NeuroTrader - The Construct™Docs
Open App

Core Concepts

The Decision Pipeline

A trade is the survivor of a gauntlet. This page traces the full path from market data to an executed order — and all the places a weak idea gets filtered out.

The eight stages

#StageWhat happens
1Analysis cycleEvery few seconds, all engines analyze every asset on the watchlist and refresh their votes and confidence.
2Vote aggregationThe five directional votes are tallied per asset. The Coherence Model scales the ensemble confidence by market alignment.
3Agreement gateDid at least your configured minimum number of engines agree on BUY or SELL? If not → recorded as a skip.
4Confidence gateIs the coherence-adjusted ensemble confidence above your threshold? If not → recorded as a skip.
5Risk gatesYour account limits are checked: daily loss cap, daily trade cap, max open positions, trade spacing, averaging-down guard, order value bounds, wave access.
6Position sizingOrder size is computed from your position-size setting and available balance. Unsizeable orders are blocked, never guessed.
7ExecutionThe order is routed — to the simulated paper engine in paper mode, or to the user’s connected exchange in approved controlled-live mode — and the result is recorded.
8Position managementA dedicated monitor manages the open position continuously: stop-loss, trailing stop, take-profit, partial exits, DCA safety orders.

Everything is recorded

Stages 3–6 each produce an auditable outcome when they stop a trade. The Decision History page shows every one with a plain-language reason — there is no such thing as a silent skip.

Outcome types

ExecutedAll gates passed; the order was placed and filled in the active mode: simulated paper execution or approved controlled-live execution.
Skipped / filteredThe signal didn't meet agreement or confidence requirements. The most common outcome by far — this is the selectivity working.
Held / blockedA risk gate stopped it: daily limits reached, position caps, trade spacing, sizing, or access gating.
No signalThe engines saw nothing actionable this cycle. Nothing to evaluate.

Why most signals die (by design)

On a typical low-conviction day, the pipeline evaluates thousands of potential opportunities and executes a handful. The ratio isn't a malfunction — it's the product. Each gate exists because of a specific, well-known way automated traders lose money:

  • Agreement gate — protects against one overconfident model in conditions that suit it but nothing else.
  • Confidence gate — protects against marginal setups where the edge doesn't cover fees and slippage.
  • Daily loss cap — protects against compounding a bad day. Losing days end early; they don't spiral.
  • Trade spacing & caps — protect against churning: re-entering the same move repeatedly and bleeding fees.
  • Averaging-down guard — protects against the classic account-killer: throwing good money after a losing position outside controlled DCA rules.

Cycles and timing

The orchestrator runs continuous analysis cycles — the cycle count on the System page is its heartbeat. Each cycle refreshes every engine on every watched asset. Decisions are made at cycle boundaries, so the platform reacts to the market within seconds, while position management (stage 8) runs continuously in between.

Paper vs. live execution

The pipeline is identical in both modes through stage 6. At stage 7, paper orders are filled by the simulated execution engine against real market prices, while live orders are routed to your connected exchange. That symmetry is what makes paper results meaningful for validating behaviour — see Paper vs. Live.