Instant replay in tennis

This afternoon, I tuned in for the last two games of the Wimbledon semifinal between Andy Murray and Jo-Willy Tsonga. Andy Murray won the match, making him the first British male finalist at Wimbledon in 74 years (no pressure, Andy!).

The winning point was a bit strange, however. Murray hit an aggressive shot that was a tough call for the umpire. The crowd exploded, but then quickly realized the ump called it out. No matter. Murray put a finger in the air and challenged the shot. Under tennis replay rules, players can challenge in or out calls, and the disputes are decided by the Hawk-Eye ball tracking system. Hawk-Eye uses high frame rate cameras to map the trajectory of the ball. The Hawk-Eye trajectory showed the ball nicking the outside of the line, and Murray was through to the finals!

I asked my buddy Tony, a staunch replay opponent, whether he would have felt bad if Murray had lost the match in the absence of replay review. He said no, because he guessed the tennis replay system faces precision issues. This is his standard “blades of grass” complaint. On extremely close plays, there is no way to tell whether a ball (or player, in other sports) is in or out, because it comes down to the deformation of the ball on impact, individual blades of grass, and lack of precision in painted lines.

I set about to prove him wrong, under the premise that a computer-driven system like Hawk-Eye would have definable precision. In one way, I was right. Hawk-Eye’s average error is 3.6 mm. But if that’s the case, why is it used to make calls that are well within it’s margin of error? Once a challenge is made, it doesn’t matter how close the ball is; Hawk-Eye makes a final decision one way or the other. Murray’s final shot wasn’t in by much more than 3 mm, and now I wonder if we should feel bad for Tsonga.

The tennis situation mirrors replay issues in many other sports. I’m a strong replay review supporter, but I think that “indisputable evidence” (or “incontrovertible,” or whatever) should be an important part of any replay system. Replay should be used to avoid big mistakes, not decide impossible calls. The challenge, then, is not going overboard as in tennis. Tony’s argument is that we will always go overboard and start deciding indeterminate calls on weak evidence.

There’s a potential solution for tennis (and soccer, as it adopts goal line review technology). Right now, Hawk-Eye shows only the most likely location of the ball. Instead, it should show the most likely location of the ball with a reasonable error range shaded around it — the area that Hawk-Eye projects the ball landed in with 99% certainty, or 99.9% certainty, for example. If the error range is too big to make the call, the original call should stand.

Note: I forgot to mention, Adrian the Canadian has discussed the Hawk-Eye technology on this blog before, albeit from a different angle — it’s use in cricket. In cricket, it is used for a judgment call (Leg Before Wicket). Replay review usually steers clear of judgment calls, but the Hawk-Eye seems to be going over very well.

16 responses to “Instant replay in tennis

  1. Interesting post on an interesting issue. One thing about Hawk Eye is that it is unbiased, which is a virtue in and of itself. If it’s more accurate than the umpire, and less likely to have its calls skew one way or another, I think it’s probably, on balance, a good thing.

    • Indeed – I agree that it does more good than harm. However, it seems odd that they would use it to change calls where it’s no better than 50-50 (and they seem to know when this is the case).

  2. So how would this be implemented this in practice? The hawkeye system gets called up, the fans start clapping, the screen zooms in, and then there’s a big question mark?

    The simplicity and speed of the current system is a virtue. It barely interrupts the flow of play, certainly far less than arguing with a chair does, certainly far less than any other sport that uses replay.

    Now for these close calls where hawkeye has an error range, is there any reason to think the best human eye would be better? That seems to be implicit in the conclusion. Otherwise who cares how you flip the coin?

  3. I take your point, but I think there’s something to be said for letting the human flip the coin rather than the computer. Or, at least I don’t like the fact the computer presents its result as certainty when it’s clearly not. There may be calls where the chair/linesman has a great view but it’s inside Hawk-Eye’s margin for error as well, so I’m not sure it’s always an even trade.

    I personally love the question mark idea, but maybe “Too Close – Call Stands” would be better for the general fan?

  4. The Dinosaur Hunter

    Your buddy Tony sounds like a lot of guys I know. If he’s anything like my friends who object to replay, it’s just because he’s a huge baseball fan and has some inconsistent and unsupportable purist hang-ups he can’t get over. Ironically, my friends who oppose replay are huge proponents of optimization in every other sphere of their lives. They’re always optimizing everything. Maybe if we called it electronic officiating instead of instant replay, these people would get on board. If a decent process is put in place (as in tennis), electronic officiating takes less time in most sports than officials conferring anyways. If computers can drive cars better than humans can drive them, you would think they would be better at calling balls and strikes too!

    • I agree – the process is extremely important for the success of replay. Also, I agree that computers can do a generally better job of calling balls and strikes or making in/out calls, but it’s a difficult philosophical hurdle to get over that computers can admit to you when a call is just a toss up (from their perspective). We want to believe that the ump can always see the right answer, even though that’s not true.

      If Hawk-Eye were used for every call, I think I would have less trouble with it deciding 50/50s. When it’s used only for a few close calls, and the errors aren’t well explained or understood, I’m not sure how much value it adds. In my post, I tried to suggest a method where it would clearly improve things at least.

  5. Also, as just evidenced moments ago, Hawk-eye doesn’t result in the award of a point if there isn’t a clean winner. Federer just hit a marginal ball that was originally called out, he challenged the call, it was in (likely w/in the MOE), and the point was replayed. This makes a lot of sense.

    • Adrian, there’s definitely a clarification that I should have made. Hawk-Eye only finalizes the point if the shot is found to be out or if the shot is clearly not returnable. If the shot was returnable and called out, and Hawk-Eye says the shot was in, then the point is replayed, since the returner may have been able to return the ball.

      However, this is different from the issue I’m raising I think. In the case you mention, the point is replayed not because the shot was within the MOE but because Murray had a chance to return it (and did not, since it was originally called out). If Fed’s shot had been unreturnable (as Murray’s was on match point in the semifinal), then Hawk-Eye would have given the point to Fed.

      However, I might be okay with replaying all challenges that fall within the MOE — including winners. I think this is what you have in mind too.

  6. Yes, I agree with that (my post, as you noted, implied something it shouldn’t have). Though I think there are practical reasons for not replaying all points w/in the MOE. At the end of the day, for efficiency’s sake, we probably need some clear rule to determine a close, unreturnable point without re-playing every close, unreturnable point. The hawk-eye is completely neutral and unbiased, even the best umpire is not.

    • Yes, I don’t think that’s unreasonable. However, it’s psychologically tough on players when the call was made one way and then it’s overturned on weak evidence. Since the confidence isn’t displayed with each Hawk-Eye projection, I’ve also gotten nervous now about how good it actually is (and thus, how unbiased). The manufacturer of course says it’s great, but there seem to be some contradictory views and not a lot of tests done publicly.

  7. That’s interesting. It’s an argument I’ve had (here and in other places) about whether the appropriate standard of review for replay should be “incontrovertible evidence” or something less stringent like “most likely in error”. It’s a bigger problem in football than tennis.

    From a psychological perspective, it seems that hawk-eye has decreased the number of player blow-ups. They seem more likely to accept the hawk-eye’s version of the facts as correct. Perhaps it’s because it’s a computer, or perhaps it’s because they don’t realize the problems with it, but players do seem to accept its judgment. It would be interesting to see how players would react to a heightened standard of review – something like “in or out AND outside of the hawk-eye’s MOE”.

  8. If the PTA wants to do instant replay/measuring, I feel like they can get to a higher level of precision by instrumenting the court. Maybe pressure strips or infrared beams (see if the ball crosses the beam)? Perhaps a Large Hadron Collider on the line and see how many Higgs-Boson particles are displaced?

    I also second Adrian’s psychological point– I feel like Hawk-Eye saying “I don’t know” is a bad thing. As long as it gives a judgement, if it’s within 3 mm, no one is going to argue that the call was wrong and the players won’t have a problem with the call.

  9. 3.6 mm is far better than a human linesperson’s margin of error. that is why they use it on calls within that MOE.

    • Sorry, but reading the rest of your silly blog shows you are just ignorant of math/statistics. Why should a call that has a margin of error over maybe 20-30 mm (the linesperson’s call) be allowed to stand if a system with a margin of error of 3.6 mm suggests that it was most likely the opposite call of the human?

      Whether it is tennis calls, or manufacturing safety equipment, you use the system with the lowest margin of error.

      And if you are going to complain about that, why not whine about how football bases first downs on a length of chains that are exactly 10 yards apart…but that and then run at least halfway across the field and eyeballed as to where to put the line of scrimmage chain? Why do they do it? Because it eliminates any uncertainty. Yes, the may be a few wrong calls, but they will tend to even out. Just like tennis. And you don’t want ot go through the whole replay process omnly to have players and fans still left with no resolution.

      • Have a look at DRDR’s comment above and a couple others (and my replies). Perhaps the repeated comment says my suggestion wouldn’t be very popular, but I will say that the point is philosophical as much as mathematical. It’s about who gets to flip the coin, computer or human. There are many situations where the most precise tool is not used (e.g., GPS ball location in soccer, ball and strike calling in baseball), with reasoning based on cost, available technology, other impacts on the game, and, most importantly, philosophy about how the game should be refereed.

        I think my main issue is simply that the uncertainty is unstated. There is a perceived elimination of uncertainty, but uncertainty is not actually “eliminated.” In many cases, the linesman might be more accurate. This is largely untested, to my knowledge, even if it’s not true most of the time. Linesman would probably be offended by your claim of 20-30mm error, though.

        Your analogy to football is interesting. I DO hate the arbitrary precision of the first down chains, and I think most fans roll their eyes when the chains come out. I’m surprised that you don’t argue for the yellow line to become official or hawkeye-like technology used to determine where the ball should be spotted. At minimum, the yellow line would be easily implementable and probably more precise than referees (in most cases).

        However, I have a different proposal, which I was planning to post soon actually. Rather than try to fix the measurement problem (with spotting the ball being the real issue), why not change the structure? Instead of four chances to get ten yards, give teams f(X) chances to get X yards, where X is the yards left to the end zone. I’m tempted to do simple division – e.g., f(X) = X/6 or maybe X/7. So if you get the ball on your 20 (80 yards to go), you get 13 (or 11 if using X/7) tries to score a touchdown. In that case, only one or two spots really matter: the first one and the one to score a TD (if you make it).

        This might lower scoring, since teams could grind out more field on offense and bury the other team deep more often, but I’m not sure that would actually happen, since teams might also get more FG chances. You would have more long drives for sure (and fewer commercial breaks). This rule change would have another big advantage: lowering the stakes of inches on most downs would reduce the incentive to really smash guys, hopefully cutting down on violent collisions.

        A nice complementary rule change would be to decrease the play clock, which would make these long drives go faster and look more like one continuous play (as in rugby, the sport I played). Bringing Chip Kelly to the NFL is a good step in this direction. Again, not only does this improve the flow of the game potentially, but it also would encourage players to lose weight to keep up, which might improve safety (unclear of course, since force is mass times acceleration, and lighter players generally accelerate faster).

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s