One of my objectives with the Linux upgrade I did a few weeks ago was to get the Python Imaging Library to work. On the old Linux (Mandrake 10) I had to build it myself and that wasn't working. On the new Linux (Ubuntu Edgy Eft) it comes preinstalled. Anyway, now that I have PIL I can modify the stats code to generate images, so I've been doing that yesterday and today (as well as getting the tables working nicely).
The first image is a scatter plot of your ratings for games against the number of times you've played them. Ideally you'd like to have played your 10s many times and your 1s hardly ever, but factors conspire to make that not the case. For example, you might have to play Sorry against your 3 year old daughter hundreds of times. In my case, I've played Lost Cities which I rate a 7 against several non-gamers.
Here's how to read the graph. Number of plays is the Y-axis. More than 25 plays counts as 25. Your rating is along the X-axis, rounded to the nearest 0.5. A red circle is for games you've recorded 0 plays of, and a cyan circle is for games you've recorded 10+ plays of (yes, the colours don't mean much for this graph). A bigger circle means more games have that rating and that number of plays.
My graph indicates that I've played lots of games with various ratings only once. Let's examine that by doing a histogram of number of plays. The Y-axis is again number of plays (capped at 25), and the X-axis is the number of games you've played that many times.
My graph shows I have far too many unplayed games, but even more games that I've played only once. My efforts to increase my Friendless metric are whittling away the unplayed, and once I'm done with that I'll start working on the 1s as well.
These graphs are prepared for all subscribers to the stats. The first graph appears on the front page of your stats, and the second graph appears on the Plays of Games Owned page.