The N types of gamers (Clustering analysis, revisited)
- Matthew Gray(mkgray)United States
Since then, many more people have created top 10 lists, and I've refined the analysis a little. On natural question is "Why 10?". Well, 10 actually yields a pretty good clustering, but there are techniques to try to pick more optimal values. These techniques involve seeing how much improvement you get for each additional cluster. You stop at the point where improvements start diminishing. Well, with gamers that's at about 10, hence why I picked that last time. This time though, I looked a bit more carefully.
There are basically 3 little plateaus. At 3 clusters, you get a substantial improvement over 2. To get the next big "bump" of improvement, you have to go up to 5 clusters, and to get the remaining bump, you have to go up to 9 clusters. A 10th cluster, helps a little, but not enough to justify it really. So, 3, 5 and 9 are the most statistically significant groupings.
Now, this is a clustering of gamers, but for convenience, I'm going to talk a lot about which cluster a game belongs to or leans toward. What this really means is that gamers who like that game tend to be in that cluster. So "ASL" isn't in a particular cluster, but the gamers who like ASL are. Hence, it's more compact to say "ASL is in this cluster", even though that's not quite right. Deal.
In the previous iteration, I named all of the clusters with descriptive names, and that was somewhat controversial. I'm giving the clusters less controversial names here and folks can discuss what, if any, label is appropriate for a cluster.
So, there are 3 partitionings: 3-cluster, 5-cluster, and 9-cluster groups. I picked generic-ish names for each of the clusters within those partitionings. The percentages are of the roughly 3500 people who have completed top 10 lists.
Core Eurogamers (62%)
Family Eurogamers (27%)
Euro Alpha (30%)
Euro Beta (20%)
Euro Gamma (16%)
Euro Delta (9%)
Euro Epsilon (6%)
Euro Zeta (6%)
Ameri Alpha (7%)
Ameri Beta (4%)
The game listed for each cluster is the single game that is most likely to be in the top 10 of members of that cluster. The "Other Games" are other games most likely to appear in the top 10 lists of members of that cluster
I'll try to get the tool up where you can classify yourself according to each partitioning sometime soon. Personally, I'm Core Eurogamer/Eurogamer-1/Euro Alpha
Feel free to add new games to the list, and I'll post info about which clusters they make prominent appearances. I may also delete some entries, or not post about some if it gets too voluminous, or the particular entries don't make any prominent appearances in the clusters.
- [+] Dice rolls