Model

AION Edge Engine

The layer that sits between market noise and your conviction.

Deployed — API Access 5-minute evidence layer Multi-underlying routing

Indian index options move fast. A 5-minute candle can look like a breakout, a reversal, or a trap — and mistaking noise for signal is the most expensive mistake a trader can make. Edge Engine exists to prevent that mistake. It does not tell you what to trade. It answers a narrower question: does the current 5-minute index and options state match historically meaningful evidence, or should you stand down? When the engine is silent, that silence is deliberate. When it speaks, it speaks with a confidence bound you can verify.

What Edge Engine Does

Edge Engine converts raw intraday market state — 5-minute index and options data across NIFTY, BANKNIFTY, SENSEX, commodities, and crude oil — into a single structured judgment: tradable or not, backed by how much statistical edge, and routed through which underlying-specific model.

It does not output a brokerage instruction. It outputs context. The difference matters.

Think of Edge Engine as a gatekeeper between market noise and your decision stack. It filters. It scores. It surfaces evidence — and when evidence is insufficient, it stays quiet.

The Output Contract

This is what Edge Engine delivers. Every field is verifiable. Nothing is hidden behind a single opaque score.

{
  "edge_underlying": "nifty",
  "edge_ev": 0.1421,
  "edge_conf_low": 0.0314,
  "edge_tradable": true,
  "edge_tradable_raw": true,
  "edge_motive": "model",
  "edge_model_key": "model:edge:nifty",
  "edge_shadow_mode": false
}

Field	What It Means
edge_underlying	Which instrument the engine evaluated
edge_ev	Expected value — the statistical edge score
edge_conf_low	Lower confidence bound — the worst-case edge estimate
edge_tradable	Enforced tradability after gating rules
edge_tradable_raw	Raw tradability before enforcement (preserved for audit)
edge_motive	What drove the judgment (model / rule / override)
edge_model_key	Which underlying-specific model made the call
edge_shadow_mode	Whether the output is live or in paper-validation

Why two tradability fields? Raw tradability is what the model actually saw. Enforced tradability is what the gating layer decided after applying session constraints, risk context, and system eligibility rules. Both are preserved so you can audit the difference. No other engine we know of exposes this.

Why 5-Minute Candles, Not 1-Minute

In US markets, 1-minute bars can carry genuine information. In Indian index options, they often carry something else:

Reflex noise — algorithmic reactions to spread changes
Expiry bursts — artificial volume spikes near contract expiration
Opening-auction spillover — leftover order imbalance from the 09:15 call auction
Retail reaction loops — cascading stop-loss triggers

Research on Indian market microstructure confirms that 5-minute intervals preserve responsiveness while filtering the transient distortions that make 1-minute signals unreliable.

We use 5-minute candles because the objective is not maximum speed. It is better-calibrated evidence. A system that looks more certain than it really is does more harm than one that waits for a cleaner signal.

Trust Posture

We optimize for calibrated trust, not blind trust. Blind trust means the system always looks confident, even when it should not. Calibrated trust means the system can say "not enough evidence" and remain useful.

Exposes expected value and lower confidence bound separately
Preserves raw tradability even when enforcement policy shadows the output
Routes through underlying-specific models — NIFTY and crude oil are different markets
Surfaces the model key and motive in every output

Limitations

It does not predict price direction. It scores whether the current state carries historically meaningful edge.
It relies on 5-minute data. A sub-5-minute structural break will not be captured in the current bar.
The engine's buckets are trained on historical Indian market data. Regime shifts can temporarily reduce edge discrimination.
Commodity and crude oil models have shorter training histories than index models.
Shadow mode outputs are for paper validation only.

What Edge Engine Is Not

Not a trade signal. It does not say buy or sell.
Not a prediction engine. It does not forecast where price will go.
Not a replacement for your signal provider. It is a pre-decision filter.
Not a black box that demands trust. Every output includes the evidence needed to verify it independently.

API Access

Edge Engine outputs are available via the AION API through subscription. All AION models require an active paid subscription — no free tier.

Subscription Options →

Choose API →

Full API Reference (TXT)

Full API Reference (PDF)

Downloads

Edge Engine documentation is available as plain text and PDF for offline review, LLM ingestion, and internal workflow onboarding.