AdaDailyThreeBarReversalLongRegimeFilter

ADA Daily Three-Bar Reversal Long with Long-Term Regime Filter

A long-only single-instrument discrete-pattern strategy on ADAUSDT perpetual futures using daily bars and OHLCV-only data. Architecturally mirrors the proven 'SUI Three-Bar Reversal' pattern explicitly cited in failure analyses as one of the successful trend-following constructions in this pipeline, applied to ADA for portfolio diversification. The strategy waits for a discrete 3-bar pullback pattern (three consecutive lower-low days) followed by a confirmed bullish reversal bar (higher high + bullish close), then enters long ONLY when the long-term regime is healthy (rising 200-day SMA + 50-EMA > 100-EMA). Exits use a tight 2× ATR initial stop combined with a wider 3.5× ATR trailing stop — the exit asymmetry deliberately structurally INVERTS the broken small-win/big-loss profile that doomed NEAR Donchian, ETH Buy-the-Dip, and ~24 other abandoned strategies. ADA has not been used in any current or pending pipeline strategy, has 7+ years of clean Binance perp data, and exhibits classic L1-token cycle dynamics (sharp rally phases punctuated by 3-7 day shake-out pullbacks within multi-month uptrends — exactly the pattern this strategy targets). Discrete pattern detection (specific 3-bar sequence) is structurally NOT a channel-position threshold (the construction class explicitly forbidden by ~6 failed channel-breakout strategies) — it requires a precise candle pattern, which is qualitatively different from 'price > Donchian high'.

Iteration 2 fix (Layer 3 — no trades). The discrete 3-bar reversal pattern and exit/sizing logic already passed static + synthetic layers and are unchanged. The zero-trade cause was the over-restrictive regime gate requiring a RISING 200-SMA slope (SMA_now > SMA_20-bars-ago) AND EMA50>EMA100 to coincide with a fresh 3-bar shake-out — that triple-coincidence never occurred in 2263 flat bars. The smallest fix loosens ONLY the regime filter to 'price above a positive 200-day SMA AND EMA50 > EMA100', which still encodes the hypothesis's healthy long-term-uptrend requirement (200-SMA + 50/100-EMA cross) but is permissive enough to fire during the multi-month uptrends ADA exhibits. Pattern detection, exit asymmetry, and capital-relative ATR sizing are untouched, so no earlier-passing layer regresses.

Signal too rare and base is net-losing on single-asset ADA daily (not a data wall — full 2,339-bar history loaded; not a code bug — the pattern is implemented correctly and all 12 signals match). The discrete 3-bar reversal + 200-SMA/EMA regime gate fired only 12 times in ~6.4 years (~2/yr), far below the walk-forward sample floor (the hypothesis itself cites ~62 ADA-daily trades as the killing level, and prior ADA-daily / rare-chart-pattern reviews abandoned at similar counts). The result is net-losing (Sharpe -0.19, total_return -71.6%, PF 0.81, expectancy -$262, win rate 33%) and its only positive contribution is a single outlier trade (kurtosis 143.7, skew 8.4, +8.8% on 2021-09-07) — strip that day and the record is decisively negative. This is NOT fixable by parameter tuning within this hypothesis: the developer already loosened the regime gate once (docstring) and still gets 12 trades, and dropping pullback_bars 3->2 reaches only ~24, still sub-floor — while loosening far enough to hit ~62 would dissolve the specific candle pattern the hypothesis tests. Same recurring factory outcome as the abandoned BTC Three White Soldiers, double-top/bottom, and BB-bounce single-asset daily patterns: a strict discrete pattern on ONE asset's daily bars is structurally sample-starved. RESEARCH-LEAD REFRAME (not a developer parameter change): the only way to make a strict discrete candle pattern statistically testable is to run it as a MULTI-ASSET cross-sectional BASKET across many liquid L1 tokens, pooling entries to reach the trade floor while keeping the pattern strict. Recommend abandoning this single-asset ADA-daily instance and, if the mechanism is still desired, re-proposing it as a cross-sectional basket.

Long-only ADAUSDT daily three-bar reversal strategy with a long-term regime filter. Waits for >=3 consecutive lower-low days (a shake-out pullback) followed by a confirmed bullish reversal bar (higher high + bullish close), and enters long only when the long-term regime is healthy: price above a positive 200-day SMA and EMA50 > EMA100. Exits via a combined rising floor of a tight 2x ATR initial stop and a wider 3.5x ATR trailing stop, structurally inverting the small-win/big-loss profile so winners run.

Verification failed (Layer 3 — sandbox backtest): No trades produced Bar type used: ADAUSDT.BINANCE-1-DAY-LAST-EXTERNAL, Bars processed: 6988 Diagnostics: should_enter() returned a side 0 times over 2263 evaluated bars -> your ENTRY CONDITION never triggered. Loosen the entry logic / thresholds. Ensure your strategy produces trades with the given data and parameters.

Code is correct and the 12 entries genuinely match the discrete 3-bar reversal pattern + 200-SMA/EMA regime gate; the exit asymmetry works as intended (avg_win $3,424 > avg_loss $2,105), inverting the small-win/big-loss profile it set out to fix.

Clean single-instrument ADA daily design on fully-available data (2,339 bars, no data wall); disciplined ATR-risk sizing and combined initial/trailing stop.

SIGNAL TOO RARE: only 12 entries in ~6.4 years (~2/yr) — far below the walk-forward sample floor (the hypothesis itself cites ~62 ADA-daily trades as the killing level). 12 trades cannot support a 3-window walk-forward + 15-day holdout; the optimizer would fit noise.

Net-losing base: Sharpe -0.19, total_return -71.6%, profit_factor 0.81, expectancy -$262, win rate 33% — no edge, and negative in every year except 2021.

Result is a single-outlier artifact: return_skew 8.39, kurtosis 143.7, one +8.8% day (2021-09-07) carries the entire positive contribution; remove it and the record is decisively negative.

The obvious fix is already spent: the developer loosened the regime gate once and still gets only 12 trades; dropping pullback_bars 3->2 would reach ~24, still sub-floor.

This strategy adapted a proven three-bar reversal construction to ADA daily futures, correctly implementing a discrete 3-bar pullback-reversal pattern with a 200-SMA/EMA regime gate and an asymmetric ATR stop designed to fix the small-win/big-loss failure mode. The code was verified correct and the exit asymmetry worked (avg win exceeded avg loss), but the strict pattern fired only 12 times in ~6.4 years and the record was net-losing (Sharpe -0.19, -71.6% total return, PF 0.81), with its only positive contribution coming from a single outlier day. At the backtest-review gate the analyst abandoned it because 12 trades cannot support a 3-window walk-forward plus holdout and the edge could not be salvaged by parameter tuning — the developer had already loosened the regime gate once, and reaching the trade floor would destroy the pattern itself. As with prior single-asset daily chart-pattern abandons, the recommended path forward was to re-propose the mechanism as a cross-sectional basket across many liquid L1 tokens.

A strict discrete candle pattern on a single asset's daily bars is structurally sample-starved and cannot support walk-forward optimization; to keep the pattern strict yet reach a testable trade count, it must be run as a cross-sectional multi-asset basket.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate: 12 trades is far below the ~62-trade walk-forward floor and the base was net-losing, so no optimization or later analyst stage was run — and it was deemed unfixable by tuning since loosening the pattern enough to reach the floor would dissolve it.

A long-only ADAUSDT daily-futures strategy that entered on a discrete 3-bar reversal pattern (three consecutive lower-low days followed by a confirmed bullish reversal bar) gated by a long-term regime filter (price above a positive 200-SMA and EMA-50 > EMA-100), using an asymmetric tight 2× ATR initial stop plus a wider 3.5× ATR trailing stop to deliberately invert the small-win/big-loss profile.

Over ~6.4 years the pattern produced only 12 trades with a 33% win rate, net-losing on Sharpe -0.19, total return -71.6%, profit factor 0.81, and expectancy -$262; the exit asymmetry did work as designed (avg win $3,424 > avg loss $2,105), but the sole positive contribution came from a single outlier trade (skew 8.39, kurtosis 143.7, +8.8% on 2021-09-07).