How it works
Every stage traceable
Clear rules at every stage. If one fails, the pipeline stops.
01 · Data
Ingest & normalize
Every decision starts with data, so data gets checked first.
- One format via adapters
- Cleaning & time alignment
- Anomalies flagged, never filled
02 · Analysis
AI & quant models
Complex conditions become comparable scores.
- Models plus factors
- Regime detection
- Structured scores
03 · Signal
Strategy decisions
Scores become explicit actions — with reasons.
- Explicit entry & exit
- Source and rule logged
- Any decision rebuildable
04 · Risk
An independent veto
Risk is a separate gate, not a parameter.
- Position & exposure caps
- Stops & abnormal pauses
- Fail means no execution
05 · Execution
Automated execution
Only fully checked signals become orders.
- Orders via adapters
- Real-time status
- Observable end to end
One signal's journey (demo)
Follow one signal through the pipeline
A walkthrough on simulated data: one signal, from the moment data arrives to being approved or blocked — what each step checks and what it records.
01 · Data
A new bar arrives
The adapter receives the latest price bar and checks that timestamps align and nothing is missing or jumping abnormally. Only validated data moves on.
Bar 64 · aligned ✓ · complete ✓
02 · Analysis
The model re-scores
The model condenses the latest market state into scores. After this bar, the momentum score rises from 0.29 to 0.36.
Momentum score 0.29 → 0.36
03 · Signal
The score crosses the threshold
The score crosses the 0.32 threshold from below. The strategy emits an “increase exposure” signal and records which score and rule triggered it.
Buy signal · exposure 0% → 25%
04 · Risk
Four rules checked one by one
Exposure cap, volatility filter, cooldown, circuit breaker — each checked in turn. Only if all pass does the signal go through.
Exposure ✓ · Volatility ✓ · Cooldown ✓ · Circuit ✓ → approved
05 · Execution
Order sent and result reported
Only approved signals become orders. The fill report is written together with the full decision record, so it can be reconstructed later.
Simulated order sent · record written
Blocked
If any rule fails
Suppose exposure is already at 50%. Adding 25% would exceed the cap, so risk blocks it and no order is created. The block is recorded too, for later review.
Exposure cap ✕ (50% + 25% > 50%) → blocked
- Data
- Analysis
- Signal
- Risk
- Execution
Threshold 0.32
Cap 50%
- DataBar 64 · aligned ✓ · complete ✓
This is an illustrative walkthrough. The 0.32 threshold, 25% steps and 50% cap come from this site's interactive demo rules and don't represent real system parameters, trades or performance.
From idea to live
Strategy lifecycle
- 01
Research
Form a testable hypothesis.
- 02
Backtest
Validate on history.
- 03
Risk rules
Set limits and pauses.
- 04
Live
Execute and monitor.
- 05
Iterate
Improve from records.
Now · Design · In development
What runs today and what comes next
Now
Running today
- Long-running live trading
- AI analysis
- Research & backtests
- Automated risk & monitoring
Design
Supported by design
- Any data source
- Pluggable execution
- Core decoupled from markets
In development
Where it's heading
- A self-improving loop
- Open free sign-up
- More data and use cases