For anybody in any doubt that India now runs world cricket, the fact that the “Decision Review System” is not being used in the current Test series against Australia is instructive. Alone amongst the cricketing nations, India objects to the use of “hot-spot” infra-red cameras, and the “Hawk-Eye” ball-tracking system. But India’s objection was enough – the International Cricket Council reversed its earlier decision to make the system mandatory for Test matches, and left it to the agreement of the participating teams.
However, there’s a more important question here – are the Indian cricketers correct? Does the technology actually work acceptably well?
Putting the case for the negative is Greg Baum in the Fairfax press today, citing a couple of cases where Hawk-eye’s predictions were inaccurate – the case of Phil Hughes in Sri Lanka. In this case, the developers of Hawk-Eye acknowledged a mistake was made, as the ball had not traveled the minimum distance (40cm) after pitching to allow Hawkeye to make a sufficiently accurate prediction of the ball’s trajectory.
For what it’s worth, it’s my strong suspicion that Hawk-eye gets it right far, far, more often than umpires do; furthermore, it has improved the game by correcting some of the more egregious umpiring errors. But my confidence in saying so has to be somewhat limited, because the detailed workings of Hawk-Eye remain proprietary, as do the testing protocols used by the ICC and the makers to demonstrate its accuracy.
The key unknown for me is how Hawkeye’s software calculates the flight of the ball after impact with the pads. Does it assume a “generic” cricket ball, and, if so, do the trajectories of a brand-new, old, or damp ball differ through the air enough to matter? How is “swing” modeled – do they assume a constant angular deviation? Does the system routinely perform any consistency checking, by making predictions based on the ball’s early trajectory and comparing them to the actual track of the ball?
My guess is that the Hawk-Eye system, over the ten years that it has been developed, has had most of the bugs ironed out of it and deals with the issues raised above in a reasonable manner. But we just don’t know. The most detailed public description of Hawk-Eye’s technology is this document from the company, which clarifies some issues – notably that Hawk-Eye’s LBW predictions are based only on what happens after the ball bounces – but doesn’t deal with the questions I’ve posed above.
Using the Decision Review System means that Hawk-Eye’s algorithms are, effectively, part of the laws by which international cricket is played. As such, they should be public, and the testing by which their accuracy has been established should also be public, so it can be subject to external review. If Hawk-Eye’s intellectual property needs protecting, they have recourse to patent law. Or, for that matter, the ICC could simply use a few million of the billions of dollars of Indian TV revenue to buy out the IP rights.
But it’s just not acceptable to have cricket matches decided by algorithms that approximate the motion of a cricket ball, when we don’t even know what those algorithms are.
