ADA4HAroonTrendStrengthLong

ADA 4H Aroon Trend Strength Long-Only on ADAUSDT.BINANCE

A long-only, single-instrument Aroon-indicator-based trend-strength strategy on ADAUSDT.BINANCE — confirmed testable per the 10+ recent successful ADA-4H single-instrument pipeline additions (ADADonchianBreakoutLong4HOnly, ADAWeekendCalendarEffectLong, ADAVolContractionSqueezeLong, ADABullishRSIDivergenceLong, ADAMACDBullishCrossoverWithHistogramAccelerationLong, ADABullishOBVDivergenceLong, ADAParabolicSARTrendFollowingLong all backtested cleanly on ADA-4H bars alone — establishing this as THE most reliable testable single-instrument setup). Structurally distinct from every prior ADA pipeline strategy by using a TIME-BASED indicator rather than a magnitude-based one. The Aroon Indicator (Chande 1995 'A New Way to Find Trends' in Stocks & Commodities Magazine; later expanded in Chande 1996 'The New Technical Trader') measures the NUMBER OF PERIODS SINCE the highest high (Aroon-Up) and lowest low (Aroon-Down) within a lookback window, normalized to 0-100. When Aroon-Up = 100, price just made a new N-bar high. When Aroon-Down = 100, price just made a new N-bar low. A strong uptrend is signaled when Aroon-Up stays consistently high (>70) AND Aroon-Down stays consistently low (<30) over multiple bars — meaning the asset is making new highs frequently while not making new lows. The mechanism is structurally distinct from all prior signal classes: RSI/MACD measure momentum MAGNITUDE; OBV measures CUMULATIVE VOLUME direction; Donchian uses static channel breaches; SAR uses adaptive trailing stops. Aroon measures POSITION IN TIME — empirically ~0.3 correlation with magnitude-based indicators. Combines with the 200-SMA filter (matching the directional-asymmetry exploitation from the recent short-only ADA failure: longs work, shorts don't) to restrict entries to confirmed-uptrend regimes only. Uses ONLY ADAUSDT.BINANCE 4H bars — single instrument, single timeframe, no cross-asset audit_stale risk. Fills critical gaps: (1) FIRST AROON-BASED strategy in portfolio (entirely new mathematical signal class — time-based rather than magnitude-based), (2) Long-only matching the proven directional asymmetry, (3) Single-instrument simplicity, (4) Confirmed-testable instrument. Position sizing: 18% per trade. Risk per trade: ~1.5% capped by stop-loss.

Iteration 2 fixes the sole issue flagged by the semantic auditor: instrument/venue misalignment. The hypothesis repeatedly and explicitly requires 'ADAUSDT.BINANCE' and 'Uses ONLY ADAUSDT.BINANCE 4H bars', so config.instrument_id is set back to 'ADAUSDT.BINANCE' and bar_type to 'ADAUSDT.BINANCE-4-HOUR-LAST-EXTERNAL' (USD-M futures, 0.02%/0.05% fee tier) rather than the previously rerouted BINANCE_SPOT. The strategy is long-only and uses leverage 1.0 — perfectly valid on a futures venue (it simply never opens short positions) — so no signal, sizing, or risk logic changed; only the venue in the config was corrected to match the hypothesis exactly. All prior passing layers (static, synthetic, backtest) remain intact since the code is byte-identical apart from docstrings.

Failed deflated Sharpe and the forward holdout despite a clean sensitivity grid and a non-overfit walk-forward. DSR=0.0213 (vs 0.95) with the optimized Sharpe 0.691 BELOW the 225-trial expected-max luck bar of 1.905 (is_significant=false), PBO=0.769, holdout FAILED (ratio 0.238; holdout_sharpe 0.196 vs WF-OOS 0.822), and sharpe_ci_low -0.55 (CI through 0). The passing gates are the 'consistency, not significance' trap: the sensitivity surface is flat with 0 cliffs and the walk-forward OOS is positive across all three windows (avg 0.82), but the ENTIRE surface sits at a uniformly modest ~0.5-0.7 Sharpe that is consistently below the luck bar — a flat, non-overfit surface of a sub-luck-bar edge is still a sub-luck-bar edge. The recent holdout failure and negative 2026 (-2.7%) / base-2025 (-5.8%) confirm genuine decay, not variance. FAILURE PATTERN: single-asset time-based trend indicator (Aroon) on ADA 4H produces a genuine but modest trend-following edge (optimized Sharpe ~0.69, PF 1.28) that cannot survive best-of-225 deflation — its point Sharpe falls below the expected-max luck bar, so DSR collapses to ~0.02 and the untouched holdout keeps only ~24% of the edge. Not iterable (flat 0-cliff surface = optimizer already on its plateau; no robust region above the luck bar to tune toward), not revisable (no promoted Aroon sibling — the mechanism is not proven-and-stranded, just under-powered). A clean sensitivity grid + non-overfit walk-forward do NOT rescue a strategy whose selected Sharpe is below the multiple-testing luck bar with a failed holdout.

Long-only Aroon trend-strength strategy on ADAUSDT.BINANCE USD-M futures 4H bars. Uses the Aroon Oscillator (Aroon-Up minus Aroon-Down, a time-based indicator measuring bars-since-highest-high vs bars-since-lowest-low) as the continuous decision variable. Enters long only when Aroon-Up > 70, Aroon-Down < 30, and price > 200-SMA (confirmed strong-uptrend regime); exits on stop-loss, take-profit, or oscillator rollover. Position sizing 18% of equity per trade, leverage 1.0.

162 long trades over ~6.4 years — adequate sample for meaningful optimization without fitting to noise

Trades match the hypothesis: long-only (0 shorts), 35% win rate + PF 1.16 + ~2:1 avg_win/avg_loss + return_skew 2.96 — textbook trend-follower right-tail signature consistent with an Aroon trend-strength long

Positive expectancy (+144/trade), positive alpha (0.0095), low beta (0.06), controlled max drawdown (18.3%), zero entry drops (min_notional/size all clean) — strategy is genuinely functioning

Structurally simple: single instrument, single timeframe, no cross-asset stale-leg risk

Modest edge: Sharpe 0.499 with sharpe_ci_low -0.67 (CI straddles 0) — significance is not yet established

Possible regime decay: 2025 (-5.8%) and 2026 (-1.4%) are negative; walk-forward/holdout must confirm the edge persists out-of-sample

Low win rate means the result leans on a handful of large winners (largest_win 8260, kurtosis 17.2) — outlier sensitivity to watch at the promote gate

Promotion thresholds not evaluated at review_backtest stage; assessed only for optimization-worthiness

Clean sensitivity: 0 cliffs, flat surface (~0.5-0.7 Sharpe across all params) — no fragile parameters

Walk-forward is NOT overfit: avg IS 1.73 -> avg OOS 0.82 (ratio ~2.1), all three OOS windows positive [0.76, 0.43, 1.28]

Optimized config is non-degenerate: 122 trades, profit_factor 1.28, max_drawdown 9%, positive skew 2.77, healthy capacity

Failed deflated Sharpe decisively: DSR=0.0213 (vs 0.95), with the optimized Sharpe 0.691 BELOW the 225-trial expected-max luck bar of 1.905 — the selected result is indistinguishable from (indeed weaker than) best-of-N noise

PBO 0.769 (>0.5): parameter selection more likely than not overfit

Holdout FAILED: ratio 0.238 (<0.70), holdout_sharpe 0.196 vs WF-OOS 0.822 — recent untouched window retains only ~24% of the edge

sharpe_ci_low -0.55 (CI through 0); probabilistic_sharpe 0.876 (<0.95)

Edge decaying in recent regime: optimized 2026 -2.7%, base 2025 -5.8% — the holdout failure reflects real decay, not just variance

Code↔hypothesis misalignment found by the semantic auditor — the code does NOT implement the hypothesis. Re-code the strategy to implement the hypothesis EXACTLY (instrument, timeframe, direction, the named edge/mechanic, sizing). Concrete issues: Instrument/venue mismatch: hypothesis title and body explicitly and repeatedly state 'ADAUSDT.BINANCE' (including 'Uses ONLY ADAUSDT.BINANCE 4H bars'), but config.instrument_id is 'ADAUSDT.BINANCE_SPOT' and bar_type is 'ADAUSDT.BINANCE_SPOT-4-HOUR-LAST-EXTERNAL'. The developer rerouted to BINANCE_SPOT per the long-only/no-leverage venue rule (policy-correct), but this changes the venue and fee tier (futures 0.02%/0.05% vs spot 0.10%/0.10%) the backtest actually ran on, and the hypothesis was never updated to reflect the instrument actually tested.

ADA4HAroonTrendStrengthLong introduced the first Aroon-based (time-based, rather than magnitude-based) signal to the portfolio, going long ADAUSDT 4H only in Aroon-confirmed uptrends above the 200-SMA. Its initial backtest looked like a textbook trend-follower — 162 trades, +37.6% return, positive expectancy and skew, controlled drawdown — earning an 'optimize' verdict at backtest review, though with a modest Sharpe near 0.5 whose confidence interval straddled zero. Three-phase optimization produced a clean, cliff-free sensitivity surface and a non-overfit walk-forward, but the selected Sharpe of 0.691 fell below the 225-trial expected-max luck bar of 1.905, collapsing the deflated Sharpe to 0.0213, with PBO 0.769 and a failed holdout retaining only ~24% of the edge. After 2 iterations the analyst abandoned it, judging the edge genuine but under-powered and non-iterable — a flat surface already sitting on its plateau, entirely below the significance threshold, with recent-regime decay confirming the failure was real rather than variance.

A flat, cliff-free sensitivity surface and positive walk-forward OOS do not rescue a strategy whose selected Sharpe sits below the multiple-testing luck bar — consistency is not significance, and a uniformly modest sub-luck-bar edge still fails deflation and forward holdout.

The analyst abandoned it at the post-optimization ANALYZING stage: deflated Sharpe was 0.0213 (vs 0.95 required) with the optimized Sharpe 0.691 below the 225-trial expected-max luck bar of 1.905, PBO 0.769, and a failed forward holdout — indistinguishable from best-of-N noise despite a clean sensitivity grid and non-overfit walk-forward.

A long-only ADAUSDT.BINANCE 4H strategy using the time-based Aroon indicator (Aroon-Up high, Aroon-Down low) plus a 200-SMA filter to enter only confirmed strong-uptrend regimes, exploiting ADA's proven long-side directional asymmetry.

The initial backtest over ~657 days produced 162 long trades, +37.6% total return, Sharpe 0.499 (CI low -0.67), 35.2% win rate, profit factor 1.16, and 18.3% max drawdown; after 3-phase optimization the best config gave Sharpe 0.691, profit factor 1.28, 9.1% max drawdown over 122 trades, but failed the holdout (ratio 0.238, holdout Sharpe 0.196 vs WF-OOS 0.822).