Signals

TRUE Signals gives every user access to six structured, AI-generated trade ideas — refreshed four times a day, tracked live from the moment they are published to the moment they hit a target, breach a stop, or expire. Each signal comes with an entry, a target, a stop, a confidence label, and a written thesis so users can evaluate the idea on its merits, not just its conclusion.

What it is

A signal is a structured, time-stamped trade idea: one asset, one direction, an entry price, a target price, a stop-loss level, a confidence label, and an AI-written thesis explaining the reasoning. Signals are decision-support tools — they are inputs to the user’s own analysis, never automatic instructions. The platform tracks every active signal in real time, updating the highest and lowest percentage moves since publication, and closes the signal when the target is reached, the stop is breached, or the signal expires.

Disclaimer. TRUE provides information, analysis, and tooling. Nothing on this platform constitutes financial advice, investment advice, or a recommendation to buy or sell any asset. Users remain solely responsible for their own decisions and outcomes.

Who it’s for

• Active traders who want a structured starting point for their own research and want to see which assets are generating directional setups.
• Market observers who want a real-time view of where AI-driven analysis is finding bullish and bearish conviction across the top of the market.
• Developers and partners who want to surface signal data in their own interfaces via the Signals API.

How it works

Generation cadence

The platform generates a fresh batch of signals four times every day: at 00:00, 06:00, 12:00, and 18:00 UTC. Each batch contains exactly three bullish and three bearish picks.

Candidate selection

The candidate pool is drawn from the top 100 crypto assets by market capitalization, filtered to those that have moved at least 1% in absolute terms over the past 24 hours. This filter ensures signals are generated on assets with active price discovery — not stagnant markets where a direction call has no informational value.

From this candidate pool, the AI selects six tickers — three in each direction — and constructs entry, target, and stop levels that are internally consistent: for longs, target > entry > stop; for shorts, the inverse. A confidence label of high, medium, or low is assigned as the model’s relative ranking among the picks in that batch.

Thesis generation

For every signal, a second model pass generates a long-form written thesis and a short tagline (under 140 characters). The thesis is stored per language so it can be served in the user’s locale without re-translation at request time. On-demand generation is available for languages not pre-generated at publication time.

Live tracking

Every active signal is tracked by a background process that refreshes prices on a short cycle and:

• Updates the signal’s highest percentage gain and lowest percentage drawdown since publication.
• Flips the signal to completed when the live price reaches the target on the correct side.
• Flips the signal to removed when the stop is breached or the signal expires after its maximum active window.
• Records the price at removal so the outcome is permanently auditable.

Signal lifecycle

Status	Meaning
Active	Currently tracked. Live price updates are running.
Completed	The target was reached. The signal remains visible for a review window before archiving.
Removed	Archived. Carries a removal reason.

Removal reason	Trigger
target_reached	Live price hit the target on the correct side.
stop_loss_breached	Live price hit the stop on the wrong side.
prediction_changed	The next generation slot picked the opposite direction for the same asset.
expired	The signal reached the end of its maximum active window without hitting target or stop.

Carry-over logic

If the next generation slot picks the same asset in the same direction it already has an active signal for, the existing signal is carried forward rather than duplicated — its timestamp is refreshed and the thesis is updated. If the next slot picks the same asset in the opposite direction, the existing signal is closed with prediction_changed and a new signal is opened. This prevents the system from holding contradictory positions on the same asset simultaneously.

The rules it enforces

• Exactly six signals per batch. Three bullish, three bearish. No batch is published with an imbalanced set.
• Internal consistency. Entry, target, and stop levels are validated for directional consistency before a signal is stored. A malformed set is rejected and regenerated.
• Exclusive slot generation. Each generation slot is claimed atomically. In a multi-replica deployment, only one process generates the batch for a given slot. There is no risk of duplicated batches from concurrent execution.
• No contradictory active signals. A single asset cannot simultaneously carry a bullish and a bearish active signal.
• Expiry enforcement. Signals that have not hit target or stop are automatically closed at the end of their maximum active window.
• Thesis in every supported language. The platform generates reasoning in every supported language at publication time. On-demand generation covers languages added after initial publication.

Safety, security & trust

No execution without user action. The signal system is read-only. No signal triggers an automatic trade. Execution requires an explicit user action via the Agentic Trading flow.

Outcome permanence. When a signal closes, the closing price is recorded alongside the removal reason. The full history of completed and removed signals is available for audit via the history endpoint. Past signal outcomes are not edited or removed retroactively.

Confidence is ordinal, not cardinal. A high confidence label means the model ranked this pick above the medium and low picks in the same batch — not that the model assigns it a specific probability of reaching its target. This distinction is important for any probabilistic reasoning about signal outcomes.

Partial batch transparency. If the candidate pool does not yield enough qualifying assets for a full batch (fewer than three movers in one direction), the batch is published as a partial set with a logged flag. Users and API consumers can see that the batch is partial rather than inferring that a full set was generated.

Data freshness. Signals are generated from live market data at the generation time. The entry price reflects the market at that moment. By the time a user reads the signal, the live price may already be past the entry level — the signal’s thesis and levels should be evaluated in that context.

Reasoning availability. Thesis content is available for every supported language either at publication time or via the on-demand endpoint. A signal is never surfaced without a thesis.

For Developers

The Signals feature exposes read-only REST endpoints for retrieving the active signal set, paging through the history of closed signals, fetching a single signal by identifier, and requesting per-locale thesis content. Full endpoint reference and authentication requirements are in the Authentication and Partner Integration docs.

Why it’s well thought through

Structured output over narrative vagueness. Every signal carries machine-readable fields — direction, entry, target, stop, confidence — alongside the human-readable thesis. This means the signal can be consumed programmatically, displayed consistently, and evaluated objectively against its stated levels.

Live tracking closes the feedback loop. Publishing a signal and never updating it creates a false picture of performance. TRUE tracks every signal from publication to outcome. Users can see how far a signal moved in their favor, when it hit target, or when it was stopped out. The history is permanent and auditable.

Three-plus-three balance prevents directional bias. Requiring equal numbers of bullish and bearish picks prevents the generation process from producing a uniformly bullish or uniformly bearish set, which would reflect model bias rather than genuine market structure.

Carry-over prevents noise inflation. Rather than generating six new signals from scratch every six hours regardless of whether conditions have changed, the carry-over system preserves existing signals when the model agrees with its previous pick. This reduces signal churn and gives users time to act on a signal before it is replaced.

Common questions

How are the six picks chosen? The AI selects from the top 100 crypto assets by market cap that have moved at least 1% in the last 24 hours. From that set, it picks the three most compelling bullish setups and the three most compelling bearish setups based on price structure, momentum, and market context.

What does the confidence label mean? It is the model’s relative ranking among the six picks in a given batch — high means the model rates this pick more compelling than the medium and low picks. It is not a probability estimate or a success-rate guarantee.

Does the platform place a stop-loss order for me? No. The stop level is a reference price for your own risk management. The platform does not place orders on your behalf unless you have explicitly configured an automated strategy via Agentic Trading.

What happens when a signal’s entry price has already passed? The signal is still visible with its thesis and levels. Whether to act on a signal whose entry has already been reached is a judgment call — the thesis is the more durable part of the signal than the exact entry price.

How long does a signal stay active? The maximum active window is seven days by default. If the target and stop are both within range and neither has been hit within that window, the signal expires with reason expired.

Can I get signal data via API? Yes. The active, history, and individual signal endpoints are available for API consumers. See the developer tab above.

Are signal outcomes auditable? Yes. Closed signals retain their closing price and removal reason permanently in the history endpoint. Outcomes are never retroactively edited.

Why do only crypto assets appear in Signals? The candidate pool is currently drawn from crypto assets. Stock and tokenized equity signals are not in scope for this feature at this time.

Does the system produce signals on weekends? Yes. The generation cadence is not sensitive to trading hours — signals are generated on the same four-times-daily schedule every day of the week.

How do I get notified when a signal hits its target? Signal transitions — target reached, stop breached — are pushed over the real-time activity stream. See Realtime API for subscription details.

Related features

• Event Catalyst — macro and on-chain catalysts used for context in signal thesis generation.
• Agentic Trading — turn a signal into an automated, enforced strategy with hard caps.
• Realtime API — subscribe to signal transitions over SSE.
• Agents — the agent system that generates signal reasoning.