I wonder if a lot of people have bought Starfinder, but haven't had the time to digest the rules enough to run a game of it yet (as of 2017). I guess the numbers for next quarter will tell.Also, I notice that for most systems there's about 4 times (give or take) as many players as there are games, but not for D&D5 - almost one game per player, about 5 times as many games as for Pathfinder. I don't know what to make of that. Lots of 'test' games maybe? 'Cause Starfinder has a similar ratio.
The Roll20 stats can be misleading in a lot of ways. I can set up a game on Roll20 and never actually play it. Moving beyond D&D and Pathfinder, it becomes even less useful as a metric. Many games, such as Call of Cthulhu, are more likely to be played theater of the mind, so online games can just use straight video chat with players rolling their own dice. The same goes for a lot of older versions of D&D. You also have the fact that some virtual tabletops are better at supporting than other. Roll20 doesn't provide the same kind of strong support for Savage Worlds that Fantasy Grounds does, to give one example. so Savage Worlds numbers will always be low on Roll20. I find the results from Roll20 interesting, but they aren't a completely accurate snapshot of what is being played.