A Developer's Guide to WikiGames

Having made a bunch of games based on Wikipedia-content, now I think I’ve learned some core lessons on what to do (and not do) when trying to make a “WikiGame”. Hopefully, sharing them here will save others time

Pick the articles carefully

There are millions of articles on Wikipedia, and, depending on your game, most of them will not fit into your game. Generally for WikiGames you are showing players Wikipedia content because you think it’ll be interesting to them (they don’t already know it but are happy to learn it) or because you want them to guess the content based on some clues. 

Unfortunately, a lot of Wikipedia content falls into a middle ground where it’s about a subject that’s not really interesting and also not something most people have heard of. A completely random example would be Colinton, Alberta. Very few people have heard of this hamlet, but there’s also nothing particularly interesting about it.

How do you fix this? Well, if the game only needs a fixed amount of content then you can just manually pick the articles. Read through one of the many lists of interesting articles and pick your favorite. I suggest the community-curated Wikipedia:Unusual articles but there are also lots of lists maintained by individuals e.g. this one or this one or this one though there tends to be a ton of overlap between these lists.

If the list of articles needs to be huge or constantly growing (e.g. for a daily game) then maybe consider sourcing from Did You Know. I uploaded a dataset of them that you can use, but DYK articles also tend to be lower quality in general.

You could also try using some ML-based approach to selecting the article. This could be especially important if you need to filter out “adult-oriented” topics. We made a model that is meant for helping with this, but you can also try using ML to identify “interesting” topics in much the same way.

One final consideration is accuracy. You want to pick articles that are high quality. Not all Wikipedia articles meet the same bar. If you just want the best articles from that perspective, you want Featured Articles. But the original warning still applies: just because an article is “featured” does not mean it will be interesting or even familiar to your audience. A larger list of articles meeting a lower bar are Good Articles.

Article paths are often unintuitive

A lot of Wikipedia-based games try to use the links between articles. The classic “Wikipedia Game” is entirely based on this. While the game is a classic, the rules of it are not totally set in stone. What links are you allowed to click varies. Can you click on categories? Lists? Links to other namespaces? Can you use ctrl-f? This confusion translates to designing a game based on paths. Depending how you ingest the Wikipedia dump you will get fewer or more links. Are you including links inside of templates? Are you handling redirects?

Wikipedia articles often contain lots of links and these links don’t always “make sense”. For example, the article on WWII links to the “Wayback Machine”. There isn’t a meaningful connection between those two topics, but one of the sources on the WWII article was archived using the Wayback Machine. If you’re computing “shortest paths” between unrelated topics you can end up with a lot of your paths containing these uninteresting links.

What’s the solution? I suggest restricting the links to only those in the introductory paragraph. Links there are the ones most central to the topic. You could also include the infobox in this. But even this approach won’t save you. For example you’ll still get a link from “Vichy France” to “Literal translation” because of a template in the lead paragraph. Depending on your game’s mechanics this might feel wrong.

Wikidata is underused

Lots of games want to use structured information from Wikipedia articles. When were they born? Who wrote this book? Who were they married to? This information is generally in the Wikipedia article. It’s often in the infobox at the top of the page, but getting it out can be challenging. The whole point of Wikidata is making this information easy to access. 

Yes, if you want everyone born in 1938 you can just use the Wikipedia category for that. But that is slow and unreliable. In more complicated cases having to recursively navigate categories can get tricky (e.g. there can be loops). Querying Wikidata is fast and reliable in comparison. Want people born in 1938 that are still alive? That’s a one line change to the query. Doing that with Wikipedia suddenly becomes much harder.

Good luck

So, those are the main lessons I've learned. While Wikipedia seems like a perfect, endless source of content, a good WikiGame depends on navigating its quirks. Hopefully, this saves you some time on your own project. Good luck!