BikeLouP Thoughts on Strategies and Tactics for FolditEdit
Foldit is an addictive game that allows for many approaches that seem to converge on about the same solution--most of the time. Most foldit players that have been doing it for any length of time find different techniques that work for us. Many of us have applied the same or very similar methods with different results because of small nuances that are hard to describe--but we try to share here in the FoldIt Wiki. Like any game, there are several distinct portions of the game, at least as it has been presented to us through the game interface:
- The initial puzzle and first moves. At this point in the game you've loaded the latest puzzle, and the structure is presented to you. You now have a large number of choices available to you, but this is where most players begin their own strategies.
- The mid-point game and rapid point runup. During this phase a lot of different tactics work to extract large point gains. When you can't easily get point runs ranging from tens to hundreds (or more on some puzzles), then this phase ends.
- The eke and tweak game. At the point where you are trying methods to get one point, or even partial points, with occasional breaks with small runs, you have entered the eke and tweak game. My guess is this is where the team "Another Hour Another Point" got their name. Ask Sirenbrian about that!
- The end game. There really is no end game--we don't know if a puzzle is complete, only that a deadline has occurred and the puzzle closes out. The results of these predictions may or may not be the best answer for the target protein predictions. Someday maybe lab results will verify them.
I distinguish between strategy and tactics. To me strategy is how you approach each of the phases of the game. Tactics are the methods and our own little techniques that we apply during the strategies. I will refer to tactics as methods.
The Inital Puzzle and First MovesEdit
The very first decision you need to make as a foldit player is whether you are going to try and noodle your solution with great technical insight or just go for it. My belief is that the beauty of this experiment lies in the fact you don't have to know anything about molecular biology or chemistry, and with lots of folks trying different methods someone may get the lucky path to the best solution. It may come from a great team effort or may just come from a savant genius or anything in between. I have no technical insight so I cannot pursue that as a strategy. I believe the bulk of the technical insight is in the algorithms that drive the solutions, so I just go for it.
Given you are just going to start, the main question is do you try anything before you do a global Wiggle? Generally, the protein you are looking at has some structure and constraints and is probably someone's best guess to start with (although I cannot confirm that.) As such, it seems logical you'd want to just start a global wiggle. However, I have come to prefer a strong pull followed by a global shake then a global wiggle often gets me several hundred points higher right off the bat than if I just to a global wiggle, shake and global or local wiggle. I have no idea why this works better for me but it usually does. You can always restore the puzzle to the beginning and try both, and early on there is not a lot invested in the structure so it makes sense to try both. The procedure I follow to implement this requires you understand how to undo, use hotkeys to stop wiggles while they are still working, and the basic moves of the game:
- Set the puzzle to stub view. Unlike some players, I like to use the stub mode to monitor the hydrophopic (orange) and hydrophilic (blue) sidechains. This allows me to see them while doing other methods on the backbone, and is faster than if you show the entire sidechain. I set the scoring to relative score view to show trouble spots.
- Pull on one end to spread the puzzle. Other players do not seem to favor this move, but I like to take the default configuration and spread it with a pull. Simply grab an end and pull the puzzle apart with a strong pull.
- Global Shake Sidechains. Other players also do not use this at this point. I find the puzzle will more likely find a sweeter spot early if you do execute this move before you do a global wiggle.
- Use Global Lock/Wiggle Strategy. Lock the entire puzzle (just hit the "l" key.) Then hit the "w" key to start a global Wiggle. Then release sections of backbone, usually 2 to 3 at a time in a section. Avoid major loops at the ends. Stop after you've unlocked about half the backbone sections.
- Shake Sidechains. Stop the wiggle with the "w" key, then unlock the entire puzzle with the "l" key. Then hit the "s" key to execute a global wiggle. If you are using an all loop structure, a right click can be used instead with a local wiggle.
- Global Wiggle. At this point the puzzle is ready for the first big point gain--the global wiggle. Wait for the score to stop rising before stopping the wiggle with either the spacebar or "w" key. This usually is the biggest move and point gain in the game, so whoop it up! Note your score after the score stops rising. It helps to wait a couple of minutes (so go get coffee, stretch, solve global warming, etc. while you wait) and let the wiggle go even thought it seems to have stopped at one spot--sometimes a major point gain all of a sudden breaks. I always prefer that to eke and tweak strategies.
- Try Alternative Methods. At this point you've only invested a couple of minutes in the puzzle, so the basic idea is to try a couple of other starts and pick the one you like the best (usually highest score but not necessarily.) Find the "reset puzzle" function and reset the puzzle. Then just do a global Wiggle and look at the score. Sometimes I'll start the wiggle, stop it about 100-150 points out from the first high score, do a global Shake Sidechains, and then finish the wiggle. This can result in a large point jump. You can also go back and try different point locations for stopping to do the shake or even try a couple of different spots with shakes. You can always go back to your highest score using the graph or the undo functions. For the less responsive puzzles doing this systematically may result in finding a dramatically higher initial starting point for the midgame.
You can also try, before you do a global wiggle, to manually move the structures with pulls, rotate sheets around, or any other method. I do not generally apply these because I have believe the basic structure has been presented and I have not found any of those techniques to get a higher score. Sometimes they do make a nicer looking protein with an all dark background, but a lower score. Scoring is our main feedback, so I generally do not use these techniques as a Pavlovian response to the score.
The Mid-Point Game and Rapid RunupEdit
Once you've had your first big point gain, I then merely pull the puzzle around and do wiggles. I often will do a global Shake sidechains before each wiggle early on, but if the score does not go up I'll undo and let the wiggle go. After several increases in score I will then pull, wiggle and stop the wiggle about 100-150 points out from my high score and do a Shake Sidechains. That sometimes frees up a sticky spot and gets a significant increase. If you don't get a new high score, undo and go back to the pull point and just wiggle. During all these phases I have learned to use the hotkeys and leave the undo window with its graph up so I can watch the progress of the method. For pulling, sometimes I find you can just grab the puzzle in pull mode and move things aroung, shake/wiggle and then keep doing it at that spot as long as points go up. Generally you are looking to "scrunch" the puzzle but you can sometimes just get in a rhythm and the puzzle seems to rock back and forth as you do this quickly. The other pulling technique I use frequently is to try and use rubber bands to pull the "orangish" sidechains inward towards each other, sometimes pulling the "blue" ones outward or using the rubber bands to shift things around. Most of the time those are quick wiggles, followed by a shake and follow-up wiggle. At any point you can undo or go back to your best solution.
Note during this phase I am indifferent to using all loops or structures. I find the loop to be handy for using the right click on the mouse with a local shake that acts as a global shake, with sometimes differen results. Some puzzles pull better when in a loop and others pull better with structure. The one advantage structures have is they pull differently than loops so you have some control over pulls, but I don't have any real preference. I usually just stick with the initial structure.
After the runups seem to be stopping, I then employ several different methods. I used to primarily use Sirenbrian's Local Wiggle Strategy as a method in between pulls, but on the more recent puzzles that seems to have been less effective. At that point I usually change the structure to all loops. I alternate between pulling and local wiggles as long as that works. You can also try using Local Rebuild Strategy instead of the Local Wiggle Strategy. The rebuilds seem to work very well on the end o the loops using 2-5 segments.
When it seems to be stuck, I then try to use rubber bands and tweaks to change the structure a bit, before pulling or using the local wiggle strategy. At this point usually the points start getting very hard to come by, and anything you do seems to drop the score.
The Eke and Tweak GameEdit
This is by far the hardest and most difficult part of the game. At this point it seems like the puzzle is "locked" or only goes to the same or lower score no matter what you do. You can see the team scores and the high player scores ahead of you so you know there is more to be done--but now nothing seems to work. For me, it seems to indicate something is "wrong" with my solution at that point--either the structure is wrong, there is something out of place, or a sidechain or two is out of place. If you can use your award points to look at a leading puzzle that may give you some insight as to what to do. So, here are a few things to try:
- Take a break. Sometimes just leaving a puzzle and coming back to it later results in some quick points when you log back on and try your normal methods. I have no idea why this occurs, other than the math in the machine is so nonlinear maybe some small initial state changes when you log off and log back on.
- Tweak sidechains. There are several different methods for messing around with sidechains. More knowledgeable players can actually look a them and know which sidechains repel or attract each other, and can move them appropriately. Note in some cases you have to spread the puzzle apart to move the sidechain. I have managed to find one sidechain on a puzzle, rotate it, and gain over 100 points near the end of a game. Another method is to just go through and move each sidechain, do a Wiggle and see what happens. One of my tricks is to do a quick pull right after that because even though you may not get a better score it helps the structure or breaks something loose that gives you some more points after a pull. Final note on this--sometimes you move a sidechain, and try doing a shake which should move the sidechain back and give you the same score--usually it doesn't but you can again pull to see if it broke something loose.
- Use Pletsch's Suggested Method. I start trying his techniques towards the end of my midgame and during the eke and tweak stage.
- Go back and reset your structure. This seems like a desperation move, but if you are a day out or so from a finish and you've peeked at some better scoring puzzles and you know you are way off it might be worth either massively moving your structure around or going back to an earlier save point or even the beginning of the puzzle and more carefully setting the structure to what you think will work better and starting all over. You kind of have a feel for the puzzle at that point and it is surprising sometimes how fast you can pass your old score.