Ratings

(Part of Loaded Dice.)

What does a rating mean?

My database contains information on 52880 games. About 17800 of those games have no ratings at all. For the rest, at least one person has expressed their opinion about the quality of that game. Ratings are given on a scale of 1-10, with 1 being the worst possible game and 10 being the best possible game. If the ratings were evenly distributed, a mediocre game would get a score of 5. A game with a rating of 6.0 would be in the 60th percentile, slightly better than average.

But the ratings are not evenly distributed. The mean rating given out is a 6.82 (median=6.95, std=0.98). People tend to rate the games they like and not rate the games they don't like.

This skews the ratings so that the average game rating is 5.8. (Median 6.0, std=1.56). A game with a six-star rating may be fun, but compared to other games on BGG it's mediocre: half the games on BGG have better ratings.

BGG Decoder Ring

Since ratings are skewed and not distributed normally, we need some way of normalizing the data. I took a tip from from Tom Moertel's work on IMDB user ratings for movies. He found that the mean movie rating was 6.2 (median=6.4, std=1.4). That's higher than the mean BGG rating of 5.8, but his IMDB rating histogram is very similar to my BGG rating histogram.

Moertel presented a "decoder ring", a table that maps an IMDB rating onto a percentile. Here's a similar decoder ring for board games. You can use a game's BGG rating to get a picture of how many games are better.

RatingPercentile
3.50 9
4.00 15
4.25 17
4.50 21
4.75 24
5.00 32
5.25 36
5.50 42
5.75 47
6.00 58
6.25 64
6.50 71
6.75 75
7.00 83
7.25 87
7.50 91
7.75 93
8.00 96
8.50 98
9.00 99

A game with a 7.5 rating sounds pretty good, but as you can see from this table, it's actually really good: in the top ten percent of games. Similarly, a game with a 4.75 rating sounds kind of bad, but it's actually really bad: in the bottom twenty-five percent.

Standard deviation of ratings over time

Although newer games are more highly rated, the range of opinion hardly changes at all, and may even be narrower for newer games.

Standard deviation of ratings over time

How many games are unrated?

Newer games are more likely to have at least one rating than older games. But even among games published within the past ten years, 25%-30% of the games on BoardGameGeek have no ratings.

Do newer games get more ratings?

For each year with more than ten rated games, I calculated the median number of ratings a game received, if it got any ratings at all.

Games published later are likely to get more ratings, assuming they're rated at all.

Are games getting better over time?

When I was a kid, most available board games were pretty bad. Have board games gotten better over time, or is it anti-nostalgia? Here's a graph that kind of address the question. For every year containing more than 5 rated games, it shows the mean game rating for that year. If games are getting better over time, we should see an upward trend. I included separate datasets for the novelty-driven genre of children's games, and the more conservative genre of wargames.

There is a definite upward trend, but I suspect this graph actually visualizes the disconnects between BGG's rating system and any notion of objective game quality:

I do think games have gotten better over time, and I think this started in the mid-1960s, with the same cultural shift that put wargames on the map, but I don't think this ratings data is very strong evidence.


Next: Categories and Mechanics


This document (source) is part of Crummy, the webspace of Leonard Richardson (contact information). It was last modified on Monday, September 05 2011, 19:14:47 Nowhere Standard Time and last built on Friday, December 19 2014, 10:00:25 Nowhere Standard Time.

Crummy is © 1996-2014 Leonard Richardson. Unless otherwise noted, all text licensed under a Creative Commons License.

Document tree:

http://www.crummy.com/
writing/
Loaded Dice/
ratings.html
Site Search: