Winning bike races meaning skirting the outer edge of sustainability. How light can you make a bike or a wheel before it breaks? How thin can you make a racer’s body before he gets sick? How fast can you take that descent without flying off the cliff? The best performances push right up to the edge of disaster.
In the future, high-tech tools could precisely measure and analyze the minutiae of performance on and off the bike, guiding the sport out to the razor’s edge of possibility. Here, we talk to coaches, nutritionists and even a mathematics PhD about what that future could look like.
As much as we romanticize the sport, winning often comes down to cold math. Develop the rider with the best power-to-weight and power-to-drag ratios and you have a champion. While doping is the quickest way to boost the engine, we’ll focus on legal options here. Similarly, the shape and weight of the bicycle is still arbitrarily held in check by the UCI with a 6.8kg minimum weight and 3:1 ratio on aero shapes.
Although that minimum weight may be changed, regardless of what is set — streaming video cameras on all riders, hyperlight bikes — all teams will do everything they can within the same set of parameters.
In racing, the difference could come in the exploitation of data — from riders’ bodies, race-day conditions and real-time mathematical analysis of race dynamics.
The body – 24-hour quantification of performance, rest and recovery
“The future is more internal metrics,” Benjamin Sharp, currently the power education specialist for power-meter company Stages Cycling and former USA Cycling national coach, tells BikeRadar. “Soon we’ll see on-the-fly, real-time blood lactate measurements. I’m already doing oxygenation measurements with athletes I coach to see how they are recovering on a day-to-day basis and as well as handling indoor workouts.”
Leading into the London Olympic Games, Sharp had the USA women’s team wear sleep monitors. “For ultra-competitive girls, it was fun for them to quantify their sleep,” he adds. “It became a competition for them, to see who could get the deepest sleep.”
While the current crop of wearables count or approximate basic things like steps taken or calories burned, elite cyclists could be measuring much more.
“I have seen development on sensors to track head trauma be incorporated into a skull cap, and soon they will be integrated into helmets,” says Vishal Patel, education and innovations leader at hydration company Nuun. “I’ve seen more and more research and patents being filed for ‘temporary tattoos’ that measure everything from the composition of your sweat to blood lactate. I’ve heard of water bottle caps that can detect what level of carbohydrates and electrolytes to release based on what internal monitoring systems that detect thirst via hormone activity in the brain.”
The future of racing science is about closely measuring the body, according to former USA Cycling national coach Benjamin Sharp
Longtime cycling coach Jon Tarkington is experimenting with new-school tools like Moxy, a muscle oxygenation meter. Sitting in his Boulder, Colorado office wearing a Fitbit, Tarkington tells us that one key to marginal gains on the bike will be measuring riders off the bike with a device they can wear without hassle for days.
The future will be “collecting data on stress, using accelerometers, gyroscopes, body temperature, heart rate and more,” Tarkington says. “At the top end, when rider X has a crazy, crappy travel day, you will have a number to quantify that, and how that affects their racing and what sort of recovery they require.”
Both Tarkington and Sharp agree that the trick is figuring out how to read the data.
“We don’t fully understand it yet,” Sharp points out. “Internal metrics are like power data 15 years ago — you get cool numbers, but we’re not sure what to do with it.”
Tarkington envisions a single stress number for each day. “Right now we have training numbers like TSS for what happens on the bike,” he says. “But we know that what happens the other 16 to 20 hours of each day is nearly as important. In the future you’ll be able to get something like a life stress score for any given day.”
The team & the conditions — measurement and pacing
Sitting at home on his couch during the 2013 Tour, optimization mathematics PhD Ryan Cooper predicted the leaders’ time trial times within 10 seconds using what he calls “a pretty rudimentary model” he made for fun. By breaking the course down into 200m segments and factoring in actual wind conditions and his educated guesstimates for riders’ power and drag numbers, Cooper realized he was on to something as riders crossed the line within seconds of his mathematical predictions.
What began as a fun experiment grew into a company — Best Bike Split — that was purchased by TrainingPeaks and is now utilized by Trek Factory Racing, among other pros. Now the model has grown much more sophisticated, incorporating a deep menu of variables that add up to literally millions of calculations. Instead of predicting a performance, Best Bike Split’s modeling seeks to improve rider performance with numerous small tweaks in pacing, strategy and gear choice.
Consider the type of calculations Cooper did for Trek rider Haimar Zubeldia ahead of the 2014 Tour de France time trial.
“For Zubeldia we took all his bike data — wheels, tires, rolling resistance on tires, helmet — his wind tunnel drag in time trial and climbing position from 0-25 degrees, positive and negative, in 2.5-degree increments, his power numbers for acceleration, threshold, aerobic and maximum power output, altitude, heat, humidity, his speed into 90-, 120- and 180-degree corners on various gradients, thespeed at which he’d descend and then ran these and other variables across the 450 segments we’d broken the course into. That’s about a million different combinations,” Cooper says.
Crunching the numbers for optimal performance, Cooper’s model made recommendations for power output and gear selection for each segment of the course. To bridge the gap between a computerized ideal and a plan a racing athlete can execute, Cooper and his team will often generalize some solutions. “We’ll create cheat sheets with things like, power of X on the flat, then add 20W for small hill, or 40W for a steeper, short hill,” Cooper says.
Sources of gadget-driven data are rapidly multiplying – but what are the most effective ways to act on this data?
Factoring in how a rider performed in the past makes for better data inputs into the model, whether that is proven threshold power or how a rider performs in the third week of a stage race. “Heat and humidity, for example, can have a big impact on some riders. Take [Fabian] Cancellara at the Vuelta, where it was really hot. His 40k power had to drop by about 10 percent. He gets affected more by heat and humidity than other riders. So we factor that into the model.”
While the gains can be most straightforward for time trials, teams are trying to determine how to use modeling in road races, too. “The future is breaking down the tactics probabilistically — determining the best places on course to attack or begin chasing a break by quantifying your riders’ abilities and the course and conditions,” Cooper says.
Instead of using rules of thumb — such as how fast a chasing peloton can typically bring back a break — teams can pinpoint exactly when to chase and how hard based on live data. The same could be done for when a rider should attack.
For fans who don’t like to see riders staring at their power meter numbers instead of racing on gut feeling, this could be a setback. But data geeks like Cooper who believe in math have a different perspective. “The holy grail is you have the model-created plan and then your Di2 just shifts for you throughout the race,” Cooper said. “I’d like it to be like a trainer workout where you just pedal and modeling does the rest.”
Even if a computer is doing the thinking, the need for live, on-the-fly information remains as important as ever, and team support staff can be found along the course of big races capturing weather data as well as measuring the drag of their riders in training using various positions and equipment.
“Most major teams are doing that kind of work. Garmin-Cannondale is doing a ton of data collection and real-world measurement,” Cooper says. “Wind tunnel data is a piece of the puzzle, but I’d say 95% of the studied drag data is real-world. It’s all about getting enough data about wind and weather conditions, putting that against the data you have on your riders, and crunching the numbers.”
Teams and some companies like Zipp are using a product called AeroStick that can be mounted on a bike to collect real-world aero data.
“The future of cycling is in this data. Being able to really tease stuff out that we really don’t know exists yet. Machine learning is a buzz phrase these days, and it refers to having large amounts of data sets, and using algorithms to find what matters. We don’t have the computation power in our brains to look at multidimensional data sets, but computers do that very well. We will have computers figure out the relationships between input and outputs. For now, we are just starting to collect all this stuff. It’s just a matter of running algorithms on this stuff to figure out what matters.”
Having data on a visor screen could be useful in certain situations