What is Survival Analysis?

The morbid world of player death

This is a non-technical introduction to Survival Analysis, which is a term that most of my readers won’t have heard before, so I’m going to sell you its usefulness first.

Players of games eventually ‘die’. Of course, they all die literally (fun to think about, I know!) but they all die figuratively. To be plain, every player will eventually stop playing every game that they start playing and they’ll never come back. Fact of life. Too bad.

In the games industry, we care deeply about this. The main stakeholders for this include:

• Live Services: When will our current players die?

• Game Design: Will players die before they reach the end credits?

• Subscriptions: When will players forcibly kill their subscriptions?

  • In this case, the player can die all they want, as long as they don’t touch their subscription packages!

  • Are you still subscribed to stuff you don’t use? I’d check now!

• Staffing: When will your employees leave your company?

  • They eventually will leave one way or another, wouldn’t it be good to have an idea when this will happen?

Survival Analysis techniques were originally developed for real-world clinical trials, but there are two main issues that real-world death and player death both share:

• We should not and can not wait for everyone to die before we make predictions about death in general.

  • Our analysis is always based on a snapshot of the present.

  • We want to intervene to keep players alive while they are living!

  • If all the players are dead, our game is dead, which is too late.

• We don’t know for sure if a given player is dead at any given time.

  • Just because a player didn’t log in today, doesn’t mean they are gone for good.

  • Just because a player *did* log on today, doesn’t mean it wasn’t their last time!

If you want to fall down a rabbit hole with a real-world example, check out this Survival Analysis of World of Warcraft players on GitHub!

See you next time, GMDQ’ers!