Gary Weinfurther
United States
West Bloomfield
Michigan
-
How is this number derived and what does it mean?
-
Richard Irving
United States
Salinas
California
-
elmonty wrote: How is this number derived and what does it mean?
In order to prevent a new or rare game with only a few high ratings from taking the top spots in the ranking, 30 average ratings (not sure if it 5.5 or simply the current average rating on the geek) are added to every rating to form the Bayesian average. As more ratings are received, the effect of these "damper ratings" is reduced to nil.
At best this a rough way to "solve" the problems with a rating system which has many problems:
- Not everyone plays every game or even a small percentage of them. I am sure I have played a realtively high percentage of the games listed, but I am also sure I have played less than 10% of the games in the database.
- People self select the games they want to rate. Games that appeal to very small, but devoted, audience may rack up a high average, but many people simply would never play the game. There is a related problem with expansions--people who dislike the base game don't play the expansion, so their negative votes don't lower the average.
- Different people are not consistent in their ratings (some give multiple 10's, others almost never.)
- People rate on their own preference which may be different than the intended audience of the game (Children's games, wargames, etc.)
At best take the ratings as a rough guide: Games within about 0.5 rating point of each other are pretty similar in preference.
-
Michael Nerman
Canada
Winnipeg
Manitoba
-
Cool. I remember trying to figure that out a while back and not being able to find the information. Thanks!
-
Gary Weinfurther
United States
West Bloomfield
Michigan
-
It seems like this information should be posted in an easily found location.
-
Ludes
United States
Honor
Michigan
-
What I'd like to know is who is/was Bayesian?
If you've ever heard of the Sagarin ratings (college basketball?), I encountered the guy about once a week - he shopped at the co-op where I worked.
-
Koert Debyser
Belgium
Brugge
-
LudesFactor wrote:
What I'd like to know is who is/was Bayesian?
If you've ever heard of the Sagarin ratings (college basketball?), I encountered the guy about once a week - he shopped at the co-op where I worked.
The probability of bumping into Thomas Bayes is rather low.
Reverend Thomas Bayes (c. 1702 — April 17, 1761) was a British mathematician and Presbyterian minister, known for having formulated a special case of Bayes' theorem, which was published posthumously. His theorem was one of the many we had to learn at high school and again at university, but don't ask me the details  .
Bayesian probability is the name given to several related interpretations of probability, which have in common the application of probability to any kind of statement, not just those involving random variables. "Bayesian" has been used in this sense since about 1950.
More info and source: http://en.wikipedia.org/wiki/Thomas_Bayes
-
Heinz Kiosk
United Kingdom
Unspecified
Unspecified
-
Bayesian probability is the extension of probability theory to non-random events.
A common modern application is spam-filters. Each recipient trains a Bayesian engine by showing it some spam and some legitimate email that you have been sent. The more the better. The engine then examines incoming mail and uses Bayesian inference to determine the probability that it is spam by simply examining the frequency that particular words commonly used by your known spammers appear in it. The method has the benefit that (for example) a doctor who genuinely researches viagra will not have viagra-related email rejected because the word will appear commonly in his genuine email and for that particular recipient the word viagra will not have high spam significance. So the engine decides the probability that an entire message is spam by reference to the spam-likeliehood of all of the words in the message, compared with its database of known spam/no-spam.
-
-
Seems like approximately 150 average ratings are added to make Geek Rating.
-
|
|
. 
