For a self-identified "data scientist", this is a pretty naive approach to "finding out what worked for others". The methodology doesn't distinguish in any way between "colors that help make apps successful" and "colors that are just used often by everybody".
It could be that there really are some colors that will help an app to be successful in the app store, but without comparing the data from successful apps with that from a control group, there is no way to identify them. If the goal is simply to make an icon that wouldn't look out of place in the top 300, this is a good formula. But that doesn't strike me as a very interesting problem.
I'm the "self-identified data scientist". I agree that there are quite a few holes in my analysis, but I was not aiming for an actual increase in sales just by designing the icon this way. I however wanted to have an icon which looked good enough and could blend in, at least from a color point of view.
So I guess it might not be an interesting problem to you, but I actually had quite some fun thinking about it. Which is the most important for a side project, right? ;)
Confused about the blending in part. One of my biggest gripes is continually missing an app icon because it's blue (or otherwise similar to another, but generally blue). We just had a conversation in a similar thread the other day - "Make Your App Icons Boring" - http://news.ycombinator.com/item?id=4219164
You also don't show the icon at the small size that we'd see on our devices. Does it translate well?
As someone else commented, your title is a bit misleading. Perhaps try the same analysis for the bottom or a random subset of 300 apps and see what comes out. Dataporn can be nice, but you should be a little more rigorous with your statements.
My guess is that while that might help with the initial evaluation of the app, and possibly even smooth the initial sale (familiarity and all that), looking like other icons is a net negative for continued usage for your app.
I routinely try to check in with Foursquare on my phone and accidentally open Facebook, because their icons are similarly colored. Meanwhile, the "Words with Friends" icon is not something that could be easily confused with anything else, and I've never had a problem finding / using it.
Once your brand is established, those problems probably go away, but I know I've dropped a number of apps before because I wasn't using them because they were too difficult to easily identify in a menu lineup.
It could be that there really are some colors that will help an app to be successful in the app store, but without comparing the data from successful apps with that from a control group, there is no way to identify them. If the goal is simply to make an icon that wouldn't look out of place in the top 300, this is a good formula. But that doesn't strike me as a very interesting problem.