ADAReturnZScoreContinuationLong

ADA 4H Bar-Return Z-Score Continuation Long (Momentum Acceleration Entry)

A long-only, single-instrument, single-timeframe strategy on ADAUSDT.BINANCE 4H bars (the most-reliably-fresh bar type per the persistent audit_stale pattern). Implements a NEW MECHANISM not yet tried in the portfolio: BAR-RETURN Z-SCORE CONTINUATION. The mechanism: compute the log return of each closed bar and standardize it as a z-score against the trailing 50-bar return distribution. When the current bar's return is in the top 15% of recent returns (z-score >= 1.0) AND the bar closed bullishly (close > open) AND price is in confirmed uptrend (close > sma_100), enter long anticipating continuation of momentum. This captures momentum ACCELERATION rather than mean-reversion — an unusually large positive bar in an uptrend signals fresh buying flow that statistically tends to extend. Critically distinct from every existing pipeline mechanism on ADA 4H: (a) NOT a Donchian breakout (uses return z-score, not channel breach); (b) NOT volume-spike MR (uses RETURNS, not volume; goes WITH momentum, not against it); (c) NOT bar-shape geometry (uses return magnitude, not high/low/wick positions); (d) NOT channel width expansion (uses single-bar return distribution, not multi-bar range); (e) NOT sequence-based (single-bar z-score statistic, no consecutive-bar requirement). The mechanism is structurally adjacent to TSMOM (Time-Series Momentum, Moskowitz-Ooi-Pedersen 2012) but applied at higher frequency with a z-score filter to fire only on UNUSUAL momentum events, not just any positive return. Expected trade frequency: ~40-60 trades/year on ADA 4H — well above walk-forward minimum without rare-event risk.

Implements the bar-return z-score continuation mechanism from the hypothesis: calculate_signal returns the raw, continuous trailing-window return z-score every bar (sign = direction, magnitude = how unusual the move is), while the binary regime filters (uptrend via close>SMA100, bullish close) are stored as state and applied in should_enter so the signal stays continuous and is not gated to a constant. This is structurally distinct from existing ADA 4H mechanisms: it uses the single-bar RETURN distribution (not a Donchian channel breach, not volume, not candle geometry, not multi-bar sequences) and goes WITH momentum (continuation) rather than fading it. Venue override: the hypothesis title names .BINANCE futures, but the strategy is long-only, uses no leverage, and its edge is not funding-based, so per the venue-routing rule it is routed to BINANCE_SPOT (CASH, leverage 1) to avoid paying futures fees for no margin benefit. Exit logic lets winners run via a trend-break MA and a wide take-profit so average per-trade return can clear the ~0.20% spot round-trip fee, while the 6% stop and z<=-1 reversal exit cut losers. Sizing is capital-relative through get_account_equity() and risk_per_trade, never hardcoded.

Fundamentally edgeless and fee-dominated on ADA 4H spot, not worth optimizing. Across 689 trades over 8 years the mechanism produces a NEGATIVE Sharpe (-0.14), profit factor 1.016, omega 0.97 and sortino -0.34 — statistical noise around break-even, not an edge. Per-trade gross edge is ~0.03% (expectancy $11 on ~$37k average positions), an order of magnitude below the 0.25% spot round-trip cost floor, so this trips the 'avg_trade_return below trading costs' abandon rule. Costs are structural, not tunable: commissions are 11.8% of gross, modeled market impact is 80.5% of gross PnL, and capacity is only $154k (untradeable at scale). The signal also fails to generalize across regimes (negative 2018/2019/2022/2025/2026; recent 2025 -13%, 2026 -14%; rolling Sharpe ends near -8). Optimization would mainly sweep entry_z to thin the trade count, but there is no evidence of a hidden profitable sub-region (PF hovers at ~1.0 throughout) and reducing sample cannot create per-trade edge where the gross edge is ~0.03% against a 0.25% cost floor. This is a code/venue-level no-edge result, not a parameterization that a 2-hour optimization can rescue. FAILURE PATTERN: high-frequency single-bar momentum/MR z-score signals on 4H alt spot produce hundreds of near-zero-edge trades whose tiny gross per-trade return is entirely consumed by commissions + market impact (impact 80% of gross, capacity ~$100k), yielding PF ~1.0 and negative Sharpe — abandon at the pre-optimization gate rather than optimizing.

Long-only, single-instrument momentum-continuation strategy on ADAUSDT 4H bars. Each bar's log return is standardized into a z-score against the trailing 50-bar return distribution. A long is opened when the latest return is an outsized positive event (z >= 1.0), the bar closed bullishly (close > open), and price is in a confirmed uptrend (close > 100-bar SMA). Exits on a hard stop-loss (6%), take-profit (18%), a trend-break below the 20-bar SMA, or an outsized negative-return bar (z <= -1.0). Positions are sized risk-relative (2% of equity over the stop distance) and capped at 95% of equity (spot, leverage 1).

Genuinely novel mechanism (bar-return z-score momentum continuation) and clean, well-structured implementation; high trade count (689) gives statistical power.

Strategy runs correctly and trades match the hypothesis (fires on outsized positive bars in uptrends).

No edge: Sharpe -0.14 (negative), profit factor 1.016, omega 0.97, sortino -0.34 across 689 trades / 8 years — indistinguishable from break-even noise.

avg_trade_return ~0.03% (expectancy $11 on ~$37k positions) — an order of magnitude below the 0.25% spot round-trip cost floor; the edge is below trading costs.

Fee/impact-dominated: commissions 11.8% of gross, market impact 80.5% of gross PnL, capacity only $154k — untradeable at scale.

Negative across recent regime (2025 -13%, 2026 -14%) and multiple prior years (2018/2019/2022); rolling Sharpe ends near -8.

deployable scale

0.25% (spot)

ADAReturnZScoreContinuationLong introduced a genuinely new mechanism — bar-return z-score momentum-continuation — to bet that outsized positive bars in an uptrend would extend, implemented cleanly with a healthy 689-trade sample. But the edge was absent: Sharpe -0.14, profit factor 1.016, ~0.03% per-trade return against a 0.25% cost floor, with commissions and 80.5%-of-gross market impact dominating and capacity of just $154k making it untradeable at scale. The analyst judged this a code/venue-level no-edge result that optimization could not rescue — thinning entry_z cannot create per-trade edge where the gross edge is below costs — and abandoned it at the backtest-review gate. It ended after a single iteration as abandoned, never reaching optimization, analysis, or risk review, confirming the documented failure pattern for fee-dominated high-frequency z-score signals on 4H alt spot.

High-frequency single-bar momentum/z-score signals on 4H alt spot produce hundreds of near-zero-edge trades whose tiny gross per-trade return is entirely consumed by commissions and market impact (impact 80% of gross, capacity ~$100k), yielding a profit factor near 1.0 and negative Sharpe — a novel, cleanly-implemented mechanism is worthless if its gross edge sits below trading costs.

It was abandoned at the pre-optimization backtest-review gate (verdict: abandon) as fundamentally edgeless and fee-dominated — failing the Sharpe, capacity, profit_factor, impact_cost, and avg_trade_return criteria, with per-trade gross edge (~0.03%) an order of magnitude below the 0.25% spot round-trip cost floor — so optimization and all later stages were never reached.

A long-only single-instrument momentum-acceleration strategy on ADAUSDT.BINANCE 4H bars (OHLCV only) implementing a novel bar-return z-score continuation mechanism: it standardized each bar's log return against the trailing 50-bar distribution and went long when the z-score was unusually large (≥1.0) AND the bar closed bullishly AND price was above its 100-bar SMA, exiting on a stop, target, MA break, or an outsized negative-return bar.

Over ~8 years (2018-07-01 to 2026-06-02) it took 689 trades with a 33% win rate but no edge: total return +39.3%, CAGR 0.95%, Sharpe -0.14 (CI [-1.50, 1.06]), profit factor 1.016, expectancy +$11/trade (~0.03%), a 24.7% max drawdown, and crushing costs (commissions 11.8% of gross, market impact 80.5% of gross, capacity only $154k).