The process

How Micro Edge actually works

A complete, no-marketing walkthrough of the pipeline behind every cycle: how data becomes a signal, how a signal becomes a trade, and how a trade becomes a transparent line in your portfolio.

~10 min readUpdated for v1.0Simulated performance disclosure
TL;DR

Three independent AI engines run on fixed 2H, 6H, and 12H cycles. Each cycle, models score the market, a strict quality filter rejects almost everything, and only high-confidence setups are executed with hard stops and pre-sized risk. You only pay a fee — 2% — on cycles that actually earn profit.

The six-step pipeline

Every engine, on every cycle, executes the same disciplined sequence.

  1. Step 01

    Ingest market data

    Each engine continuously pulls real-time price action, orderbook depth, funding rates, open interest, on-chain flows, and the volatility surface across major centralized venues. Data is normalized, deduplicated, and timestamped before any model ever sees it.

    We treat data quality as a first-class problem. Bad ticks, exchange outages, and stale feeds are detected and quarantined automatically — a model is only as honest as the inputs it receives.

  2. Step 02

    AI scoring

    Models evaluate dozens of features per candidate setup — trend structure, momentum, volatility regime, liquidity, microstructure, and cross-asset context — and output a calibrated confidence score with directional bias.

    Confidence is not a marketing word here. Each engine has a target hit-rate and expected payoff, and the score is calibrated against historical out-of-sample outcomes so a 70% score actually means roughly 70%.

  3. Step 03

    Quality filter

    Only setups above the engine's confidence threshold survive. Most cycles end with zero trades — and that is the point. Forcing a trade is the single fastest way to lose money in crypto.

    Thresholds are tuned per engine. The 2H engine is more reactive and accepts a higher trade frequency at smaller targets; the 12H engine waits for fewer, cleaner setups with larger expected moves.

  4. Step 04

    Defined-risk execution

    Each accepted trade enters with a hard stop, a defined target, and a position size that is risk-budgeted to a fixed fraction of capital — typically 0.5% to 1% of the engine's allocation per trade.

    There is no averaging down, no widening of stops, and no 'hoping it comes back'. Once a trade is live, the only outcomes are: target hit, stop hit, or time-based close at cycle end.

  5. Step 05

    Realized P&L & reporting

    Cycles close cleanly. Realized profit and loss is written to your portfolio with a full audit trail: entry, exit, size, fees, and timestamps. The 2% performance fee is computed only on net positive cycles.

    If a cycle ends flat or negative, no fee is charged on that cycle. You only pay when the engine actually makes you money.

  6. Step 06

    Learn & adapt

    Outcomes feed back into model retraining on a scheduled cadence. Engines adapt to new market regimes — bull, bear, chop, high-vol — instead of memorizing the last one.

    Retraining is governed: every candidate model must beat the live model on a held-out, walk-forward backtest before it is allowed to take real risk.

The three engines

Same pipeline, different timeframes, different risk profiles. Pick one, mix several.

2H Engine
CadenceEvery 2 hours
Target0.6% – 1.5% per cycle
ProfileLower risk · higher frequency

Best for investors who want frequent, smaller realized cycles with tighter risk per trade.

6H Engine
CadenceEvery 6 hours
Target2.0% – 3.5% per cycle
ProfileMedium risk · balanced

Balanced middle ground — fewer trades than the 2H engine, larger expected moves per cycle.

12H Engine
CadenceEvery 12 hours
Target4.0% – 6.5% per cycle
ProfileHigher risk · selective

Patient, highly selective. Many cycles result in no trade. When it does trade, it expects a larger move.

Operating principles

Capital protection first

Every trade has a predefined maximum loss. No exception, no override.

Probability over volume

We do not chase trade count. A cycle with zero trades is a valid, often correct outcome.

Discipline by clock

Engines run on fixed cycles. Decisions are scheduled, not emotional.

Aligned fees

2% performance fee on profits only. If you don't earn, we don't earn.

Lifecycle of a single trade

What actually happens between one cycle tick and the next.

  1. T+0:00
    Cycle opens
    Engine ingests fresh market data and rebuilds its feature snapshot.
  2. T+0:01
    Scoring
    All candidate setups are scored. Most are discarded immediately.
  3. T+0:02
    Decision
    If a setup clears the threshold, position size, stop, and target are computed.
  4. T+0:03
    Execution
    Order is routed to the venue with the best executable price.
  5. T+...
    Monitoring
    Trade is monitored until target, stop, or cycle expiry — whichever comes first.
  6. T+cycle
    Close & report
    Realized P&L is booked to your portfolio. Performance fee is calculated only on net profit.

What's under the hood

Engine core

  • • Python 3.12 trading core with FastAPI control plane
  • • CCXT for unified exchange connectivity
  • • pandas, NumPy, TA-Lib for feature engineering
  • • APScheduler driving fixed 2H / 6H / 12H cycles

Modeling & ops

  • • Gradient-boosted ensembles + sequence models for scoring
  • • Walk-forward backtests gate every model promotion
  • • Per-trade audit log with entry, exit, size, and fees
  • • 24/7 monitoring with circuit breakers on anomalies

More detail on infrastructure and security: see the Technology page.

A note on performance

Targets shown for each engine are simulated, derived from historical and walk-forward backtests. Crypto markets are volatile and losses are possible at any time. Past or simulated performance is not indicative of future results.

Read the full risk disclosure

Frequently asked

How often do engines actually trade?+

It varies by cycle and regime. The 2H engine may take several trades per day; the 12H engine often takes none for stretches. Inactivity is a feature, not a bug.

Can I run more than one engine at the same time?+

Yes. Engines are independent and can be allocated to in any combination. Many investors split allocation across all three to diversify cadence and risk.

When am I charged the 2% fee?+

Only on net profitable cycles. If a cycle closes flat or negative, no performance fee is taken on that cycle.

What happens if a cycle is interrupted?+

Open positions are managed by their hard stops and targets. If infrastructure detects an anomaly, circuit breakers close exposure conservatively.

Is my capital ever pooled with other users?+

No. Each investor's allocation is tracked individually and trades are reported per account.

Ready to see the engines?

Explore live cycle targets, cadence, and minimum allocation for each engine.