
Digital stopwatches can make rounding errors of up to 0.11 seconds, which could be enough to change the winner of close races. Although the devices record time extremely accurately internally, they often make errors when converting that raw data to human-readable values.
Most digital stopwatches measure time by counting the predictable oscillations of a quartz crystal. The crystal will pulse many thousands of times a second and the final count is converted into seconds with decimal places, in a format known as a floating-point representation that is commonly used by digital devices.
This floating-point number is extremely accurate, but must be converted again for display on the limited screen of a stopwatch. and at the University of Surrey, UK, have found that this process introduces significant rounding errors.
Advertisement
Faux says that as many as two-thirds of stopwatch-timed results have tiny discrepancies added, but in some cases, this difference between reality and the display can be as high as 0.11 seconds. That margin is wider than that between gold and silver places in many Olympic Games events.
Faux and Godolphin took 647 race times from two swimming competitions in which a variety of digital stopwatches were used. In theory, every possible pair of final two decimal places in a race time should appear equally, but they found that three particular digit pairs – 00, 50 and 75 – together accounted for more than one-eighth of results. There were also eight pairs entirely missing from the data.
The pair believe that a wide range of devices from many manufacturers are affected, partly because the problematic method of converting floating-point numbers used is set out in an industry standard. Similar issues have been .
To check their findings, the researchers created a computer model of a stopwatch and recorded over 3 million random times. They found that each pair of final numbers showed an even distribution. But when they passed the times through an algorithm similar to those used to display numbers in stopwatches, they once again saw a skewed distribution.
Faux says that the obvious solution is for stopwatches to have large data tables for converting the numbers of oscillations into the corresponding time in seconds, rather than using an algorithm. The researchers now hope to investigate whether the errors are found at large sporting events.
“If you take the fraction of swims where there are pairs of swimmers within 0.11 seconds, yes, I think there’s a real issue there,” says Faux. “There’s going to be a significant fraction of those [results] that are incorrect.”
èƵ asked the International Olympic Committee if timing at the Olympic Games, which began today, would be affected by this phenomenon. The IOC passed the request for comment to Omega, the official timekeeper of the Games, which didn’t respond.
American Journal of Physics