Bernardo Silva described Pep Guardiola’s centre-back experiment against Arsenal as “uma loucura” — madness. Guardiola then moved the experiment to left-back, and still called the first-half version “horrible.”

In that first plan, Nathan Aké was at left-back and Bernardo was next to Rúben Dias. After another training session, Guardiola altered the setup. Aké moved inside to centre-back, Bernardo went to left-back, and the most extreme version stayed on the training pitch. ¹

Bernardo was the left-back, Aké was inside him, and Manuel Akanji started on the bench. The change affected more than one position. It changed how City built from the back, how they defended Arsenal’s right side, and who had to cover the space when Bernardo moved inside or stepped out.

Arsenal’s right side shaped the first half

Guardiola said afterwards: “I tried something new [at first] and it was horrible. [In the] second half, we were more like we are.”

The issue was visible on Arsenal’s right. Bukayo Saka held the width. Martin Ødegaard worked inside. Bernardo was often caught between the two. If he stayed narrow, Saka could receive near the touchline. If he moved out early, Arsenal had space inside.

That is a difficult full-back problem in any match. It was even more difficult for a midfielder playing the role as part of a one-off plan.

At half-time, Guardiola changed the shape. Akanji replaced Riyad Mahrez, Bernardo moved to the right wing, and City moved closer to their usual 4-3-3 in possession. ²

Guardiola later said he talks with his staff, reads the training sessions, looks at the problems his players are having, and accepts the criticism when a tactical change is described as overthinking. In this case, the training week had already reduced the experiment before kick-off. ³

The training picture

As a centre-back, Bernardo would have been receiving the ball under pressure in the first line. As a left-back, he had to deal with Saka outside him, Ødegaard inside him, and Arsenal’s pressure after City’s first pass. In both versions, the same questions followed the ball: who gave him the next option, who covered behind him, and what happened when City lost possession?

The FA’s coaching resource on constraints lists the parts a coach can change in a session, including the pitch, the numbers, the rules, the rewards, and the way players are arranged. In this case, those details would decide how close the rehearsal came to the Arsenal match. A clean passing drill would not recreate Saka receiving wide or Ødegaard finding space inside. ⁴

Browne et al.’s Australian football study found that match simulations looked closer to competition than grouped small-sided games when the researchers measured how the practice conditions worked together. The sport and sample limit how directly that can be applied to Premier League football, but the connection is still useful: a role change gives better information when the pressure, teammates, and opponent actions look closer to the match. ⁵

Machado et al. found in youth soccer that changing the rules and setup of small-sided games changed how teams behaved. They also found that adding more rule changes appeared to reduce the variety of solutions players used. For a role like Bernardo’s, that matters because a heavily conditioned drill can become too artificial. It may show whether the player follows the rule, rather than how he deals with the winger, the inside pass, and the counter-attack. ⁶

The running part of the role

Bernardo at left-back was not just Bernardo standing 20 yards deeper. The role changed the runs he had to make.

A midfielder moved to full-back may have to defend longer sprints toward his own goal, recover into the back line, or chase a winger from a wider starting position. The player can be comfortable on the ball and still face a very different game without it.

A Bundesliga study found that when players switched positions, their running changed with the role. The authors estimated that the position itself explained roughly 44–58% of the change in total distance, high-intensity distance, sprinting distance, and accelerations. The study covered 25 players from the 2019/20 season who had completed at least four full matches in at least two positions, and the authors still found large individual differences. ⁷

That helps explain why a role can look possible in possession but still be difficult across a full match. Bernardo could receive, combine, and help City move the ball. The harder part came when Arsenal went the other way. A turnover could leave him recovering toward his own box, defending Saka in space, or trying to get back into the defensive line from an unfamiliar position.

The cover around the experiment

City’s use of John Stones later offered a different version of a positional change. Stones started as a centre-back, stepped into midfield when City had the ball, and dropped back when possession was lost. The movement was unusual, but City kept cover behind it. They usually had three centre-backs behind the ball and at least one of Stones or Rodri protecting the middle.

After the switch, direct attacks conceded fell from 1.2 to 0.6 per game, opposition progression speed dropped from 1.47 metres per second to 1.18, and goals conceded fell from 0.9 to 0.5 per game. ⁸

The 2021 Champions League final showed a different kind of change. City left out both natural holding midfielders, Fernandinho and Rodri, and used Bernardo with Kevin De Bruyne and Ilkay Gündogan. They “never really managed to get a grip on midfield” and finished with one shot on target in 97 minutes. ⁹

Those examples sit either side of the Bernardo experiment at Arsenal. Stones moved into midfield with protection behind him. The Champions League final removed the usual midfield protection. Against Arsenal, Bernardo’s left-back role changed the balance on City’s left side and forced a half-time correction.

By the second half, Akanji was on, Bernardo was back on the right wing, and City were closer to their normal structure. The most radical version of the idea had stayed in training. The softened version reached the match, and Arsenal’s right side still made it uncomfortable enough for Guardiola to change it.