Handicap races were particularly attractive to me for the simple reasons that I’m no longer young and, even if I were, my running success would be limited by a distinct lack of talent. Over the past few years I’d run in handicap race series organised by different clubs and been amazed by both the popularity of these races and the anger and venom you’d get when people felt unfairly treated.
Among the travesties I witnessed were a couple of five mile races where the winner in the first beat his handicap time by four minutes and followed it up two weeks later beating the target by five minutes. Several people who ran significant PB’s in the second race were, understandably, unimpressed with the handicap targets. Even worse was series of twelve (handicapped) races where 10 in a row were won by the same person. No adjustments to runners’ targets had been made at any time during the series even though it was clear that some were easy to beat while other people were having a tough time getting remotely close.
The secretary of my current club (Beverley AC) asked me to look into ways of improving the means of calculating handicap targets. In these days where you can find just about anything on the internet I was amazed to find that a Google search for “running handicap” turned up virtually nothing. No software. No discussion forum. No methods. There was plenty of material concerning horse racing and equal amounts for maintaining and recording golf handicaps. But for running it was an information desert.
It seemed to me that any viable method for arriving at targets had to meet a number of criteria…
1. Runners should be able to understand how their target had been arrived at
2. The method should apply equally to all
3. The targets should be verifiable
4. The targets should reflect the runner’s current level of ability
And that if these criteria could be met we’d stand a chance of at least pleasing most of the people most of the time.
I’d heard about any number of “methods” used for arriving at targets some involving little more than a group of guys in a huddle (the handicap committee) trying to estimate (guess?) people’s finishing times and others based on runners’ PB’s for an arbitrary distance – usually 10K. The problem with the PB approach being not least when do you regard an individual time as being no longer relevant? Then what do you replace it with?
Most handicapping methods also used the Riegel formula as a means of adjusting times from a race of one distance to another distance. This is a formula devised by Peter Riegel from research into the performances of elite and semi-elite athletes. It takes the form t2 = t1 * (d2 / d1)^1.06 and, in plain English, says that if the distance run is doubled then the speed declines by 6%. This formula is widely used by the various running calculators available on the internet. A much more complex formula (the Cameron formula) tends to give quite similar results even though the reasoning is very different. The predictions only begin to vary from the Riegel formula when you, say, predict a marathon time using a much shorter race, like 10K, as the base time. Under these circumstances the Cameron formula predicts slower times.
Another problem with trying to use PB’s as a basis for calculating future targets is that you can guarantee that they would have been run under a variety of different conditions. Some hot, some cold. Some windy, others still. Some round in a circuit others point to point where the effects of wind and elevation change would be even more marked (viz Boston Marathon which loses nearly 900 feet between start & finish)
I came up with a simple hypothesis – what if we took into account, for each runner, their last 3 recent races then used the Riegel formula (or a variant) to adjust each race to an equivalent 10K distance. Then, if there was also a way to factor out the effects of elevation changes on the course, in effect producing a “flat 10K” time we’d have all runners on approximately the same basis. Taking the average of the 3 most recent races as the base time for the target in the next race you’d then adjust the base time first for distance then for any known elevation in the course for the target race.
The hypothesis was easily tested by using sample data from past seasons and initial calculations with a Microsoft Excel spreadsheet were encouraging. The elevation change calculations were based on some work reported by Dr Tim Noakes (the author of the monster work Lore of Running) which proved the notion that, on a course that goes up as much as it goes down, most runners definitely loose time when compared with a course that’s completely flat (ie you don’t get back on the downhill all the time you lost going up). So, for meaningful time comparisons you have to have some compensation when all runner’s base times are not sourced from exactly the same races.
Trying to apply the system to a live situation several things became apparent…
1. For many runners the system worked well
2. A spreadsheet could be made to work but it was very easy to make significant mistakes that affected the accuracy of the results and were hard to prevent or even spot. This is a common feature on many spreadsheets.
3. Even with a spreadsheet the system needed a lot of work to maintain once you had more than a few runners. Beverley AC had 160 members at least half of whom were active in a 10 race handicap series.
4. You can’t just take into account races in the handicap series – if you want an accurate assessment of a runner’s current capability you need to log all races run by each individual
5. There were some runners and circumstances where, to be fair to everyone, you had to make adjustments. The challenge being to come up with a way of doing it that wasn’t arbitrary or open to criticism if someone were to challenge what you’d done.
The special circumstances that needed another look were…
Runners who hadn’t run a race for a while, say, six months or more.
Runners picking up an injury or short term illness
Runners turning in a performance way out of the norm
New runners with no race history
Most clubs running a series of handicap races work out the series results with some kind of points system. We used to operate a sliding scale that went something like 4 minutes under target got you 10 points, three minutes under scored 9 points and so on. Using the program to do the work we’ve now modified that so the scoring goes better than 4% under gets you 10 points, 3% under earns 9 points etc. This percentage system balances out much better between long and short races and also between the high flyers and average runners. In the past it’s been harder for the fast guys to do at all well in a handicap competition partly because they tend to be very consistent (therefore it’s harder to beat the target by much) and also because it’s easier to be 2 minutes under target if you run 10 K in 55 minutes that it is if you’re a regular 35 minute finisher.
We came up with the following solutions…
1. Where a runner is new or hasn’t run a race for a while we don’t try to predict a time. For the first race we simply assume that the target time is equal to the time run and award points from the middle of the table (ie six points using our system). If the next race they run isn’t wildly out from the target based on just one race then we let the target stand otherwise it’s six points again until a sensible average base time is established.
2. For runners with a short term injury or illness it’s normally simplest to just exclude the bad result if the race they run is true to previous form.
3. If the drop off in performance is greater and looks like lasting longer then we treat the runner like a new entrant and establish a new performance standard for that runner.
Sometimes runners put in a performance that’s way off what’s expected. If it’s way better then it feeds back into their average for the next race and they, effectively, pay a penalty approximating to a third of the improvement (assuming the averaging is over 3 races). Conversely sometimes you’ll get someone who runs a race but doesn’t try to run a good time. “Who would do such a thing?” you might ask. Well someone who’s in training for an important event and just using the race as another training run – it happens quite often. In these circumstances it’s just a matter of asking the runner why the time was so slow then excluding it from the next calculation on the basis of it being unrepresentative of current form.