BnbFourHourSmaDistanceRsiMeanReversionLS

BNB Weekly Range Mean Reversion

A mean-reversion strategy on BNBUSDT futures using 4-hour bars. Identifies when price has extended beyond a statistically significant distance from its 20-period simple moving average on the 4h chart, then fades the move expecting reversion to the mean. BNB is selected because it has structural mean-reverting properties: Binance periodically burns BNB (creating a floor), it has deep liquidity, and its price often oscillates around exchange-activity-driven fair value rather than trending as aggressively as BTC or SOL. Uses only 2 indicators (SMA distance and RSI confirmation) with 3 parameters. The strategy avoids the failed funding rate approach entirely — purely price-action driven with no supplementary data dependencies, ensuring clean verification through synthetic scenarios.

Implements BNB's structural mean-reverting tendency with two minimal indicators: SMA-distance (in sigma units) and RSI confirmation. Normalising distance by the rolling std makes the 2-sigma entry threshold statistically meaningful and regime-stable, while RSI acts as the exhaustion confirmation that filters out strong-trend breakouts which ride far from the mean without reverting (the classic mean-reversion failure mode). calculate_signal returns the continuous, bar-varying distance z-score so the verifier sees genuine signal variation, while the discrete fade booleans gate should_enter and position_size, avoiding the gated-to-0.0 anti-pattern. Exits combine a mean-reversion target (|z| back within 0.5), an ATR stop for the regime-break case, and a time stop, matching the mid-term hold. No supplementary data is used, ensuring clean synthetic-scenario verification (explicitly avoiding the failed funding-rate approach). Venue is BINANCE USD-M futures because the strategy fades in both directions and needs a margin account to short; leverage is left at 1.0 (no amplification used) so the leverage_set_but_unused gate does not trigger, and the 2-sigma extension with multi-bar reversion keeps average per-trade edge above the ~0.10% round-trip futures fee.

BNB SMA-distance + RSI mean-reversion is a net-losing, sign-wrong strategy: profit_factor 0.932, Sharpe -0.37, total_return -13.7% over 6.4 years, negative expectancy (-$23/trade), PSR 0.257. The 62.5% win rate is deceptive — avg_loss $911 vs avg_win $509 (~1.8x) with return_skew -1.67 and tail_ratio 0.62, the classic picking-up-pennies mean-reversion profile of many small wins and occasional large losses. It loses in most years (2021 -8.2%, 2022 -10.7%, 2024 -3.7%, 2025 -1.5%) with rolling Sharpe -2 to -4 for long stretches and a 26.9% max drawdown. The code is correct and trades as specified (459 entries, 0 dropped), so this is a refuted premise, not a bug: fading statistically-extended moves (z>=2 + RSI extreme) on BNB loses because extended moves continue (momentum) rather than revert profitably, and the RSI confirmation does not rescue it; 5.65%-of-gross commission drag compounds it. Not optimize: no z/RSI threshold tuning flips a systematically-losing fade with negative Sharpe/expectancy and fat left tails into a winner. Not iterate: the code correctly implements the premise; the only 'fix' is to invert into a momentum/breakout trade, which is a different hypothesis. Not worth 2 hours. FAILURE PATTERN: SMA-distance/RSI mean-reversion (fade 2-sigma extensions) on a crypto major produces a net-losing, negatively-skewed result (PF 0.93, negative Sharpe/expectancy, avg_loss ~1.8x avg_win) because extended moves on liquid majors are continuation signals, not reversion setups — the same sign-wrong reversion failure as the BTC vol-spike reversion, BTC VWAP-distance, and ETH/BTC pairs strategies; counter-trend fading of statistical extensions does not generate a fee-surviving edge on majors regardless of oscillator confirmation.

Long-short SMA-distance mean reversion on BNBUSDT.BINANCE USD-M futures, 4H bars, pure OHLCV single feed. Measures how far price is extended from the 20-period SMA in rolling-std (sigma) units; when extension reaches +2 sigma with RSI >= 70 it SHORTS (fade the up move), and at -2 sigma with RSI <= 30 it goes LONG (fade the down move). Exits when the distance reverts to within 0.5 sigma of the mean, on a 2x ATR stop, or after an 18-bar time stop. The signal is the continuous distance z-score (varies every bar); the z-threshold AND RSI confirmation gate entry. Sized at 20% equity notional, capped so ATR-stop risk <= 1.5% of equity. leverage 1.0.

Implemented correctly and trades as specified: 459 entries (212 long / 247 short), 0 dropped — no code bug; healthy sample

Net losing: profit_factor 0.932, Sharpe -0.37 (CI -1.50 to 0.85), total_return -13.7%, expectancy -$23/trade, omega 0.92, Sortino -0.34, PSR 0.257

High win rate (62.5%) is deceptive: avg_loss $911 vs avg_win $509 (~1.8x), return_skew -1.67, tail_ratio 0.62 — classic pennies-in-front-of-a-steamroller mean-reversion with fat left tails

Net-negative in most years (2021 -8.2%, 2022 -10.7%, 2024 -3.7%, 2025 -1.5%), rolling Sharpe -2 to -4 for long stretches, max_drawdown 26.9% (CI to 47%), max_drawdown_duration ~1970 days

Premise refuted: fading 2-sigma extensions on BNB loses because extended moves continue rather than revert; RSI confirmation does not rescue it; commission 5.65% of gross compounds it

This strategy faded ≥2-sigma extensions from BNB's 20-SMA with RSI confirmation, betting on reversion to a burn-supported fair value. The implementation was clean (459 entries, none dropped) and the win rate looked attractive at 62.5%, but the payoff geometry was inverted — average losses ran ~1.8x average wins with a heavy left tail — producing profit factor 0.93, Sharpe -0.37, -13.7% total, and a 26.9% drawdown, losing in most years. The analyst abandoned it before optimization, concluding the premise is simply backwards: extended moves on a liquid major continue rather than revert, so fading them is a systematically-losing trade no z/RSI tuning can fix short of inverting it into a momentum trade. It was recorded alongside the other sign-wrong reversion failures on majors this session.

Counter-trend fading of statistical extensions does not generate a fee-surviving edge on crypto majors regardless of oscillator confirmation — extended moves are continuation signals, and a high win rate paired with avg loss ~1.8x avg win and a -1.67 skew is the classic picking-up-pennies profile, the same sign-wrong reversion failure as the BTC vol-spike, BTC VWAP-distance, and ETH/BTC pairs strategies.

The analyst abandoned it at backtest review as a refuted, sign-wrong premise rather than a bug — the code traded exactly as specified, but fading 2-sigma extensions on BNB loses because extended moves on a liquid major continue (momentum) rather than revert profitably, RSI confirmation does not rescue it, and 5.65%-of-gross commission drag compounds the negative edge.

A long-short mean-reversion strategy on BNBUSDT 4H perpetual futures (pure OHLCV) that faded statistically-extended moves — shorting when price was ≥2 sigma above its 20-SMA and RSI ≥70, longing when ≤2 sigma below and RSI ≤30 — using RSI as exhaustion confirmation and exiting on mean reversion (|z|≤0.5), an ATR stop, or a time stop.

Over a 459-trade sample (212 long, 247 short, none dropped) it was net-losing: profit factor 0.932, Sharpe -0.37 (CI -1.50 to 0.85), total return -13.7%, expectancy -$23/trade, Sortino -0.34, PSR 0.257, with a 26.9% max drawdown (CI to 47%); the 62.5% win rate was deceptive because average loss ($911) was ~1.8x average win ($509) with return skew -1.67 and tail_ratio 0.62, and it lost in most years.