These are only a couple of points. You could add so much to it... srcset, http/2, minification etc.
Saying that, why create the post at all? I don't mean that in a harsh way, I mean to say that if you are testing with google page speed, it tells you exactly what you can improve on and how to do it!
I recommend to use Crunch for PNG compression. The result is so impressive I don't see myself using JPG again.
I appreciate the effort but mostly I don't see the point for sites that are already scoring relatively high (80+). It feels masturbatory.
High level optimisation is masturbatory?!
I dig that there's a diminishing return at a point above 80ish but if you've got the time or resource to do it, I don't really see a downside.
Absolutely. Unless you are getting enormous amounts of traffic, which the vast majority of sites do not, then what you should be optimizing is different. I would say focusing on doing split traffic tests, focused on layouts and optimizing for flow is waaaaay more important than shaving 0.25s off your loading time.
I partially agree, but on the other hand — 0.25s is quite a lot.
And, keep in mind that reducing load time by 0.25s for a decent network/browser will be an even larger (and noticeable) win for slower networks and/or devices.
I too am used to super-fast network connections, so when travelling somewhere with crappier connectivity, everything seems to take forever to load. The finest A/B tested copy and button colors will not help you if I close the page because I had to stare at a blank loading state for too long :)
I'm curious – why is there diminishing returns above 80ish?
What so many of those optimization guides really miss is that it is also about making content choices. If your visual direction is going to be image heavy, let alone video, or you have long pages with much varied content, that level of optimization is often not your main concern in a quest to get better load performance.