Back when I played a lot of original Half-Life online (robots versus scientists on Gasworks), I got real good at anticipating lag. On a server with 200 ping, you would click to fire your grenade or crossbow rail, and then, a fifth of a second later, it would actually do that. If you couldn’t lead your shots correctly or calculate the time until a rocket would hit, you’d be in pieces mighty quick. They’ve since changed how netcode handles that stuff, but it’s interesting to see that controller lag is still at large despite everything else having advanced so far.
Eurogamer’s Digital Foundry decided to take a semi-scientific approach to find out which games were the slowest to respond to player input. The minimum time for a game to respond turns out to be machine-limited to three frames, or 50ms. That’s an eternity for PC gamers inured to the <10ms response times of 1000Hz mice and gaming LCDs. Sadly, most games ended up having double or triple that, which is pretty sad in my opinion. The method they use seems suspect, though; they really should be using a more high-speed camera than a Zi6.
It sounds like gamer uber-pedantry, but really, a 100ms difference in input can make all the difference between the controls of a car feeling snappy or mushy. Or, say, a fighting game — why should your character's reflexes be so much worse than your own? In the end, everyone is playing under the same handicap, so it comes to nothing, but who hasn't raged thus after a hotly-contested race or battle: "What the hell?! I pressed the button but it didn't go!"
I'd be interested to see how an SNES or something else fares. I remember the controls for Super Mario Kart and Street Fighter II being pretty damn responsive, but maybe that's just nostalgia.