[lossolo]

There was an article some time ago by someone from Chrome/google in which you could read that they optimized for benchmarks before but benchmarks are not real world usage so they stopped and started to optimize for real world usage. They did it because they optimized for some benchmarks while making real world average usage less efficient. Benchmarks do not show real browser performance.

[fastball]

Ah, thanks for that, it seems you are spot on.

https://blog.chromium.org/2017/04/real-world-javascript-perf...

I actually just ran their "real-world" test (http://browserbench.org/Speedometer/), and the results were:

  Chrome 62: 39.6 ± 3.4 rpm
  Firefox 59: 27.0 ± 5.6 rpm

So there may definitely be something to this theory. Would be good to have a larger sample size though.