BtcBnbPullbackToSmaContinuation4hTrendLong1H

BTC and BNB 1H Pullback-to-SMA Continuation with 4H Trend Confirmation Long Portfolio

A long-only, multi-timeframe, multi-instrument continuation strategy on BTCUSDT.BINANCE and BNBUSDT.BINANCE perpetual futures. Combines proven elements from the session's only confirmed-winning strategy (ETH+SOL 1H Breakout with 4H Trend Confirmation, Sharpe 1.73) with a DIFFERENT entry style (pullback rather than breakout) and a DIFFERENT instrument set (BTC + BNB rather than ETH + SOL). The thesis: when the 4-hour trend is up (price > 50-period 4H SMA), brief intraday pullbacks to the 20-period 1H SMA represent low-risk re-entry points in a confirmed uptrend. The 4H filter restricts entries to genuine trending regimes only (avoiding the 'breakout + trend filter' tension that killed prior single-timeframe Donchian strategies); the 1H pullback entry captures the well-documented intraday-rebound effect at moving-average support. Critically: this strategy does NOT rely on the 4H pure-breakout class which has been confirmed-failed multiple times this session — it's a fundamentally different mechanism (pullback continuation, not breakout). The pullback-into-trend mechanism failed at the DAILY single-instrument level (AVAX 1D Trend-Pullback), but the failure was statistical (insufficient sample, 1.11x avg-win/avg-loss ratio combined with 35% win rate). At 1H bars × 2 instruments × multi-year history, the sample size is ~17,520 1H bars per instrument × 2 = 35,040 evaluation points, generating ample trade events. The multi-timeframe gate ensures entries fire only in confirmed trend regimes, avoiding the whipsaw periods that destroy single-timeframe strategies. Uses ONLY BINANCE USD-M perp 1H + 4H bars (proven clean for BTC + BNB; multi-timeframe verified via the recent success). Long-only, fee-aware, 3 parameters maximum. Fills under-represented buckets: multi-instrument scope (8.4% → grow), 1H horizon (currently underrepresented), and adds proven-mechanism extension. Position sizing: 15% of equity per instrument when entry conditions met (max 30% gross exposure if both fire). Risk per trade: 1.5% of equity capped by ATR-based trailing stop.

Reuses the proven element of the session's only confirmed winner (1H entry gated by a 4H SMA trend filter) but swaps the entry STYLE (pullback-to-support rather than breakout) and the INSTRUMENT set (BTC+BNB rather than ETH+SOL), giving a genuinely different, de-correlated mechanism that sidesteps the repeatedly-failed 4H pure-breakout class. The 4H gate confines entries to confirmed trending regimes, fixing the whipsaw/sample problems that sank the single-instrument daily pullback attempt; at 1H × 2 instruments × multi-year history the trade count is far above the walk-forward stability floor. Wiring follows the validated BtcBnbPullback pattern: the base's extra-bar routing is mirrored explicitly, each instrument keeps its own 1H/4H buffers and is evaluated on its own 1H bar so there is no cross-symbol index misalignment, and a continuous tanh regime-fit signal keeps Layer-2 synthetic scenarios non-frozen while real entries are decided per symbol. Risk is controlled with three exits (1H trend break, ATR trailing stop sized to the hypothesis's ATR-based risk cap, and a hard position stop ≈1.5% equity at 15% sizing). Three core tunables (sma_1h_period, sma_4h_period, pullback_proximity_pct) keep the surface small; ATR settings are execution-only. Leverage stays at 1.0 — USD-M futures is chosen purely for the ~0.10% round-trip fee versus ~0.20% spot, so sizing references no margin leverage and avoids the inert-leverage gate while remaining functionally long-only.

No edge and structurally fee-destroyed — not optimizable. The strategy returns Sharpe -0.006, profit_factor 1.05, alpha -0.0001, CAGR 1.28%, omega 0.999, with sharpe_ci centered on zero (-0.90 to 0.76) — essentially no risk-adjusted edge. The decisive killer is trading costs: impact_cost_pct 51.7% (impact eats HALF of gross PnL) plus commission 20.5% of gross means ~72% of gross is consumed, and capacity_usd is only $373k (toy scale). This is the classic high-trade-count + thin-edge cost-domination pattern: 2015 trades at ~8.6h holds with a 23.4% win rate; the 3.4:1 avg_win/avg_loss let-winners-run payoff nets only to ~breakeven at that hit rate and fees then erase it. NOT optimize: there is no edge to tune toward (Sharpe 0, alpha 0), and the gross per-trade edge is too thin to clear 1H fees at any parameter setting. NOT iterate: reducing trade frequency cannot manufacture a continuation edge that does not exist, and the pullback-to-SMA continuation mechanism has now failed across timeframes — the hypothesis's own cited AVAX daily version failed (35% win, 1.11 payoff) and this BTC+BNB 1H version is worse (23% win), so the 4H-trend filter did not rescue it. FAILURE PATTERN: pullback-to-SMA continuation on crypto majors does not predict intraday continuation — at 1H it generates thousands of low-win-rate (23%) churning trades whose razor-thin gross edge (PF ~1.05) is annihilated by fees+impact (~72% of gross, capacity $373k), yielding Sharpe ~0; the mechanism fails at both daily (AVAX) and 1H (BTC+BNB) and across the let-winners-run payoff shape, indicating an absent edge rather than a tuning problem.

Long-only, two-instrument (BTCUSDT + BNBUSDT USD-M perp), two-timeframe pullback-continuation strategy. Per instrument: when the 4H trend is up (4H close > SMA(50)), a 1H pullback that holds just above the 1H SMA(20) — close > SMA1h and within pullback_proximity_pct above it — triggers a LONG re-entry into the confirmed uptrend. Exits on a 1H close below SMA(20), an ATR trailing stop (trail_atr_mult·ATR below the in-trade high), or a hard per-position stop. Sizes 15% of equity notional per instrument (≤30% gross if both fire). BTC trades via the primary path; BNB and both 4H series are handled in on_extra_bar with independent per-symbol state. Pure OHLCV, leverage 1.0.

Clean execution, long-only as designed (2015/2015 entries, no artifacts, metrics_reliable=true), low drawdown (6.0%); let-winners-run payoff (avg_win/avg_loss 3.4:1) is structurally sound

Genuinely low market exposure (beta 0.047, exposure 31%)

No edge: Sharpe -0.006, profit_factor 1.05, alpha -0.0001, CAGR 1.28%, omega 0.999, sharpe_ci centered on zero (-0.90 to 0.76)

Fee-destroyed: impact_cost_pct 51.7% + commission 20.5% of gross (~72% of gross consumed); capacity only $373k (toy scale)

Over-trading churn: 2015 trades, 23.4% win rate, ~8.6h holds — the 3.4:1 payoff nets to ~breakeven at 23% win and is then eaten by costs

Mechanism failed across timeframes: AVAX daily pullback-to-SMA (cited) AND this BTC+BNB 1H — pullback continuation on majors does not predict continuation

viable with payoff

deployable

BtcBnbPullbackToSmaContinuation4hTrendLong1H tried to extend the session's one confirmed winner — the ETH+SOL breakout with 4H trend confirmation — by keeping the 4H filter but switching to a pullback entry and a BTC+BNB instrument set, betting that ~35k 1H evaluation points would give the daily pullback mechanism (which had failed on AVAX for sample reasons) enough events to work. It didn't: the strategy returned essentially zero risk-adjusted edge (Sharpe -0.006, PF 1.05, CAGR 1.28%) and was then annihilated by costs, with impact and commission consuming ~72% of gross across 2,015 churning trades at a 23% win rate and only $373k capacity. The reviewer abandoned it pre-optimization, concluding there is no edge to tune toward and that reducing frequency cannot create a continuation edge that does not exist. The decisive finding was that pullback-to-SMA continuation now fails at both daily (AVAX) and 1H (BTC+BNB) — the 4H filter did not rescue it, marking the mechanism itself, not the parameters, as the problem.

Pullback-to-SMA continuation on crypto majors does not predict intraday continuation — at 1H it generates thousands of low-win-rate (23%) churning trades whose razor-thin gross edge (PF ~1.05) is annihilated by fees and impact (~72% of gross, capacity $373k); a sound let-winners-run payoff and a 4H-trend filter cannot manufacture an edge that is absent at both daily and 1H.

The analyst abandoned it at the backtest-review gate, before optimization, because there is no edge to tune toward (Sharpe ~0, alpha ~0) and the gross per-trade edge is too thin to clear 1H fees at any setting — a high-trade-count cost-domination pattern — and the pullback-to-SMA continuation mechanism had now failed across timeframes (the cited AVAX daily version at 35% win/1.11 payoff, this 1H version worse at 23% win), so the 4H filter did not rescue it.

A long-only, multi-instrument, multi-timeframe pullback-continuation strategy on BTCUSDT and BNBUSDT.BINANCE USD-M perps: when the 4H trend is up (price > 50-period 4H SMA), buy 1H pullbacks to the 20-period 1H SMA as low-risk re-entries, exiting on a 1H SMA break or ATR trailing stop — combining the winning ETH+SOL breakout strategy's 4H-trend filter with a different (pullback) entry and a different instrument set, at 15%/leg (max 30% gross).

It had no risk-adjusted edge and was destroyed by costs: Sharpe -0.006, profit factor 1.05, alpha -0.0001, CAGR 1.28%, 23.4% win rate over 2,015 trades (~8.6h holds) — with a structurally sound 3.4:1 payoff but impact_cost_pct 51.7% plus commission 20.5% of gross (~72% of gross consumed) and capacity only $373k; execution was clean (metrics_reliable=true, beta 0.047, 6% max drawdown).