Accuracy of timing in sports (Olympic Games)

[aro]

Hello all,

If you have been watching the olympic games lately you could have seen that in the past week it has occurred twice that an athlete won (and another lost) a gold medal by a time difference of a few 1000th of a second. Both occurrences where in speed skating (i.e. the 500m and 1500m men's events).

My question is: is it possible with state of the art electronics to measure the times so accurately? What equipment would be necessary for that? And what would be the errorbar on such a measurement?

Generally it is assumed that a high speed camera at the finish line that can record (and presumably time-stamp) easily more than a 1000 frames per second, ensures such accurate timing. But I would say there are many other factors to consider.

Also take into account that in some sports (i.e. 100m sprint final) contestants are compared within one race, so they hear the same start signal and the photofinish observes all runners in the same heat and measures relative time. However in other sports, such as speed skating, or bobsleigh, athletes times are set in different runs, so timings of these runs must be compared.
Then also try to take into account all other possible influences, for a start I could think of: position of the starter (holding a starter gun), speed of sound differences with temperature/pressure of the air, non ideal shape of the ice staking ring (also there is an inner and an outer lane)
What is the error in the time-stamp of the start signal for different runs? Probably there is lots of other stuff...

Omega (the company taking care of timing at the olympics) claims that they can measure with accuracy of 1 millionth of a second and that the measurements are flawless (no errorbar).

I'm not doubting that they can measure so accurately, but maybe they make an error in what they are actually measuring.

What do you think?

 

[russ_watters]

The timing is completely electronic, so none of the errors you list are relevant to the timing, only the reaction of the athlete. For example, if you are speed-skating and the the skaters are oriented parallel to the gun and 4 feet apart, then each receives the sound 4/1000ths of a second later than the last. For 6 skaters, that's 0.02 sec. Enough to matter, so I'd think they'd make the sound come from directly in front or behind.

Since the timing is electronic, when the starter presses the button, the sound is sent and the timing started at basically exactly the same time. I don't know how they detect the different athletes finishing in a multi-athlete event, but I did see one speed-skating event where they used the high-speed camera to adjust a race result after the fact, resulting in two finishers swapping gold and silver medals.

The error in time is whatever the framerate of a high speed camera is for a multi-athlete event. If they wanted to, they could go to the photos and separate an equal-time finish using the photos.

For skiing or bobsled, since it is entirely electronic from start to finish (using a switch at the start and some sort of electronic eye at the finish), the accuracy is microseconds at worst.

Some details here:
http://entertainment.howstuffworks.com/olympic-timing1.htm

 

[rfdave]

An off the shelf GPS receiver will give you a timepulse that's accurate to within 30 nanoseconds (30*10^-9), so measuring a lap time accurate to a millisecond is dead slow. Getting to a Microsecond is a little more work, but still not particularity difficult.

 

[Baluncore]

When the difference in times comes down to the millisecond, it is probably the attitude or stance of the athletes body at the line that makes the difference.

With advancing technology, there is a technical enthusiasm to produce timing systems that are more accurate than is really required. If one athlete is significantly better than another, then it will be obvious and they will be rewarded. There is no draw for the gold medal. If accurate times cannot separate the athletes, then it does not matter, it is a lottery. At some point, better timing does no more than call the lottery between equal athletes.

Remember, it is taking part that is important, a medal is a bonus, but only if drug free.

 

[aro]

http://www.omegawatches.com/press/press-release-detail/2232

 

[aro]

Previous link shows Omega's implementation.
But, here is a photo of the position of the speakers: https://twitter.com/robbertderijk/status/435021551809732608

Distance difference to athlete seems to be approximately the width of the lane, i.e. 3m or 0.01s. IMO, if you want to use advanced technology you should do it properly :)

 

[AlephZero]

The reaction time after hearing the start signal is orders of magnitude longer than the time for sound to travel a few meters.

In any competition that is run in lanes, there is an element of luck about which lane you compete in, especially if the track is curved. If you are in the outside lane, you can't see the other competitors who are behind you. If you are in the inside lane, the radius of the turns is smaller.

In a race situation, unless you are going for a world record, you are not trying to beat the clock. You only have to beat the other competitors!

 

[russ_watters]

With advancing technology, there is a technical enthusiasm to produce timing systems that are more accurate than is really required. If one athlete is significantly better than another, then it will be obvious and they will be rewarded. There is no draw for the gold medal. If accurate times cannot separate the athletes, then it does not matter, it is a lottery. At some point, better timing does no more than call the lottery between equal athletes.

Remember, it is taking part that is important, a medal is a bonus, but only if drug free.

I doubt the athletes would agree with that.

 

[russ_watters]

The reaction time after hearing the start signal is orders of magnitude longer than the time for sound to travel a few meters.

In any competition that is run in lanes, there is an element of luck about which lane you compete in, especially if the track is curved. If you are in the outside lane, you can't see the other competitors who are behind you. If you are in the inside lane, the radius of the turns is smaller.

In a race situation, unless you are going for a world record, you are not trying to beat the clock. You only have to beat the other competitors!

The reaction time at the start isn't the issue: discriminating the finish is. A few milliseconds can change the outcome of a race, so at least in track, each athlete has their own speaker.

 

[russ_watters]

This year, there is a controversy in the women's downhill, with a tie and two gold medal winners. The timing is measured to hundredths of a second*, vs luge, which is measured in thousandths. Both go about the same speed and the difference is on the order of one inch vs ten. In my opinion, they owe it to the athletes to measure accurately enough to discriminate.

*According to this, it was measured to 1/10,000th but reported at 1/100th.
http://www.pri.org/stories/2014-02-13/two-downhill-skiers-tie-olympic-gold-medal-did-one-actually-win

 

[D H]

In my opinion, they owe it to the athletes to measure accurately enough to discriminate.

I disagree, for two reasons. Time can be measured very precisely, but whether those precise differences represent a real difference is a different matter. Downhill is a good example. The race conditions change over the course of the event. The conditions typically get worse as more and more skiers carve trenches in the snow. A skier with a slower time than another may actually have skied better than that other skier because of degrading course conditions. At some point, the officials have to give up and say that overly precise differences in timing are just that -- overly precise.

The other reason comes from measurement theory. You typically want a digital measurement to be at least an order of magnitude, and preferably two orders of magnitude, more precise than the reported value. To reliably report times down to the hundredth of a second, the times should be measured down to the 1/10,000 second -- which is exactly what they are doing.

This year, there is a controversy in the women's downhill, with a tie and two gold medal winners.

This is not the first Olympic tie. People have short memories. There was a tie in women's luge in 1972. The luge committee decided to switch to 1/100th second timing after this tie. Men's bobsled saw two ties in 1998, for the gold medal in the two man contest and for the bronze in the four man contest. Unlike luge, bobsledding has not switched to more precise timing.

 

[.Scott]

Time can be measured very precisely, but whether those precise differences represent a real difference is a different matter. Downhill is a good example. The race conditions change over the course of the event.

So have the contestants bid on starting position. For example, if I believe that the track will be best after two skiers have preceded me and will start to deteriorate after that, I might offer a 1.94 millisecond advantage in order to get that position.

So award the positions from the highest bidder on down. Then measure the time as accurately as practical.

 

[Borek]

In ski jumping they already measure the wind and add/subtract points, to make the result "more objective". Is it more objective? I doubt, it is just more convoluted. There is, there was, and there will be an element of lottery, that's inevitable.

 

[D H]

In ski jumping they already measure the wind and add/subtract points, to make the result "more objective". Is it more objective? I doubt, it is just more convoluted. There is, there was, and there will be an element of lottery, that's inevitable.

I agree. I'm not thrilled with that added twist to ski jumping (added in 2010). To make matters worse, the wind compensation is arguably invalid from a physics standpoint.

 

[Baluncore]

What is the fascination of physicists and engineers with the games? Is it the technological challenge of measuring something to a ridiculous accuracy? Or is it the alchemical practice where bronze can be converted into gold by the injection of certain biologically active substances?

I think this comes down to the detection of a signal in noise. The sources of noise are so significant that it is hard to know what is actually being measured.

To base a decision on one measurement only is unscientific. The result must be shown to be repeatable. Multiple measurements make it possible to study the distribution of measurement noise. It also makes it possible to eliminate obvious mistakes.

To some extent, the pattern of competition leading up to an event final will do a good job of ensuring that the best athlete will actually be in the final. But unless the final is repeated at least three times we can never know that the result is correct, or even relevant.

Until the International Olympic Committee are prepared to run the experiment to determine the placings more than once every four years, this will remain unscientific and so, should not be a topic for a Physics Forum.

 

[AlephZero]

Until the International Olympic Committee are prepared to run the experiment to determine the placings more than once every four years, this will remain unscientific

I have a more cost-effective proposal. Don't increase the frequency of the experiment. Reduce it to zero.

 

[russ_watters]

I disagree, for two reasons. Time can be measured very precisely, but whether those precise differences represent a real difference is a different matter. Downhill is a good example. The race conditions change over the course of the event. The conditions typically get worse as more and more skiers carve trenches in the snow. A skier with a slower time than another may actually have skied better than that other skier because of degrading course conditions. At some point, the officials have to give up and say that overly precise differences in timing are just that -- overly precise.

The issues of luck and conditions are common in all sports and are accepted by athletes as part of the game.

Worse, your suggestion requires deciding exactly how big of an issue luck/conditions is, even though it changes all the time. On one day a course might be .1 seconds faster for the first race than the last. On another day it might be .05 seconds faster. How do you choose what accuracy is acceptable?

The only objective thing to do is to make the measurement accuracy as high as necessary/possible.

The other reason comes from measurement theory. You typically want a digital measurement to be at least an order of magnitude, and preferably two orders of magnitude, more precise than the reported value. To reliably report times down to the hundredth of a second, the times should be measured down to the 1/10,000 second -- which is exactly what they are doing.

Can you explain why that is? For the record, though, they can do much, much better than that. The link I gave quotes the Seiko rep saying they can go down to microseconds if they choose.

This is not the first Olympic tie. People have short memories. There was a tie in women's luge in 1972.

I have no memory of that since I wasn't alive yet. This has nothing to do with history, it has to do with having the ability to do better.

 

[D H]

How do you choose what accuracy is acceptable?

I'm not the one who is doing the choosing. It's the governing bodies of the various sports that make that determination. I'm assuming they have more information at hand than do I.

The only objective thing to do is to make the measurement accuracy as high as necessary/possible.

There's a big difference between precision and accuracy. Those down to the microsecond results you have talked about -- that's the precision of the measurement. The accuracy is much less than that.

Consider bobsledding. The time starts when a photodetector detects the nose of the bobsled crossing a line just after the start point and ends when another photodetector detects the nose of the bobsled crossing the finish line. The precision of that timing might be very high, but the accuracy is limited by the ability of those photodetectors. The one at the start is particularly problematic because the sled is moving relatively slowly at that point. The accuracy of the timing is limited by the accuracy of the start equipment.

Consider speed skating. The starter (a human) is responsible for judging whether a skater did a false start. Apparently (I can't find the exact rules), that means the skaters have to remain stationary for about 1/10 second *after* the gun fires. Can a trained human judge that down to the 1/100th second? Maybe. Down to a 1/1000th of a second? No. The accuracy of the timing is once again limited by the accuracy of the start.

The other reason comes from measurement theory. You typically want a digital measurement to be at least an order of magnitude, and preferably two orders of magnitude, more precise than the reported value.

Can you explain why that is? For the record, though, they can do much, much better than that. The link I gave quotes the Seiko rep saying they can go down to microseconds if they choose.

You are conflating precision and accuracy.

By way of analogy, suppose some engineering student comes to this site and complains about being docked for stating an answer of 2.3456789 Newtons to some question when the numbers going into that calculation only had three places of accuracy. The correct answer is 2.35 Newtons. Those other digits are superfluous (arguably, they're worse than superfluous).

From a measurement theory perspective, you should want the raw measurements to be highly precise so the imprecision of those measurements doesn't come into play in the accuracy of the result This is particularly so for digital measurements (e.g., Swiss timing) because pink digital noise is an absolute bear to deal with.

Every single one of the timed sports at the Olympics has a number of confounding issues that affect the accuracy of the timing. Accuracy is what should limit how times are reported. You can take the times measured by clocks out of the picture by making those clocks much more precise than the accuracy limits. Just because those internal times are very precise doesn't make them real. The accuracy of those highly precise internal times is limited by other factors.

This is not the first Olympic tie. People have short memories. There was a tie in women's luge in 1972.

I have no memory of that since I wasn't alive yet.

Ahem. You cut short my quote. That's not nice. I also said "Men's bobsled saw two ties in 1998, for the gold medal in the two man contest and for the bronze in the four man contest." Presumably you were alive in 1998 and old enough to be aware of that event (if you watched it).

This has nothing to do with history, it has to do with having the ability to do better.

You're an engineer. You know better than to conflate precision with accuracy. Given all those confounding factors that limit accuracy of the timing, I'd say we're pretty close to the limits of doing better.