Football Player Ranking System

Ranking players across dozens of attributes is something spreadsheets get quietly wrong. Add up all the scores and the player who is average at everything beats the one who is genuinely exceptional at two or three things and unremarkable elsewhere. In football the second profile is often the more valuable - the player who does the one thing the team needs very well.

Positions aren't comparable either: a goalkeeper's attribute scores cannot be read on the same scale as a winger's.

What was needed was a ranking that doesn't dissolve standout ability into an average, and stays fair within a position.

I built a scoring and ranking engine based on a statistical threshold rather than a sum.

1. Comparison is confined within a position. Each position is evaluated on its own: a goalkeeper only against goalkeepers, a centre-back only against centre-backs. Meaningless cross-position comparisons are ruled out from the start.
2. An "exceptional" line is drawn for each attribute. Instead of a fixed score I used a statistical threshold derived from that position's own distribution: mean plus one standard deviation. If a player's attribute sits above that line, they genuinely stand apart on it - relative to their peer group.
3. Rank is driven by count of standouts, not total. Each player scores by how many attributes they clear that line on. A player scoring 67 across the board no longer blends into one scoring 89 on three attributes and 45 elsewhere.
4. The output reads directly. A list ordered by player and "how many attributes they stood out on" - a ranking a scout can interrogate, not a single black-box score.

The real benefit of the threshold approach: "good" isn't defined by hand, it emerges from the data. If the overall standard of the pool rises, the line rises with it, and the system needs no recalibration.

Stack: Python, pandas; position-segmented pivot tables, mean + standard deviation threshold, per-attribute rank aggregation.