Active Trader Model™ Methodology

This page documents the model behind Estimated Active Traders™. Shared rules that apply to every PTC estimate — modelled universe, publication gating, baseline governance — are documented in the Estimate Methodology.

What is estimated

PTC estimates the number of people trading with a prop firm in the observed period — the firm's current active trading population. The unit is people, not accounts, and not cumulative sign-ups.

Population: The current active trading population of a firm, covering evaluation and funded traders together. The model does not split evaluation from funded populations, and it does not attempt to separate simulated-funded from other funded arrangements.
Unit: People. Where a firm reports accounts rather than people, the figure is converted before it is used (section 03), because one person may hold several accounts.
Not estimated: Cumulative registrations, lifetime customers, accounts ever opened, or dormant registered users. Those figures exist in the evidence set, but they constrain the estimate rather than define it.
Definitional limit: "Active" is defined by what firms disclose and what public scale signals can support. PTC does not observe individual account activity, so there is no fixed per-trader activity test such as a minimum number of trades in a period.

Firm-level values are modelled estimates, not company-reported figures, unless the page explicitly labels them as evidence-led.

Model universe

Being tracked and being modelled are two different cohorts. The public Active Trader estimate covers the Forex and Futures prop markets.

Cohort 01

Tracked market

Firms PTC actively researches and maintains as operating. Tracking a firm does not produce an estimate for it.

Cohort 02

Modelled universe

Tracked firms with a covered market assignment and at least one usable calibrated scale signal. Every modelled firm receives a central value and a range.

Cohort 03

Firm-level published

The subset of modelled firms whose firm-level value passes the publication gate (section 07). Always smaller than the modelled universe.

Public coverage: Tracked Forex prop market and tracked Futures prop market. Each is calibrated as its own cohort.
Excluded from public: Crypto, stocks and multi-asset cohorts are modelled internally only and are excluded from the public Active Trader estimate and its aggregates.
Closed firms: Firms confirmed closed are excluded from current estimates and current aggregates rather than carried forward.

Current cohort counts are published on the estimate page itself and change as coverage expands; they are deliberately not written into this methodology.

Evidence and inputs

The model uses absolute scale signals — counts and volumes, never 0–100 scores — plus the firm's own disclosures. Individual data providers are abstracted in public documentation.

Scale signals: Branded search volume, organic search value, public review volume and site traffic scale. Each is an absolute quantity, each carries a model weight, and each is read at its latest available snapshot.
Disclosed counts: Firm-published trader, customer and funded-population figures. Depending on wording, freshness and internal consistency, a disclosure can calibrate the model, replace the modelled value, bound it, or be excluded from all three.
Cumulative populations: Disclosed lifetime customer or account-creation totals are treated as a registered base. Active traders are a share of that base and never a multiple of it, so the base supplies a ceiling.
Funded populations: A disclosed current funded population is a subset of active traders and supplies a floor. Cumulative funded, paid or rewarded populations are historical context and do not floor the current estimate.
Market Presence™: Applied as a small bounded adjustment to the modelled value, not as a source of scale (section 04).
Not used: Percentile or index scores as headcount drivers, firm-supplied unpublished figures, and any figure PTC cannot attribute to a public source.

Evidence gating

Not every disclosure is usable

A trader count below the plausibility floor is discarded. Two current active-trader claims by the same firm that disagree materially disqualify both until the sources are reconciled. A claim that exceeds the firm's own disclosed registered population is rejected as internally inconsistent.

Accounts to persons

Conversion is an assumption

Where the disclosed wording describes accounts rather than people, the figure is scaled down by a stored conversion assumption and additional uncertainty is added to that firm's range. The assumption is a model parameter versioned with the model, applied to account-worded evidence only, and identical across cohorts — it is not observed and it is not a measured industry average.

Calibration

Calibration converts observable scale into people using firms that disclose real trader counts as anchors.

What anchors: Only wording that describes a current population of active people. Cumulative, lifetime and marketing-language claims are reclassified as registered-base, subset-floor or contextual evidence and cannot anchor the model.
Ratio construction: For each signal, PTC takes the ratio of disclosed active traders to that signal across the anchor firms of a cohort and uses the median. This is a robust cohort ratio, not a fitted regression.
Cohort separation: Forex and Futures are calibrated separately whenever the cohort has enough anchors. The minimum anchor requirement is a model parameter and can be set per cohort.
When anchors are thin: If a cohort falls below its minimum anchor requirement for a signal, that signal is priced against the pooled anchor set across all cohorts instead. Pooling is a recorded fallback, it is stored with the estimate, and it lowers confidence.
No calibration at all: A signal with no cohort ratio and no pooled ratio is dropped from that firm's calculation rather than assigned a default.

Cohort calibration is the reason the same signal level does not imply the same trader count in every market: Futures and Forex audiences convert differently, and the ratios reflect that.

Estimation process

Each available signal implies a trader count. The implied values are combined, adjusted once for market presence, and then constrained by the firm's own disclosures.

01Signal reading
Latest available absolute scale signals are read per firm from public sources.
02Cohort calibration
Signal-to-trader ratios are derived from firms that disclose current active-trader counts.
03Implied values
Each available signal implies a trader count; the implied values are combined into one modelled value.
04Presence adjustment
Market Presence™ applies a small bounded adjustment to the modelled value.
05Evidence constraints
Disclosed populations act as floors and ceilings; fresh direct disclosure replaces the modelled value.
06Confidence, range, publication
Confidence is scored, a range is built around the central value, and publication is classified.

Combination: Implied values are combined as a weighted geometric mean, which suits multiplicative scale signals and prevents a single unusually large signal from dominating.
Presence adjustment: Market Presence™ v2.0 shifts the modelled value within a small bounded span set by a model parameter. It cannot move a firm by an order of magnitude and it never supplies scale on its own. Because Market Presence already contains demand and engagement signals, it is a corroborating adjustment rather than independent evidence.
Evidence constraints: The adjusted value is then pushed inside the constraints implied by disclosed evidence: a share-of-registered-base ceiling, a current funded-population floor, and — for reasonably recent direct disclosures — a bounded window around the disclosed figure. Constraints are recorded when they bind.
Fresh direct disclosure: When a firm's own current active-trader figure is inside the freshness window, that figure becomes the estimate and the modelled value is retained only as a cross-check. These estimates are labelled evidence-led.
Basis labels: Every estimate stores how its value was produced — PTC Estimate modelled, modelled and constrained by disclosure, or anchored on a fresh disclosure — and the estimate page shows the corresponding public basis label.

Confidence and uncertainty

Confidence is scored from the evidence actually available for that firm. It is not inherited from any other PTC model or index.

Confidence inputs: How much of the signal weight was available, whether the firm discloses a usable figure and whether that figure binds the estimate, how much validation evidence exists, whether calibration came from the firm's own cohort or the pooled fallback, and how current the inputs are.
Bands: The component scores produce a single score that maps to High, Medium or Low. Only the band is published; the score components remain internal.
Range construction: The published range is a symmetric band around the central value whose width is set by the confidence band and the value basis: an estimate anchored on a fresh disclosure carries the narrowest band, modelled estimates widen as confidence falls, and account-to-person conversion adds further width when converted evidence drives the value.
What the range is not: It is a model uncertainty band expressing how much the estimate could reasonably move, not a statistical confidence interval, and it is not derived from sampling error.

Firm-level publication

Calculation, confidence and publication are three separate decisions. Every modelled firm has an estimate; not every estimate is shown at firm level.

Decision 01

Model calculation

Produces a central value and range for every firm in the modelled universe, regardless of how the value will be presented.

Decision 02

Confidence

Grades the evidence behind that value as High, Medium or Low. Confidence alone does not decide publication.

Decision 03

Publication class

Decides whether the firm-level value is shown with its range, shown range-first, or withheld. The class is stored with the estimate.

Shown with range: High or Medium confidence, a covered public market, a usable central value and a complete range.
Shown range-first: Low confidence estimates that still rest on usable cohort calibration. The range leads and the central value is presented as approximate.
Withheld at firm level: Estimates that are outside public market coverage, lack a usable value or range, or combine Low confidence with pooled-fallback calibration on every signal. Withholding is a presentation decision — the estimate itself is unchanged.

Aggregate estimates

Aggregate Active Trader totals cover the full modelled universe, including firms whose individual value is withheld.

Universe totals and the combined modelled total sum every modelled firm in public coverage, not only the firms listed individually. This is why the modelled firm count exceeds the number of firm estimates on the page.
Firms confirmed closed are excluded from current aggregates.
Internal-only cohorts are excluded from both the aggregates and the firm list.
Aggregate uncertainty ranges are not published. Summing firm ranges assumes errors are independent, which PTC does not claim, so only central aggregates are shown.
Any figure derived from an aggregate — a share, a ratio, a run-rate — is labelled as derived rather than modelled.

Observed: Directly measured or recorded data.
PTC Estimate: Modelled values based on a disclosed PTC methodology.
Derived: Calculated from PTC datasets, rankings or published snapshots.

Versioning and baseline

Each production run is dated, versioned and preserved. Published industry figures come from a locked baseline rather than a live recalculation.

Methodology version: Published Active Trader values are produced by Active Trader Model™ v2.0 and public queries filter on that model and version explicitly. Earlier draft methodology remains in the archive and is never mixed into public figures.
Run history: Every production run is stored with its parameter set, the firms processed and its calibration thresholds. Estimates are appended: a new run supersedes the previous value for a firm and the prior row is retained as history rather than being edited in place.
Observation period: Each estimate carries an as-of date describing the period of the inputs behind it, not the date a reader loads the page.
Locked baseline: Public industry figures are drawn from a dated baseline snapshot. Once a period is locked or published its values are immutable; corrections are issued as an explicit restatement of that period rather than a silent overwrite.
Comparability: A change in methodology version breaks direct period-to-period comparison until the earlier period is restated under the new version. Baseline-to-quarter movements are labelled as such.

Runs are executed deliberately when inputs, evidence or calibration change materially. PTC does not operate an automatic fixed-interval refresh of this model, so each published figure should be read against its as-of date rather than an assumed cadence.

Limitations

These are modelled estimates, not audited trader counts. No regulator or auditor verifies them.
Disclosed firm figures use each firm's own definitions and periods, which are rarely identical to one another.
Converting disclosed accounts into people is an assumption and introduces uncertainty that PTC carries into the range rather than hiding.
Search, visibility and review signals measure attention and audience scale. They correlate with trader populations but are proxies for them, not counts of traders.
Calibration quality depends on how many firms disclose credible current figures; cohorts with few anchors rely on pooled calibration and score lower confidence.
The tracked market is not a census of every prop firm in existence, and coverage expands over time.
Public coverage is limited to the Forex and Futures cohorts; other market cohorts are not represented in the published Active Trader figures.
A firm-level estimate can be withheld while still contributing to the aggregate totals, so the firm list is not a complete view of the modelled universe.
Signals can move for reasons unrelated to trader numbers — a marketing campaign, a rebrand, or a change in search behaviour — and the model cannot always separate those effects from real growth.