How to Build a Data-Informed Scouting Shortlist

A scouting shortlist built on data is not a list of the highest-scoring players in a database — it is a filtered set of profiles worth a human scout's time, produced by narrowing a large pool down using statistics before anyone watches a single minute of video. Done well, it saves scouting departments from spending scarce viewing hours on players who were never realistic fits. Done badly, it produces a list that looks rigorous but is quietly shaped by bad assumptions baked into the filters. The workflow below is the difference between the two.

Before You Start: What Data Scouting Can and Can't Do

Statistical data is excellent at narrowing a pool and terrible at making a final decision. It can reliably tell you which players in a given position, age range, and competition tier are outperforming their peers on a defined set of metrics. It cannot reliably tell you how a player will adapt to a new league, a new tactical system, or a new dressing room, and it struggles badly with qualities that resist counting — leadership, coachability, decision-making speed under pressure that doesn't show up cleanly in the event data. Treat the process below as a funnel that gets a large pool down to a manageable one, not as a system that outputs a final answer.

It also helps to be honest about what "data scouting" replaces and what it doesn't. It replaces the earliest, most time-consuming stage of traditional scouting — watching hundreds of players across dozens of leagues just to find the handful worth a closer look — which is exactly the kind of broad, repetitive filtering a database handles far faster than a human eye ever could. It does not replace the scout's eye once the pool is small enough to watch properly. A shortlist workflow that tries to make the data do both jobs at once usually ends up producing a list that is either too large to be useful or too narrow to be trustworthy.

Define the Profile Before You Touch a Database

Start by writing down the specific job the player needs to do, not just the position. "Central midfielder" is too broad to filter against usefully; "a central midfielder capable of playing as a single pivot who can also progress the ball under pressure" is specific enough to translate into metrics. This step matters more than any technical filtering that follows, because a precise profile turns into a precise metric selection, while a vague profile turns into a shortlist padded with players who fit the position label but not the actual role.

Set Statistical Filters Wide, Then Narrow in Stages

• Start with hard constraints that are not really about ability at all — age range, contract length remaining, competition tier, availability signals such as recent injury history — since these eliminate players who are not realistic targets regardless of how well they perform statistically.
• Apply role-specific metrics next, chosen to match the profile written in the previous step rather than a generic set of position averages. A ball-progressing pivot should be filtered on progressive passes and pressure-resistant retention numbers, not on tackles won, which measures a different skill entirely.
• Use percentile rankings within a comparable population — same position, same competition tier, similar minutes played — rather than raw totals, since raw totals reward players who simply play more minutes rather than players who perform better per minute on the pitch.
• Resist narrowing too aggressively in one pass. A shortlist that goes from thousands of players to fifteen in a single filtering step has usually applied thresholds too strict to survive scrutiny; narrowing in two or three looser stages catches players who are strong on the metrics that matter most even if they fall just outside an arbitrary cutoff on a secondary one.

Adjust for League and Competition Context

The same raw number means different things in different leagues, and skipping this adjustment is one of the most common ways a data shortlist misleads. A progressive-passing rate that looks exceptional in a league with a slower average tempo and more space to exploit may be closer to average once measured against a league with tighter, faster defensive structures. The practical fix is not to discard cross-league comparisons entirely — that throws away useful signal — but to weight them, treating a statistic as more impressive the tougher the competitive environment it was produced in, and to flag explicitly when a metric's meaning is likely to travel differently once the player moves up or across competition tiers. This is also where age enters the picture alongside league quality: a strong statistical profile produced by a younger player in a lower tier carries a different kind of promise than the same profile from an established player already near his peak, since the younger case still has development runway that the numbers alone cannot quantify.

Check Sample Size Before Trusting Any Number

A standout percentile built on eight matches is not evidence of a skill; it is evidence of a short run of matches. The threshold for trusting a statistic should scale with how much year-to-year and match-to-match variance that particular metric typically shows — counting stats like tackles stabilise faster than shot-quality or expected-goals-based numbers, which need a larger sample before the noise settles into something closer to signal. A shortlist compiled without checking underlying minutes will systematically overrate players who happened to have a hot handful of matches and underrate players whose full-season numbers are strong but less flashy in any single stretch.

Cross-Reference With Video Before Anyone Commits Time

Data narrows the pool; video confirms what the numbers are actually describing. Before a name moves from the statistical shortlist to an active scouting target, a short video pass should check for the things statistics routinely miss or misrepresent — body shape on the ball, decision-making speed, off-ball movement patterns that don't register cleanly in event data, and any tactical role nuance that a single stat label like "central midfielder" can't capture. This step is not optional scepticism; it is where a large share of statistical false positives — players whose numbers were inflated by a favourable role, a weak surrounding cast, or a run of soft fixtures — get caught before they consume real scouting resources.

Common Mistakes That Undermine a Data Shortlist

The most frequent error is filtering on metrics that don't match the actual role definition, which produces a technically rigorous-looking shortlist built on the wrong criteria from the first step. The second is ignoring sample size and treating a short hot streak the same as a full season of consistent output. The third is skipping the league-context adjustment and comparing raw numbers across competitions of meaningfully different quality as if they were interchangeable. The fourth is treating the data shortlist as a finished product rather than a funnel, and skipping the video cross-check because the numbers already look convincing on their own.

A fifth mistake, subtler than the rest, is anchoring the entire search on one standout metric because it happens to be easy to sort a database by. Progressive-carry totals or expected-goals numbers are useful, visible, and simple to rank, which makes them tempting to lean on as a stand-in for the full profile. But a role defined by five or six requirements collapsed into a shortlist built off one sortable column will systematically miss players who are strong across the full profile but do not happen to top the leaderboard on that particular number. The fix is not to distrust the standout metric, but to treat it as one filter among several rather than the entire search.

Checklist Before the Shortlist Goes Out

• Profile defined in role-specific terms, not just a position label.
• Filters applied in stages, from hard constraints to role-specific metrics to percentile comparisons.
• Every included metric checked for league and competition-tier context.
• Sample size verified for every standout number, with short-sample outliers flagged rather than trusted outright.
• Video cross-check completed before any name is treated as an active target.

Platforms such as RubiScore publish the per-90 and percentile data that this kind of workflow depends on, filterable by position, competition, and age — the raw material for the funnel, not a replacement for the judgment applied at each stage of it. That underlying dataset is available at rubiscore.com.