The Geometry Trading Research Report ¶

A Quantitative Audit of Retail Price Action, Smart Money Concepts, and Macro Regime Gating

Executive Summary ¶

Over the course of intensive backtesting and algorithmic vetting, every mainstream retail trading concept—including Smart Money Concepts (SMC), ICT Fair Value Gaps, Order Blocks, Asian Session Breakouts/Sweeps, VWAP Reversion/Momentum, Round Numbers, and Oscillator Divergences—was subjected to strict institutional quantitative testing standards.

Using direct tick data from MetaTrader 5 (MT5) with institutional broker friction modeling (raw spreads and commissions), each concept was tested across 100,000 to 200,000 continuous tick candles (2 to 4+ years).

The research yielded two profound discoveries:

1. The Retail Geometry Manifesto (Continuous Failure): When executed as static, always-on single-asset strategies, all intraday chart geometry mathematically collapses into a random walk. The market efficiently counter-balances the Win Rate against the Risk/Reward ratio, ensuring that total Expectancy converges to 0.0 - Broker Spread.
2. The Regime Gating Breakthrough (Conditional Alpha): When tested across discrete rolling 3-month blocks over 3 to 5 years, intraday geometry strategies were proven not to be fundamentally useless, but strictly regime-dependent. Certain geometric patterns (like FVG mitigations or Inside Bar breakouts) yield massive positive expectancy, but only when gated by macro-environmental filters (e.g., ADX > 40 or extreme distance from the 200 SMA).

1. Quantitative Vetting Standards Applied ¶

To eliminate heuristic bias and curve-fitting, all strategies were coded into a standardized StrategyCore interface and executed via a vectorized simulation engine (FastEngine):

• Sample Depth: Intraday M5 and M15 timeframes required a minimum of 50,000 to 100,000 continuous candles (~2 to 4.5 years).
• Sample Count ($N$): A minimum of $N \ge 300$ executed trades to satisfy the Law of Large Numbers.
• Expectancy Metric ($EV$): Calculated in R-multiples:
  $\text{Expectancy } (R) = (\text{Win Rate} \times \text{Avg Win } R) - (\text{Loss Rate} \times \text{Avg Loss } R)$
  Pass threshold: Minimum $+0.15\text{ R}$ to $+0.20\text{ R}$ after full friction deduction.
• Profit Factor ($PF$): Minimum $PF \ge 1.40$.
• Risk Modeling: Full raw spread matrix, latency buffer, and commission modeling.

2. Continuous Macro Test Results (The Falsification) ¶

A. ICT Fair Value Gaps & Order Block Mitigation ¶

• Module: fvg_order_block_core.py
• Timeframe: M15 (100,000 candles, ~4 Years)
• Mechanics: Identified 3-candle price voids where Candle 1 High and Candle 3 Low did not overlap. Required Candle 2 body to exceed 1.5x ATR (institutional displacement). Placed limit orders at the gap retest with a 2:1 RR (Stop Loss below origin candle).
• Macro Results:
  • EURUSD: 1,208 Trades | Win Rate: 32.70% | Expectancy: -0.09 R | PF: 0.97
  • GBPUSD: 1,265 Trades | Win Rate: 31.94% | Expectancy: -0.12 R | PF: 0.94
  • USDJPY: 1,271 Trades | Win Rate: 34.38% | Expectancy: -0.02 R | PF: 1.05
• Macro Filter Test: Adding an H4 200 EMA directional filter reduced trade count from ~1,250 down to ~750, but expectancy remained negative (EURUSD: -0.06 R, GBPUSD: -0.10 R).
• Conclusion: Mechanically catching pullbacks into visual price voids holds zero standalone statistical edge.

B. Asian Session Volatility Breakout ¶

• Module: volatility_breakout_core.py
• Timeframe: M15 (100,000 candles)
• Mechanics: Measured Asian session range (00:00–08:00 UTC). Entered on London open breakout if today's range was 10% tighter than the 20-day rolling average (volatility compression).
• Macro Results:
  • Wide Stop (Bottom of Asian Range): Generated a 42% Win Rate, but Expectancy was -0.05 R.
  • Midpoint Stop (Tighter SL to double R:R): Win Rate immediately fell from 42% to 32% (GBPJPY 32.69% WR, -0.10 R; GBPUSD 26.02% WR, -0.05 R; EURUSD 24.45% WR, -0.15 R).
• Conclusion: The market dynamically neutralizes risk/reward adjustments. Compressing stop losses simply increases stop-out frequency at an identical mathematical ratio.

C. London Stop Hunt (Asian Sweep / Judas Swing) ¶

• Module: asian_sweep_core.py
• Timeframe: M15 (100,000 candles, ~4 Years)
• Mechanics: Waited for price to break beyond the Asian session high/low during London open, then entered a fade if the M15 candle closed back inside the range (false breakout/liquidity run).
• Macro Results:
  • EURUSD: 855 Trades | Win Rate: 30.64% | Expectancy: 0.00 R
  • GBPUSD: 868 Trades | Win Rate: 26.38% | Expectancy: -0.16 R
  • GBPJPY: 780 Trades | Win Rate: 23.85% | Expectancy: -0.16 R
• Conclusion: Over 855 trades on EURUSD, the expectancy flatlined at exactly 0.00 R before spread. Stop-run fades are efficiently counter-balanced by high-frequency market-making algorithms.

D. VWAP Mean Reversion & Inversion ¶

• Modules: vwap_reversion_core.py, vwap_momentum_core.py
• Timeframe: M5 (200,000 candles, ~2.6 Years)
• Mean Reversion (Fading 2.5 SD Bands):
  • EURUSD: 344 Trades | Win Rate: 29.65% | Expectancy: -0.07 R
  • GBPUSD: 360 Trades | Win Rate: 30.56% | Expectancy: -0.10 R
  • USDJPY: 336 Trades | Win Rate: 22.02% | Expectancy: -0.26 R
  • Diagnosis: Fading standard deviation bands on intraday timeframes is stepping in front of institutional trends.
• Momentum Inversion (Buying 2.5 SD Breakout with 2:1 RR):
  • EURUSD: 669 Trades | Win Rate: 32.44% | Expectancy: -0.08 R
  • Diagnosis: Inverting a losing strategy does not produce a winning strategy. Because forex ranges ~70% of the time, breakout entries trigger on wicks and get chopped up by spread and variance.

E. Daily-Anchored M15 Momentum ¶

• Module: daily_anchor_momentum_core.py
• Timeframe: M15
• Mechanics: Evaluated M15 VWAP pullbacks strictly in the direction of the Daily 50 SMA.
• Macro Results:
  • With 16:00 London close exit: Win rate was ~40%, but expectancy was -0.03 R to -0.07 R because time-stops truncated runners while taking full 20-pip losses.
  • With 4:1 RR (80-pip TP / 20-pip SL): EURUSD produced 355 trades, 19.44% Win Rate, -0.07 R Expectancy. Moving averages are lagging indicators; by the time the Daily trend is confirmed, M15 pullbacks frequently align with macro trend exhaustion.

F. The Timeframe Paradox (Bollinger Double Bounce) ¶

• Module: bb_double_bounce_core.py
• H4 Timeframe: Generated verified alpha (+0.22 R, PF ~1.50).
• H1 Timeframe: Tested over 8 years (1,600+ trades) to increase frequency. The edge immediately evaporated to -0.05 R.
• The Law of Timeframe Physics: A 2.5 Standard Deviation extension on H4 represents global macroeconomic liquidity exhaustion (central bank level re-balancing). A 2.5 SD extension on H1 or M15 represents localized retail stop runs (random noise). High frequency and standalone structural geometry are fundamentally incompatible.

G. Other Quarantined Intraday Concepts ¶

• Round Number Bounce (M1/M5): 24 trades in 100,000 candles (-0.12 R). Excessive constraint stacking and negative expectancy.
• RSI Divergence Reversal (M5): -0.09 R on EURUSD, -0.18 R on GBPJPY. Oscillators fail to predict the duration of institutional blow-off tops.
• London Close Trap (M15): Profitable over 8 months (favorable regime), but collapsed to -0.09 R to -0.22 R across a full 3-year multi-regime sample (220,000 candles).

3. The Regime Orchestrator Breakthrough ¶

Rather than discarding all failed geometry strategies, we tested the hypothesis: "Do these strategies work consistently under specific, identifiable market weather?"

All 18 strategies were re-evaluated across rolling 3-month blocks over 3 to 5 years, tracking four macro sensors:

1. ADX (14): Trend strength and momentum velocity.
2. ATR (14): Realized market volatility.
3. Dist200 (%): Absolute percentage distance between price and the 200 SMA (Macro Extension).
4. Bollinger Squeeze (20): Volatility compression.

The Findings: Regime Separation ¶

4. The Universal Failures (Permanently Quarantined) ¶

Four strategies failed to generate a single profitable 3-month block across the multi-year history, regardless of market conditions:

1. Round Number Bounce (M5): 0 for 5 blocks profitable. Psychological round numbers offer no quantifiable edge over spread.
2. ATR Trend Follower (D1): Flawed ATR trailing-stop logic caused unrecoverable equity drawdowns.
3. ATR Fade (M15): 0 for 10 blocks profitable. Intraday fading based on ATR bands bleeds capital continuously.
4. VWAP Reversion (M5): 0 for 4 blocks profitable. Mean-reverting to VWAP intraday is mathematically negative across all regimes.

5. Architectural Implementation & Strategy Portfolio ¶

These findings led to the construction of a dynamic Regime Orchestrator, establishing a two-tier portfolio architecture:

Tier 1: Universal / Global Strategies (Always-On) ¶

Strategies that do not rely on intraday chart patterns and are immune to single-asset regime collapse:

• Statistical Arbitrage (Pairs Cointegration): H1/D1 mean-reversion between cointegrated assets (EURUSD/GBPUSD, SPGI/MCO, GS/MS). Market neutral.
• Connors RSI-2 Mean Reversion: D1 equity pullbacks on S&P 500 / Nasdaq stocks above the 200 SMA.
• Macro Carry Trades: EOD central bank interest rate differential harvesting with a Daily 200 EMA trend filter.

Tier 2: Regime-Gated Geometry Strategies ¶

Intraday strategies dynamically toggled by the Orchestrator via real-time macro telemetry:

• London ORB: Active only when Dist200 < 1.5%
• Inside Bar Momentum: Active only when ADX > 35.0 AND Dist200 > 4.0%
• Asian Sweep: Active only when ADX < 35.0
• FVG Order Block: Active only when ADX > 40.0
• Daily Anchor Momentum: Active only when ADX > 40.0 AND Dist200 < 2.0%
• RSI Divergence: Active only when Dist200 < 2.5%
• BB Double Bounce: Active only when Dist200 < 2.0%
• Weekend Gap Fill: Active only when Dist200 < 4.0%