Richard Moore: Cycling can be a difficult sport to predict

Who could have predicted Caleb Ewan's stage win?  Picture: AFP/Getty Images
Who could have predicted Caleb Ewan's stage win? Picture: AFP/Getty Images
Share this article
Have your say

A year ago, when Geraint Thomas claimed the yellow jersey of Tour de France leader in the Alps, I confidently dismissed his chances. I said it was “fanciful” to even imagine that somebody who had never even finished close to the podium could win the Tour de France.

Two weeks ago, writing in this very same slot, I suggested that Egan Bernal was well placed to become Colombia’s first winner. And a week ago, also here, I wrote about the 33 years since a Frenchman last won the Tour and said that Thibaut Pinot had a good chance of ending the drought by winning this year.

This must be a kind of reverse superpower that enables me to curse someone by tipping them or to enhance their chances by writing them off. Maybe I was correct that the French wait for a successor to Bernard Hinault could end this year, but incorrect to pick Pinot. While Pinot, who won at the Col du Tourmalet on Saturday, is not yet out of it, Julian Alaphilippe, currently in yellow, is better placed – though in writing this I may have cursed him, too.

If nothing else, my hopeless forecasting demonstrates that cycling can be a difficult sport to predict. It’s not just me. Even sophisticated machines get it wrong with reassuring frequency.

In Toulouse on Wednesday I spent the afternoon in the NTT data truck, which each day is parked in the Zone Technique, close to the finish.

Inside it resembles an air traffic control tower, its two storeys full of screens, whirring computers and people staring at the screens and tapping on keyboards, crunching numbers and other pieces of information that come from the 170 or so sensors, one attached to each bike.

On the first floor is the predicting machine: a grey box beneath a screen showing its picks for the day. On Wednesday, a sprint stage, the machine gave Caleb Ewan a 17 per cent chance of winning, ahead of Elia Viviani (6 per cent) and Peter Sagan (5 per cent).

Ewan, an Australian sprinter riding his first Tour, had never won a stage. Viviani had already won a stage in this Tour, as had Sagan, adding to the 11 he has won in his career.

Ewan was a contender, but such a strong favourite? It seemed eccentric that the machine should rate his chances so much higher than those of Viviani and Sagan. And yet in Toulouse a couple of hours later, when the bunch came heaving into town, who should cross the line first, by about a centimetre, than Ewan? On this occasion the machine was right.

But the most interesting thing is that it is not always right. In its daily predictions game against journalists, the journalists are narrowly ahead. And although there is still a long way to go, and a lot of high mountains in the final week, the machine’s pick for the overall win, Bernal, currently looks a long shot (despite what you might have read here a fortnight ago).

In the bus I spoke to Peter Gray, NTT’s vice president of global advanced technology. He’s been coming to the Tour since 2015, when his firm first hooked up with ASO, the Tour organisers, to become official data partners.

Cycling and the Tour de France is behind the curve when it comes to collecting and using data, said Gray, though some teams are making major gains.

“There’s an opportunity for data savvy tems to continue to get a competitive advantage, but we’re seeing this even more in other sports,” he said. “You’ve got football clubs going and buying analytics companies: that tells you that they’re seeing value there.”

I was curious about the predicting machine. Gray told me that it uses six years of data, from all races, not just the Tour de France, to create a profile of each rider that takes into account their recent form, the nature of the course, who they’re racing against, the strength of their team, among other bits of information.

From all this, it generates a prediction, one that is right roughly as often as it is wrong. “It is generally able to make predictions in line with the experts,” said Gray, “though sometimes it surprises us.” Forecasting Ewan’s win in Toulouse was a good example of a leftfield prediction that caught out many of the experts and proved right.

But there are as many days like last Monday, when crosswinds ripped the race asunder. In these situations, when the peloton splits into echelons in crosswinds, bike racing is hugely complicated and unpredictable: it is less about physiology and form than about concentration, positioning and team support.

On Monday Pinot and Rigoberto Uran were caught out, largely because their teams were badly positioned. Team Ineos and Deceuninck-Quickstep, on the other hand, packed the first echelon with their riders and drove home their advantage, allowing their leaders, Thomas, pictured, and Alaphilippe, to gain serious time on Pinot and Uran, perhaps fatally damaging their chances of winning the Tour.

It was a great tactical move by these two teams, but the losses of Pinot and Uran owed little to any lack of form and fitness on their part. Pinot in particular is in the form of his life. But it is one of the great misconceptions about cycling that it is purely a test of endurance and strength. There are simply too many other variables: weather, road surfaces, luck, skill, to name a few. Otherwise, you could hold the Tour de France in a laboratory.

As Gray said, “The uncertainty is what makes it fun to watch. We couldn’t have predicted those splits on Monday. The machine didn’t predict that. None of the commentators predicted it. If the machine could predict everything 100 percent it would be really boring.”

For the moment the machine is sticking with Bernal for the overall win. Interestingly it has not, from the start of the race, rated Thomas’s chances very highly. Thomas cracked towards the summit of the Tourmalet, losing some time. He remains second overall, and is far from out of it. But perhaps the machine knows more than we think.