Hi guys, Sergey Galyonkin here. Thanks for using SteamSpy and thanks for letting me it, I've been reading NeoGAF for years
I just want to clarify some things.
Effects of free weekends on owners data are worse than you'd think.
Steam works this way.
1. Game goes on free weekend.
2. Everyone online during the weekend get the game added to his account - it's around 25M users.
3. The promotion ends.
4. Only people that are currently online will get the game removed from their accounts.
Because of this
free weekend might affect data for several months.
Check out Europa Universalis IV. It sold around 700K, but it shows 1.4M because of free weekend a full month ago!
There is no SteamGate
Guys that contacted me thought that Steam Spy shows Steam sales. It's not. It shows owners of the game on Steam and that particular game had a lot of retail sales and has been in at least one bundle.
Geo data is self-reported
If you have set a location in your profile - great, Steam Spy will use it. If you did not - also great, Steam Spy won't use it. But if you lied - we have a problem.
Fortunately people don't lie that often, but it still affects geo data on Steam Spy.
And, most importantly
Owners!=Sales
You could get a game in a bundle, as a gift or try it on free weekend and never log in.
Some technical data
I'm collecting roughly 150 profiles per minute, but because half of those are empty it gives me around 100K of valid open profiles per day. It turns to roughly 1M data points (games per user).
Today I've made some changes to algorithm and should be able to collect 150K user profiles. Or server will meltdown at midnight
I'm using a rolling sample for last three days, that's why data is a bit lagging.
Anything below 30K frankly isn't reliable - be it owners data or players data.
But on the other hand AppAnnie might be off by 200% to 700% and people still pay tens of thousands of dollars to access it.