Replay Mode

Case number:699969-986177
Opened by:ChristianK
Opened on:Friday, May 22, 2009 - 21:30
Last modified:Saturday, November 30, 2013 - 17:17

After a big chess tournament the moves that each player made in a game are usually published. That allows people to learn from each other.
The same goes for real time strategy games like Starcraft.
I think that also needs such a mode to let people learn from the way the best people and teams solved past challenges.

(Fri, 05/22/2009 - 21:30  |  7 comments)

Joined: 06/17/2010
Topic: Game: Tools » General

It can be done from data send to server? Kind of movie?
It would be really interesting.

Joined: 05/19/2009
Groups: Go Science

I would note that if you want to copy human strategies into algorithms, it's very helpful for it being easy to record human strategies.

As far as space issues go, you don't have to safe an image every minute. You can just safe all human input and 10-20 random protein images that are sent to the server. At the beginning maybe every 10 minutes and afterwards every hour.

Joined: 05/18/2008
Groups: None

Chess and Starcraftt have not the goal to find new ways to solve real life problems.

The goal of Foldit is to explore the maximum number of strategies.
If you only copy a good strategies or a few that give high scores, you will not add anything to the goal of the game.
If you like the game, find your own strategy, create a new team or at least join an existing team for inspiration.
Note that some players choose to share a picture their solution at the end of each puzzles and some players choose to share their recipes, we wait to see yours.
In the wiki, you have videos in 101 Puzzl Levels and Types and in Puzzle 668 and many tips.

Joined: 05/19/2009
Groups: Go Science

The goal of Foldit is to explore the maximum number of strategies.

That would be a pretty stupid goal. The actual goal is:

We’re collecting data to find out if humans' pattern-recognition and puzzle-solving abilities make them more efficient than existing computer programs at pattern-folding tasks. If this turns out to be true, we can then teach human strategies to computers and fold proteins faster than ever!

Humans' pattern-recognition skills simply work better if you give the human specific examples of how other humans solve proteins.
Chess is a field where we have a lot of research about how humans develop good pattern-recognition skills related to chess.

There no reason to ignore the way humans get good at pattern matching if your goal is to have humans beat computers at pattern matching.

Note that some players choose to share a picture their solution at the end of each puzzles

The picture doesn't help that much to get humans to get better.

It's that old chestnut: high scores vs viable proteins.

No, it actually hasn't anything to do with "high scores vs viable proteins". Seeing replays that produce "viable proteins" that don't achieve maximum scores could be just as valuable as watching replays that try to maximize scores.

Joined: 05/18/2008
Groups: None
Joined: 04/20/2012
Groups: Go Science

I like the idea of replaying structural optimizations. When using Recipes on large proteins, it is often hard to tell what is changing as the score rises. Perhaps watching the replay would give more insight into what is occurring.

Perhaps this movie option could be implemented for each user as follows:

(1) The user saves key structures/steps on his computer as he/she works on a puzzle.
There could be a check-box for each structure saved to say if you want to include it in the movie.
Perhaps re-orient the protein when you save it so that its center of mass is always in the same position (the origin), alpha-carbon #1 always points up (along the z axis), and the Nth alpha-carbon always lies in the xz plane (where the x axis points to the right).

(2) The user sets the view options within the Foldit GUI as desired.
The user can use tabs and label key things. These options will hold for all frames of the movie.

(3) Pressing a special replay button goes through all these saved structures like a flip show.
It could have rewind/fast-forward/pause/stop/record buttons. Perhaps let the user save the flip
show in a gif file to share the flip show with teammates or globally.

(4) It could loop through the structures repeatedly and
let the user rotate/reposition the structure as the flip show goes.

Joined: 04/20/2012
Groups: Go Science

The video at shows how the structure changed as the score rose for a certain puzzle. Can such videos be made for more puzzles?


Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons