FairValue

Checking model

Esc

  • EnvironmentOverviewWhat FairValue measures, what it found, where to go next.
  • EnvironmentTerminalExplore, compare and track players; market, league and model views.
  • EnvironmentStoriesValuations, head-to-heads and research findings, told as stories.
  • TerminalPlayer ExplorerSearch, filter and sort the player export.
  • TerminalComparison StudioTwo to four players on one scale.
  • TerminalWatchlistsNamed collections, kept in this browser.
  • TerminalMarket LandscapeHow observed value is distributed, and where the model drifts.

↑ ↓ to move · Enter to open · Esc to closeLoading player index…

João

Costa

CFPortugal · CD Santa Clara

№ 856423Emerging €703k −0.56

Player dossierLiga Portugal

JoãoCosta

CFCentre-ForwardAttack

Club
CD Santa Clara
League
Liga Portugal
Nationality
Portugal
Age
26
Height
185 cm
Foot
Right
Valued
28 May 2026

Reading the valuation

The market values Costa at €400k. Model A, which never saw Costa during training, estimates €703k: the market value is 43% below that estimate.

That is a factor of ×1.76. Model A's typical held-out miss is ×1.50 either way, so this gap is larger than its usual error.

Without league and club, Model B estimates €1.6M, further from the market. Model B is usually further off (typical miss ×2.07): league and club explain much of how the market prices players.

A gap describes the model's arithmetic for this player's inputs, not what the player is worth.

▼ −0.56A €703k▼ −1.37B €1.6M
Pitch length is a log value scale fitted to this player's figures. ● observed · ◇ Model A · dashed ◇ Model B · each arc is a residual, log(observed) − log(estimate).
Apps
35
Minutes
1,034
Goals
3
Assists
1
Started
11/35
Min / app
30
Cards
3Y 0R

Club matches, two years to the valuation date

The Valuation Journey

What moves Model A

Asking the prediction API to explain Model A for Costa…

SHAP describes the model's arithmetic for this player's inputs, not what causes market value. Model B can be explained only by the live service; Model A also has a published explanation for the research players.

Model A vs Model B against the observed €400k

Checking API
EmergingEstablishedEliteWorld Class€2M€10M€30M€100k€1M€100MMarket €400kModel A €703k

Asking the prediction API…

Held-out estimates come from cross-validation, with this player outside the training data: the honest comparison. Served models were refitted on every player, this one included, so they sit closer to the market. Shaded band: Model A's typical miss (×1.50 either way).

Player traitspercentiles vs 1,583 attackers

João Costa's percentiles among attackers: Goals per 90 49th, Assists per 90 26th, Minutes, two years 20th, Starts 12th, Minutes, last 365 days 8th, European club minutes 81st49thGoals26thAssists20thMinutes12thStarts8thRecent81stEurope

Dashed ring: the attacker median (50th percentile)

European club minutes

Share of minutes played in UEFA club competitions.

18% this player0% attacker median81st percentile

0%peer deciles · median line · this player ●100%

No passing, progression or defensive actions exist in the dataset, so no trait claims to measure them.

Valuation history every Transfermarkt valuation on record

First valuation
€50k
Jun 2021
Peak
€500k
May 2025
Model target
€400k
May 2026
Valuations
6
2 club changes
May 2026€400kCD Santa ClaraValuation the model is trained to reproduce

May 2026: €400k, CD Santa Clara. Dotted lines mark the €2M, €10M and €30M tier boundaries.

Show as a table
DateMarket valueClub
2021-06-29€50kCF Estrela Amadora B
2024-06-20€350kFC Alverca
2024-12-19€350kCD Santa Clara
2025-05-30€500kCD Santa Clara
2025-12-18€400kCD Santa Clara
2026-05-28€400kCD Santa Clara

Audit context group-level, not about this player

Audit group
UEFA
Population gap, Model A
+0.093 95% CI +0.063 to +0.124
This player, market vs held-out A
−0.564

The group comes from first-listed citizenship and is never a model input. Across the population, non-UEFA players' market values sit about 10% above Model A's held-out estimates, relative to UEFA players in the same leagues. One player's residual says nothing about that pattern (typical miss ×1.50), and none of it is evidence of discrimination: the target may itself be biased. Open the Fairness Audit

In context observed value among comparable players in this build

All players in the Terminal →

Liga Portugal423 players · median €1M

Valued above 12% of the 423 Liga Portugal players.

€20k€200k€2M€20M€400kSee them in the Explorer →

Attackers, all leagues1,583 players · median €2.5M

Valued above 11% of the 1,583 attackers.

€20k€300k€5M€75M€400kSee them in the Explorer →

Where Costa appears

Watchlists and the comparison tray are kept in this browser only.

Near this valuation closest observed values: same league, then same position

Counts and ranks of observed Transfermarkt values only; model errors are never ranked here. Full player export: Every player in the modelling population (D-011).