Moneyball meets cycling: How underdog team NTT is using data to turn the tide

Don’t miss out on the latest CyclingTips updates.

0
Jump To Comments

Team NTT will be the first to admit it hasn’t set the WorldTour alight. In its first three seasons at the highest level, the team formerly known as Dimension Data finished last on the team rankings. Last year, its fourth in the big leagues, Dimension Data finished second-last among WorldTour teams, securing fewer UCI points than five second-division teams.

In this context, it makes sense that Team NTT is looking to improve; to force its way into the top 10 on team rankings. But even without attaching a timeframe to that goal — at least publicly — it’s an imposing target. Based on last year’s rankings, the team will need to double the number of points it accrues to land in the top 10. So how will NTT do that exactly?

For those involved with the South African team, one source of hope comes via the team’s title sponsor, Japanese telecommunications company NTT. Over the past 12 months NTT, in conjunction with the team bearing its name, has been working on a data analytics solution that, if all goes to plan, should help Team NTT climb the rankings.

ProCyclingStats on steroids

Peter Gray is the Senior Vice President of NTT’s Advanced Technology Group: Sport. He’s a key conduit between NTT and its cycling team — besides working in the data analytics space, he knows and loves his bike racing and is a keen cyclist (and VeloClub member) himself.

As Gray told CyclingTips in late 2019, the foundation of NTT’s analytics-based push to move up the rankings is a classification of just about every professional male cyclist on the planet.

“Over the last few years and together working with the team, [we] have developed out a number of statistical and analytical models for basically classifying different riders, studying their career trajectories and their form,” Gray explains.

Built upon data from ProCyclingStats — which the team has a partnership with — NTT’s analytics package allows the team to distil every rider down to a matrix of strengths and weaknesses.

“It’s basically a process that we have running in our cloud platforms that actually is running every day, scraping data from their various places and calculating all of the key metrics,” Gray said. “And then we basically … encapsulate that into a dashboard that we’re able to give to Doug [Ryder, team principal] and the team management to sit on their laptop.

“They can literally sit there with the laptop and go and look at different scenarios and different groups of riders, deep-dive into an individual rider to look at their detailed profile, their form trajectory, all of their race results, etc.

“So it’s kind of like, I guess, going to a rider page on ProCyclingStats, but then with a bunch of additional information that we’ve calculated out and created all these additional metrics that then the team can use as part of that decision-making process.”

NTT’s dashboard allows team management to find riders based on particular characteristics they’re looking for.

Rider recruitment

So what do you use a detailed database of every rider’s strengths and weaknesses for? Rider recruitment, for one thing. As Gray explains, detailed rider metrics help team management work out which riders they should be speaking to, seeking power data from, and ultimately considering as potential recruits.

“Doug Ryder, who’s the team owner, he’ll say, ‘Give me the list of the 20 riders that I need to be talking to. These are the characteristics and the criteria of the types of riders that we’ll be looking to fill in our roster. Who should we be talking to?,'” he said.

Identifying the best riders in a certain category is relatively trivial — anyone with even a passing knowledge of the sport knows that, say, Egan Bernal is a great climber. What NTT’s solution allows the team to do is look beyond the big names, to the true “moneyball” performers — the riders who have previously been undervalued but who tend to punch above their weight.

“The challenge comes with the guys who are less-well-known,” Gray said. “[They] might be the up-and-coming riders or, in the case of some of the guys that we’ve recruited — Ben Dyball is a good example somebody who has probably been a bit overlooked.”

NTT’s solution looks both at what the riders have done in the past, and what their form says about the trajectory they might be on.

“Is it [the rider’s form] highly variable? Has it been a consistent growth pattern? Is it on the incline? Is it on the decline?,” Gray said. “Mapping that against some statistical averages … starts to give us a bit of a feel for riders who we think have a good chance to succeed in the types of roles that we’re looking to fill within the team.”

And ultimately that’s what it comes down to: finding riders for specific roles, to help fill out the roster.

“If you’re looking for an U23 rider who has strong climbing capabilities and is OK at time-trialling, we can quickly, within a couple of clicks, go ‘OK, here are the 10 people that we need to be talking to.’” Gray said. “You look at the top three and you go, ‘Well, yeah, probably Mathieu van der Poel and Egan Bernal: we’re not going to hire those guys this year.”’ But then, once you start looking down that list you start to identify … the guys who start to become opportunities.”

NTT’s recruits for 2020 and their strengths and weaknesses, as per the team’s data analytics solution.

More than numbers

Of course rider recruitment is about much more than looking at how riders have gone in the past.

“It’s one thing to have a data scientist — they can go away and look at a whole bunch of numbers and come up with stuff,” Gray said. “But what you really need to be able to do is then bring the knowledge and the expertise of the people who’ve been working in cycling for years, year after year after year, to then help validate and refine the models.

“You’ve got a set of sports directors who have more years than I’ve been alive working in professional cycling. They know the sport inside out. You’ve got coaches and sports scientists who have PhDs in their field working on this. And so then this data really is there to add and supplement the process and also to give them the information to make faster decisions as well without having to do any unnecessary legwork.”

Not just rider selection

NTT’s metrics solution isn’t just about recruiting riders for the new season. It’s also used for optimising the team’s calendar, making sure that Team NTT is using its resources most effectively.

“I think it’s well known that the team struggled this year and it really needs to win a much greater percentage of the UCI points that are available,” Gray said in late 2019. “So given our team, the riders within our team, where are the best opportunities for us to win those points?

“So firstly, which races should we be targeting? And then secondly, which riders should we be targeting those races with? So there’s a bunch of work going on in the background around statistically — and through analytics — identifying who are the riders that are best targeted to which races. And also, what are the team profiles that you target for different races.”

A post on the NTT website explains the process in more detail.

“The algorithm that the team uses takes into account the points available by race, calculates the predicted field strength based on the caliber of riders who have competed in the past, and creates a points-per-race-day weighted by the field strength indicator,” the post reads. “This shows them where the best value races are – where they are likely to achieve the best results and therefore earn the most points.

“From there they can rank the races best suited to the logistical capabilities of the team and from that build out the race calendar. Coupled with our rider and team composition analyses, they can pinpoint who they should be sending to which races.”

NTT has found a way to quantify the biggest “bang for buck” on the racing calendar.

One of the main findings from this analysis is that, according to the team, UCI points are four times easier to come by in Asia Tour .HC (now ProTour) races than in European WorldTour races. It’s for this reason that Australia’s Ben Dyball joined the team.

For years the Australian was regarded as a WorldTour-quality rider but, for one reason or another, the now-30-year-old was frequently overlooked. In late 2019 Dimension Data announced that it had signed Dyball on a one-year contract.

Dyball slotted into the team as NTT’s third-highest UCI points earner from the past 12 months, courtesy of winning the Tour de Langkawi and finishing third at the Tour of Qinghai Lake — the Asia Tour’s two biggest stage races. Only Michael Valgren and Ryan Gibbons scored more points than Dyball in 2019. NTT’s hope was that Dyball could help the team secure a raft of UCI points in Asia Tour races throughout 2020.

Sadly for Dyball and NTT, Dyball was diagnosed with Epstein-Barr Virus in December 2019 and subsequently missed both the Tour Down Under (where significant UCI points were a real possibility) and Tour de Langkawi (where he was the defending champion). Dyball will be back racing again soon, but the team will be ruing the missed opportunity to get some valuable points on the board early.

A unique strategy?

It would be naive to think that NTT is the only team looking at gaps in its line-up and trying to recruit to fill those gaps. It’s also unlikely to be the only team with a data analytics solution of some kind to help with the process. But according to Gray, NTT is ahead of the curve in the road cycling space.

“We’ve gone, I think, a long way beyond what other teams are doing,” he said. “I’d be very surprised if there aren’t teams out there who are doing some sort of analysis and using data as part of their recruitment process. But I’d also be pretty surprised if there are any other teams that have access to the kind of capabilities that this team has. Certainly there’s nobody that we’ve spoken to or identified who seems to be doing anything at that sort of level at this stage.”

It will happen though, according to Gray. Data analytics has been heavily embraced in other sports and it’s perhaps just a matter of time before cycling heads in the same direction.

“Baseball is obviously the most high profile example of where analytics has been heavily used,” he said. “European football — Arsenal, for example, acquired an analytics company to work with them. If you’re Arsenal you can afford to do that; not every sporting team can.

“But it’s certainly something that, across other sports, is becoming more mainstream. And I think this is the opportunity to start bringing some of that capability into the professional cycling space.”

A ranking of every NTT rider for the 2020 season.

Things are looking up

Team NTT’s quest for more UCI points in 2020 is off to a good start, thanks in no small part to Giacomo Nizzolo. The Italian sprinter won a stage of the Tour Down Under; reached the podium at Race Torquay, a stage of Tour de La Provence and Kuurne-Brussels-Kuurne; and just this week won a stage of Paris-Nice. Add in a stage win at Etoile de Besseges for Ben O’Connor, plus two stage wins at the Tour de Langkawi for Max Walscheid, and as of the last teams ranking update, NTT is up to a more-respectable 17th place. Were it not for Dyball’s absence, the team would almost certainly be higher up the list.

As things are tracking, Team NTT will probably struggle to break into the top 10 on the World Ranking in 2020. But the season is still young, and with coronavirus running roughshod over the racing calendar of late, just about anything is possible. For now, the team will be encouraged by the fact that, after less than three months, it already has twice as many WorldTour wins as last year and only one less victory overall.

Editors' Picks