How We Calculate Market Values
Unlike men's football, where transfer fees are public and frequent, women's football relies heavily on free transfers and undisclosed fees. Theno.10 uses a transparent, deterministic model — a documented formula whose every coefficient is published and versioned, not a black-box algorithm. Every factor that goes into a valuation is inspectable, and the model is recalibrated against real transfer fees and external reference values.
1. The formula
Each player starts from a positional base value which is then multiplied by a set of context factors:
- Performance — per-90 metrics (goals, assists, xG, xA, progressive passes) and minutes, normalized to percentiles against positional benchmarks and weighted differently per position. Small samples are shrunk toward the mean so a couple of goals in 300 minutes can't spike a valuation. Metrics we don't yet hold for a player are excluded and their weight redistributed — never faked.
- Age — position-specific curves (goalkeepers peak later than wingers), with a bonus for U-21 talents already performing.
- League — competition strength multipliers (WSL and NWSL strongest, then Liga F, D1 and Frauen-Bundesliga), plus a UWCL participation bonus.
- Contract — remaining duration: a club with years of control commands a premium; expiring deals reduce leverage. Free transfers are the norm in the women's game and are treated as such, not punished.
- International — national-team tier derived from the live FIFA ranking.
- Availability — share of possible minutes played, with a discount for active injuries.
- Reputation — major honours (Ballon d'Or, World Cup, UWCL…) with time decay, capturing the brand and scarcity value a pure stats formula misses.
Two of those factors are close to dormant given what we currently hold. On 27 July 2026 only 13 players in the whole database carried any honour, and there was one active injury on record — so reputation and the injury discount move almost nobody's valuation today. They are in the formula because the data will fill in, not because they are shaping the numbers you see. Coverage of the performance inputs is uneven too: 835 of 16,311 season rows hold an xG figure (xA 834, progressive passes 986), which is why missing metrics are excluded and their weight redistributed rather than treated as zero.
2. Anchoring to the real market
The formula's output is then anchored to real market evidence where it exists:
- Confirmed transfer fees — the strongest evidence we hold. A recent confirmed fee heavily outweighs the model (and acts as a floor for fees under a year old): a player who just moved for €1M cannot be listed at €300K. The anchor's weight decays with time.
- Reference values — editorial and external market references nudge the value softly. The model leads; references are primarily used for calibration and QA, and are ignored for players without an established track record.
3. When we don't know enough, we say so
Players with too little data (few minutes, incomplete profile) enter cold start: their value is capped low regardless of what the formula would say, and only real market evidence on a player with an established record can lift it. Players with no meaningful minutes, no honours and no market evidence are not valued at all — their profile shows "valuation pending" instead of a made-up number.
4. Confidence intervals
Every valuation ships with a confidence interval (e.g. €250K – €350K). The base interval is ±20% and it widens when data is missing — unknown contract, no xG/xA, a single data source — up to ±50% for cold-start profiles. The interval is honest about what we don't know.
5. NWSL trades vs. European transfers
Since the 2024 collective bargaining agreement the NWSL has no college draft and every player reaches free agency when her contract ends; trades between clubs still exist (with player consent) and allocation money is being phased out. Historic trades from the draft/allocation era are kept as recorded and labelled as trades. We show an amount only when a source confirms it — we do not convert trade packages into an artificial euro equivalent.
6. Statistical catalogue and derivations
Season stats merge multiple providers with per-metric provenance: each metric records which source wrote it and with what confidence. A lower-confidence source never overwrites a higher-confidence value, and a source that lacks a metric never blanks another source's data. Profile pages show a source badge per stat block.
Confidence is set per metric, not per provider, and the distinction matters: StatsBomb writes its event metrics (xG, xA, progressive passes, pressures) at 90, but its volumetric counts — appearances, passes, duels, tackles — at 60, because other providers cover a full season while StatsBomb covers only the matches in the open dataset. So the ordering is 90 (StatsBomb events) > 70 (Sportmonks) > 60 (StatsBomb counts) > 30 (API-Football). Measured on 27 July 2026, 15,422 of StatsBomb's 21,901 writes are at 60 — the majority of what it writes loses to Sportmonks, which is the intended behaviour and not a demotion.
Some advanced metrics are derived from StatsBomb event data with documented heuristics:
- xA — the xG of the shot a pass created (StatsBomb Open Data has no native xA field).
- Key passes — passes flagged as creating a shot (shot_assist / goal_assist).
- Progressive passes / carries — actions gaining ≥12 / ≥10 field units toward the opponent goal.
- xG against (goalkeepers) — the xG of opponent shots faced, attributed via the shot freeze-frame.
- Clean sheets, starts, appearances — derived from official lineups and match scores.
- Duels, tackles, interceptions, recoveries, pressures — counted from the raw event stream.
Provider match ratings (0–10 scale) are a single provider's opinion, always displayed with attribution and never blended across sources. Percentile ranks compare a player only against peers of the same position family (GK / DEF / MID / FWD) in the same season, and are published only when at least 10 peers hold that same metric. Below that the value is still shown but the percentile bar is not: with five peers a single transfer swings any bar by twenty points.
7. Versioning and calibration
The model's coefficients are versioned and every recalibration is documented. Production currently runs multiplier config 2.6.0, calibrated on 9 July 2026, against model contract `deterministic-v2.0` — the identifier stamped on 18,080 of the 18,130 valuations on record. Calibration compares the model's output against real disclosed fees and external references; when the model drifts from the market, the coefficients change — not the individual numbers.