Skip to content
v3 redesign is live — welcome to the trading cockpit.
Skip to content
nexural
build.logtraders.online=2,166trades.60s=74swings.ranked=308edge.latency_ms=42ms
// surface 01 · swing desk

The desk decides what’s worth your size.

A live ranked board of every viable setup across futures, equities, and crypto. Each candidate is regime-gated, gauntlet-tested, and scored with the same composite-z used inside the SageQuant research stack. No signals. No alerts. Just the trades that pass.

regimetrend·long·candidates7/412
Open the Swing Desksee the architecture
  • 412
    candidates / day
  • 97.4%
    filtered out
  • 5s
    refresh tick
AI Nexural Swing Desk
--:--:-- UTC
filters·score ≥ 70regime ≠ shifthorizon ≤ 5d7 of 412 candidates
symsetupscoreside
ESPOC bounce94long
NQDemand zone92long
CLVAL bounce89long
GCNaked POC84watch
RTYSupply rejection78watch
BTCORB72skip
AAPLGap fill68watch
ranked by composite_z · gauntlet ✓ · DSR / PBO / WFE / SPArefresh 5s

real layout · sample data · gauntlet ✓

// why this exists

The three problems a real desk solves.

If you've traded long enough you already know these. The Swing Desk closes them as code, not as advice.

  • 01the watchlist problem

    Too many tickers, too little signal.

    Most retail tools surface every chart with a hammer pattern and call it a setup. Real desks scan thousands of names, then publish a shortlist. We replicate the shortlist — the AI Nexural Swing Desk filters 400+ candidates a session into seven that actually pass.

    97.4%filtered out
  • 02the regime problem

    Setups don't work in every regime.

    A demand-zone bounce is gold in trend·long, garbage in regime·shift. Without a regime gate, you take the same playbook into every tape and wonder why your hit rate decays. The desk flips state with the market — long candidates literally stop printing when the HMM trips.

    6regime states
  • 03the conviction problem

    Knowing what's good isn't sizing what's good.

    When five names look identical, traders pick at random. The desk ranks them by composite-z — a single number combining regime fit, setup quality, location, and risk — so you know which one deserves full size and which one deserves a starter.

    1rankable score
// how it works

412 candidates in. 7 trades out. Five stages.

The same pipeline runs every five seconds. Nothing in the path is optional. Nothing in the path is hidden.

  1. step 00ingest

    Universe scan

    Tick-level prints across futures, equities, and crypto, plus the active fundamentals frame. Same data feed that powers the dashboard charts.

    in raw ticks412 candidates
  2. step 01regime gate

    Regime fit filter

    An HMM tags the live regime (trend·long, vol·rising, rotation, …). Setups whose historical edge is < 0 in the current regime are dropped on the spot.

    in 412168
  3. step 02score

    Composite-z scoring

    Each surviving setup is scored on regime fit, setup quality, location, risk-clarity, and historical edge. Z-normalized so a 92 in NQ means the same thing as a 92 in CL.

    in 168168 scored
  4. step 03rank

    Composite ranking

    Names are sorted by composite-z, then capped per-symbol so you don't see five flavors of the same trade. Result: one ordered shortlist, every session.

    in 168 scored12 ranked
  5. step 04action

    Side & size hint

    Each ranked candidate gets an action — long, short, watch, or skip — plus a starter size suggestion based on your account profile. You decide if the trigger fires.

    in 12 ranked7 actionable
// inside the score

Composite-z, decomposed. Five inputs. Published weights.

The number on the desk is a weighted sum of five components, normalized so a 92 in NQ means the same thing as a 92 in CL. Here are the weights.

  • regime_fitRegime fit

    Does this setup historically edge in the live regime? Pulled from a 5-year rolling window of regime-tagged outcomes for the same setup family.

    30%
    weight
    example·Demand-zone bounce + trend·long → +0.82σ edge
  • setup_qualitySetup quality

    How clean is the structural pattern itself — touch count, rejection wick magnitude, volume confirmation, time spent in zone. A model graded on labeled examples.

    25%
    weight
    example·POC bounce, 3 prior touches, vol-spike on entry → 0.91
  • locationLocation

    Where the price is relative to value — distance from VPOC, proximity to HVN/LVN, pivot context. Trades closer to high-conviction levels score higher.

    20%
    weight
    example·Entry within 0.4 ATR of session VPOC → 0.78
  • risk_clarityRisk clarity

    Is the invalidation level obvious? A wide invalidation, or a fuzzy one, drags this component down even when everything else is clean.

    15%
    weight
    example·Stop within 1 ATR · R:R ≥ 2.5 → 0.82
  • edge_persistenceEdge persistence

    How stable has this setup's edge been across the gauntlet — DSR, PBO, WFE, SPA. A high score here means it didn't decay in walk-forward.

    10%
    weight
    example·DSR 0.71 · PBO 0.18 · WFE 0.62 · SPA p<0.05 → 0.74
// regime gating

Six regimes. Six different playbooks.

An HMM tags the live tape. The desk re-weights its candidate set every five seconds based on what the market is actually doing right now — not what it was doing yesterday.

  • trend·longstate

    Trend, long-side

    Higher highs, expanding ATR, breadth confirms. Trend setups carry edge.

    enables
    • Demand-zone bounce
    • POC continuation
    • Pullback to VWAP
    disables
    • Mean-reversion shorts
    • Range fades
  • trend·shortstate

    Trend, short-side

    Lower lows, broadening, breadth weak. Short setups carry edge.

    enables
    • Supply-zone rejection
    • Lower-high fail
    • Bear flag continuation
    disables
    • Long breakouts
    • Demand-zone bounces
  • rotationstate

    Rotation

    Range-bound, mean-reverting, ATR compressed. Edges are tactical.

    enables
    • VAL/VAH fades
    • Range mean-reversion
    • ORB fade
    disables
    • Trend-continuation longs
    • Breakout chases
  • vol·risingstate

    Vol, rising

    Realized vol expanding faster than implied. Risk widens, sizing shrinks.

    enables
    • Breakout follow-through
    • Gap-and-go
    disables
    • Tight-stop scalps
    • Low-vol mean-reversion
  • vol·crushstate

    Vol, crushing

    Realized < implied, mean reverting fast. Premium-selling regime.

    enables
    • Premium-selling structures
    • Tight-stop continuations
    disables
    • Breakout chases
    • Vol-expansion plays
  • regime·shiftstate

    Regime shift

    Statistical regime changing. Historical edge no longer applies.

    enables
    • Nothing. The desk goes flat.
    disables
    • Every prior playbook until the new regime stabilizes
$ regime is published live at /api/regime — same endpoint the desk reads. Every transition is logged with timestamp, prior state, and trigger covariance.
// anatomy

What you’re actually paying for. Five surfaces, one cockpit.

The desk is built around a single screen. No hopping tabs, no losing the candidate while you check the chart. Here is the layout, annotated.

05regimetrend·long·held 47mAI Nexural Swing Desk
01filters·score ≥ 70regime ≠ shifthorizon ≤ 5drisk ≤ 0.5R
02ranked candidates
symsetupscoreside
ESPOC bounce94long
NQDemand zone92long
CLVAL bounce89long
GCNaked POC84watch
RTYSupply rejection78watch
04session planner2 in plan · drag rows to add

real layout · annotated · sample data

  • 01
    Filter strip

    Score floor, regime exclusions, horizon, and account-aware risk caps. Every preset is shareable as a URL.

  • 02
    Ranked candidate table

    Each row is a candidate that passed the gauntlet. Sortable columns, side chip, composite-z bar.

  • 03
    Detail panel

    Click a row to expand: live chart, score breakdown, similar historical setups, and the trade plan template.

  • 04
    Session planner

    Drag candidates into your plan. The desk persists it to the journal so tomorrow's review starts with last night's intent.

  • 05
    Live regime band

    Always visible. Tells you the active state and how long it has been in force.

// anti-pitch

What it isn’t. In writing.

If any of these would have been a deal-breaker, better to find out here than after a refund.

  • It's not a signal service.

    We do not push you trades. The desk ranks; you decide. There is no chat-room "BTO NQ at 22184" message.

  • It's not auto-execution.

    No order routing. No broker integration that fires fills on your behalf. Sizing is a hint; pulling the trigger is yours.

  • It's not a black box.

    The composite-z weights are public, the regime model is documented, and the gauntlet (DSR · PBO · WFE · SPA) runs on every release.

  • It's not institutional-grade indicators with marketing copy.

    It's a single ranking pipeline. The institutional-grade indicators feed it — they aren't the product, they're the inputs.

  • It's not a guarantee.

    Past performance does not predict the future. The desk reduces noise. It cannot protect you from a bad regime call or your own sizing.

// faq

Questions people actually ask.

If yours isn't here, email sage@nexural.io and we'll add it.

  • Most retail scoring stacks one indicator on top of another and normalizes at the end. Composite-z is a weighted sum of five z-normalized components — regime fit, setup quality, location, risk clarity, and edge persistence — refit quarterly on out-of-sample data. The weights and the refit history are published in the changelog.
  • Yes. The candidate universe spans futures, US equities, and the major spot crypto pairs. The same pipeline runs on each — the regime model is asset-class aware, so an equity setup is graded against an equity-regime tape, not a futures-regime one.
  • The ranked board ticks every five seconds on the production tier. The regime classifier re-evaluates each minute. Composite-z weights are re-fit quarterly; refits ship behind a feature flag and are announced on the changelog.
  • Yes — the production weights are the default, but Pro and Automation accounts can clone the formula and override any component weight. Cloned formulas run alongside the default so you see both rankings side by side.
  • Regimes mis-classify; that’s why we publish the transition log. When the live regime is mid-flip, the desk widens its score thresholds and biases toward ‘watch’ over ‘long’ or ‘short’. You won’t see a full long set in regime·shift even if the prior state was trend·long.
  • No. There is no broker integration that fires orders for you. The desk produces ranked candidates with a recommended side and a starter size; you place the trade. This is intentional — see the anti-pitch.
  • Yes. It’s the same family of overfit-resistance tests institutional desks run before deploying a strategy. We publish the gauntlet result on every released indicator and on the composite formula itself, so you can audit whether what shipped held up out of sample.
  • Start in the academy and on the playbooks page first. The desk is most useful when you already understand the setups it surfaces — otherwise the score just becomes a number you trust without context.
// the rest of the cockpit

Where this fits in the system.

Every surface compounds the others. The Swing Desk feeds the journal. The journal grounds Sage AI. Sage AI cites the indicators. The indicators feed QuantFlow.

desk · open·regimetrend·long

Stop scanning. Start trading what already passes.

The AI Nexural Swing Desk is included on the Pro tier at $47/mo. Seven-day money-back, no retention call. If it isn’t making your prep faster, full refund — we’d rather keep the relationship than the $47.

Open the Swing Desksee plansread the track record
$47/mopro tier · swing desk included
14 dayspro trial · cancel anytime
7 daysmoney-back, no questions