This paper was published in 2014. Here are links to previous discussions on HN (from the same year) that might be interesting. It seems the employed methodology might have been biased.
* https://news.ycombinator.com/item?id=8797002
* https://news.ycombinator.com/item?id=8807497