Tuesday, March 10, 2009
Benchmarks and Forecasting
I woke up this morning and looked out the window to find a good accumulation of snow on the ground. I was a little surprised. I didn’t remember my local Fox meteorologist, Crystal Egger (pictured) saying anything about an upcoming snow storm. Usually, she gets it pretty right. In fact, if I think back a couple of decades to my childhood, it sure seems that weather forecasting has improved a whole lot. It seems that back in the day, it was a running joke that the weather that we would wind up with was basically the opposite of what the meteorologist would predict.
Of course, like most things, I am sure technology has played a part in the improved accuracy of weather forecasting. Scientists are now able to sample, minute by minute, a myriad of benchmark #’s, from barometric pressure to humidity to minor fluctuations in temperature and furthermore, they are able to summarize this data to create accurate computer models to predict future weather behavior.
If you’re reading this blog, chances are that you’re seeing where I’m gonna go with this.. there are parallels between weather forecasting and performance forecasting in the world of athletics. Similarly, there are those who are understanding and embracing these new technological tools to better forecast their athletes’ performances and there are those who still see this new science as ‘hit and miss’ at best. I want to chat a little, in this blog, about how to go about selecting appropriate benchmarks and how to use them to accurately forecast and plan improvements in performance.
Regular benchmarking is necessary and useful for a number of reasons on a number of levels….
1. Is the training program working?
Clearly, from peak to peak, we want and expect an improvement in performance from the preceding training program. It would be a foolish athlete who would commit to a multi-year training plan without some ‘checking in’ along the way to make sure the program is working. This ‘check in’ may take the form of racing or testing but it is clearly an important component of the feedback loop when it comes to training. You would probably be surprised by the number of athletes that I come across who don’t have regular effective ‘check ins’. It is common for Ironman athletes to race much less frequently than short course athletes, focusing more on work than testing out whether the work is working Additionally, athletes will often choose different events from year to year in the name of variety and miss the benefit of having a standard check-in at a local event at a given point in each training year. The problem is compounded by the fact that many athletes do not undergo regular field or lab testing.
Clearly, this represents the first and most important level of feedback for the athlete. Are they getting better? In my meteorological metaphor, this level of understanding as akin to having a general sense of what the weather is like in a given region of the Country. When I first moved from Florida to Colorado, I had a sense that the weather would be considerably colder than what I was used to, however, I had limited experience or information to understand how the differences may manifest over the course of a year or to predict the specific temperature on a given day. Which brings us to the second level of understanding…..
2. Is the athlete where they need to be in the context of the training cycle?
Better coaches will have a good general understanding of where they expect their athletes’ performance to be through the course of the training year. This is an important upgrade on the first level of understanding.
Experience has taught us that it is necessary to get ‘unfit’ at certain points in the season in order to get ‘more fit’ at other points. This ability represents a critical distinction between very good coaches and most self-coached athletes. On a lot of levels, many self-coached athletes expect continuous, steady improvement. They do not understand the performance ebb and flow that comes with the training seasons. Good coaches have a good general idea of how ‘out of shape’ they want their athletes to get and what level they expect their training sets to be at through the course of a training season.
It is really important that the athlete and coach establish seasonal benchmarks that are best correlated to their peak performance.
In weather terms, this level of understanding is akin to having local knowledge of the annual weather. For example, when I spent my first year in Colorado, I learned just how warm it would be in Summer, just how short the Spring and Fall seasons could be and just how varied the winter could be, from 60 degrees and Sunny one day to a 2 day snow storm the next. Even with this local knowledge, however, my ability to forecast the actual temperature on any one day is limited at best. Which brings us to the third level of understanding…
3. How can we ‘forecast ahead’ to determine the performance effect of different training strategies – different work:rest cycles, different peak training loads, different season structures.
If an “old school” coach wants to determine if an alternative training strategy is beneficial, they do so via long term trial and error. I have been fortunate to work with a number of truly world class coaches over the course of my career and, to a man, they are very structured, methodical individuals who rarely stray from the well tracked path. Occasionally, however, they will conduct an ‘experiment’ and will slightly modify their core training plan in accordance with the result. I remember during my coaching apprenticeship with Ian Thorpe’s coach, Doug Frost, Doug decided to change from Ian’s usual 3:1 work:rest cycles to a 2:1 structure. I asked him how he would determine whether this strategy was effective and he simply said – by Ian’s race results at the end of the season.
Simply put, the example above illustrates just how limited is the number of ‘experiments’ that the coach can conduct. Best case, an athlete may work with a given coach for 10 years. At most, this represents 10 or so alternative experiments that can be tailored to the individual. What if, however, we had a specific understanding of the fitness and fatigue dynamics for each individual athlete? What if we could predict relatively accurately how a given athlete will respond to a 2 week taper vs a 4 week taper without throwing away a whole season in the name of an experiment? What if we could determine whether an increase in load of 2hrs a week would lead to a new PR or overtraining for a given athlete, without compromising their health and athletic career in the process?
By assessing the performance dynamics of an athlete over the course of their training, we are able to get a firm handle on how quickly a given athlete acquires and loses fitness in response to a given training load.
I have presented the following figure a number of times now – Selye’s GAS curve.
Usually when this curve is referenced, it is done so in a general, theoretical sense of how a general athlete responds to training stress. However, like any curve, we can generate an actual, real world, mathematical expression for the curve and by using criterion performances, or Benchmark tests, we can change the constants of the formulae to create a curve that best fits the athlete’s specific response to training load.
When this is done, we can provide a relatively accurate performance forecast for a given athlete in response to different training protocols.
However, generating an accurate performance curve for the athlete demands that we have a good number of accurate criterion performances or ‘benchmarks’ for each athlete. We have a couple of options on this front (each with it’s own strengths and weaknesses):
a) “Flat Out” Criterion Performance Tests
In ‘pure’ terms, it is hard to beat regular races or time trials to answer the first of the questions listed above: Am I getting better?
However, there are a couple of problems with the use of time trials that compromise their effectiveness in answering the other questions. The first of these is assessing fitness at various points in the season. While theoretically possible to run time trials throughout the training year, many coaches prefer a progressive ramp up in the preparatory phase before including any ‘flat out’ work. It can be biomechanically and physiologically risky to throw a time trial in in the very early season. Therefore, we can miss out on assessing how fit the athlete is when ‘kicking things off’, or more importantly, how much fitness has been maintained and carried across from last season.
Additionally, the frequency that would be demanded to create sufficient samples for an accurate forecast model is prohibitive if the athlete wants to actually perform some training in addition to the testing trials. Or, put another way, flat out tests tire you out and wind up compromising total training load.
b) Critical Power/Pace monitoring.
Wko+ provides the weekly and monthly metrics of power and pace ‘bests’ for a given duration. IMHO, using these as a benchmark is a mistake because they do not take into context whether these were a 100% effort, an 80% effort etc.
Take for example, an athlete who performs a regular 2x20min session at a moderately-hard effort on a weekly basis. This is prescribed as a session at roughly the power than the athlete could sustain for 3-5hrs. Week after week, the athlete completes a pretty standard basic week, with a tolerable load and a similar (though slightly increasing) power level for this session. Then, 8 weeks out from their peak for the year, the coach decides to up the ante and rapidly increase the load, in addition to adding some races. With the addition of more load, the athlete’s fatigue #’s go through the roof (and their TSB sinks to -50) but, with the addition of a flat out race, they experience their highest 20min power for the season. A less than astute coach may attribute the highest power of the season to the increased training load, when in fact the 20min power best is actually the result of a change in training content rather than fitness.
Of course, the coach could include a critical power ‘best’ set each week however this brings with it the same issues as the ‘flat out’ time trial mentioned above.
c) Power or Pace vs. Heart Rate.
My preferred primary method of benchmarking is using power or pace vs heart rate for given aerobic sets throughout the training week.
The primary advantage to this method is the pure frequency of sampling. Several times per week I can look at the power:HR relationship for a given set and I can establish an improvement curve without compromising or affecting total training load. Additionally, I can assess performance throughout the training year, testing at times of fatigue as well as peak form to establish both fatigue and fitness components of performance with respect to a given training load.
Of course, this method has it’s shortfalls also. Heart rate is a physiological measure that is affected by many non-training related environmental factors. However, when it comes down to it, the pure number of data samples that I am able to accumulate with this method outweighs the standard error associated with the method.
A couple of key sessions that I like to use for the Bike and Run:
Run (Track Workout):
2mi Easy (@130bpm)
2mi Steady (@140bpm)
2mi Mod-Hard (@150bpm)
Record 400m splits for all.
Bike (Long Workout)
1hr on the trainer (30min easy w/up, 30min @ 135bpm) Record power
2-4hrs on the road
1 hr on the trainer (30min @ 135bpm, 30min cooldown) Record power
As mentioned, these workouts should be done regularly and during hard weeks as well as easy weeks so that the both the athlete’s fitness and fatigue responses to a given training load can be determined
Stay tuned in coming weeks for more on how you can use your training benchmarks to ‘calibrate’ your CTL and ATL constants in wko+