Backtesting

01Methodology

The Falcon AI backtesting methodology applies the live signal engine logic retroactively to historical bar data, recording every signal that would have been generated and its subsequent outcome based on whether price reached the take-profit or stop-loss level first.

• Data Source — CME historical tick and OHLCV data as provided through TradingView's data feed. Data quality and availability vary by contract and time period.
• Testing Period — 3-year walk-forward from May 2023 → May 2026 (three consecutive 12-month periods, each tested independently) with Bar Magnifier enabled. This window spans the elevated-volatility normalization following the 2022–2023 Fed rate-hike cycle, the 2023–2024 AI-driven equity rally, and the 2024–2026 cycle of rate cuts and consolidation.
• Instruments Tested — MNQ (15-min primary / 30-min alternate). Both timeframes verified profitable. 15-min produces more signals with higher win rate; 30-min is simpler to monitor. Secondary testing on MES, US30, and BTC on matching timeframes.
• Entry Assumption — Signals are recorded as entered at the open of the bar following signal confirmation, representing the first available execution opportunity.
• ATR-Based Exits — Exits are sized off a rolling Risk ATR (length 21 on 15-min / length 14 on 30-min). Stop-loss is placed at 0.5 × ATR from entry, take-profit at 5 × ATR, with a fixed 15 / 30 point fallback when ATR compresses. A trailing stop of 0.3 × ATR tightens behind price once the trade moves favorably. The trade closes on whichever of these is reached first within the signal's validity window. All other parameters are identical between timeframes.
• Signal Filtering — Only signals with a confidence score of 2/12 or above (the engine display minimum) are included in the backtested results, consistent with the default threshold in the live indicator.

02Performance Metrics

The following metrics are tracked and displayed in the Falcon AI performance dashboard, accessible to subscribers via TradingView's strategy results panel:

• Win Rate — Percentage of signals that reached TP1 before stop-loss. A secondary win rate to TP2 is also reported.
• Total Trades — Count of all signals taken during the backtest period, used alongside win rate to evaluate sample-size confidence in the results.
• Maximum Drawdown — The largest peak-to-trough decline in the hypothetical equity curve, expressed as a percentage of starting capital.
• Sharpe Ratio — Return per unit of risk, annualized. Calculated using daily P&L under fixed 1% per-trade risk assumptions.
• Profit Factor — Gross profit divided by gross loss. A profit factor above 1.0 indicates a net-positive strategy; above 1.5 is considered strong.
• Consecutive Losing Trades — Maximum number of back-to-back losing signals recorded during the test period. Important for evaluating psychological and capital drawdown risk.

03Key Results

Falcon AI is one signal strategy, backtested across 4 chart configurations to demonstrate robustness. Subscribers pick ONE configuration matching their trading style and run that single setup. Validation method: 3-year walk-forward on MNQ tick data covering three consecutive 12-month periods from May 2023 → May 2026 (TradingView Strategy Tester, Bar Magnifier ON, 2 contracts, $1.50 commission/contract/side, 1 tick slippage, 2/12+ confidence signals). Each period was tested independently — the edge has strengthened year over year, the opposite of a curve-fitted result.

3-Year Walk-Forward Summary (Combined Across All 4 Configs)

The table below shows Period 3 (May 2025 → May 2026 · the most recent 12-month window) broken out by each of the 4 chart configurations — Aggressive style (Trail 0.10), the Elite-default pre-loaded tuning:

Hypothetical backtested results from Period 3 (May 2025 → May 2026). Past performance does not guarantee future results. Apply a 20–30% haircut when projecting real-world performance. Period 3 combined across all 4 configurations: $503,662 · 91.60% avg WR · PF 3.78 avg (system-robustness stat — individual customers run ONE configuration, with results depending on the configuration chosen).

The Aggressive style (Trail 0.10) is the Elite-default pre-loaded tuning. In the most recent walk-forward period, the 15M OPTIMAL configuration delivered 90.98% win rate, PF 3.10, and +$152,315 net. The 30M PRIME configuration is the highest-profit-factor option (PF 4.65) at only ~0.92% max drawdown on a $100K account — the recommended config for prop firm evaluations. All four chart configurations share an identical signal engine, confluence rules, and risk model — only the timeframe, session window (OPTIMAL 8 AM–2 PM ET vs PRIME 9 AM–12 PM ET), and trailing-stop multiplier change. Subscribers pick the configuration matching their style (e.g., scalpers → 15-min OPTIMAL, swing traders → 30-min PRIME, prop-firm conservative → 30-min PRIME).

04Limitations

Backtesting provides useful directional insight into signal methodology but has significant inherent limitations that all subscribers must understand before placing any live trades based on Falcon AI signals.

• No slippage modeled — Backtested entries and exits assume fills at the exact signal price. In live markets, particularly around news events or during pre-market/overnight sessions, actual fill prices may differ materially from the displayed level.
• No commissions included — Exchange fees, NFA fees, and broker commissions are not deducted from backtested P&L. At typical retail commission rates ($0.50–$2.00 per contract per side), high-frequency signals can see meaningful performance degradation.
• Look-ahead bias risk — Despite rigorous controls, Pine Script-based backtesting environments carry a residual risk of unintentional look-ahead in complex multi-timeframe scripts. Results should be treated as optimistic estimates.
• Survivorship bias — The instruments tested remain highly liquid today. Contracts or markets that experienced structural changes or delisting are not represented in the data set.
• No execution psychology — A backtest assumes every signal is acted upon exactly. In live trading, traders frequently miss signals, enter late, exit early, or skip trades — all of which affect actual results.
• Changing market conditions — The SMC patterns and AI confidence thresholds used in the backtest reflect historical effectiveness. Future market regimes, regulatory changes, or shifts in institutional order flow behavior may reduce the signal engine's effectiveness.