m still trying to precisely define the parameters that cause infobot to fail so that we may then ask why. Once I’m done with that, though, I think I’m going to try and more accurately define the number of bonded to unbonded nucleotides which cause failure in Vienna and SSDbot.