Recommend
1 
 Thumb up
 Hide
7 Posts

BoardGameGeek» Forums » Gaming Related » General Gaming

Subject: (Practical) voting systems rss

Your Tags: Add tags
Popular Tags: [View All]
Curt Carpenter
United States
Kirkland
Washington
flag msg tools
badge
Avatar
mbmbmbmbmb
This is only somewhat tangentially related to gaming, but it's related enough that I hope people will humor me for asking here.

I'm not a statistician, but I have been thinking about voting systems. And specifically one that can be administered immediately after tallying the votes. For this audience, you could think about it in terms of voting system to choose the best game. The BGG Gold Geek awards use the Schulze method. One critical aspect of that is that votes are based on relative rankings, which is an element I want to keep, but also based on the all relative rankings being considered as outranking all unranked options. That aspect is one I don't want. For example, consider a game design competition where instead of comparing games that are already biased in terms of quantity by measures of marketing, user feedback, etc., we instead have a set of games that no one has played. Then assume that a random number of people play each game, and each voter plays a random number of games. In this case, assuming that ranked options exceed all unranked options is a false assumption.

Can anyone suggest a good voting system that can actually be carried out in practice (people physically together casting votes, and then resolving on the spot), that uses relative rankings, but does not bias toward the options that get the most ratings?

Ok, just to make sure that folks don't take this too abstractly, the real reason I'm asking is that our church group is planning a chili cook-off, and I want to make sure the voting system is fair, and that quantity of chili brought is not a deciding factor.

Thanks!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
David Molnar
United States
Ridgewood
New Jersey
flag msg tools
Avatar
mbmbmbmbmb
my first thought was, what if you can rate a chili +2, +1, 0, -1, -2. But, since people tend not to give negative ratings because they think that's "mean", that could still end up being biased in favor of the big chili providers (BCP). How many chilis is someone liable to try? What if you keep the same scale, but your total ballot has to add to 5? You'd still have to have a fall-back policy for what to do with ballots that don't add to 5. But, that could force the negative votes that, really, you want.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
'Bernard Wingrave'
United States
Wyoming
Ohio
flag msg tools
Avatar
mbmbmbmbmb
I think one method that is sometimes used is to have a select jury sample all of the items and give one prize ("Jury's or Expert's Choice"), and for everyone else to pick and choose whatever they want to sample (and maybe give a separate prize for that: "People's Choice"). So if you have a jury of 6 people, and the whole jury and all the contestants are there to give samples to the jury first, then each contestant only needs to have 6 tastes' worth in order to get a fair shot from the jury.

I know this doesn't directly address your question, but it still seems applicable to the situation.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Enrico Viglino
United States
Eugene
OR
flag msg tools
Slowed - BGG's moderation policies have driven me partially from here
badge
http://thegamebox.byethost15.com/smf/
Avatar
mb
molnar wrote:
my first thought was, what if you can rate a chili +2, +1, 0, -1, -2. But, since people tend not to give negative ratings because they think that's "mean", that could still end up being biased in favor of the big chili providers (BCP). .


You could not tell people that they're giving negative ratings.

I like the numeric limit (even of zero) to ratings though. makes 'em think
harder about the ordering.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Russ Williams
Poland
Wrocław
Dolny Śląsk
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
What about a radically simple idea (which might be good since it sounds like the voters are a bunch of ordinary non-geeky people):

For each chili a voter tried, they say whether they would willingly pay to eat a second bowl. (Or some equivalent "thumbs up or thumbs down" binary question.)

Then compute (# yeses) / (# who tried it) for each chili (so each scores between 0 and 1, but there's no explicit advantage for more people trying a particular chili), and the high score wins.

The risk is that people might not answer, instead of answering "no", for chilis which they tried but weren't enthusiastic about (so as not to be "mean") (even though that's being "mean" to the better chilis, but people don't think that way...)

It could be wise to make a policy (and clearly publish it) that you won't report the specific scores of all chilis. That way, people might feel less "guilty" giving an honest thumbs down.

---

On the other hand, if a chili does get sampled a lot more, maybe it's because it is good, and by word of mouth more people are wanting to try it. In that case, simple Approval voting might be appropriate.

(But if there's a risk of some chili coming in insufficiently large quantities and thus running out, then no...)
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Donald X.
United States
flag msg tools
designer
curtc wrote:
Can anyone suggest a good voting system that can actually be carried out in practice (people physically together casting votes, and then resolving on the spot), that uses relative rankings, but does not bias toward the options that get the most ratings?

Ranked pairs. People list chilis by how much they liked them - no numbers, just relative rankings. Chilis can be at the same level, that's fine. You look at each pair of chilis and see how many people preferred one and how many people preferred the other (make a master list of pairs and go through the ballots making tick marks on your list). Find which pair has the most decisive victory and drop the loser from the contest. Find the most decisive victory among remaining pairs; repeat until one chili remains.

If only one person tries Joe's Terrifying Chili, and that person puts it ahead of all others, then Joe's Terrifying Chili will win. This demonstrates that having a smaller quantity of chili doesn't lose you the contest, but is bad. For your low-key situation you can reasonably deal with this by requiring a minimum number of rankings for a chili to actually compete.

With a small number of voters you may easily get a tie. If you really want a single winner, try to have some chili left over at the end for tiebreaking tests.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Curt Carpenter
United States
Kirkland
Washington
flag msg tools
badge
Avatar
mbmbmbmbmb
Thanks for the ideas guys. Ranked pairs does seem to be the best, but a bit tricky to implement while people are standing around waiting for results. Maybe some drama can be created by working through the process. Would require some practice.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.