Nadal's victory against Big Data: how to come back when mathematical models give only 4% options

• José Perlas The best Nadal arrived in its original version

• Chronicle Nadal displays all his mystique against Medvedev

• Champion of 21 'majors' Full-fledged among the greatest in history

• Emilio Sánchez-Vicario The victory in the battle of the initiative

• Reactions Roger Federer's congratulations

• Analysis The law of the eternal return of Nadal

• Opinion A greatness that Spain had never seen and will never see again

• Data All the numbers of the best Spanish athlete

The clock at the Rod Laver Arena marked 12:35 p.m. and the final three hours had not yet passed, but

already had a two-set lead and was leading 1-0 in the third.

At that time, according to Smarkets, a British betting exchange company, the Russian's chances of victory reached 98%.

At that very moment, according to the official Australian Open tool,

's comeback options did not exceed 4%.

After all, in his previous 62 appearances at Grand Slams he had only gone 0-2 against

(Wimbledon, 2006) and

(Wimbledon, 2007).

In addition, Medvedev had a record of 33-0 in

played on hard court after winning the first two sets.

The Muscovite's power from the back of the court, who dominated 24-19 in points decided by nine or more shots, seemed to have decided the fate of the match.

Or at least so the computers believed.

Since 2018, the advanced statistics of the Australian Open have been signed by the Game Insight Group (GIG), a prestigious group of computer scientists and engineers, under the umbrella of the Australian Tennis Federation and the University of Victoria, commanded by Dr.

. "Our mathematical models cover a wide spectrum. Not only the tennis player's state of form, nor the entity of his last opponents, but also the surface on which he plays or the percentage of points won with the service", they explain to EL MUNDO GIG sources. Obviously, the most intricate elements of its algorithm, the ones that allow it to stand out from the rest, are reserved with extreme caution.

In any case, the work of GIG is complemented by that of Infosys, a company in charge of transmitting the data to millions of fans, not only during the first

of the season, but in 60 more tournaments.

Including the last Roland Garros, where they enhanced their tools, based on artificial intelligence, at the service of tennis players and coaches.

After seven years of collaboration with the ATP, no one considers the Indian computer giant a newcomer to the circuit.

However, Nadal's enormous reaction shattered all his figures.

How to explain the triumph of the will against the power of the machines?

Kyrgios' former coach

Adrià Arbués

, a doctor from Pompeu Fabra University, offers some answers.

"A model of the probability of victory or defeat can be built based on the result and the remaining time. And from there, improve it little by little, with what we call

, the way in which each point and each game is broken down.

, begins this expert in artificial intelligence applied to sport.

"However, what is really interesting comes later, through the

data , facilitated by the previous work with tracking cameras. With them you can contextualize what happens, for example, with fatigue. And through there, how evolves the strategy on the track".

In the case at hand, the Australian Open maintains an alliance with Swing Vision, an application that analyzes each stroke through a camera installed on a mobile phone.

It is one of the innovations introduced by

, head of the Australian Federation's Sports Science department.

And to process that brutal volume of information, the GIG has recently brought in

, former trainer of

and

.

However, in light of what was seen on Sunday, there is still a way to go.

"The objective is to find generalizable models, but at the same time they are also interpretable. Because sometimes there is a risk of giving too much information to the model. I am afraid that if in the final it had been applied more personally to Nadal's case, they would have refined more," says Arbués. "With Nadal you always have to take into account the key factor of his mental strength. That way in which he sees himself, even well behind the scoreboard, capable of completely turning the fortunes of the match. This is difficult to capture with data, but yes, it can be complemented through history", completes the data scientist at Zelus Analytics.

Arbués' words update, in a way, that conclusive judgment issued by

's coach

, after the 2018 Roland Garros final. "We are talking about the best competitor of all time. And I don't just stick to tennis "said the Austrian coach.

With a favorable 5-4 in the final round, after letting the resolution escape with his service, Nadal overcame another of those moments that would have taken away anyone's morale.

"There I thought: 'Fuck, I'm going to lose like in 2012 and 2017.'

Agreement with Barça Innovation Hub

Nor should we forget that before the final the GIG estimates gave Medvedev an advantage of 64% and that this figure remained almost intact at the start of the fifth set (63%), despite the fact that Rafa arrived with the momentum of the heroic and the breath of the stands.

Those balances, of course, did not leave in a very good place the predictions that Robertson himself had pronounced the previous weeks.

"Our new projects will add another layer of detail to understanding what will happen on the track."

In recent months, one of the achievements of the GIG was its agreement with the Barça Innovation Hub, the laboratory promoted in 2017 by

to apply technological advances to improve the performance of Barça football.

There is always the risk of something happening to break the mold.

Adrian Arbues.

"I don't think this is negative publicity for the companies involved, although it does draw a lot of attention to the general public. If there was no resounding failure within the model, all this will serve as learning to develop other types of tools. I am convinced that that everything must go for the better, with increasingly precise models", points out Arbués, who prefers to focus on the inescapable tenacity of the Majorcan.

"It seemed very exaggerated that it only had 4% and with a higher percentage there would not have been such a stir. Better models will come out, but that does not mean that what is seen on Sunday will continue to be something that is out of the sample. There is always the risk of something happens that breaks the mold", he concludes.

So far, after his success at Melbourne Park, Nadal's Grand Slam winning percentage stands at 87.9%, better than

's 86%

.

Only

, with 89.2%, is ahead in the Open Era.

The digits expand with 55% of points won in

, out of a total of 66,663.

Perhaps next March 4, during the 16th edition of Sloan Sports Analytics, organized by the Massachusetts Institute of Technology (MIT), some expert will find an explanation for this excess.

Or maybe Nadal is still inscrutable to the machines.

Conforms to The Trust Project criteria

Know more

• sports

• tennis

• Rafael Nadal

• Novak Djokovic

• HBPR

Australian Open 2022Nadal holds Berrettini and stands in the final, where he will seek his twenty-first Grand Slam title

Australian Open Nadal's longest battles: "Imagine someone behind you with a rifle"

Australian Open Djokovic, a player marked forever

See links of interest

• Last News

• Translator

• What

• Work calendar 2022

• Home THE WORLD today

• Master investigative journalism

• Ibiza - Real Saragossa