Winning the Training-race
Yet again I am trying to get into decent shape so that I can run a fast marathon. But, at the age of 52, running a sub-2:45 marathon is looking increasingly unlikely. I was not a runner as a kid and only started running a bit less than 10 years ago. I have completed 41 marathons but have yet to break 2:45. I know that to do it I will first need to win the training-race. By training-race I don't mean win a race in training, or even beat someone else or a previous time of mine on a training run. The training-race is the long period leading up to a marathon where the performance (hence the term race) is the main determinant of the marathon race time.
Redefining the meaning of race
A race is usually an event with a single fixed parameter. Most commonly races are of a fixed distance (e.g. 5K, 10K, half marathon etc). A variant of the fixed distance race is a race of poorly defined distance but all runners complete the same course. Cross-country races are a good example of that. Other races are of a fixed time where the aim is to cover as much distance as possible such as a 12 hour or 24 hour race. The training race is a race of undefined distance and undefined time that starts about 6-8 weeks before the 'real' marathon race. Whilst the training race seems a hard one to win, because neither the distance nor time are defined, your performance in that race is critical - it will determine how fast you can run the marathon you are training for. Most people don't realize that training is a race because the rules of the race aren't obvious - few coaches ever explain what the rules are. There are some parallels here with other races or competitions where the rules are unclear. The caucus-race (Alice in Wonderland) is such a race where everyone runs around in circles for an undetermined length of time before everyone is declared the winner. The rules here are simple, as long as you are 'in the race' and running around you will be one of the winners. In other competitions the rules are more complicated and less obvious, but nonetheless exist. Mornington Crescent is such an example. Here the players submit to the pretence that rules exist - which, formally, they don't. Any player can win, at any time by saying; "Mornington Crescent". But, to achieve a real win the player must delay saying Mornington Crescent until it can be done so with great comic effect. The suspense generated by prevarications has to be carefully calibrated, in response to audience and player reactions, for the comic effect to work. The rules here are complex but the great comedians have an innate understanding of them.
Performance in one race predicts performance in another
We know that many types of human 'performance' can be predicted or estimated from other similar performances. This is a wide ranging observation. Someone who has an organized desk at work is also likely to have an organized filing cabinet, someone who wears expensive suits is likely to wear expensive shoes. Of course, there are going to be contradictions - but, the saying; "If You Want Something Done, Ask a Busy Person To Do It" has a good deal of truth to it. Someone who does a lot is likely to get a lot done. In many endurance sports we use this form of extrapolation to estimate race performances over different distances. In rowing the obvious example is Paul's Law whilst in running it is the Riegel formula. They are both equations that allow performance at one race distance to predict performance over a different distance. This form of performance-equivalence has become embedded within running in the form of age-graded performance tables. These tables work slightly differently from the Riegel formula in that they are based on the World Record performances at different distances of men and women of different ages. By way of example, as a 52 year old male who has recently completed a 5km road race in 20 mins, I would be ranked as a 73.3% age-grade. Assuming that the age-grade is some form of measure of my aerobic fitness then it would predict that I could run a 3:13:42 marathon - since that would also give a 73.3% age-grade for a 52 year old male. The Riegel formula is a little bit more aggressive, predicting a 3:11:49 marathon from a 20 min 5km. There are plenty of other examples of such predictive formulae that take one race effort and estimate performance over another distance. But, these races are single efforts - they take place at one point in time and are over relatively quickly. Would a formula that takes race performance over a longer period not be more accurate? This is where the training race becomes useful. It is not just a race we all do before a marathon (whether we intend to or not), it is also the training stress that determines our performance.
Defining the training race
Many scientists have tried to define what we might refer to as the training race - that is the combined set of parameters that can be used to predict marathon performance. Many of these predictors are just correlates of performance not causative. For instance, amongst the elite field having dark skin seems to correlate relatively well with performance. Some mistake correlation with causation and then attempt to justify the correlation with spurious assertions. The dark skin issue is just such an example where dark skin becomes the assertion that some athletes are genetically-gifted. Science seems to differ. There are very few genes that have been associated with endurance performance. Of course genes are important - but, the probably only determine a very small fraction of performance, and the number of important genes is likely to be very large indeed.
There are some factors that are both correlative and causative. Weight is a decent example of that. We know that most fast marathon runners are fairly light - and of course being light does not mean you will be able to run a fast marathon. But, if you are training for a marathon losing excess fat - again within reason - will probably result in better performances. It is very simple energetics - the less you weight the faster you travel for a given power output. So, weight probably belongs in the 'formula' that makes-up the definition of the training race. Marathon training will involve some optimization of weight, but it is not the primary determinant.
Of the attempts to correlate training performance with race performance the best that I have come across was a relatively short paper by Giovanni Tanda in 2011. In that paper, which I discuss elsewhere, he proposes that a combination of the number of miles run in the 8 weeks before the marathon and the time it took to run those miles is a good correlate of marathon performance. So, here we have a definition of the training race. It is not a fixed distance, or a fixed time but the output of a formula which includes both distance and time. This is not as complicated as Mornington Crescent, it is a simple collapse of two variables (distance and speed) into a single metric which is the marathon prediction. The winner of this training race is the person that can push this formula to the best marathon performance prediction. By that I mean, when training we need to understand that formula and stop thinking about either just speed or just distance and think about the way in which the two are mathematically combined to produce the marathon prediction.
The Training-Race - an eight week event of undefined distance and duration
So, all of this might sound a bit like waffle, but it isn't. There is a race to be won in training. It is real and we are all taking part. It is a race that determines the performance in the marathon. The winner of the training race is also, most likely, going to be the winner of the marathon. The training race is won by running as far as possible and as fast as possible for 8 weeks before the marathon. You can calculate who has won by putting the average distance and speed into the Tanda formula. The Tanda formula contains within it the mathematical relationship that governs how speed and distance trade. The person with the fastest predicted marathon time is the winner. They have pushed to two parameters (speed and distance) as far as possible. Of course, how you do that is the key question - how do you trade speed in training with distance. Is it better to run another 1 km fast or 5 km slower? These things can be calculated.
[Hastily written - due to be edited 19th Oct, 2018]
