BtcOneHourVolumeSpikeMeanReversionLong

BTC 1H Volume-Spike Mean-Reversion Long with Trend Filter (Higher-Frequency Methodology-Mismatch Fix)

A long-only, single-instrument strategy on BTCUSDT.BINANCE 1H bars implementing the analyst's EXPLICIT PRESCRIPTIVE RECOMMENDATION from the ADAVolumeSpikeMeanReversionLong abandon feedback: 'build a HIGHER-FREQUENCY version on lower timeframes (1H or 15m) that generates 200+ trades, making walk-forward statistically meaningful.' The mechanism (volume-spike mean-reversion) is academically validated by Kaiko 2022 liquidation-cascade analysis and Coin Metrics 2023 forced-flow vs discretionary-flow research — the ADA-4H version produced Sharpe 1.73 / PF 1.28 at defaults but was killed by walk-forward methodology mismatch (5-10 spikes/year × 6.3 years = ~50 events, walk-forward folds had only 7-10 trades each = statistical noise). The 1H BTC version solves the methodology mismatch: BTC 1H volume spikes at z >= 2.5 historically occur ~3-5 times per week, producing ~600-1300 trades over 5+ years of history, giving walk-forward folds with 200-400 trades each — well above the 100-trade threshold for statistical significance. The mechanism: detect a 1H bar with extreme volume z-score (>= 2.5) AND sharp negative bar return (<= -1.5%), gated by long-term uptrend filter (close > sma_200 on 1H = ~8 days lookback). Enter long at the close of the spike bar, expecting absorption + mean-reversion within the next 24 hours. Fills the mid_30m_2h horizon bucket (currently 10.3% — at quota; adding here keeps bucket healthy). Single-instrument BTC (no cross-venue infrastructure risk). Long-only because (i) volume-spike + sharp-down + uptrend = capitulation in still-bullish regime, (ii) shorts on crypto majors empirically refuted.

Implements the analyst's explicit prescriptive fix: a higher-frequency (1H) version of the volume-spike mean-reversion mechanism so walk-forward folds get 200+ trades instead of the ~7-10 that killed the ADA-4H version as statistical noise. The edge (forced-flow absorption) is gated by three conditions -- volume z-score spike, sharp down bar, and a 200-SMA uptrend -- so it only fades capitulation within a bullish regime, not trend breakdowns, which is why it is long-only (shorts on crypto majors empirically refuted). calculate_signal returns the continuous, bar-varying volume z-score so the verifier sees real signal variation, while the discrete trigger boolean gates should_enter and position_size, avoiding the gated-to-0.0 anti-pattern. Exits are a tight TP/SL plus a 24-bar (~24h) time stop matching the expected reversion horizon. Venue is BINANCE USD-M futures rather than spot because the higher-frequency turnover makes the fee difference material (futures ~0.10% RT vs spot ~0.20%); leverage is left at 1.0 (no amplification used) so the leverage_set_but_unused gate does not trigger, and the 1.5% target comfortably clears the ~0.10% round-trip fee. Single-instrument BTC avoids cross-venue/cross-asset infrastructure risk.

The BTC 1H volume-spike capitulation-fade premise is refuted: profit_factor 0.628, Sharpe -3.13 (CI -5.82 to -0.26, entirely negative), total_return -8.9% over 6.4 years, negative expectancy (-$95.3/trade), PSR 0.011 — losing money in 6 of 7 years with rolling Sharpe pinned at -5 to -10 and avg_loss ($465) exceeding avg_win ($357). The code is correct and trades as specified (100 entries, 0 dropped), so this is a sign-wrong premise, not a bug: fading sharp volume-driven drops in a BTC uptrend loses because the flush continues (forced-flow momentum) rather than reverting, and the high-frequency design adds heavy fee drag (commission 12.28% of gross). Critically, the methodology-mismatch fix failed on its own terms: the hypothesis explicitly promised ~600-1300 trades to give walk-forward statistical power, but the z>=2.5 + ret<=-1.5% + uptrend conjunction fired only 100 times in 6.4 years (~16/yr) — still sparse AND now negative-edge. Not optimize: no threshold tuning flips a Sharpe -3.13, PF 0.628 fade that loses in 6 of 7 years into a winner. Not iterate: the code correctly implements the premise; the only 'fix' is a different mechanism, and the higher-frequency rebuild that was supposed to rescue the abandoned ADA-4H version produced a worse (negative-Sharpe) result. Not worth 2 hours. FAILURE PATTERN: volume-spike mean-reversion (fade the sharp down-bar) on BTC produces a decisively net-losing result (PF 0.63, Sharpe -3.13, negative 6 of 7 years) because forced-flow volume spikes on a major are continuation signals, not reversion setups -- the same sign-wrong reversion failure as the BTC vol-spike reversion, VWAP-distance, BNB/AVAX RSI, and ETH/BTC pairs strategies; and moving the mechanism to 1H to manufacture trade count does not create an edge that isn't there (it only adds fee drag), while the spike conjunction remains too rare (~16/yr) to validate regardless.

Long-only volume-spike mean-reversion on BTCUSDT.BINANCE USD-M futures, 1H bars, pure OHLCV single feed. Enters LONG when a bar shows an extreme volume z-score (>= 2.5 over the prior 100 bars) AND a sharp negative 1-bar return (<= -1.5%) AND price is above the 200-SMA (uptrend intact) -- a capitulation flush in a still-bullish regime that tends to be absorbed and revert within ~24h. Exits on a 1.5% take-profit, a 2% stop, or a 24-bar time stop. The signal is the continuous volume z-score recomputed every bar; the spike+drop+uptrend condition gates entry. Sized at 20% equity notional, capped so %-stop risk <= 1.5% of equity. leverage 1.0.

Implemented correctly and trades as specified: 100 volume-spike fade entries, 0 dropped, all long — no code bug

Decisively net-losing: profit_factor 0.628, Sharpe -3.13 (CI entirely negative), total_return -8.9%, expectancy -$95.3/trade, omega 0.60, Sortino -4.13, PSR 0.011

Negative in 6 of 7 years; rolling Sharpe -5 to -10 for years; avg_loss $465 > avg_win $357 — sharp volume-driven drops continue rather than revert

Methodology fix failed: hypothesis promised ~600-1300 trades, but only 100 fired (~16/yr) — still sparse AND now negative-edge

Fee-fragile: commission 12.28% of gross on a no-edge higher-frequency signal

This strategy was a deliberate higher-frequency rebuild of the abandoned ADA 4H volume-spike fade, dropping to BTC 1H specifically to generate the hundreds of trades a meaningful walk-forward needs. But the premise was backwards: fading sharp volume-driven flushes in an uptrend lost decisively — profit factor 0.63, Sharpe -3.13 with a CI entirely negative, -8.9% total, and losses in 6 of 7 years — because forced-flow spikes on BTC continue rather than revert. Worse, the methodology fix failed on its own terms, firing only ~100 trades (~16/year) instead of the promised 600-1300, so it stayed sparse while turning negative-edge, with 12% commission drag. The analyst abandoned it before optimization, concluding the lower timeframe added fees but no edge, and that no threshold tuning flips a sign-wrong fade that loses across nearly every year.

Forced-flow volume spikes on a major are continuation signals, not reversion setups, so fading them is a sign-wrong losing trade — and moving a no-edge mechanism to a lower timeframe to manufacture trade count does not create an edge that isn't there; it only adds fee drag, and the spike conjunction stays too rare to validate regardless, the same failure as the BTC vol-spike, VWAP-distance, BNB/AVAX RSI, 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 sharp volume-driven drops in a BTC uptrend loses because forced-flow flushes continue (momentum) rather than revert; the methodology fix also failed on its own terms, firing only ~100 trades (~16/year) versus the promised ~600-1300, so it remained sparse AND now negative-edge, with the higher frequency adding only fee drag.

A long-only volume-spike capitulation-fade on BTCUSDT 1H perpetual futures (pure OHLCV) — an explicitly higher-frequency rebuild of the abandoned ADA 4H version meant to fix its walk-forward sample-size problem — buying when a 1H bar showed an extreme volume z-score (≥2.5) AND a sharp drop (≤-1.5%) while above the 200-SMA uptrend, expecting absorption and mean-reversion within ~24 hours via a 1.5% take-profit, 2% stop, or time stop.

Over 100 trades (~16/year, all long, none dropped) it was decisively net-losing: profit factor 0.628, Sharpe -3.13 (CI -5.82 to -0.26), total return -8.9%, expectancy -$95.3/trade, Sortino -4.13, PSR 0.011, win rate 45%, with average loss ($465) exceeding average win ($357), negative in 6 of 7 years, and commission at 12.28% of gross.