Methodology

BinaryBias model outputs are produced by BinaryBias.ai. The legal entity identified for service disclosures is BinaryBias.ai. Each output should be read together with the displayed publication time, pair, timeframe, model version, threshold version, source description, assumptions, and conflicts disclosure.

This page is intended to support objective presentation by separating factual display metadata from model estimates, describing material source categories, and identifying assumptions that may affect later observations.

BinaryBias provides non-personalised FX probability analytics for informational and research use. The service is designed to describe a model output and related historical observations in a consistent format. It is not designed to assess any person's financial position, objectives, knowledge, experience, risk tolerance, tax position, or legal obligations.

Model outputs are probabilistic estimates about a defined currency pair and timeframe. They are not instructions, recommendations, offers, solicitations, or guarantees of any result.

A published output may include the currency pair, timeframe, publication timestamp, model version, threshold version, directional classification, probability score, and related display metadata. The probability score is an estimate produced by the relevant model at the time of publication.

Publication timestamps identify when the service completed or published the displayed output in UTC. A timestamp is factual service metadata; the probability score and directional classification are non-factual estimates produced by the model.

The public page does not describe proprietary implementation details, including feature design, training process, internal weighting, infrastructure, or vendor-specific processing rules.

BinaryBias currently applies fixed display thresholds to convert a probability score into a directional classification. A score of 0.35 or below is classified as downward. A score of 0.65 or above is classified as upward. Scores greater than 0.35 and less than 0.65 are classified as neutral. The current threshold version is fx-probability-0.35-0.65.

A neutral classification means the score did not reach the current upward or downward threshold. It is not a view that price will remain unchanged, that risk is low, or that no material movement can occur.

The service uses configured third-party FX market-data feeds and internal model-processing pipelines for FX market data and processes that data into candle windows for supported currency pairs and timeframes. A candle is constructed from provider data for the relevant interval and may include open, high, low, close, and related fields used for display or observation.

Timestamps are shown in UTC unless the interface states otherwise. Publication time, candle time, observation time, and archive time may refer to different events. Users are responsible for reviewing the timestamp label shown in the interface before comparing values across pages, exports, or archives.

Source data is treated as indicative unless the interface states otherwise. Where source reliability is uncertain because of provider outages, late data, corrections, symbol mapping issues, or market disruption, the affected display may be delayed, unavailable, corrected, restated, or removed.

Displayed data is indicative and may differ from prices or data shown by another venue, broker, platform, or data provider. Provider corrections, missing records, delayed delivery, connection outages, maintenance windows, symbol mapping changes, or late-arriving data may affect charts, model inputs, archives, observations, and displayed metrics.

BinaryBias may recalculate, correct, remove, or mark affected records when a data or processing issue is identified. Historical displays can therefore change after initial publication.

A model output is a forecast-style estimate made at or near the publication time shown. An observation is a later measurement of what was recorded for the relevant currency pair and observation window. An archive is a stored record of outputs, observations, and related metadata available to the service at the relevant time.

Forecast-style estimates, subsequent observations, and archived records serve different purposes and are not the same type of information.

Directional accuracy is a historical monitoring measure. It compares a published non-neutral directional classification with the subsequent observed direction of the relevant currency pair over the configured observation window. The calculation is based on eligible archived outputs for the selected pair, timeframe, model version, threshold version, and display period.

Accuracy figures are displayed with the relevant sample size. A small sample size may not be representative. Neutral outputs are not counted as upward or downward outcomes unless the interface expressly states another treatment. Records may be excluded where the required publication data, observation data, timestamps, or provider records are missing, delayed, inconsistent, or later corrected.

Directional accuracy does not include spreads, commissions, fees, slippage, taxes, financing costs, margin requirements, execution availability, position sizing, or any personal constraint. It is not a measure of profit, loss, suitability, or expected future performance.

Model-output displays may differ from earlier outputs because of a new model version, threshold version, supported-pair set, data source, processing rule, or display rule. Where the service shows a model version or threshold version, those identifiers apply to the output record or view where they are displayed.

BinaryBias does not currently publish a public 12-month list of all output changes as formal investment-research updates. Historical archives are product records for model-output review and may be incomplete where data, access, retention, or correction limits apply.

Live or current views show outputs generated by the operating service using the data and model configuration available at publication time. Historical views show archived records and later observations. Where a backtested or reconstructed view is shown, it is based on historical data and assumptions and may not reflect conditions that existed for a live publication workflow.

Live results, historical observations, and backtested or reconstructed results are not directly interchangeable and may differ because of data availability, provider changes, timing, corrections, model version, threshold version, or processing rules.

BinaryBias may update models, thresholds, supported pairs, supported timeframes, data providers, data processing, user interface labels, archive logic, and observation rules. Material output changes may be identified through a model version, threshold version, timestamp, release note, or other service metadata where appropriate.

Unless the interface states otherwise, a new model or threshold version applies from its effective publication time and does not guarantee that older records will be restated. BinaryBias may deploy, rollback, pause, or adjust model and threshold versions as part of normal service operation.

Model outputs are limited by the data, assumptions, model design, and processing configuration available at the time they are generated. They may be affected by market news, scheduled announcements, unexpected events, high volatility, low liquidity, wider market stress, regime shifts, model drift, data provider changes, and technical failures.

Past observations may not describe future conditions. No methodology can remove uncertainty or ensure that a probability estimate will match a later observation.

BinaryBias is a paid SaaS product. Commercial interests connected to paid access are relevant when reviewing any methodology description, display metric, archive, or other site material.

Unless expressly disclosed elsewhere, BinaryBias does not execute user orders, route transactions, or receive compensation based on a user entering into an FX transaction through the service.

BinaryBias does not provide investment advice, financial advice, legal advice, tax advice, portfolio management, order execution, brokerage, or dealing services. The service does not personalise outputs for an individual user and does not determine whether any FX transaction or other financial decision is suitable for any person.

Users remain responsible for their own independent review and for obtaining professional advice where appropriate.