Looney Labs Icehouse Mailing list Archive

Re: [Icehouse] IGDC Winter 2009?

  • FromDale Sheldon <dales@xxxxxxxxxxxxxxxxx>
  • DateTue, 24 Feb 2009 14:58:28 -0500 (EST)
On Mon, 23 Feb 2009, David Artman wrote:

* Should we have one?


I would go with a less-emphatic "sure--why not?" But that's still a vote for "yes".

So I reckon it's clear where I stand: we haven't done terminological; and I
think it would rock. At the least, we'd get some strongly themed games...

I also agree with this (but probably also not as strongly :)

Categories: I think they would not only make it easier to "focus" judges and build their own confidence in how they are ranking (hopefully leading to more judges!),

I still think that if they're _required_, they'll just be seen as more work, and lead to fewer judges participating. But certainly an expanded page on judging guidelines that *suggest* these sorts of categories would be worthwhile.

I think, further, that folks respond well to the "poll" nature of categories, rather than having to just sort of say "Well, I guess I like this one better."

This seques into your next point, but there has been significant debate (and some research) on the differences, cognitively, between ranking (as in "put these 10 things in order from best to worst") and rating (as in "assign these 10 things a score from 0 to 99). With more than 3 or 4 items, rating is easier, i.e., faster, i.e., would encourage more judges to participate.

In fact, maybe some sort of "combined" determination using Condorcet, Average, and Total would be best

I can't help but evangelize.  In advance, I'm sorry :)

Assuming any strategic voters, score voting is more likely to pick the true Condorcet beats-all winner than any commonly used actual "Condorcet" method. (And even if everyone is perfectly honest, score voting has been shown to pick better (using a yardstick called Bayesian regret) winners than any commonly used Condorcet method, and a new paper suggest that, with more than 3 candidates, score voting is mathematically guaranteed to be better than _any_ ranked-order voting method that exist or could exist.)

Trying to combine the two would be either redundant or incredibly awkward. And, while this is still being researched, score voting with "soft zeroes" is currently the recommended way (by the Center for Range Voting) to combine "average score" with "total score" (with the goal of avoiding poorly-known candidates with fanatical followings from getting surprise, and undeserved, victories.)

Anyhow, I'd collect the votes and do the wiki-work (and the math, if not
Condorcet; I'd call for an "impartial math head" to handle that).

I am more than happy to compute the Condorcet votes if we stick with that for this contest. (Incidentally, that's another advantage of score voting; it doesn't require a dedicated math-head to compute it, just lots of addition and a little division. Although, given the abbreviated time line, perhaps we should just stick with what we've been doing for this time.)

Dale Sheldon