How the system works.

What follows is the architecture at the level of detail we are willing to publish: pipeline shape, layer roles, gate categories, learning loop. Specifics — scoring weights, signal thresholds, classifier names, per-filter parameters — are intentionally not disclosed. The published idea is the output, not the recipe.

The pipeline at a glance

Multiple stages, each with its own veto power. A candidate has to clear all of them to reach a subscriber.

pipeline / public view

  UNIVERSE        Daily filter on liquidity, OI, spread, event proximity.
                  A curated set of names survives into the day's universe.

  SIGNAL GEN      Independent generators run in parallel, looking at the
                  market from fundamentally different analytical angles —
                  flow, catalyst, and structure.  Continuous intraday
                  scanning, pre-market through close.

  STRATEGY STACK  A diversified set of strategies pick up candidates
                  appropriate to their opportunity class.  Strategies are
                  independent — one bad regime cannot wipe the whole stack.

  QUALITY GATE    A multi-rule gate enforces minimum signal score,
                  liquidity floors, premium-efficiency bands, and a daily
                  trade cap.  Most candidates die here.

  AI CONSENSUS    A multi-agent AI consensus layer reviews every survivor
                  and must reach sufficient consensus before publication.

  RANK & PUBLISH  Final ranking layer orders survivors and publishes the
                  top of the stack at curated points in the session.  Each
                  idea ships with a hard pre-set stop and a multi-tier
                  scaled profit ladder.

  CLOSED LOOP     A forensic learning layer reads from realised outcomes
                  and trims strategies that under-perform.  Compounds
                  with every market day.

The four conceptual layers

Conceptually, every signal that reaches you has navigated four layers in order.

Layer 1

Universe construction

Each morning, before the open, the day's investable universe is rebuilt from scratch. A candidate name has to clear liquidity, open-interest, spread, and event-proximity floors before it qualifies.

Curated set of liquid U.S. equities and ETFs across all major sectors
Names that thinned out overnight drop out automatically
Event windows (earnings, FOMC) bias the screen toward catalyst names

Layer 2

Pattern detection

Independent signal generators look at the market from fundamentally different analytical angles. Each generates candidates separately; convergence between them is one of the highest-conviction signals the platform produces.

Flow-based signals from institutional positioning
Catalyst windows (earnings, macro events, scheduled releases)
Structural signals from compression and mean-reversion patterns

Layer 3

Conviction & quality gates

Each candidate has to pass a multi-rule gate covering liquidity, premium efficiency, regime-aware sizing, and source-specific risk. Most candidates fail at this step — and that is the design intent, not a bug.

Liquidity, premium-efficiency, and contract-quality floors
Regime-aware sizing keeps risk per name bounded
Per-day trade cap and per-strategy sizing budget

Layer 4

Retrospective learning

Every published signal is tracked against its peak realised P&L. A forensic learning layer reads from the loss tape and adjusts the gate. Strategies that under-perform are sized down or paused; the system improves itself with every market day.

Per-strategy attribution every session
Rolling forensic review on every meaningful loss
Adversarial sizing on patterns that historically under-perform

A diversified strategy stack

Each strategy is tuned to a different opportunity class. No single style works in every regime — diversifying across opportunity classes is how the system stays useful across bull, bear, sideways, and high-vol environments. Specific strategy names, parameters, and triggers are intentionally not published.

Class · 01

Short-term tradable structure

Setups on liquid names with a clear pattern over a multi-day horizon. Capital preservation comes first; every position has a hard pre-set stop and a multi-tier exit.

Class · 02

Ticker-specific downside

Bearish setups on names showing controlled, ticker-specific weakness. Sized small; aimed at controlled-decline scenarios rather than 'shorting the market'.

Class · 03

Asymmetric upside in calls

Convexity-favouring positions where a small premium has a meaningful chance of an outsized return. Multi-agent confirmation is mandatory at this risk profile.

Class · 04

Lottery-style asymmetric payoff

Small-allocation positions in deep-OTM contracts. Hard per-trade dollar cap; designed so even a string of losses cannot meaningfully damage the book.

Class · 05

Cross-strategy convergence

Fires only when independent strategies and analytical angles converge on the same name. Historically the highest-conviction setup the platform produces.

Risk first, always

Every strategy starts with the same question: what can we lose? Capital preservation enables long-term compounding; everything else is downstream.

Hard pre-set stop on every position. No exceptions, no overrides.
Multi-tier scaled profit ladder so paper-gains turn into realised P&L.
Daily loss circuit breaker — if session P&L crosses the threshold, all strategies stand down.
Portfolio-heat cap and correlation gate limit concentrated risk.
Per-strategy sizing budget so one strategy cannot dominate the book.
Forensic learning layer vetoes high-loss-probability candidates before they reach you.
Trade-quality gate enforces liquidity, premium-efficiency, and signal-score floors.

Profit ladder · architecture

Multi-tier scaled exits

Every published idea ships with a multi-tier profit ladder. Position size is scaled out across several price levels, with a trailing-stop runner on the final tranche so paper-gains turn into realised P&L. Specific breakpoints, allocation fractions, and trail parameters are intentionally not disclosed.

Multiple scaled-exit tranches across the move
Trailing-stop runner for the final tranche
Break-even floor engages once the position runs
Hard pre-set stop sets the worst-case downside

What we will never do

—Trade in your account. TradeNova is a research and signal service, not a managed account or broker-dealer.
—Sell your data. We do not share, rent, or sell user data to third parties for marketing.
—Promise specific returns. Past performance is not indicative of future results, and we will say so loudly and often.
—Hide losses. The performance dashboard reports every published signal — winners and losers.
—Publish the recipe. The output is the product. Scoring weights, classifier names, thresholds, and per-filter parameters stay private.

See how the system reports performance

What we measure, how it is computed, and how to read the dashboard.

Performance reporting →