Quest to Native

[zoran]

We have recently noticed that for some puzzles, althought the scores are significantly improved, the native is in a slightly different place then where the game community has been exploring. Currently we would like to find out of the reason why the native protein structure remains elusive in those puzzles. There are two main possibilities:

• the native structure is not reachable with current set of game tools
• we did not provide sufficient intuitiion behind the game mechanic to direct the game exploration in the direction where the native structure is hiding.

To find this out we will be posting a few Quest to Native puzzles, where the native structure will be shown as a ghost.  The challenge for the community is to see if you can find the significantly better score that is around the ghost.  In the process we would also be interested in whether this process gives more intuition about how native structures can pull a much higher score.

Stay tuned,

Zoran

[Nicky666]

question though, about this part:
"In the process we would also be interested in whether this process gives more intuition about how native structures can pull a much higher score."

I'm assuming with "this process" you mean the "ghost puzzles"/Quest to Native puzzles.
But I'm curious about the part I quoted…what do you mean with intuition (the folders?), and how are you going to test that?

Anyway, I'm curious how these puzzles will work out, as my experience thus far is that a ghost of a better scoring puzzle doesn't help too much…..ooooh, which brings up another question: how will you be able to decide if the current set of rules is the limiting factor for reaching the best solution?

so many questions…lol
Thanks for the heads up though.

[zoran]

we're curious to see if seeing a few native guides, the players might develop additional intuition about the best structures. if so, we may be able to see it in the subsequent puzzles (which would not have guides). For example, we're hoping that seeing multiple quite different protein configurations that have equally good scores, might motivate greater diversification of the structures. Currently, once a relatively good structure is found, we're seeing almost all game effort focused on just that solution.

with the Quest to Native puzzles, we're hoping that to either see the guide matched well (which would imply that tools are sufficient). If not, we'll work to understand this better so that we can provide better tools.

[Susanne]

when I work on a new puzzle I will automatically go with the move that yields maximum score at that point in time and then build on that which might mean that there is a sudden stop in any progress of the current shape. At present I would not be aware at the beginning that some moves which may be low-yielding at that stage would then lead to the correct ones further down the line and the better shape as I have no knowledge of the native stage at that point. Even if I could see the native 'shadow' it would not tell me how it came to be that way, ie by moving sidechains, reshaping the backbone, nudging etc, and I would still have to opt for the 'pack the orange ones inside' option initially and see where it takes me and then try to adjust it to fit its native shape later. I am already hopeless at gaining advantage from Peekaboo solutions and am usually glad when the shadow of the best solution has disappeared. It would be a help if we could turn the native shadow off in 'View' as that gives the option of consulting the native state from time to time without constantly distracting from ones own puzzle shape.
The only possibility would be to release a puzzle in several 'completed' stages over and over again where solo or team solutions were nearest during development and then get all of us to improve on it again and again by going back to those solutions, ie the neverending All Hands puzzle……, maybe with no rankings as that distracts a lot.
I do think showing the native shape is a good idea but what will become of Peekaboo since the native state will always be the best option? If there is no Peekaboo what will be the gain from playing duels?

[Nicky666]

Thanks for the answer Zoran!
It would be great if you keep us updated on the findings :)

And Susanne…normally you can turn off the guide view, don't know about the peekaboo guides though (haven't used those). But I don't think being able to turn off the guide to the native structure in the Quest to Native puzzles would be helpful, as that kinda messes with the purpose of those puzzles in the first place.
Oh, and seeing your last two questions, I think you missed the fact that this will only be for a couple of puzzles.

[LennStar]

"Why I got points?" is the most asked question you ask yourself. That I think is the problem.

You have a number of methods that you use in different combinations, and then you shake/wiggle.
But you never know why the score increased - or why a configurations makes such a bad score.
The only thing you can work for is the score.
Intuition doesn't comes out from the blue sky. It also needs input.
So, if you want people to know why they should do smth., they need to see what it changes.

And there is a definitive lack of tools.
The note mode gives you the scores for one point. But if you change a sidechain, scores change on half of the protein. More or less invisible.

What is propably useful would be to open these score-notes on all parts. With numbers given in 3 columns: current score, score from last change (red if decreased, blue if increased) and point change from a user choosen point. (to see changes between 2 decidedly different configurations from big changes)

With that, the users get more feedback. That doesn't necessarily makes different configurations that are lacked, but very likely would train people to "see" where a change could be good, propably even in the very big picture.

If I may compare that to Go ^^ : If you first play this game, you have to think if you place your stones one point away. Then you learn to "see" what happens. You see what happens in the next moves. Then you learn to see what happens if you place the stone two points away… and so on. For a beginner it is impossible to see why a stone 3 points away is necessary. He propably can reason it out after 10 minutes, but only propably. A very advanced player may put down the stone without even bothering to think about it. Thats intuition. (And widely discussed - you propably should have a look at http://senseis.xmp.net/?HowDoesReadingAheadCorrelateToRank about this process of "cutting down thinking-work" to see if it brings something to your intuition :D)
But that is what comes after playing, training and reflecting about your past errors. The best way to do this is together with a bit better player that can point out errors and better ways that you can follow. A player that is a lot better is a bad choice, he shows you things you can't understand. And I think we lack the intermediate player. We only have the professional (the score), that says "put a stone at c4 to prevent a gap at f9", but no one who says the beginner "with d6 you get a better shape".

---

A point in diversity: I think the current save mechanism is bad for different approaches. I already had thought a bit about a better structure for the all hands mode.
We can take the same thoughts for single players:
A visible tree with savegames would make it easier to try different starting positions. So you have position A, go it down to Point B and then go back to A, make a change and have a go to point C.
But to go through this all takes time. Even big teams would have problems with that.

There is one point you should try: Start a puzzle 2 times. From the first time the players know it a bit. The second time they will have a different solution - how much is the question, but it will be different. AND because of the fresh start they won't say "If I try a different way, other will get better points in the needed time".

[Susanne]

To Nicky666: - thanks, I feel properly told off now! I am well aware that I can turn off the guide (ie team solution) in View but was unsure whether that would also be possible with the native guides, as they seem to be more significant. I would prefer to switch them on and off at will. I don't work well with a shadow protein in the background which needs to be realigned regularly. The whole point of Peekaboo is of course to take a peek and I will usually try to alter the shape of my puzzle in the time given to match the shown solution but find that even though my puzzle is on the move with the changes I am making, the peekaboo solution is not ('realign' does not work, thereby creating a visual mess). I couldn't work a whole puzzle like that.
Additionally you pointed out that there may not be many native puzzles for us to do and therefore I will withdraw my question re the peekaboo/duel conundrum.

To LennStar: - you have got near the essence of what I was trying to say re: - not knowing how to obtain the native shape if I am solely guided by points gains which may lead me to fold the protein the wrong way.

[phi16]

Folding proteins is sometimes like trying to untaggle a ball of yarn covered in honey. Sometimes I want to try moving structures to a new spot, e.g. take a helix running up one side of a protein and have it go down the other side instead. The tools we have currently make such a task something like wrestling or cow-punching.

I would like to have the ability to turn down (relax) the attractive/repulsive forces just long enough to do some manipulations without everything flying to pieces. After moving the structure, the forces could be made to come back into effect.

[CharlieFortsConscience]

As a noob, part of the Boing Boing recruits, I'm still not entirely sure what I'm doing, but I start with the assumption that each puzzle is basically almost there, and just needs optimising, with no major structural unravelling and rebuilding involved.

I understand primary, secondary and tertiary structure, but the tutorials could have a level 5, where more complex small proteins are introduced, maybe with the ghost guide, and the focus would be tertiary structure. Or have a results section, where we get to see one of the top 3 solutions when the puzzle ends, and how it relates to the original. I know I can join a team, and get a lot of this kind of feedback, but for most players, it seems very very easy to fall into the trap of just chasing decimals, without understanding why your decimals have gone up.

A nice tool to have would be to see each puzzle in its native primary form and then animate how it folds into the starting position, maybe first time you load it up.

[LennStar]

You can't have the native form, since its the goal of the game to find it - in unknown proteins. I don't know if the current Rosetta ones are known or not, but normally we work on known ones and the foldit team want to see if we can find the native - which hadn't happen enough from this blog entry.

To your suggestions, have a look here:
http://fold.it/portal/node/985644
If I remember correctly, this is all included there. If not, add it ;)