I have started a document on ways to deal with voting for lab. I am sharing what I currently do, in hope it will help others too, while you are still busy with the voting for the current lab. It is primarely a method on how to deal with second or later round lab voting. But it can be helpful for round 1 to and lab designing as well.
For myself, I do it by finding a high score or design I like and make it my comparison template, then when I find another by side by side comparison that is better, it becomes the new template and gets promoted and maybe a vote. I quit after eight votes. Look for any others that catch my random eye and sometimes they end up on my list as voted winners–again by comparison.
Your last comments in the above “Ways to deal with lab voting”–to vote for a candidate that already has more votes than just one (yours)–is a good strategy for points award but not a good way to figure out what people really think of designs on their own merits.
That said, I do it all the time–to keep my points award decent and keep some EteRNA ‘cred’ by maintaining my rank somewhat. Again, not a good way to find out what I really think of designs on their merits. but I have to operate within the limits of the game.
I think it a much worse form of crowd source selection than “ranked voting” would be. But I’ve said this many times and either nobody understands it, or somebody thinks it too much trouble to implement. So, I live in the ecosystem as it is.
Having a powerful filtering system, combined with a powerful side by side visual comparison of relevant charts like dot-plots or SHAPE charts would be enormously helpful. I spend most time collecting the data in ways I can process rather than thinking about and evaluating the actual designs. I posted some screenshots of how it might be done in the forum, FWIW.
“Having a powerful filtering system, combined with a powerful side by side visual comparison of relevant charts like dot-plots or SHAPE charts would be enormously helpful.”
Totally agree.
Janderson came up with an idea on how to integrate a filter for voting.
And yes my strategy isn’t bad for points award. What I perhaps forgot to mention in the strategy is that for second rounds, I am primarely interested in looking at the best scoring desings from results and designs that specificly mod those high scoring desings. I am interested in anything that attempt to fix the problem bases in a good design. I won’t just vote for a mod of a highscorer, unless I see the mod not just changing random stuff, but changing something at the problem spot or in nearby area. (I know that sometimes changing an already working neck with another, can actually improve both the score and have a stabilizing effect elsewhere in the design).
I liked the screenshots you posted on sorting the information. I like the idea of having shape data and design besides each other.
Having a tool or option to find similar designs to another (presumably high scoring) design would be helpful. For example, pick a sample design and then compute a modified Hamming Distance (by number of bases–not bits–changed) from that reference design and sort on that column.
Hi Janderson. I like your idea very much. I think you should talk with Omei. He is planning on programming such a tool.
I love these ideas too, and have higlighted them to the dev team.
We are currently refactoring the sequence browser to better visualize and handle the larger data throughput. So your comments are well-timed.
I had another idea – if we could allow the ‘strategy collection’ devised by players to act as custom sortable columns, that might help in visualization & voting – we should definitely use more than the free erngy , Tm, and other features since you and the players have come up with much better and predictive features!
"…we should definitely use more than the free erngy , Tm, and other features since you and the players have come up with much better and predictive features! "
Hmm. Any chance of adding somewhat generic “scriptable” columns? That might allow say running some of the StrategyMarket ™ scripts as predictive features, then performing a “weighting function” or other meta-script to generate a PlayersRanking of the strategies.
It could engender a whole new competition: who’s Ranking Strategy performs better on Lab X?
Yes, that’s exactly what we would want. The challenge for the devs now is how to create this experience and ensure that the resulting knowledge would be harnessed and written up… if you have ideas on that, please share!
It depends on how open you want the cookie jar to be. Having a RESTful interface via HTTP that returns application/json or even application/csv query results will help non-dev folks to prototype/develop stuff, possibly in JavaScript which would make it accessible to lots of users/browsers. HTML5+JavaScript seems to make adding browser-based functionality relatively easy.
Are you still using Flash exclusively? I don’t have any experience with that so I don’t know how well it would integrate, nor how difficult to migrate.