The Uncertain Eye

Tuesday 28th June 2011

As I was watching the match last night between Nadal and Del Potro, I started to wonder just how accurate the Hawk-Eye system is. Turns out that Hawk-Eye reckon that the average error of the system is 3.6mm. Which is actually pretty good, but is it good enough?

For most calls, yes, probably. When it’s clear that the ball landed on or off the line, the system is probably adequate; a few millimetres won’t change the call. But we often see it showing really close calls, where a few millimetres would make a difference. But as far as I can tell, in those cases the system is still trusted, the uncertainty of the measurement isn’t taken into account. Which seems a bit bizarre, really.

So given that for close calls Hawk-Eye isn’t accurate enough, should it be trusted as much as it is? Perhaps not. Or at the very least, the margin of error of the calculation should be displayed whenever the system is used.

On the Hawk-Eye website, there are several documents which look at controversial Hawk-Eye calls. For example, from the 2007 Wimbledon final where Hawk-Eye showed a shot to be in by 1mm (3.6mm uncertainty, remember), which looked from the camera to be out. Here’s a short clip:

The manufacturers recreated this shot and filmed it with high-speed cameras, to show that Hawk-Eye might not have been wrong in this instance. But that seems to me to be a bit backward; why not just use the high-speed cameras to check line calls in the matches? Then there would be a definitive record of what the ball actually did, rather than a back-calculated approximation of its movement.

Posted at 3:03 pm | Posted In: Geek Tagged:



Thursday 30th June 2011, 12:02 pm

It’s probably much more accurate than a human line-judge, though. In that sense, a bad Hawk-Eye call is still probably preferable to a bad human call.

I suppose they could change the rules so that if the call is within the margin of error that it be called indeterminate and they play the point again, but that seems somehow even less satisfying.


Friday 1st July 2011, 11:23 pm

Yeah, they could change the rules to take account of the margin of error, but as you say that’s a pretty unsatisfactory solution. Especially when compared to the one I suggested at the end of the post…

Write a comment: