After every match, apps like FotMob and Sofascore show a rating out of 10 next to every player's name. Fans argue about them constantly, but few know how they are actually produced. The short answer: most modern ratings are not opinions at all. They are mechanical scores, calculated from match events by an algorithm, then adjusted for position and context.

The most common system starts every player from a baseline, typically 6.0 or 6.5, and then adds or subtracts points for what happens on the pitch. FotMob's model, which builds on the WhoScored rating engine, uses roughly 30 or more inputs. Goals are large positives, assists are medium-large positives, and key passes, completed dribbles, successful tackles, interceptions, won aerial duels, accurate crosses and won penalties all add smaller amounts. On the negative side, missed clear chances, lost duels, dispossessions, errors leading to a shot or a goal, fouls, cards, missed penalties and own goals all pull the score down.

Crucially, the same action is not worth the same amount for every player. The weights are position-specific. A goal scored by a centre-back is worth more than a goal scored by a striker, because it is a rarer event. Tackles are weighted more heavily for defenders than for forwards, while pass completion percentage matters most for midfielders. Some providers build from 51 individual statistics covering every position, so a centre-back and a striker are never judged against the same checklist.

Sofascore takes a different route: a machine learning model that assigns a value to every offensive, defensive and goalkeeping action based on its context and impact. Actions are grouped into five categories, shooting, passing, dribbling, defending and goalkeeping, and each action is scored by how much it changed the team's chance of scoring or conceding. That is why not every pass is equal: a progressive pass that moves the team toward goal is worth more than a safe sideways pass, even if both are completed.

Context also feeds into the numbers. Opposition strength and game state can adjust the value of events, and ratings are often updated in real time during the match rather than calculated once at full time. Substitutes get special treatment: most systems only rate a substitute after a few minutes on the pitch, unless they immediately register a high-impact event like a goal, an assist or a red card.

What ratings cannot measure is just as important. No algorithm sees positioning, movement off the ball, leadership, pressing intensity or tactical discipline well, and defenders and goalkeepers often score lower than attackers simply because of the stats involved. A quiet, perfectly disciplined defensive performance can look ordinary in numbers, while a forward who scores once and does little else can look outstanding. Journalist and pundit ratings, by contrast, are subjective and can capture those intangibles, but they carry their own biases, including team bias and recency bias.

The sensible way to use player ratings is as a starting point, not a verdict. A 6.0 usually means an average, unremarkable performance. Anything above 7.5 suggests a strong game, and 9.0 or above is reserved for genuinely outstanding displays. Compare players in the same position, look at the underlying stats behind the number, and remember that a single match rating is a snapshot, not a judgement of a player's quality. Used that way, ratings are a useful tool for understanding a match, not an argument to end one.