The term “line ball” is sometimes used to indicate an argument that could go either way. In tennis, though, any part of a ball that does hit any part of the line, is “in.” At the highest level of tennis, players flirt with the lines all the time, and there are many “line balls,” or, as they now seem to be called, close calls. The latter I discovered when watching broadcasts of the Australian Open matches. I also discovered, to my great surprise, that there were no longer any linesmen, or even any lineswomen. The Hawk-Eye system is automatically making the calls. And if anyone has questions about the call, there is an instant replay of the bounce of the ball, and super closeups of the impact mark of the ball on the court. Isn’t technology wonderful?
Looking for information about Hawk-Eye turns up these sources, among surprisingly few others:
and the wiki entry, Hawk-Eye.
The replay in “instant replay,” is not, strictly speaking, a “replay.” It’s actually a “preplay.” There are 10 (in some reports 12) cameras in the Hawk-Eye system. None of them are anywhere near the lines. The cameras are arrayed around the court in high positions looking down on the play. They record at up to 340 frames per second, and feed that information to the associated computers. [What they are looking at is the ball strike and the subsequent trajectory of the ball. On that basis, the entire trajectory of the ball, including the bounce, is predicted. The ball could, in fact, be called “out” before it crosses the net.] As was pointed out in comments, the previous sentence is, as far as can be determined from inadequate documentation, incorrect. It is only the ball tracking on the approach to the bounce that is taken into account. [See comment by Sancho Panzer.]
That explains why Hawk-Eye never shows any actual footage of the ball in the vicinity of the line. It’s a simulation. This will be obvious to cricket lovers. Hawk-Eye for tennis is a development of the original product for cricket. Those balls do not go through the batman’s leg and on to, or past, the wicket.
For something upon which so much rests, there is a paucity of public information about the system. The software is proprietary, and not even the mathematical modelling has been revealed. Not many questions about calibration seem to have been answered, and things like the number of cameras differ in different reports. In an article in New Statesman from 2015, Harry Collins discussed the accuracy of the system, and the secrecy surrounding it.
So we telephoned the firm to talk about [accuracy] and we hit a wall. As sociologists of science we had spent decades chatting with scientists about this kind of thing but suddenly we were told this information was private and lawyers were on call.
Hawk-Eye had been creeping in through the outside courts for a few years, but in 2021 Covid-19 was the rationale for its full-scale introduction across all matches in both the Australian and US Opens. Think how many Covid deaths were prevented by getting rid of all those linesmen around the court.
Let’s suppose that there is a lot of money to be won or lost in betting on the results of major tennis matches. Now imagine that some disreputable but technically adept persons manage to gain access to the software of the system. These nefarious people introduce into the system a bias factor that can be invoked from the control system, or better yet, remotely. The bias is only applied to very close calls to reverse them, with the direction of the reversal depending, of course, on the player.
Who would notice? Players are being conditioned to mistrust their own perceptions and go with the Hawk-Eye call. For example, at the Aussie Open in 2021, Dominic Thiem said, “If the electronic call is out, the ball is out, so there’s no room for mistakes. I like it.” Spectators, live and remote, are shown the “definitive” animation almost immediately. A major match between the world’s top players will often come down to matters of millimetres, and those are the very players who are aiming close to the lines more often.
One of the videos referred to above mentions International Tennis Federation testing of the system in 2006. “Results showed the system to have a mean error of only 2.6mm when compared to a high-speed camera located on the playing surface.” So why not just use the high-speed camera? A competitor, Foxtenn, has done just that. Their high-speed cameras watch the lines, and their replays feature not just the simulation of point of impact, but the actual footage of the ball. But the imagery is coming from a computer, so anything is possible.
That’s the general problem. The world is presented to us as digital information mediated by computers, and all the virtue can be hacked out of the virtual. Once you accept that you must take on faith the video imagery that you see, just as you must the digital photo images, destination Dystopia is much closer.
19 thoughts on “How To Cheat At Tennis”
For something upon which so much rests, there is a paucity of public information about the system. The software is proprietary, and not even the mathematical modelling has been revealed. Not many questions about calibration seem to have been answered, and things like the number of cameras differ in different reports.
Also could be said of those US voting machines.
On Hawkeye, for years the other half has disputed its use in cricket – saying that once the ball hits the ground it could go anywhere, given that the wicket surface deteriorates.
Are human linesmen any less likely to be involved in rigging a call? As (I think Dr Turf) said: “Don’t bet on anything that can talk”.
I like the Foxtenn system idea better. Actual measurement beats estimation.
In these days of PhotoShop and computer generated (or altered) imagery, it is no longer possible to say that “the photo never lies”.
You forgot the Aussie special:
Ban the best from playing as ‘not vaxxed’
What a lovely phrase
We are already conditioned to believe women are men and men are women (and millions do accept this without question). What is a little more conditioning here or there?
Never believe those lying eyes of yours, always accept what you are told, and obey.
I’m sure anything that is “modelled” is going to get a lot more scrutiny over the coming years.
It has been my opinion for years that cricket’s Hawkeye is complete and utter bullshit.
There are so many unpredictables once a cricket ball strikes the pitch that trying to then accurately track the ball is nothing more than electronic voodoo.
Speed, spin, swing, seam condition, moisture, pitch condition, wind, etc etc.
Not all the electronic trickery is useless though. The HotSpot and Snicko are excellent umpire aids as is the slow motion replay of run out and stumping appeals.
The unfortunate part of promoting all this wizardry is that Grade cricketers (and umpires) never see any of it on the ground and must rely on human decisions.
All the times I was out lbw or caught behind. Brilliant centuries denied me because no hawkeye.
Err, I think this post is based on fallacious assumptions.
Hawkeye does, in fact, track the ball’s trajectory accurately and uses that data to track flight. The video implies that it takes a snapshot off the racquet and works out the entire flightpath from that. The estimation algorithm comes into play in two areas:-
Firstly, in estimating the behaviour of the ball between frames. For a 200 kmh serve (55 mps) and at 340 frames per second, the ball will travel about 160mm between frames. So a ball might travel, say 6″ from the last frame before hitting the ground. Given the number of data points prior to the last frame, the prediction will be highly accurate.
Secondly, the behaviour of the ball where the ball contacts the ground. Could there be a difference between the compression behaviour of a brand new ball on a freezing cold day and a six game old ball on a sweltering hot day? Sure. Is it significant in measuring whether a ball is in or out? Not by more than a couple of millimetres which is accurate enough for this purpose, and far more accurate than the human eye.
The argument about prediction using Hawkeye is relevant when talking about LBW in cricket, but totally insignificant when talking about tennis.
No tennis player disputes a line call made by Hawkeye, because they know it is highly accurate.
Why do they show a simulation rather than raw footage? Because the essence of the system is triangulation which is hard to show on TV and would be dull as dishwater for TV audiences.
I wonder about the claimed 2.6mm accuracy. Is that one, two or three or more Standard deviations?
I’ll bet there is a Kalman filter lurking in their algorithms which is fine IF you have really good physical model. I suspect that even on a tennis court there are uncontrolled physical variables.
Dunno. It tracks the ball before and after it strikes the pitch, so it is able to track the post-stike flight path. I suppose the accuracy might be dependent on how full the ball was.
For full toss lbw decisions, I believe they use the original flight path and disregard any effect the pitch may have. In the case of a ball likely to stike the pitch between hitting the pads and the wicket, it may be less accurate, particularly on a spinning pitch.
You can satisfy your hankering for the old fashioned honesty system still operating in club tennis. Many of those midweek ladies have the competitive streak equal to that of any world No. 1 and deteriorating eyesight. Its never a pleasant outcome for their opposition.
Bring on the virtual, I say.
I wonder how John McEnroe would have dealt with this.
@ Sancho Panzer
Thanks for that. Where are you getting this information?
My wording was inadequate. I did not mean to imply that is was only the strike, but the strike and a series of position points following the trajectory. However, if the predictive element is only for predicting the court contact over a trajectory range of less than 160mm, the chances of actual physical error are minimised. And, in that case, the claimed 2.6mm error is triangulation error. I’m with Eyrie, too, in asking what the variance was.
(Foxtenn’s claims of superior accuracy are based on line cameras (40 all up) running at 2,500fps, equivalent to 22mm at 200kph.)
However, all that aside, the fact that we are never shown the actual images from the triangulating Hawk-Eye cameras leaves the system open to the possibility of manipulation as I indicated.
And the actual LBW law says just that. The umpire must assume the ball would not deviate even if it would likely have bounced between striking the batterer and reaching the stumps.
You cannot be serious!!!
January 30, 2023 at 4:20 pm
I wonder how John McEnroe would have dealt with this.