DATA

After a prolonged hiatus from blogging, one of my favourite topics has pulled me back to the keyboard... data.

I’ve always been about of a stats geek. As a kid I’d spend weeks deliberating over the previous season’s points totals or studying the progression of young players before submitting a fantasy football team. That only got worse once I discovered the Championship Manager game on my PC!

My life in track hasn’t been any different and I’m a big fan of collecting as much training performance data on athletes as possible, which is why a couple of recent events prompted me to write this.

New toys!

The first of those is that I have a shiny new toy! I’ve had a freelap timing system for probably around a decade now - the old (original?) “candlestick” system. Well I’ve now upgraded to the “yellow pyramid” system that has gained a lot of attention in recent years.

I’m no electrical engineer, but my crude understanding is that the freelap system uses a radio field from the yellow pyramid to trigger a timer unit when it enters that field. It’s not a full photo finish system, but it offers a high level of accuracy without occupying 3 lanes as a light gate system would.

For longer intervals of over 150m I tend not to worry too much about having high levels of accuracy and will settle for a stopwatch. However, when timing sprints over shorter distances the odd tenth here or there can have an enormous impact on the overall time. Timing a flying 30 timed with a stopwatch is like trying to perform surgery with a bread knife - it's just not a sufficiently accurate tool. I feel sorry for athletes who come away from these sessions with training times which are considerably faster than the times of which they are truly capable. They invariably end up disappointed when their race performances don’t measure up. In my mind, if you are not in a position to accurately time shorter reps, it’s best not timing at all.

Predicting performance

The second thing to encourage me to get back to the keyboard was a Twitter exchange around predicting race performances from training performances.

Whilst there is a lot of rubbish to wade through on Twitter, it can be a wonderful place in terms of coaches sharing thoughts and experiences.

Before going into the merits of whether race predictors are a useful thing or not, having quality data is a prerequisite (which brings me back to freelap...) Even with the best prediction formula in the world would only ever be as good as the quality of the data entered.

There are a lot of race prediction calculations around, especially online. Personally I tend not to use them. If, for example, I want to find out what sort of 100 meters shape an athlete is in I find that having them race 100 meters is usually the best way!

I typically keep recordings of athletes' times from starts to 10,30,50 and 80 metres, in addition to flying 10, 20 and 30 meter times. These are all times that can be plugged into the predictor tools, but I don’t really know what I would do with that information. I use them to compare against their own best performances In order to monitor rather than to try and extrapolate other, less accurate data.

One argument I have seen put forward Is that they can be used to identify other events that an athlete may be good in. I think this may have some value for less experienced coaches (eg a teacher who gets allocated a sprint group despite having no experience in the area). However, if I were to have an athlete who is only running 3.3 for a flying 30 meter sprint, yet managing a 36 second 300m, I don’t need any formula to know the athlete is likely to fare better in the 400 than the 100 meters. Perhaps a novice coach to the sport may not have that appreciation of what a good time is over those distances to make that judgement.

Another argument is that it can help identify areas that need addressing. For example, if an athlete's start to 30m time and flying 30m time suggests the athlete should be running sub 11, yet they are regularly recording race results of 11.3, it may well be speed endurance that is lacking. Again, I can see the logic in this, I just don’t know how I would use it. If I’ve got 80m times with a 50m split, I can see that there is going to be a drop off in the second half of the race and work is needed in that area.

Even without the data, this can usually be seen just by watching a race. One of my athletes raced a month or so ago and looked superb for 70 meters, before fading. It is obvious that the speed endurance we need by the peak of the season isn’t there yet just from watching the race. Given how I structure the group’s training year, it’s also exactly where I would expect the athlete to be - the speed endurance work is always the last piece of the puzzle for us.

Then of course there is the issue of accuracy of the prediction methodology itself. Before writing this I plugged a few athletes’ data into some of the online calculators. Whilst they were predictably fairly accurate for distances closest to the data submitted (eg an accurate flying 30 is a good predictor of a flying 20) they were often excessively optimistic in terms of some distances and pessimistic in terms of others. There is so much variation from one athlete to another that I’m just not convinced any predictor is going to predictably reliable across a group of athletes.

I'm yet to be convinced of any circumstance in which I'd ever really have a true use for a race prediction model, regardless of the accuracy.