TOUR 20 1940 | July 25, 2019 | Cycling Weekly
NT T’s @letourdata Twitter account is
using data from a new machine-learning
model that ’s being trialled at this year’s
race for the first time. The model, called
Le Buzz — the result of an in-house
innovation competition — analyses the
shape of the peloton to try and predict
key events in the race such as a split in
the bunch, or a crash, for example.
Even if you’ve not felt the need to
check that out you may have seen its
new ‘augmented reality ’ overlays being
shown on T V (highlighting a rider in the
bunch, supplemented with speed data, for
example). A lso new for 2019 is a feature
on the Tour de France app which gives
fans at the roadside an ETA for the race —
handy for anybody planning a quick trip
to the bar, or the loo.
It appears this is exactly the sort of
information that cycling fans want,
judging by the fact that, according to a
recent NTT/Brighttalk.com webinar,80 per cent of sports fans access sports
content on their smartphones when
watching the event on T V, and even live
at the roadside.Pixel power
But what ’s behind those fancy pixels?
For the Tour coverage, it ’s all made
available by three small numbers —
latitude, longitude and speed — pinged
off a transponder, which is fitted under
the saddle of every rider. This data is
bounced off a helicopter and on to a
mobile data centre, where it ’s fed live to
T V and commentators and the Tour’s
‘race centre’ website, as well as examined
by high-tech computers and ‘enriched’
— for example by feeding in
historical data to aid analysis.
Once this is all done, those three
data points per second become
53 per second, which ultimately
produce an incredible 100
trillion digital ‘decisions’ made
across the whole Tour de France.
What the Tour does lack,
however, is data from the riders
themselves, such as wattages,
cadence and heart rates.
That data that is available at
some other races from Velon,
the organisation owned by11 WorldTour teams
including Jumbo-Visma,
Deceuninck-Quick Step
and Ineos to, in part,
make the most of this
k ind of information. That
system was first used
at the Tour de Suisse in
2016 and more recently at
this year’s Giro d’Italia.
Wade says: “The
current agreement
between [Tour owner]
ASO and the teams for
the Tour de France means
we currently cannot
capture other data. Our
research shows that
our audience are most
interested in the data we
already capture as well as
A I [A rtificial Intelligence]
predictions which we
already deliver as well, so
that ’s our focus.”
While the Tour de France doesn’t use
Velon’s system, the two do work closely
together when it comes to on-bike
cameras and have done for several years.
Velon cameras have been responsible,
for example, for some of the most
popular video clips to come out of this
year’s Tour — Bahrain rider Jan Tratnik
bunnyhopping onto the kerb to avoid a
crash on stage one being a prime example
(see boxout).
The 11 Velon teams have aggregated
the rights to such footage, meaning,
essentially, that such clips are owned
not just by one team but all of them and,
equally, that if a race wants one team to
have on-bike cameras, then all must be“NTT’s Twitter
account is using
a machine-
learning model”
Helicopters receive and relay
the peloton’s data trailReal and virtual viewing is
merging for today’s spectators