Methodology
Simulation Before Deployment
A diagnostic protocol for observing decision behavior in controlled environments — producing a forward performance record that enables qualification assessment without capital risk.
Objective
Produce a forward performance record by running a validated strategy against live market data in a simulated environment with isolated capital. The record enables qualification assessment by revealing behavioral characteristics — trade frequency, exit patterns, drawdown behavior, regime sensitivity — that historical backtests cannot measure.
Scope
Covers
- Paper trading account setup with isolated capital
- Tick-based signal evaluation against live market data
- Position lifecycle: entry, management (breakeven, trailing, partial TP), and exit
- Exit attribution recording (MAE/MFE, drift, mechanism)
- Performance accumulation over the observation window
- Behavioral signature detection (fee-bleeder, drawdown patterns, regime sensitivity)
Does Not Cover
- Historical validation or backtesting (covered by Validation-First Qualification methodology)
- Qualification assessment logic (covered by Decision Lifecycle methodology)
- Live execution, order routing, or exchange connectivity
- Capital sizing or portfolio construction
- Strategy generation or parameter selection
Preconditions
These inputs and assumptions must be satisfied before the methodology can produce valid results.
| Input | Requirement |
|---|---|
| Validated strategy specification | StrategySpec that has passed historical validation (backtest) with at least 3 trades and a passing composite score |
| Live market data feed | Real-time OHLCV candle data at 1-minute cadence from the target venue (Hyperliquid) |
| Paper trading infrastructure | Tick-based execution engine with isolated capital accounts, position tracking, and exit attribution |
| Portfolio assignment | Target portfolio and workspace for the paper account (determines diversity constraints) |
| Initial capital | Simulated capital allocation (default $1,000 for paper accounts) |
Procedure
Create Isolated Paper Account
Provision a paper trading account with dedicated capital. The account is fully isolated — its own equity, positions, and performance history. No capital sharing across accounts. Record the account creation as a strategy event. Link to the strategy version and portfolio allocation.
Decision Criteria
If bot capacity is exhausted (system-wide or per-user) → queue for admission. If diversity constraints are violated (asset/family/timeframe caps) → reject or queue. Each account must have a unique allocation — no duplicate strategies in the same portfolio.
Configure Tick Evaluation
Set up the tick cadence and data source. Paper ticks run at 1-minute cadence via Cloud Scheduler. Data source is Hyperliquid candle data (matching the live execution venue for consistency). For higher-timeframe strategies (4h, 1d), mark-to-market uses 1m klines while signal evaluation occurs at the strategy's native timeframe candle close.
Decision Criteria
If TICK_MODE is event_driven → tick evaluation occurs on candle close events from the kline listener (Pub/Sub). If polling → tick evaluation occurs on scheduler cadence. Hybrid mode supports both.
Evaluate Entry Signals
On each tick, evaluate the strategy's entry conditions against the current candle. Check all indicator conditions (up to 24 supported indicators) with configured operators (gt, lt, crosses_above, crosses_below, between). Apply cooldown period between trades. Respect position limits (one open position per account).
Decision Criteria
If all entry conditions are met AND no open position AND cooldown period has elapsed → open position. If any condition fails → no action. Duplicate position guard: pre-write check + unique index prevents double-entry from event/catch-up races.
Manage Open Positions
For open positions, evaluate trade state management on each tick. Exit evaluation order: partial TP → stop loss → take profit → exit signals → time decay. Breakeven activation moves stop to entry when profit exceeds threshold. Trailing stop tracks high-water mark and trails by configured width. Partial TP closes a fraction of the position at the trigger level. Exit profile v2 uses ATR-relative thresholds for all management levels.
Decision Criteria
Stop level is max(fixed_sl, breakeven, trailing). SL/TP evaluate against intra-candle extremes (high/low) and fill at the stop/TP level. Signal exits evaluate on close. Time decay fires only when bars_held >= max_bars AND breakeven never activated AND position is negative.
Record Exit Attribution
On every position close, capture exit quality metrics in the position_exit_audit table. Record: close reason (exit_signal, stop_loss, take_profit, trailing_stop, time_decay, manual), MAE (maximum adverse excursion), MFE (maximum favorable excursion), post-exit drift (price movement after close), and mechanism attribution. This data is the raw material for the Evidence Loops methodology.
Decision Criteria
Every closed position must have an exit audit record — no exceptions. Post-close hooks fire asynchronously: qualification refresh and micro-backtest divergence analysis.
Accumulate Behavioral Data
Over the observation window (minimum 7 days), the 6-hourly evaluation batch aggregates behavioral data: paper ROI, win rate, trade count, average hold duration, drawdown profile, regime-conditional performance. Fee-bleeder detection runs on a rolling 20-trade window. M6 anomaly detection (SPC: level shift, trend, volatility) runs in shadow mode. M5 correlation scoring computes portfolio diversity.
Decision Criteria
After 7 days with 3+ trades → eligible for qualification assessment. After 30+ days with 10+ trades → high-confidence qualification. Fee-bleeder alert if rolling avg return < 0.3%. Divergence flag if paper-backtest p < 0.05.
Expected Outputs
Forward Performance Record
Complete paper trading history: every trade with entry/exit prices, hold duration, P&L, close reason, and regime context at entry
Exit Attribution Trail
Per-trade MAE/MFE, post-exit drift, mechanism attribution — raw material for evidence loops
Behavioral Assessment
Aggregated behavioral signatures: trade frequency, win rate, drawdown profile, regime sensitivity, fee efficiency
Divergence Metrics
Statistical comparison of paper trading performance vs historical backtest over matched periods
Equity Curve
Time-series of account equity for Sharpe calculation, correlation analysis, and visual assessment
Interpretation Guide
How to read the outputs this methodology produces.
Paper ROI > Backtest ROI
Forward conditions may be more favorable than historical period, or strategy benefits from current regime
Action: Positive signal but verify regime is not unusually favorable — performance may revert
Paper ROI < Backtest ROI (within tolerance)
Normal variance. Forward performance rarely exactly matches historical
Action: Continue observation. Qualification criteria account for expected variance
Paper ROI << Backtest ROI (p < 0.05)
Statistically significant divergence. Strategy assumptions may not hold in current conditions
Action: Flag for review. May indicate overfitting or regime mismatch. Continue observation — may resolve
Zero trades after 14 days
Entry conditions not triggered in current market
Action: Not a failure. Strategy may be designed for a specific regime. Continue observation.
Fee-bleeder detected
Strategy generates trades but net returns are consumed by transaction costs
Action: Advisory alert. Strategy trades frequently with small edge — may not be viable for live deployment
High MFE with low capture
Strategy enters correctly but exits too early, leaving profit on the table
Action: Exit profile may need adjustment. Evidence feeds back into Continuous Optimization methodology.
High MAE with recovery
Strategy endures significant drawdown before recovering to profit
Action: Stop placement may be too wide. Risk-adjusted return suffers even if gross return is positive.
Worked Example
Scenario
A breakout strategy on SOL/USDT with 1h timeframe, Bollinger Band entry, ATR trailing exit profile (v2), is admitted to paper trading after passing backtest validation with composite score 62.5.
Create paper account with $1,000 capital in user's default portfolio. Diversity check: 5/15 SOL slots used.
→ Paper account #1847 created. Allocation linked to strategy version #392.
Configure: TICK_MODE=event_driven, data source=Hyperliquid, native timeframe=1h, MTM=1m klines.
→ Tick evaluation triggers on 1h candle close events from kline listener.
Day 3: BB lower band breakout detected. SOL price crosses above lower band after squeeze. All entry conditions pass.
→ Long position opened: 10.5 SOL at $142.30. Cooldown timer starts.
Day 3-5: Position management. Breakeven activates at ATR% x 1.5 (day 4, +$3.20). Trailing starts at ATR% x 2.5 (day 5, +$5.80).
→ trade_state: {breakeven_activated: true, trailing_active: true, trailing_hwm_return: 4.1}
Day 6: Trailing stop hit. Exit at $148.10. Exit audit: close_reason=trailing_stop, MAE=-1.2%, MFE=+5.1%, post-exit drift=+0.8%.
→ Position closed with +$60.90. Exit audit record written. Qualification refresh fired.
After 21 days: 5 trades completed. Paper ROI +3.8%. Win rate 60%. Backtest ROI over matched period: +4.2% (divergence p=0.41).
→ Qualification eligible: 21d > 14d min, 5 trades > 3 min. Divergence within tolerance. Regime proof: trending_up 3 trades +2.1% avg.
Outcome
Paper trading produces a forward performance record over 21 days with 5 trades. Behavioral assessment: healthy trade frequency (1 trade/4.2 days), good MFE capture (trailing stop working), acceptable MAE. Strategy is eligible for qualification assessment via the Decision Lifecycle methodology.
Failure Conditions
How to recognize when the methodology is not producing valid results.
Tick cadence disruption (>5 min delay)
Paper tick scheduler is delayed or stale
Check scheduler health. Stale ticks degrade simulation fidelity. Live-tick-status endpoint reports delayed/stale thresholds.
Data source inconsistency
Paper trading data source differs from live execution venue
Verify data source is Hyperliquid (PAPER_DATA_SOURCE=hyperliquid). Mismatched sources invalidate divergence assessment.
Position isolation breach
Capital or position state leaks between accounts
Critical failure. Each account must be fully isolated. Check for shared state in tick evaluation.
Missing exit audit records
Closed positions without attribution data
Breaks the evidence chain. Exit attribution is mandatory for every closed position. Fix instrumentation before continuing.
Extreme drawdown (>20% of account)
Strategy experiencing severe adverse movement
Advisory alert. Paper trading does not auto-pause (unlike live RiskGuard). Continue observation — drawdown behavior is itself useful data.
Duplicate position opened
Race condition between event-driven and catch-up ticks
Should be caught by pre-write fetch + unique index (409 handler). If occurring, investigate tick overlap.
Quality Checks
| Check | Passes When | Fails When |
|---|---|---|
| Tick cadence matches configured interval | Paper ticks arrive at 1-minute intervals with < 5s jitter | Ticks are delayed, skipped, or arrive in bursts |
| Data source matches live venue | Paper trading uses Hyperliquid candle data (same as live execution) | Paper uses a different data provider, introducing venue-specific price divergence |
| Position isolation is complete | Each paper account has independent equity, positions, and P&L calculations | Any shared state between accounts (capital pool, position limits, etc.) |
| Exit attribution is complete | Every closed position has an exit audit record with MAE, MFE, drift, and mechanism | Closed positions without attribution data — breaks evidence chain |
| Mark-to-market reflects current price | Open positions show unrealized P&L based on latest 1m kline | Unrealized P&L is stale or computed from wrong timeframe |
Handoff
Where the outputs of this methodology go next in the protocol chain.
Qualification Pipeline
Evidence Loops
Allocation Intelligence
Divergence Predictor (M1)
Termination State
Produces
A forward performance record that either supports qualification (sufficient trades with acceptable performance) or documents why qualification was not achieved (no trades, divergence, negative ROI). Both outcomes are valuable — negative evidence is still evidence.
Confidence
Minimum viable: 7 days, 3 trades. Moderate confidence: 21 days, 6+ trades. High confidence: 30+ days, 10+ trades. Confidence also depends on regime coverage — a strategy observed through a regime transition provides richer behavioral data than one observed in a single regime.
Next Step
Forward performance record feeds into qualification assessment (Decision Lifecycle step 6). Exit attribution feeds into Evidence Loops. Equity curve feeds into allocation intelligence scoring.
Replication Notes
What another team would need to reproduce this methodology.
- Simulation must use the same data source as the target live execution venue to enable meaningful divergence measurement
- Tick cadence must be fast enough to capture intra-candle extremes for stop/TP evaluation (1-minute recommended)
- Exit profile v2 (ATR-relative thresholds) requires ATR computation from the same candle data used for signal evaluation
- Position isolation is a hard requirement — shared capital or shared position limits invalidate the simulation
- Exit attribution must be recorded at close time, not backfilled — post-exit drift is measured from the actual close candle
- Minimum observation period (7 days) should be calibrated to the strategy's expected trade frequency — high-timeframe strategies may need longer windows
- The simulation environment must support regime-aware position tracking (entry_regime_label on every position) for downstream qualification and learning
Knowledge Lineage
Introduced In
Investigated In
Depends On
Related Research
Frameworks
See this methodology implemented in Orqis