Hacker News Reader

From the Ars Technica link I posted:

dougg3 said his setup only successfully completed the level 1 to 5 percent of the time, an oddly low success rate for a system that's supposed to play with robotic precision.... "What I found is although the actual electrical pulses I'm sending to the controller are exactly the same length each time, verified on an oscilloscope, sometimes Mario doesn't jump," dougg3 said.

View on HN

wouln't that suggest some sort of additionaly applied poll rate?

Or a desync between the game and the display.

In Super Smash Bros Melee the game runs at 60 Hz but the display runs at 59.94 Hz... one frame in 1001 is simply not displayed.

Even if it’s perfectly synchronised, there’s likely no guarantee of deterministic playback. Perhaps a background process is dealing with some network activity or other OS task and throws the timing off fractionally. Or maybe a disk access (even for something in the background like streaming music) takes fractionally different amounts of time per run and affects the precise time at which the input is polled.

Ah, that’s another interesting possibility. I definitely observed that the actual HDMI refresh rate is 59.94 Hz while playing Super Mario Maker, but no idea what the game’s internal frame rate is.

Reply on news.ycombinator.com