on picking the next game

[McLeopold]

It would be nice to have some official word on when and what the next contest is going to be. Correct me if I'm wrong, but I believe than means getting a decision from Jeff?

It would like to see a description of the process that will be used to pick the game. (Vote of the community or Jeff will decide, no complaining.)

It would also be nice if there was a commitment to a timeline of when the game will be announced. If we all knew the game will be announced on Feb 1st, that would be great. Even if the announcement is to narrow the list of choices down to a concrete few on a specific date, that would be useful too.

[jeff.cameron]

We've got two choices at this point.

Choice 1: multiplayer ants. The prerequisite here is that somebody has to sit down and build a multiplayer ranking system that actually works. There's been lots of ideas, but somebody has to build something that works before we can do this.

Choice 2: come up with a two-player game. We could do ants as a two-player game, but it seems like that would be a waste since it would be so awesome as a multiplayer game.

Choice 2 would be the fastest since it involves minimal new development. But we need a game idea!

[dmj]

Jeff,

I did find a python version of Trueskill. It doesn't handle teams, at the moment, but it handles multiple individuals playing against each other, and appears to have the same results as the Trueskill webpage.

The current version requires SciPy for normal gaussian functions, but I have a copy that includes other normal functions from Wikipedia.


https://github.com/dougz/trueskill


Dave

[jeff.cameron]

Dave that looks good. Perhaps we could hack together for a few hours sometime soon and try to get your trueskill library integrated into the codebase for the last contest. The objective would be to launch an actual contest instance using your ranking library as quickly as possible without having to do a complete rewrite of the backend. If we can get this working then we might be able to launch an ants (or zombies) contest really soon, instead of months from now.

What do you think?

[dmj]

Sure,

Let me clone his repository, and add the code for the normal functions that avoids needing scipy. I'll also try to see where it would fit into the old framework while I wait for us to find time to get together. If you want to try to coordinate, send me an email.

Dave

Ok, added the normal code: https://github.com/dmj111/trueskill

Can a git repo be spliced into another? I.e., can this be spliced into the contest repo? Git isn't my flavor of choice, so I am not sure what makes sense with it.

Also, we probably don't want to use it like bayeselo. One advantage of trueskill, for the contest at least, is that it is iterative. We can just update the ratings based on the new game, we don't have to run the whole history every time the ratings are updated.

[McLeopold]