Percentiles: why raw numbers lie across leagues
Here's a number: a midfielder completes 88% of his passes. Is that good? You genuinely cannot know. Not without knowing his position, his league, his team's style and what everyone else like him posts. Raw football numbers don't carry their own meaning, and that's the single biggest trap in scouting with data.
Percentiles are the fix, and they're the spine of everything on FDscout. Here's how they work and how to read them.
The problem with raw numbers
Football stats lie by context in three ways at once:
- Position. 88% passing is routine for a centre back playing safe sideways balls and outstanding for a winger receiving under pressure in the final third. The same number describes opposite achievements.
- League. Leagues differ wildly in tempo, pressing intensity and how their data even gets collected. Three take-ons a game against passive defenders is not three a game in the Premier League.
- Team. A dominant side gives its players more touches, more passes and more of everything countable. Volume stats partly measure the badge on the shirt. (This is also why normalisation matters; see per 90 vs per 100 touches.)
Compare two raw numbers across any of those boundaries and you're comparing apples to weather.
What a percentile actually says
A percentile converts a raw number into a rank against a defined group of peers. If a striker's shot volume is in the 90th percentile, he attempts more shots than 90% of the players in his comparison pool. That single translation carries all the context the raw number was missing, provided the pool is built honestly. On FDscout the pool is always: players in the same position group, across all 30 leagues, over the same 12-month window, above minimum-minutes floors. Every percentile you see answers the same precise question: among current players who do his job, where does he rank?
Three details make the numbers trustworthy:
- Floors first. Players without enough minutes never enter a pool, so nobody posts a 99th percentile off four substitute appearances. (Why sample size matters so much is its own topic.)
- Lower-is-better stats are flipped. Getting dispossessed rarely is a strength, so for stats where small numbers are good, the percentile is inverted. High is always good, everywhere on the site.
- Missing data is honest. Some leagues don't record certain stats. Rather than guessing, FDscout shows N/A and excludes the player from that stat's pool only.
How to read them quickly
FDscout colours every percentile by band: under 35 is a genuine weakness, 35 to 65 is par for the position, 65 to 95 is a real strength, and 95+ is elite, the top handful of players in the world doing his job. A profile radar is exactly this, drawn in a circle: long bars are what he's for, short bars are what he isn't.
Two habits make you a sharper reader. First, read shapes, not spikes: a winger with 95th-percentile dribbling and 20th-percentile chance creation is a specific player with a specific role, not simply "good at dribbling". Second, remember that a percentile ranks within the pool, not across levels of difficulty: opposition quality still varies, which is why every FDscout ranking (like our live 1v1 wingers table) states its pool and floors up front.
Why this beats eye-test arguments
The eye test answers "does he look good?", which depends on which games you watched and what you hoped to see. A percentile answers "is this output unusual for his job?", which is the actual scouting question. The two work best together: the data tells you where to look and what's abnormal; the video tells you why and whether it translates. Skip the first step and you're scouting on anecdotes; skip the second and you're signing spreadsheets.
Every stat on every FDscout profile is a percentile with the raw value alongside, so you never have to choose between the number and its meaning. The full list of what we measure lives in the stats glossary.