How We Operate

1. Publish everything

Every signal detected is published — wins, losses, and misses. Nothing is hidden, deleted, or edited after the fact. Paper trading is always labelled as paper trading. Live trading will be labelled too. Signal IDs are sequential, so gaps would be visible.

2. Market-neutral by design

The bot does not bet on market direction. Every strategy — cross-exchange arbitrage, triangular arbitrage, mean reversion, flash crash harvesting — is structured to capture spreads, not take directional exposure. Profit comes from inefficiency, not prediction.

3. Verifiable from signal #0001

Every signal has a unique ID and a verification record. Every signal is hashed with SHA-256 and anchored to the Bitcoin blockchain using OpenTimestamps. Anyone can verify any signal independently — without trusting us. The track record is the track record. There is no separate curated version.

4. Real data only

Every price, spread, and fee number comes from real exchange APIs in real time. The system connects to 6 exchanges via WebSocket feeds and monitors 550+ trading pair combinations. Backtests are labelled as backtests. Simulated data is labelled as simulated.

5. No frozen assumptions

Every performance parameter is a hypothesis under continuous test — spread thresholds, timing windows, z-score triggers, pair weightings. These improve automatically as data accumulates. Risk parameters are different: loss limits, circuit breakers, and leverage constraints are fixed by design. The system optimises for performance within a risk framework it cannot modify.

6. Closed-loop learning

Nothing changes without a hypothesis. Nothing stays changed without evidence it worked. Every change is logged with rationale, a measurable expected outcome, and a review date. Changes that don't improve the tracked metric after 7 days revert automatically. The system learns from every signal, every trade outcome, every market regime change, and every piece of subscriber feedback. When evidence contradicts an assumption — the assumption loses.

7. The subscriber is a signal

Subscriber behaviour — what they engage with, what they ignore, what makes them stay, what makes them leave — is treated as a data source of equal importance to trading performance. This feedback runs through the same improvement loop as trading data — classified, measured, and acted on. A system that ignores its subscribers is optimising in the wrong direction.

8. Risk limits are non-negotiable

Circuit breakers and daily loss limits are hard-coded into the system and cannot be overridden during a trading session. If a loss limit is hit, the system stops. These values are set deliberately before deployment and changed only by a human through a formal deploy process — never reactively, never autonomously.

9. Compliant exchanges only

The system operates through a Singapore entity and trades only on MAS-compliant exchanges. All API keys have withdrawal permissions disabled. The current exchange set is Kraken, Gemini, Crypto.com, Coinbase, OKX, and Bitfinex.

10. Infrastructure stays lean. This is not financial advice.

Infrastructure costs are capped and tracked against revenue. Every new tool requires a cost-benefit gate before deployment. The system runs on a single VPS at $26/month. Complexity is added only when justified by measurable outcome. Proof Systems documents what an autonomous trading system does. It does not recommend trades, suggest positions, or direct action. Signals describe bot activity. What you do with that information is your decision.