BnbFourHourVolumeSurgeMomentumContinuationLongDailyRegime

BNB 4H Volume-Surge Momentum-Continuation Long with Daily Bull Regime Filter

A long-only single-instrument VOLUME-CONFIRMED MOMENTUM-CONTINUATION strategy on BNBUSDT perpetual futures using 4-hour bars and OHLCV-only data, with a daily-bar bull regime filter. This proposal fills BNB's mechanism gap (only BnbFourHourVolumeBreakoutLong currently — 1 strategy vs 3-4 for every other major asset) while consciously AVOIDING the documented cross-asset failure patterns: (1) NOT an RSI(2) mean-reversion replication (analyst proved this is BTC-specific and fails on altcoins with -22.9% return on ETH); (2) NOT a daily trend strategy on mid-cap altcoin (altcoin_daily_minimum_param_trend pattern killed AVAX/SOL Donchian); (3) NOT options (framework unsupported). The mechanism is mechanistically DISTINCT from BnbFourHourVolumeBreakoutLong: the existing Vol Breakout requires Donchian high break AS WELL AS volume confirmation, firing only on rare regime-change events. This proposes a pure volume-surge momentum-continuation that fires more frequently DURING established trends — capturing intra-trend momentum-acceleration events rather than only breakouts. The trigger: 4H bar with volume > 2.0× 14-bar avg AND bullish close (close > open) AND daily close > 200-SMA AND prior 4H bar did NOT meet the volume+bullish condition (edge detection). This signals a momentum-acceleration event within an established uptrend — the kind of capital-flow event that produces continuation over 4-12 hours. Economic mechanism for BNB specifically: (a) BNB has quarterly BNB-burn events that produce systematic buying pressure with elevated volume — these are catalyst-driven momentum events; (b) Binance ecosystem traders systematically buy BNB during exchange utility-cycle phases (token utility tier upgrades, BNB Chain integration milestones); (c) BNB has medium-leverage retail open interest that responds to volume-confirmed breakouts with cascade-buying. Each of these flows produces volume-confirmed bullish bars that mark momentum-acceleration, with 4-12 hour continuation typical. This is the SAME 'altcoin momentum-continuation' economic argument that justifies the ETH/SOL overbought RSI strategies, applied via a DIFFERENT signal mechanism (volume + price, not RSI velocity). Cross-asset transfer is justified per the analyst's filter: BNB has the leverage + staking dynamics that produce continuation on momentum signals (same as ETH/SOL); the volume-burst trigger captures these flows directly. Only 4 explicit parameters (volume multiplier, volume lookback, regime SMA period, stop pct).

Smallest fix for the Layer-3 'No trades produced' (should_enter returned a side 0/13938 times). The volume-surge edge (2x-volume bullish bars) is frequent, so a structural zero implicated the regime gate. Diagnosis: the 200-day regime relied on a 1-DAY same-instrument extra feed because the base caps self._bars at 500 (~83 days of 4H); in the sandbox that feed never accumulated >=200 daily closes, so _regime_ok() was always False and no entry fired (Layer 2 passed because it only feeds primary bars and merely checks the continuous signal varies). Fix: derive the regime from a self-maintained 4H close history (regime_sma x bars_per_day = 1200 closes) so it activates ~200 days into the sample with no external-feed dependency; BNB spends long stretches above its 200-day SMA (2020-21, 2023-25) during which volume-surge edges occur, so the corrected backtest will produce the expected dozens+ of trades. Only the regime computation and its config (extras removed) changed; signal, edge detection, exits, and capital-relative sizing (50% equity notional, leverage 1.0) are intact, so earlier layers stay green.

Failed deflated Sharpe on attempt 1: DSR=0.0098 (vs 0.95 bar), is_significant=FALSE, with the optimized Sharpe 2.597 FAR BELOW the 225-trial expected-max luck bar of 5.8312 (PBO 0.5701 >0.5, sharpe_ci_low -0.3274 straddling 0) — after multiple-testing correction the selected config is statistically indistinguishable from best-of-N noise. The high probabilistic_sharpe (0.9696) is the classic PSR-vs-DSR trap: PSR ignores the 225-trial count that DSR corrects for. The walk-forward is not flagged overfit only because its OOS windows are [0.0, 5.97, 0.0] — TWO zero-trade windows with the entire avg OOS (1.99) resting on a single window, i.e. a starved signal, not a robust one. The holdout 'pass' (ratio 5.75, holdout_sharpe 11.45) is a tiny-window artifact: the optimized config trades only 84 times at 2.15% exposure, so an 11+ Sharpe on a handful of holdout trades is not validation. Sensitivity FAILED with 2 cliffs on the CORE signal parameters (vol_mult, vol_lookback) — Sharpe collapses to ~0/negative at vol_mult 1.8/1.6, so the defining trigger is fragile. The base is weak and impact-fragile: Sharpe 0.395, profit_factor 1.08, negative alpha -0.049, IR -0.83, max_drawdown 44.2%, and impact_cost_pct 27.7% at only $1.3M capacity. Decisively, optimization 'improved' Sharpe only by thinning the strategy to near-non-trading (218->84 trades, 9.7%->2.15% exposure) — a low-activity curve-fit, not a stronger edge. Not iterate (attempt 1 of 2): two zero-trade OOS windows, cliffs on the core trigger params, DSR ~0.01, and an optimized config thinned to non-trading mean there is no robust region to tune toward — a second best-of-225 sweep re-finds the same fragile, starved config. Not revise_hypothesis: BNB volume-surge momentum-continuation is not a proven mechanism stranded on a dead target; it is an under-powered, impact-fragile single-asset trigger that fails deflation. FAILURE PATTERN: a differentiated, low-parameter single-asset 4H volume-surge momentum-continuation long on BNB produces a weak base (Sharpe 0.40, PF 1.08, negative alpha, 44% DD, 27.7% impact) whose best-of-225 optimization lifts Sharpe to 2.6 only by thinning to 84 trades / 2.15% exposure — far below the 5.83 luck bar (DSR 0.0098) with two zero-trade OOS windows and cliffs on the core signal params. A non-overfit-flagged walk-forward and a spectacular tiny-window holdout Sharpe (11.45) do not rescue an edge that fails deflated Sharpe; the flattering forward numbers are artifacts of a starved, low-activity signal.

Long-only BNBUSDT.BINANCE USD-M perp, 4H bars, OHLCV-only. Fires a momentum-acceleration entry when the current 4H bar has volume > vol_mult x its trailing vol_lookback-bar average AND a bullish close, the prior 4H bar did NOT meet that condition (edge detection at the onset of the surge), AND price is above the 200-DAY regime SMA. Exits on a 6% hard stop, a ~24h continuation-window time stop, or a trailing stop below the min low of the prior 5 bars. Iteration 2 makes the long-term bull-regime filter SELF-CONTAINED: instead of a fragile same-instrument daily extra feed (which delivered too few closes in the sandbox, so the regime never engaged and zero trades were produced), the strategy keeps its own uncapped 4H close history and computes the 200-day SMA from regime_sma x bars_per_day 4H closes. The base ring-buffers self._bars to 500, so this private history is required for a 200-day filter on 4H. Volume-surge edge logic, exits, and sizing are unchanged.

Verification failed (Layer 3 — sandbox backtest): No trades produced Bar type used: BNBUSDT.BINANCE-4-HOUR-LAST-EXTERNAL, Bars processed: 13954 Diagnostics: should_enter() returned a side 0 times over 13938 evaluated bars -> your ENTRY CONDITION never triggered. Loosen the entry logic / thresholds. Ensure your strategy produces trades with the given data and parameters.

Functional and correctly implements the hypothesis: 218 long-only volume-surge-onset entries with bull-regime filter, entries fire cleanly with zero drops

Healthy trade sample (218) over ~6 years supports statistically meaningful parameter sensitivity analysis

Iteration-2 self-contained regime-SMA fix resolved the prior zero-trade failure — regime filter now engages

Sane exposure (9.69%) and no liquidation; positive total return (23.3%)

Clear 4-parameter tuning surface with a plausible improvement path (higher vol_mult -> fewer, higher-conviction trades -> less impact drag)

Thin edge: profit_factor 1.081, Sharpe 0.395 — barely above breakeven

High trading-cost drag: impact_cost_pct 27.7% + commission 5.51% consume ~1/3 of gross PnL; capacity only ~$1.3M

Negative alpha (-0.049) and information_ratio (-0.83) vs a meaningful buy-hold benchmark — underperforms holding BNB

Return decay: 2023 -18.9%, 2025 -7.4%, 2026 -2.8%; profit is front-loaded in the 2020-2021 bull, rolling Sharpe negative through much of 2023/2025

Elevated max_drawdown (44.2%) for a long-only single-name strategy

Coherent, differentiated volume-surge momentum-continuation thesis with edge-detection to avoid firing every elevated bar

Only 4 explicit tunables; self-contained regime filter (iteration-2 fix produced trades)

Optimized config is capital-light (capacity $23M, low impact 6.6%) if the edge were real

Failed deflated Sharpe: DSR=0.0098 (vs 0.95), is_significant=false, optimized Sharpe 2.597 FAR below the 225-trial expected-max luck bar of 5.8312

PBO=0.5701 (>0.5) and sharpe_ci_low -0.3274 (CI straddles 0)

Walk-forward OOS windows [0.0, 5.97, 0.0]: two zero-trade windows, the avg OOS of 1.99 resting entirely on a single window — a starved, not robust, signal

Holdout 'pass' (ratio 5.75, holdout_sharpe 11.45) is a tiny-window artifact — optimized config trades only 84 times at 2.15% exposure

Sensitivity FAILED with 2 cliffs on the core signal params vol_mult and vol_lookback (Sharpe collapses to ~0/negative at vol_mult 1.8/1.6)

Base is weak: Sharpe 0.395, profit_factor 1.08, negative alpha -0.049, IR -0.83, max_drawdown 44.2%, impact_cost_pct 27.7% at only $1.3M capacity

Optimizer 'improved' Sharpe only by thinning to near-non-trading (218->84 trades, 9.7%->2.15% exposure) — a low-activity curve-fit artifact

all > 0

This strategy filled a mechanism gap for BNB with a deliberately differentiated, four-parameter volume-surge momentum-continuation long, guarded by a 200-day bull-regime filter and edge-detection to fire only at the onset of a surge; an iteration-2 fix made the regime filter self-contained after an earlier zero-trade failure. The base backtest was weak but functional (Sharpe 0.395, profit factor 1.08, 44% drawdown, negative alpha, 27.7% impact cost), enough for the reviewer to approve optimization. Optimization produced a spectacular-looking Sharpe of 2.6 and an 11+ holdout Sharpe, but only by thinning to 84 trades at 2.15% exposure, and it collapsed under scrutiny: deflated Sharpe 0.0098, PBO 0.57, cliffs on both core signal params, and two of three zero-trade out-of-sample windows. The analyst abandoned it on attempt 1 as an under-powered, impact-fragile single-asset trigger whose flattering forward numbers were artifacts of a starved, low-activity signal that fails deflation.

A big jump in optimized Sharpe is meaningless when it comes from thinning the strategy to near-non-trading (218→84 trades, 2.15% exposure) — future iterations should treat collapsing trade count/exposure and zero-trade walk-forward windows as evidence of a starved curve-fit, and screen against the deflated Sharpe luck bar rather than the flattering probabilistic Sharpe or a tiny-window holdout.

The analyst abandoned it after optimization on attempt 1 of 2: it failed the deflated Sharpe test (DSR 0.0098 vs 0.95, not significant), with the optimized Sharpe 2.597 far below the 225-trial expected-max luck bar of 5.83, PBO 0.5701, and a Sharpe CI straddling zero (low -0.327). Sensitivity failed with 2 cliffs on the core signal params (vol_mult, vol_lookback), the walk-forward OOS windows were [0.0, 5.97, 0.0] with two zero-trade windows, and the flattering holdout (ratio 5.75, Sharpe 11.45) was judged a tiny-window artifact of a near-non-trading config.

A long-only single-instrument strategy on BNBUSDT 4H futures that entered at the onset of a volume-surge (volume > 2× the 14-bar average on a bullish bar, using edge-detection so it fired only at the start of the surge) while a 200-day SMA bull-regime filter confirmed an established uptrend, betting BNB burn events and Binance ecosystem flows produce 4-12 hour momentum-continuation.

The base backtest was thin: +23.3% total return, Sharpe 0.395, profit factor 1.081, 47.2% win rate over 218 trades, but with a 44.2% max drawdown, negative alpha (-0.049), and heavy cost drag (27.7% impact plus 5.51% commission) at only ~$1.3M capacity. Optimization lifted the headline Sharpe to 2.597 (+15.9% return, 5.5% drawdown) but only by thinning the strategy to 84 trades at 2.15% exposure.