BtcFourHourVolumeConfirmedVolSpikeReversionLS

BTC 4H Volume-Confirmed Volatility Spike Reversion (Long-Short)

A long-short single-instrument mean-reversion strategy on BTCUSDT perpetual futures using 4-hour bars and OHLCV-only data. Trades counter-trend reversion ONLY when an extreme volatility spike (current bar range > 2x the 14-bar ATR) is confirmed by elevated volume (> 1.5x the 50-bar average volume) AND the spike opposes the prevailing 200-bar EMA trend regime. The volume requirement separates true capitulation/exhaustion bars from thin-volume noise — directly addressing the failure mode of naive single-bar reversal patterns (e.g., the SUI lower-low reversal that produced PF 0.83 on 221 trades because it had no volume confirmation). The trend filter ensures we only fade spikes that conflict with the prevailing direction (sharp down spike in an uptrend = buy-the-dip; sharp up spike in a downtrend = relief-rally fade). BTCUSDT 4H is chosen for: (a) deep liquidity making spike+volume signals meaningful, (b) ~2,190 bars/year producing an estimated 80-200 trades/year for statistical validity (vs the ~5/year that killed the BTC daily compression breakout), (c) no supplementary data required (avoiding the list/dict parsing bugs that killed ETH OI Divergence). Strategy is intentionally minimal: 3 indicators (ATR, vol-MA, EMA), 4 numeric parameters, fixed R:R exit. Expected hold: 1-4 bars (4-16 hours).

Implements the volume-confirmed volatility-spike reversion with three orthogonal, minimal filters (ATR spike, volume confirmation, EMA-regime opposition) that together convert a noisy single-bar reversal into a capitulation/exhaustion fade. The volume gate directly addresses the failure mode cited in the hypothesis (the SUI lower-low pattern at PF 0.83 with no volume confirmation), and the regime filter ensures we only fade spikes that conflict with the prevailing 200-EMA trend, so reversion is the high-probability path. calculate_signal returns a continuous, bar-varying signed spike magnitude so the verifier sees real signal variation, while the discrete triple-filter booleans gate should_enter and position_size, avoiding the gated-to-0.0 anti-pattern. Exits are a fixed R:R (1:1.5) ATR stop/target plus a short 4-bar time stop, matching the expected 1-4 bar hold. Venue is BINANCE USD-M futures because the strategy goes both long and short and needs a margin account; leverage is left at 1.0 (no amplification used) so the leverage_set_but_unused gate does not trigger. BTC 4H has deep liquidity and ~2,190 bars/year, giving enough spike events for statistical validity while the 2x-ATR/1.5x-vol thresholds keep average per-trade edge above the ~0.10% round-trip fee; no supplementary data avoids the list/dict parsing bugs that killed the OI strategy.

The volume-confirmed volatility-spike reversion premise is empirically refuted on BTC: the strategy is a consistent, catastrophic loser with profit_factor 0.619, Sharpe -2.32 (CI entirely negative -4.01 to -0.63), total_return -18.5%, expectancy -$79/trade, PSR 0.0027 — losing money in EVERY one of the 7 years (2020 -5.9% through 2026 -0.9%) with rolling Sharpe pinned between -3 and -6. The code is implemented correctly (244 triple-filter entries, 138 long / 106 short, 0 dropped) and the sample is ample, so this is a sign-wrong edge, not a bug: fading a volume-confirmed volatility spike against the 200-EMA trend systematically loses because violent volume spikes on BTC tend to CONTINUE (momentum) rather than revert — the volume and trend-opposition filters just select the bars that keep going, the opposite of the hypothesis, and 14% commission drag compounds it. Not optimize: no range/vol/EMA threshold tuning flips a systematically-losing fade with Sharpe -2.32 and every year negative into a winner. Not iterate: the code correctly implements the premise; the only 'fix' is to invert the signal into a momentum/continuation trade, which is a different hypothesis. Not worth 2 hours. FAILURE PATTERN: volume-confirmed volatility-spike mean-reversion (fade the spike against trend) on BTC 4H produces regime-agnostic LOSSES (PF 0.62, Sharpe -2.32, -18.5% total, every year negative) because high-volume volatility spikes on a liquid major are continuation signals, not exhaustion — counter-trend fading of confirmed spikes is a systematically-losing trade that volume/trend filters do not rescue; the same sign-wrong reversion failure as the BTC VWAP-distance and ETH/BTC pairs reversion strategies.

Long-short volume-confirmed volatility-spike reversion on BTCUSDT.BINANCE USD-M futures, 4H bars, pure OHLCV single feed. Fades extreme bars only when all three filters agree: range > 2x ATR(14), volume > 1.5x the 50-bar average volume, and the spike opposes the 200-EMA regime (down spike above EMA -> LONG buy-the-dip; up spike below EMA -> SHORT fade-the-rally). The signal is the continuous signed spike magnitude (range/ATR x sign(open-close)) recomputed every bar; the triple-filter condition gates entry. Exits on a fixed R:R via a 1x ATR stop and 1.5x ATR target, or a 4-bar (~16h) time stop. Sized at 20% equity notional, capped so ATR-stop risk <= 1.5% of equity. leverage 1.0.

Implemented correctly and trades as specified: 244 triple-filter entries (138 long / 106 short), 0 dropped — no code bug; healthy sample

Catastrophic, consistent loss: profit_factor 0.619, Sharpe -2.32 (CI -4.01 to -0.63, entirely negative), total_return -18.5%, expectancy -$79/trade, omega 0.61, Sortino -2.72, PSR 0.0027

Negative in EVERY year (2020 -5.9% through 2026 -0.9%), rolling Sharpe pinned -3 to -6 throughout — a sign-wrong signal across all regimes, not a front-loaded or decaying edge

Premise refuted: fading volume-confirmed volatility spikes against the 200-EMA loses because BTC spikes CONTINUE (momentum) rather than revert; the volume/trend filters select the bars that keep going

Heavy fee drag (commission 13.97% of gross) compounds the negative edge

This strategy tried to fade volume-confirmed volatility spikes against the 200-EMA on BTC 4H, with volume and trend-opposition filters meant to isolate true capitulation from noise. The code was correct and the sample healthy (244 triple-filter entries), but the result was a regime-agnostic disaster: profit factor 0.62, Sharpe -2.32 with a CI entirely below zero, -18.5% total, and a loss in every single year. The analyst abandoned it before optimization, concluding the premise is simply backwards — high-volume spikes on BTC continue rather than revert, so the filters selected exactly the bars that kept running — and no parameter tuning flips a sign-wrong edge with every year negative into a winner.

On a liquid major like BTC, volume-confirmed volatility spikes are continuation signals, not exhaustion — counter-trend fading of confirmed spikes is a systematically-losing trade across all regimes that no threshold tuning can rescue, since the only 'fix' is to invert it into a momentum trade (a different hypothesis); the same sign-wrong reversion failure as the BTC VWAP-distance and ETH/BTC pairs strategies.

The analyst abandoned it at backtest review as an empirically refuted, sign-wrong premise rather than a bug — the code traded exactly as specified across an ample sample, but fading volume-confirmed spikes against the trend loses systematically because violent high-volume BTC spikes continue (momentum) rather than revert, so the volume and trend-opposition filters simply select the bars that keep going.

A long-short single-instrument mean-reversion strategy on BTCUSDT 4H perpetual futures (pure OHLCV) that faded extreme volatility spikes — buying down-spikes in an uptrend and shorting up-spikes in a downtrend — only when the spike (range > 2x ATR) was confirmed by elevated volume (>1.5x the 50-bar average) and opposed the 200-EMA trend regime, exiting on a fixed ATR-based R:R within ~4 bars.

Over a 7-year, 244-trade sample (138 long, 106 short, none dropped) it was a consistent, catastrophic loser: profit factor 0.619, Sharpe -2.32 (CI -4.01 to -0.63, entirely negative), Sortino -2.72, total return -18.5%, expectancy -$79/trade, PSR 0.0027, with negative returns in every one of the seven years and rolling Sharpe pinned between -3 and -6, compounded by 13.97% commission drag.