Steam: Taking behavioral based profiles to the next level

Lots of folks talk about profiles and personas when trying to perform User Centered Design. Obviously these techniques can help designers (and usability folks) make better decisions and recruit relevant participants in order to verify that they are serving their client needs.

I love (love) the fact that Valve are providing so much game play data via their Steam website. Obviously the data aren't presented in as comprehensive a fashion as the folks do at Bungie or on the WoW stat site but it still provides a great service to players who want to study the game and improve.

Looking specifically at the Team Fortress 2 stats, you can begin to break down behavioral profiles that can potentially be unpacked to reveal what motivates different gamers. There's already a natural experiment in that the class based system allows for multiple gaming experiences. Take a look at the top 3 most played classes (Scout, Engy, and Soldier) and the 3 least played classes (Spy, Pyro, and Medic) and then take a look at the subsequent charts that reveal behavioral differences amongst the classes:

• Point scoring isn't the only form of motivation for players:
  
• Note that one of the most popular classes (Engy) is the lowest average point scoring class.
• Note that the 3 highest scoring classes (Sniper, Heavy, and Spy) are not amongst the most played classes.
• That said, Flag Captures are obviously hugely rewarding:
• Scouts are the most popular class, dominate in flag caps, but score the least overall kills and overall points.
• Assists (which contribute to total points) are not hugely rewarding:
• Note that Medics dominate the assists category, but they are the least favored class. This may be compounded by the fact that Medics seem to be getting ONLY assists and not any other points (even though they dominate assists, they are in the bottom tier of overall scoring).
• Engy's are, simply, a different breed:
• They are one of the most popular classes, yet score middle to lowest in almost all reported scoring categories.
  
• However, they do tend to live longer than all other classes. Maybe there are just some folks who like the slow plodding support function that the Engy class offers.

Of course, there are many structural variables that complicate these assessments. Game play is competitive and this means that teams generally have a set number of slots for each class (no one wants an all-Medic team) so there are some artificial constraints. There are no explicit "satisfaction" measures (e.g., reports on how "happy" or "sad" people are after playing a session as one particular class) and there are also no data on class monogamy/polygamy (e.g., which players select only one class, which players rotate through a few, and which players will play 'em all).


What I'd hope to unpack (if I were able to tap the data and add some additional instrumentation) was what motivated people to choose certain classes when they could freely decide? And, what can be done to make less desirable classes more fun for folks to play when they need to take one for the team in order to fill a session and get a game started? In other words:

• Tease out the core play style motivations and preferences in order to generate a nice set of classes and gameplay mechanics that will be welcoming to a wide diversity of gamers.
• Tease out the social reward structure (points, kudos, etc) that will encourage gamers to pick up and play and have a good time even if they need to play a less desired class.