Research's job is to notice patterns the market hasn't priced out — a structural edge in options premium decay, a momentum regime in a specific instrument, a mean-reversion tendency in correlated pairs. The hypothesis doesn't need to be proven at this stage. It needs to be specific enough to test.

Once Research proposes an idea, the Managing Director scopes it: instruments, timeframe, capital estimate, risk budget. The idea is registered in the initiative registry with a unique ID before any analytical work begins. An unregistered idea cannot advance — the registry is the source of truth for everything Wesden has ever considered.

Full backtesting across at minimum three distinct market regimes: trending, ranging, and high-volatility. No cherry-picked start dates. No parameters tuned to peak past performance. Walk-forward validation is required — the parameter set is fixed on historical data, then run blind on an out-of-sample period. If it doesn't hold, it doesn't pass.

Wesden's anti-overfit protocols are non-negotiable: minimum 100 trades in backtest, no parameter adjustment after out-of-sample begins, regime break analysis required. Competitive benchmarking against known external strategies in the same category is required before R&P handoff. Research must exhaust all angles before bringing conclusions to council — the council sees a final recommendation, not an open question.

Risk & Performance runs an independent audit of Research's output — not a peer review, but an adversarial check. R&P has authority to reject any strategy that fails its thresholds, regardless of how promising the backtest appeared. A single metric breach is a hard stop.

The audit covers maximum drawdown and Calmar ratio, options Greeks exposure (delta, theta, vega, gamma), portfolio-level impact analysis across concurrent strategies, and walk-forward re-validation on R&P's own data copy. A FAIL verdict cycles the strategy back to Stage 02 with documented blocking conditions — it cannot return without addressing every listed condition.

Engineering builds the production system — not a prototype, not a script requiring human intervention. The standard: screener runs on schedule, signals process through the webhook server, orders route to IBKR, positions are monitored continuously, exits fire automatically on trigger conditions, and nightly reconciliation catches any drift between broker state and internal records.

Execution reviews the integration path at this stage — their operational knowledge of live trading constraints is a required input before the build is complete. End-to-end testing on the paper account is the final gate: the system must complete a full trade cycle without manual intervention before Stage 04 closes and the council package is assembled.

The council receives the complete package: Research backtest memo, R&P review memo with any conditional requirements, Engineering E2E test results, and Execution's integration sign-off. Each seat reviews independently — no pre-vote discussion, no lobbying. Each seat casts a documented vote with written reasoning.

The voting rule is strict: unanimous PASS required for promotion. One FAIL blocks regardless of other votes. Abstain is treated as FAIL. Conditional R&P passes must be resolved before the vote is called. MD aggregates results and communicates the outcome. A failed vote triggers a blocking conditions memo — the strategy can only return when every listed condition is formally resolved.

The strategy runs on live market data with paper account fills. This is not simulation — signals are generated by the same production system that will run with real capital. The difference is fills route to the IBKR paper account, not the live account. P&L, drawdown, and all exit triggers behave identically.

Automated monitoring tracks every position: 50% profit target triggers a close, 2× stop triggers a protective exit, EMA trend break triggers a strategic exit. The monitor runs on a 60-minute cycle during NYSE hours. MD oversees pilot performance and prepares the Owner review package when sufficient data has accumulated. A quiet pilot with no signals is a research gap, not acceptable behavior.

The Owner reviews the complete pilot package: realized P&L across all pilot trades, drawdown envelope, signal quality metrics, exit behavior analysis, and the council's promotion recommendation. Position sizing and capital allocation are set at this gate. On approval, the strategy goes live with real capital on the first valid signal after the gate.

This is not a rubber stamp. The Owner has rejected strategies at this stage before, and will again. The standard is not "did it work in pilot" — it is "would I commit real capital to this system, run by this team, at this moment in the market cycle." Only strategies that survive that question go live. This gate cannot be delegated, bypassed, or accelerated.