Board Game Geek logo

Finding fun games with Board Game Geek.

Karim
5 min readApr 10, 2023

Boosted by the lockdowns of the recent pandemic, the popularity of board games is at an all time high. The Board Game Geeks website is the premium resource for all things board game. The website’s API let’s you access a treasure trove of data on board games, from user ratings, to reviews, to game mechanics, to game designer.

So let’s see if we can use this data to find good board games.

1. What type of games are there?

Let’s start of by looking at the different types of games.

Board Game geeks offers different labels like subdomain, game mechanic and game category. It’s important to note that they are user generated, so there can be some overlap, also the labels are not mutually exclusive.

Game Mechanic Counts
Top 30 game mechanics by number of games

It should come as no surprise that most games use well established mechanics like dice rolling (Risk, Monopoly, Parcheese, Catan), hand management (Rummy, Canasta), roll/spin and move (Monopoly, Parcheese, Game of Life) and set collection (Rummy, Canasta). Yet, other well established mechanics like bluffing (Poker) and trick-taking (Bridge, Tarot, Skat) are surprisingly missing from our Top30 list.

To the initiated board game aficionado, exploring new game mechanics is part of the fun, which is reflected in the more than 100 different game mechanics listed in Board Game Geeks.

Game Category Counts
Top 30 game categories by number of games

Game category is a more flexible label, and thus overlaps a lot with subdomain and game mechanic, but also includes subjects like World War II, settings like Fantasy, skills like Action/Dexterity, time periods like Medieval, themes like Movies/TV, etc.

Game Subdomain Counts
Subdomain by number of games

Subdomain loosely translates to game genre and is a tight label with only 8 different instances. Some are obvious like war, children’s, family and party games, while other’s need a bit of explanation.

Abstract games include games like Chess, Go, Mah-Jong and Mancala.

Strategic games is the subdomain mostly associated with modern board games, as it includes more complex games with elaborate rules and game mechanics. The most well known is probably Catan, but others like Ticket to Ride have become more widely spread in recent years.

Thematic games focus on atmosphere and narrative, and include games like Cluedo or Munchkin.

Customizable games include trading cards games and the like and are customizable in the sense that the game play changes with each further expansion of the initial card/dice/miniature set. The most prominent member by far is Magic The Gathering.

Bottom line, games come in all sorts of shapes and form.

The only surprising insight is how dominant wargames are as a genre. What’s up with that?

What IS up with all the war games?

War games were a very popular board game genre with the late seventies considered as the ‘Golden Age of Wargaming’. Let’s check if we can see this in our dataset.

Number of games by subdomain and publishing year
Number of games by subdomain and publishing year

Going by number of games published, wargames did indeed outcompete other game genres during the golden age, but even though other game genres like strategy games have caught up, they still enjoy a high popularity today.

How do I find a good board game?

Users of Board Game Geek can rate board games — let’s see if we can build a model that predicts those ratings and analyse the model to find hints on how to pick nice games. To do that, we have to preprocess the dataset, by dropping irrelevant columns (e.g., accessories and versions), one hot encoding game categories, game mechanics, subdomains, game families and publishers, as well as adding counts for expansions and honours (i.e., awards). (full project on GitHub).

Table of Coefficients
Coefficients of Linear Model

The leading features are “usersrated”, “owned”, “numcomments”, “wanting”, “wishing”, “numweights”, “trading”, accounting for 95% of the total explainability.

It’s fairly obvious that the underlaying factor is user engagement, which basically comes down to recommendations from the community.

As an avid board gamer, I already know that talking to other aficionados is a great way to find new games, so let’s get rid of those features and see if we can uncover something deeper.

Table of Coefficients
Coefficients of Linear Model without engagement

After getting rid of the “engagement” features, “hon_count” is the predominant feature. The “hon_count” represents the myriads of board game awards listed in the Board game geek dataset.

Conclusions

In a nutshell, the best way to find a good game is to get recommendations from you local board game geek, or by tracking gaming awards.

Next steps

It might be worthwhile to analyse the honours columns in detail, to identify specifically which gaming awards correlate with user ratings the best. Also, spending more time on reducing the number of one-hot-encoded columns in more elaborate ways than using an arbitrary cut off and binning, might improve the prediction.

Resources

GitHub repository

https://boardgamegeek.com

https://www.kaggle.com/datasets/seanthemalloy/board-game-geek-database

--

--