“A lot of people are drowning in data without making any discoveries from it. Data can be so overwhelming that you don’t have the capacity to tackle its implications. You’re trying to find one key thing that can have an impact.”
Team Sky performance analyst Robby Ketchell hits the nail on the head.
He isn’t saying that the volume of data now available to riders makes it incomprehensible or necessarily confusing — it’s not a problem of too much information so much as too little interpretation. To pick the patterns that truly matter can require teamwork.
“Riders have key indicators that they are used to monitoring and believe will lead to good performance. These may not fall in line with what the coach or the team tracks. A big part of data analysis is the communication between the team and the athlete.”
This relationship between coaches and riders and the sharing of data is crucial.
Co-founder of TrainingPeaks Dirk Friel explains that performance data has become so mainstream at the highest level of the sport that teams are now allocating more of their budgets towards data analysts and backroom staff.
This allows them to analyse ride and training data, and also facilitates talent identification while monitoring up-and-coming riders. Friel believes that teams’ embracing of the best data analysis is reflected in their success.
“If you were to plot the usage of training software to a team’s WorldTour ranking, you’d probably find a very close parallel,” he says. “The days of hiring riders and just letting them decide how to train on their own are long gone.”
Using data the right way
For some riders, average power is the be-all and end-all of data analysis. In reality, spotting trends and anomalies over time is the key to making improvements.
“Quantifying the data in a comprehensible manner, which can then be interpreted and acted upon is the goal,” says Friel.
“[Using software] we have found by plotting a Training Stress Score [TSS] of each individual workout over time can give valuable information that shows variance in overall fitness, fatigue and form.
“Taking those TSS scores, you can look at trends over the last seven to 42 days to calculate a Chronic Training Load [CTL], Acute Training Load [ATL] and finally a Training Stress Balance [TSB] (see glossary at the bottom of the page).
“With these three values, an athlete can see how their training evolved and where they may have been too fatigued or not fit enough.”
Advancing the case for data has meant convincing sceptics who at first were hesitant to listen to advice from performance analysts.
Watch now: Critical power: the hardest test you can do on a bike?
Having spent eight years within the sport homing in on which data points really matter, Ketchell believes that discovering emerging patterns helps map out training for a particular rider for a certain race.
“It comes down to managing the demands of an event, and doing so for each particular rider — not just for their target or what their job is going to be for the team but also what their capacity is. You need to discover that to plan what your strategy is.”
Comparing and amalgamating training data values with race results provides context that then informs future training.
This detail within performance data is no longer reserved for privileged professionals teams with specialist staff. Amateurs now have greater access to data and analysis tools than ever before.
One amateur rider who used performance data to excel when it really mattered was multiple national time trial champion Matt Bottrill. Specifically, he used his TSS to accurately work out his fatigue levels and how a certain race was likely to pan out.
“You can never know with 100 per cent certainty what is going to actually happen, but I knew when I was at my best in a time trial,” he says.
“It was so controlled that once I factored in my aerodynamic drag, I was able to predict my time almost to the second.”
One of the risks with sharing data — for example, via Strava — is that riders can become competitive over training statistics.
This can be harmless but Bottrill warns against pointless heroics and urges using data strictly for the purpose of improving your performance.
“People get carried away with, ‘What is your 20-minute power? What is your FTP?’ What goes wrong for a lot of people is that they’ve got all these cracks in their training in spite of recording impressive data.
“It might be that they’ve got good threshold power but they have got a 40 or 50-watt drop-off all round.”
It’s not all about the data
Despite many riders’ love of statistics and their implications, data shouldn’t be seen as the Holy Grail. Multiple factors come into play and tracking data is no guarantee of success, admits Friel:
“Improving as a cyclist isn’t all about numbers. This is why training is an art and a science. Each athlete and coach has their own balance of the two. There are many variables that have to be balanced, such as recovery, mood, diet, sleep, tactics, bike-fit, equipment selection, pacing.
“Training data is simply a tool to be leveraged as part of the overall development of an athlete.”
This is true in racing too, as Ketchell explains: “Sports are unpredictable, especially cycling. Data analysis is never 100 per cent correct. That is why it should never be the only consideration that goes into making a decision.
“It’s great to be evidence-based, but always challenge the evidence and look deeper into the data to discover new insights. This means accepting you’re wrong sometimes.”
Data mistakes and misuses
It can be easy to fall victim to oversights in data-processing and analysis. Rob Kitching from Cycling Power Labs highlights some of the most common data analysis mistakes he has seen.
“One of the major benefits of using power is objectivity: 200 watts is 200 watts, regardless of what a rider’s heart rate is doing or what their perception of exertion might be on a given day.
“If we’re not to lose that objectivity, and a rider’s faith in the data, then issues that can lead to misleading data need to be avoided. We don’t want a rider reverting to guesswork because the data doesn’t seem to make sense or because the level they are able to achieve isn’t consistent.
“Bad data could lead to over or under-training relative to what you have planned, but the real cost can be in terms of motivation.
“A moderately or well-trained cyclist, in the short term, will target levels of improvements you can measure in single-digit watts.
“Bad data can easily be misleading up to and beyond that level and could potentially lead a rider into a rollercoaster of false emotions, either with new PBs or inexplicable failures to hit targets.
“You want to avoid that at all costs in a sport where hard work that leads to improvements depends so much on motivation.”
Data mistakes to avoid
– Failing to bear in mind whether average power includes freewheeling time, which dramatically affects the number.
– Confusing simple average power with normalised power. Normalised power accounts for intervals and efforts that have occurred over the entirety of the ride. Average power will simply average out the training session as a whole, which can lead to misinterpretation of more intense sessions that may have been more fatiguing.
– Forgetting to ‘zero offset’ a power meter before riding and then relying on inaccurate data. Think of zero offsetting a power meter as the same process as resetting a set of measuring scales. Air pressure, ambient temperatures and other things can alter power meter readings in between rides. Therefore ‘zeroing’ your power meter before each ride clears the residual torque and sets an accurate baseline to work from.
– Failing to acknowledge the differences between indoor and outdoor riding: the former involves zero coasting and no air resistance, whereas outdoors there are many variables such as wind and drafting gains.
Training data glossary
Many power meters and training programs convert training effort into stress, load and fatigue scores. Here’s what the key metrics mean…
Training Stress Score (TSS): The number that relates to the intensity of a single training session. The higher the number the more strenuous it has been.
Acute Training Load (ATL): The short-term fatigue number that is accumulated, estimated over a seven-day period.
Chronic Training Load (CTL): The longer-term fitness accumulation rating based over a 42-day period of time. Rides that are completed more recently will be more weighted towards this number.
Training Stress Balance (TSB): This number is the difference between CTL and ATL and addresses whether a rider may be approaching top form. When this number is positive it indicates a good performance is approaching following a decent block of training combined with a low recent value of fatigue. This is where the tapering effect comes to fruition.