installations probably exceed downloads
The actual number of installed instances is probably somewhat higher the number of downloads as some people - myself, for example - downloaded it once but installed it on multiple machines.
@statistics - It's speculated that users of Windows 7 RC1 have been more likely to try MSE out than people running XP or Vista that already have an AV running on them.
Arstechnica had an interesting article about a false positive MSE flagged that was generated by a custom script written by one of its forum contributors, which indicates that MSE looks for suspicious code in addition to suspicious behavior.
The script in question was a quick .vbs hack (admitted its author), some sort of logging script, that wrote its data to itself and then modify it (rather than write the data to a separate data file and modify it there, which is best accepted practice apparently), a common feature of malware, though this particular script had no malicious purpose. Changing the extension from .vbs to .txt, MSE still flagged it as malware, reinforcing the impression that MSE was examining code text rather than just code behavior.
They ran both versions of the script through that testing website that runs code through 17 (or is it 23?) different virus scanners and MSE was the only one that flagged the .vbs script, though Dr Web flagged it in it's .txt form.
While the author of the Ars article took MSE to task for generating this false positive, the vast majority of the commenters disagreed, saying this was exactly the type of thing good AV software *should* be looking for and taking all the other AV vendors to task for missing it.