Head Race Timing

Timing Accuracy

There is a peculiar question about how accurate the times in a Head race should be. The obvious answer is: as accurate as possible; better than one hundredth of a second; like they have with photo finish. But a little thought shows that this is unrealistic, and that in fact as a timekeeper you are forced to choose what level of accuracy you can afford to provide.

The best human reaction times are around two tenths of a second. Most people (unscientific guess) are around three or four tenths. You can test yours here: Human Benchmark. I know some people are faster than that. But in the Olympics a reaction of less than one tenth is counted as a false start, because it is humanly impossible.

People have been timing athletics with stopwatches for a long time, and doing better than that. The way they achieve this is by: using reaction times, not anticipation times; and doing the same at the Start and the Finish. That way, the accuracy is the difference between reaction times, and not the reaction time itself. So with good concentration and good technique you may be able to get to an average accuracy of plus or minus a tenth.

But then you have to add the clock accuracy. I won't go into it again here, but all clocks drift, and only expensive clocks drift less than one or two tenths in the course of a race. A normal stopwatch is not sufficiently accurate. Nor is the clock in a video camera. And a computer or iPad clock is way off, even with time correction.

I am happy to accept that races run with different levels of accuracy. I am sure there are many races where one second is perfectly adequate. Even if, as a competitor, you would like it to be better, can you honestly say that it really matters? What I am not happy about is for competitors to believe that the times are accurate, when they are not. For me, the first rule is that times should not be given to a level of precision that is not warranted by the timing system. If the times are given to the full second, and if competitors are fine with that, then that's OK. But if times are give to the hundredth, and competitors "win" or "lose" by a few hundredths when in fact the times are accurate only to a second, then that is not OK.

I have tried to find out what competitors think is a reasonable margin to win or lose by. I have not been able to get a good answer. I was lucky enough to share the cab of a truck with an Olympic rower a few weeks back. I asked him about this. The immediate answer was "as accurate as possible". I think that competitors just don't engage with the idea that times are not absolute, but only plus or minus something. In regattas they are as precise as the photo finish, as we saw in the Olympic single sculls result. Maybe Head races just aren't that serious, and so the question is not asked.

I asked a top coach about this as well. His concern was more about time trials and crew selection. Here the times really do matter; and they may well be inaccurate. As we have more time trials, maybe we need to take more note of how accurate the timing is.

I find it is quite common to hear the view that manual times are only accurate to a second anyway, so why bother with trying to make the system more accurate than that. But then, if that's the case, don't publish times with a precision of less than a second; and see what competitors think.

One of the outcomes will be many more ties. I did an analysis of different races with the distribution of time differences: time distributions. It is quite remarkable how many competitors finish inside the same second as another. In some races, but not all, it affects the category winners too. I have wondered why this is. I think the underlying reason is that times are based on power and drag, and for most rowers the effective power band is fairly narrow. 

So another way to look at the question is: would competitors be happy to have a lot of ties? Ties for first place overall, ties for place in category, ties for top ten. Personally I don't think so. For me, every result is the outcome of a lot of hard work. I want to take from it everything I can get! If I have a third place, then I would like to know it was third, and not equal with several others. it is quite obvious watching crews go by on the water that I can discriminate between them by much less than a second. One second is five to ten meters. It is a long way. We really should be able to do better than that. 

But if that is the conclusion, then you need to have a timing system that is demonstrably more accurate. And many of them are not.

Veteran Fours Head 2016

The Veteran Fours Head 2016 took place on Sunday 13 November, starting in the morning at 10:32 and running on the flood tide, with a course from Barn Elms to Chiswick Staithe.

It was an eventful race from a timing perspective, with some new and unforeseen problems. We had real time results, as normal for the last few years, but for the first time we had a long delay before getting out the results. What happened?

We had an excellent timing team this year, with Kings College London (KCL) Boat Club playing a big part.  We had real time results flowing in nicely, and being published straight onto the website using SkyDrive. After the race the Start and Finish timing devices came back and we started to go through them, as normal, to update the missing crew identifications. This is where we can't make out the crew number by eye, so we put in a dummy number in real time and correct it later. It's quite normal to have crews like this. We make a note of the blades and fill in the number afterwards. But it quickly became clear that some of the identifications at the Start could not be right. 

I forget the exact order of event, but we had a very polite enquiry asking if we were sure about one time; and we went through the crews and found some that could not be right. So we pulled the published provisional times off the web, put up a note to say that we had good times but needed to check the crews, and started work.

The basic problem was that we knew that some crews had been mis-identified, but we did not have an authoritative source of the correct crews. I should explain how this works. There is always a risk that the timers will get something wrong. They have to work quickly to keep up with the crews, and it is easy to make a mistake, or simply not have time to get the correct data. It is not too bad at the Start, but much harder at the Finish. We cover this by having: an independent sequence with just the crew numbers and no attempt to match them to time; and a video. Like most races we don't have an abundance of volunteers, and we have to choose what roles are most important. We normally have a video at the Finish, but not at the Start.

This year it went badly wrong. We (especially I) did not foresee that, running on the flood tide in the morning, the Sun would be low and in the eyes of the timers at the Start. The low Sun and the reflection on the water made our new yellow numbers unreadable. The timing team literally could not see the number. Looking at the data after the race, I can see that they made a fantastic and largely successful effort to get the crew identities, by recognising them and looking up the number.

This is where our problems compounded. Because the numbers could not be read, the sequencers conferred to agree who the crew was. This meant that we had only one sequence, and not several to compare. And we did not have a video because it was at the Finish.

The good news, before you leap to a conclusion, is that the times were good. Times and crew numbers are two different data streams. In our system we use a separate button presser, who is not distracted by trying to identify a crew. We did the usual thing of comparing the Timy times with the stopwatch to see if there were any big discrepancies, and there were not. This may be a little hard to understand if you are a competitor. You imagine that if a time is "wrong" then the timing team must just guess at what is right, but that's not the case. The time is exactly right: it is just matched to the wrong crew. By elimination you can eventually get to the right crew, giving the right result.

We had about three or four places in the results where the identified order at the Start was clearly wrong. We were helped by some very constructive and supportive information from crews. We pieced together information from different sources about who was behind and who was ahead; who overtook and was overtaken. 

We managed to piece it together, and I am confident of the result. We ended with, in one case, a three way tie between three crews that we had trouble identifying. They may have thought this was a bit of a coincidence, but it was in fact one of the most validated results of the race. They had a thrilling race, and their results were corroborated by: the coxes view of who was ahead and behind at the Start; intermediate places at Chiswick; BigBlade photos at three places; and an observer on Barnes Bridge.

It was an unforeseen problem, and it ended our run of several years of good quick results. It reinforces the view that you absolutely must have an independent backup of some kind, and you must test it as if you had to rely on it alone.

What will we do to stop this happening again? Number one is that you can't use yellow numbers with a low Sun behind. Number two is to use photographs at Start and Finish as the independent sequence. But where to find the volunteers is another question.