Requirements

 

What level of accuracy can competitors reasonably expect; and can we achieve it? How accurate is accurate enough?

Expectations

Rowing over a course at say 6 meters a second, a competitor will cover 0.6 meters in one tenth, or 6 centimeters in one hundredth of a second. If two elite competitors were side by side, they might consider 0.6 meters to be significant, but 6 centimeters, over ten or twenty minutes, not relevant. A win is a win in a side-by-side race, but if you row over the course at different times, in different wind and tidal conditions, with different competitors around you, it does not seem reasonable to measure the difference by less than 5-10 centimeters or less than one hundredth of a second.

Rowing regulations are that, in a lane race, the order of finishing can be determined by any means possible. With a photo finish, the gap between competitors can be measured in the one thousandth of a second. But in a Head race there is no gap to measure. It is just a time. FISA racing regulations say that times should be given to one hundredth of a second, for the purposes of records. So it is reasonable to assume that elite competitors would expect to be timed with an accuracy of one hundredth.

Another factor affecting accuracy is the duration of the event. If two competitors race close together, for example Elite and Elite Lwt, they will experience similar conditions and may expect the gap to be measured accurately. If two competitors race far apart, for example Elite and Novice, they will experience different conditions and may not expect the times to be comparable. If a stopwatch gains or loses time during the race, then in the first case it may not be material and in the second case it may not be relevant. But other things being equal (e.g. cost) then it would be preferable to ensure that all the times are within a level of accuracy and to let competitors make their own mind up about the conditions without being concerned about the accuracy of the timing.

From the times of previous Head races it is clear that several competitors in the same category can finish within the same second. It seems realistic overall to expect to measure them accurately to one tenth, or ideally one hundredth. Somewhere between these two seems about right.

Consistency

Times also need to be consistently accurate. For example, when using a stopwatch and writing down time, some times may be accurate to one tenth, but others may be wrong by several seconds through a transcription error (e.g. a 5 instead of an 8), or mistaken identification.

Consistency requires that there are no sudden jumps in the accuracy of timing. Jumps would occur if:

  • the timing devices (e.g. a computer) were re-synchronised
  • times were mixed from different devices or timers
  • a time is assigned to the wrong crew
  • lapse of concentration of a manual timer.

Such a jump could easily account for one tenth of a second, and would affect direct competitors.

Competitors need to know that their times are consistently accurate. It would be reasonable to think that in a Head Race 99.9% of times were accurate to within one second, and 99% to within one tenth, if that were achievable. Of course we are already assuming that there are some timing constraints - otherwise why not just photo finish the whole race and have an accuracy of one thousandth?

Reliability

The accuracy also needs to be highly reliable. It is not possible to run the race again. If you think what failure rate would be acceptable, would it be once every hundred races, every thousand races? Computers can lock up, stopwatches can fail, batteries can run out, hands can get cold. The system needs to work accurately every time.

Requirement

So in summary, the minimum accuracy requirement of a Head race timing system is:

  • accurate to a few tenths of a second
  • for 99% of competitors
  • with no errors of more than a second
  • and a failure rate of less than one in a thousand races.

Timing System

The total timing system needs to provide accuracy to the requirement that we specify. It might be acceptable to time with an accuracy of one full second in a club time trial. It might be a requirement to time with an accuracy of one tenth in a major Head race (assuming it can be done).

The total timing system comprises:

  • the response time of the timers, whether automatic or human
  • the accuracy of the timing devices
  • the variations between timers and timing devices.

With manual timers the best outcome will be in the region of one tenth of a second precision. Therefore the timing system should be more accurate than that if possible.

With video, the images have a resolution of 0.04 seconds (25 frames per second) and therefore the timing system should be more accurate than that.

With photo finish or photocell the timers have a resolution of a few thousandths of a second, but this may exceed the accuracy of the timing system.

Given the above, a timing system will only be acceptable if it: a) exceeds the accuracy of the timer or b) is sufficiently accurate to meet the requirement of the race.

The reason this matters is that timing systems may well not be acceptable. Stopwatches may well not have an accuracy of one tenth of a second per hour. Video camera clocks will probably not be accurate to 0.04 seconds per hour. Computer clocks corrected by NTP may well jump a few hundredths when they correct. Times in different devices (e.g. two stopwatches, or one stopwatch and one video camera) may well not be synchronised to within one tenth. The accuracy of the whole timing system needs to be known to assess the accuracy of the result.

Actual results

It is interesting to consider how often places depend on a time difference of less than one second. Here are the results for the Head of the River 2012, and the Head of the Charles 2012. The HoRR results are for all categories combined. The HoCR results are for the Championship Open classes.

Head of the River 2012

Time difference between crews

Number of crews: 394
Crews within 1 second: 310
Crews within 0.1 second: 63
Mean of difference: 00:00:00.99
Standard deviation of difference: 00:00:02.47

Distribution of time difference

Difference s No. % Cum
0.10 63 15.9% 15.9%
0.20 55 13.9% 29.7%
0.30 44 11.1% 40.8%
0.40 27 6.8% 47.6%
0.50 28 7.1% 54.7%
0.60 25 6.3% 61.0%
0.70 18 4.5% 65.5%
0.80 19 4.8% 70.3%
0.90 19 4.8% 75.1%
1.00 12 3.0% 78.1%
1.10 6 1.5% 79.6%
1.20 6 1.5% 81.1%
1.30 11 2.8% 83.9%
1.40 8 2.0% 85.9%
1.50 4 1.0% 86.9%
1.60 5 1.3% 88.2%
1.70 1 0.3% 88.4%
1.80 1 0.3% 88.7%
1.90 1 0.3% 88.9%
2.00 5 1.3% 89.9%
  357 90%  

Head of the Charles 2012

  CSM CSW CDM CDW CFM CFW CEM CEW
Finishers 27 18 20 20 19 20 28 40
Where difference <= 1s 7 3 2 1 2 2 6 5
Percent 25.9% 16.7% 10.0% 5.0% 10.5% 10.0% 21.4% 12.5%
Average difference 0:00:07.081 0:00:10.461 0:00:14.480 0:00:18.776 0:00:05.809 0:00:09.488 0:00:07.787 0:00:05.079
Standard deviation 0:00:11.588 0:00:09.860 0:00:13.833 0:00:17.605 0:00:03.753 0:00:08.047 0:00:19.042 0:00:05.910

C = Championship
S, D, F, E = Singles, Doubles, Fours, Eights
M, W = Men, Women

Comparison

It is striking that the results from combined categories (HoRR) show a far higher proportion of close times, and a much lower average difference, than for single categories (HoCR). The combined times show a small deviation around the average, meaning that most crews are separated by very little. This could be because most crews are of a very similar standard, or perhaps because the format encourages crews to match those around them.

It is also striking that the Championship crews show a much higher average difference, but nevertheless produce a significant number of close times. For example, in the Championship Men's Single the average time difference is 7 seconds, but a quarter of the finishers are within 1 second of the crew ahead.

In terms of timing, this suggests two conclusions:

  1. Timing to sub-second accuracy is necessary to separate crews in the larger classes (this is true, for example for the J18s in the Scullers Head and Pairs Head). Without it, there would be a large number of joint placings, although you can debate whether the sub-second differences are meaningful.
  2. In the Championship classes the results are more clear cut, but there are still a significant number of close finishes, which the competitors may well regard as significant.

Conclusions

  1. Competitors should have a reasonable expectation to be timed to one tenth of a second or better.
  2. Actual results show a high proportion of placings depend on a difference of less than one second, and a significant proportion less than one tenth.
  3. Timing systems (methods and hardware) do not have this level of accuracy unless meticulously organised.