Top Gun Fast 1
- エキスパート
- バージョン: 8.0
- アップデート済み: 27 6月 2026
- アクティベーション: 5
Top Gun Fast 1 (Update v8.00): Combining Trend-Following and Mean-Reversion with an Auto Regime Switch
Introduction
Top Gun Fast 1 started out as a single trend-following XAUUSD (gold) trading system (Group 1-2 entry logic plus Group 4 capital-based recovery orders). Through dozens of iterations — adding, removing, and re-tuning various sub-systems — it eventually settled on version 8.00, the best-performing build across every metric tested (Net Profit, Profit Factor, Drawdown) over a 5-year backtest (2021-2026) on XAUUSD M5.
This article summarizes the core idea, the development journey, and the final results of Top Gun Fast 1 version 8.00.
Core Idea: Use the Right Tool for the Right Market Regime
Trend-following systems make money in trending markets but lose by design in sideways markets. Mean-reversion systems do the opposite — they thrive in range-bound conditions but get crushed fighting a strong trend. No amount of parameter tuning can make either approach profitable in both regimes at once.
Top Gun Fast 1's solution is to not pick one system, but let both run, automatically switching control between them based on the market regime detected live — never on a fixed calendar schedule.
Top Gun Fast 1 Architecture
Group 2D — Original Entry System (Trend-Following)
TopGunFast's original system: an entry signal from Group 1 (indicator is configurable), confirmed by two independent trend filters (H1+M15 or M30+M10, OR'd together), plus an optional Volume Delta filter.
Group 2E — Mean-Reversion Range Day-Trading System
A new system built to replace several earlier attempts (PSAR Trend-Flip, Opening Range Breakout, MACD Momentum Reversal — all of which were eventually scrapped for failing to hit the goal). It's built on Bollinger Bands + RSI:
- Buy when price touches/breaks the lower band and bounces back inside it, with RSI in oversold territory
- Sell when price touches/breaks the upper band and bounces back inside it, with RSI in overbought territory
- Filtered through three layers: ADX (H1) must be low (not strongly trending) + the Bollinger Bands basis must not be drifting short-term (M15) + a longer SMA must not be drifting over the medium term (H4, ~8 days). All three are needed because ADX alone can't distinguish a genuinely sideways market from a slow, grinding trend that drifts one direction for weeks.
- Take Profit targets the basis (middle band) instead of a fixed ATR distance — matching mean-reversion's actual "revert to the average" logic.
Auto Regime Switch — the brain that decides who's in control
Uses ADX on the Daily timeframe (deliberately slow/smooth so it doesn't flicker), with two different thresholds (hysteresis):
- ADX drops to ≤ 20 → declares "RANGING" → Group 2E is enabled, Group 2D is blocked
- ADX rises to ≥ 30 → declares "TRENDING" → Group 2D is enabled, Group 2E is blocked
- If ADX sits between 20-30 (the hysteresis gap), no switch happens — the current state holds
This gate only blocks new entries from the deactivated system. Positions already open from the system that just lost control keep being managed normally (SL/TP/Trailing Profit Lock) until they close on their own — they are never force-closed on a regime flip. This means positions from both systems can coexist in the basket during a transition.
Lot Sizing — continuous percentage-of-balance (shared by both systems)
Both Group 2D and Group 2E use the same formula: Lot = current balance × a fixed factor — replacing the old $30-per-step staircase formula. This lets lot size track the account continuously instead of sitting flat for a stretch and then jumping, so risk per trade always matches the account's actual size.
Lessons From What Didn't Work
Top Gun Fast 1 wasn't a straight line. Several approaches were tried and then scrapped or reverted once backtest data proved they didn't work:
- PSAR Trend-Flip / Opening Range Breakout / MACD Momentum Reversal — every attempt to find "one accurate entry per day" was eventually abandoned, because none of them solved the core problem of losing money during sideways markets (2021-2023), no matter how many filters were layered on.
- Equity High-Water-Mark Protection — a sound idea (stop giving back too much profit from the peak), but the first version had a deadlock bug (it waited for equity to recover, but with everything closed, equity could never move on its own). Fixed with a time-based cooldown instead, but it's still unproven as a net benefit, so it's left off by default.
- Max Risk Per Trade Cap (tried as both a flat $ amount and a % of balance) — tested across three rounds (V16, V17, V18), each one significantly reduced overall profit by choking the lot growth the system depends on to compound. Left off by default.
- Fixed 0.01 lot for Mean-Reversion — made Group 2E's contribution to the overall result negligible, even though its standalone win rate was 65-75%. Fixed by switching it to the same lot formula as Group 2D.
Backtest Comparison (5 years, $60 starting capital)
| Metric | Group 2D alone (no Regime Switch) | Top Gun Fast 1 (2D+2E via Regime Switch) |
|---|---|---|
| Net Profit | 3,216.75 | 4,579.49 (+42.4%) |
| Profit Factor | 7.04 | 9.18 |
| Total Trades | 295 | 373 |
| Largest Loss | -38.25 | -28.10 (26% smaller) |
| Avg Win : Avg Loss (R:R) | 2.32 : 1 | 3.01 : 1 |
| Max Drawdown (Balance) | 72.09% | 44.28% |
| Win Rate (Long/Short) | 75.4% / 75.0% | 75.1% / 75.6% |
Key observation: combining both systems through the Auto Regime Switch wasn't the usual "trade safety for profit" compromise typical of risk management. It improved both at once — because it fixed a structural weakness in Group 2D alone (trading in the wrong regime some of the time), rather than simply shrinking position size.
Known Limitations Before Going Live
- Group 2E's risk:reward is still below 1:1 on its own (Avg Win 4.10 vs Avg Loss -4.88) — it survives purely on a high win rate. There's room to improve (e.g., widening the take-profit target beyond the basis line).
- Profit is concentrated in the later years of the test — a natural consequence of lot size compounding with the balance. Early-year trades, when the account was still small, will look tiny in dollar terms compared to later years.
- These are historical backtest results and don't guarantee future performance. Forward/demo testing is strongly recommended before risking real capital, especially in market conditions not covered by this test window.
- The Equity HWM Protection and Max Risk Per Trade Cap features are still present in the code but disabled by default. Anyone wanting tighter risk controls should understand the profit trade-off already demonstrated in testing before re-enabling them.
Conclusion
Top Gun Fast 1 version 8.00 is the product of dozens of rounds of test-measure-discard-improve. What it converged on is simpler than expected: two opposite-natured systems (trend-following and mean-reversion) take turns trading, switched by a hysteresis-smoothed Daily ADX signal, sharing the same percentage-of-balance lot sizing, with no forced closing of positions on a regime switch — each trade is simply left to finish on its own terms.
The result is a system that makes more money while cutting drawdown nearly in half compared to running the trend-following system alone.
