Arrow.AlgoNexta: Multi-Regime MACD Correction & Hidden Divergence Indicator

TL;DR

• Function: Identifies trend continuation setups by mapping fast MACD correction-releases and hidden divergences against a slow MACD directional regime.
• Target Market: Intra-day and swing traders utilizing hourly (1h) or similar intermediate timeframes.
• Key Edge: Quantifies correction depth using an ATR ratio, preventing premature entry signals during shallow pullbacks, layered with strict ADX and EMA-separation filters.

Not financial advice. Backtested performance does not predict future results.

Context & Thesis

Trend continuation strategies typically suffer from the "premature trigger" problem: entering on what appears to be a pullback, only for the price to continue its counter-trend correction and stop out the position. Arrow.AlgoNexta attempts to solve this by untangling the trigger from the regime.

Rather than relying on a single oscillator to define both trend and timing, it isolates the directional bias using a slow, heavily smoothed MACD. It then waits for a faster, standard MACD to cross below a zero-line buffer (in a bullish regime) to confirm that a meaningful correction has occurred. By forcing the correction to achieve a minimum ATR-adjusted depth, the script filters out market noise and isolates structural pullbacks prior to their release.

Methodology

The indicator logic is built on two primary MACD engines and a sequence of stringent filters:

1. Directional Regime (Slow MACD): Defaulting to 48/104/36, the slow MACD establishes the structural trend. If the value sits above the Slow MACD Zero Buffer (0.0 by default), the regime is bullish. If below, it is bearish.
2. Correction Release (Fast MACD): Using a standard 12/26/9 calculation, the script looks for exhaustion. For a long signal, the fast MACD must dip below zero, reach a minimum depth defined by the Minimum Correction Depth (ATR Ratio) (default 0.05), and then cross back above its signal line while still below zero.
3. Hidden Divergence: The script independently scans for hidden divergence. A bullish hidden divergence occurs when price forms a higher pivot low while the fast MACD oscillator forms a lower pivot low.
4. Filtration: Before an arrow is plotted, the signal must pass a 60/240 EMA trend filter (EMA 60 > EMA 240 for buys), a slope filter on the 240 EMA (measured over a 10-bar lookback), and a Range Filter (requiring ADX > 20 and minimum EMA separation).

Pine Script v6 Implementation

Below is a conceptual Pine Script (v6) block demonstrating the core regime and trigger logic utilized by the indicator:

// [Pine Script v6] Core logic approximation for Arrow.AlgoNexta
//@version=6
indicator("Arrow.AlgoNexta Core Logic", overlay=true)

// MACD Inputs
fast_ema1 = ta.ema(close, 12)
fast_ema2 = ta.ema(close, 26)
fast_macd = fast_ema1 - fast_ema2
fast_signal = ta.ema(fast_macd, 9)

slow_ema1 = ta.ema(close, 48)
slow_ema2 = ta.ema(close, 104)
slow_macd = slow_ema1 - slow_ema2

// ATR Depth Filter
atr_val = ta.atr(14)
min_depth = atr_val * 0.05

// Regime
bullish_regime = slow_macd > 0
bearish_regime = slow_macd < 0

// Correction Conditions (Long)
fast_macd_below_zero = fast_macd < 0
correction_deep_enough = fast_macd < -min_depth
trigger_long = ta.crossover(fast_macd, fast_signal)

// Signal Execution
buy_signal = bullish_regime and fast_macd_below_zero and correction_deep_enough and trigger_long
plotshape(buy_signal, title="Buy Release", style=shape.triangleup, location=location.belowbar, color=color.aqua, size=size.small)

Backtest & Vision Grade Results

Note: The following metrics are derived from an AI-reconstructed approximation of the script's logic via AlgoNexta's reference engine on TradingView OHLC data, not live execution. Results are evaluated on EURUSD (1h timeframe).

Confidence Tier: E2 (Low Confidence) Sample Size: 72 total valid signals

Because the sample size (72 signals) falls below our robust statistical threshold of 100, these results must be treated as directional, not statistically reliable.

Walk-Forward Signal Performance

AlgoNexta Vision Grade evaluates signal-based performance (holding for a fixed dynamic horizon) split into in-sample (IS) and out-of-sample (OOS) sets:

• In-Sample (49 signals): 46.94% Win Rate, 0.0001 average return per signal.
• Out-of-Sample (22 signals): 54.55% Win Rate, 0.0002 average return per signal.
• Overall Signal Metrics: 42.25% Win Rate, Profit Factor 0.656, Expectancy -0.00022.

Reference Execution Approximation

When run as a continuous sequential strategy (72 closed trades), the reference execution yielded a nominal positive net profit of +0.019 (approx 1.9% on un-leveraged base) with a maximum drawdown of 0.0098 (0.98%).

Limits: The dual-MACD logic relies heavily on a trending environment. In ranging markets, the minimum correction depth filter mitigates some noise, but the low overall signal profit factor (0.656) indicates that without active trade management and optimized take-profits, the raw signals alone experience high drag from false positives.

Key Takeaways

• Tune the ATR Depth: The default 0.05 ATR ratio for the correction depth is highly sensitive to your traded pair's volatility. Optimize this parameter first to avoid triggering on superficial intraday noise.
• Session Filters Matter: This indicator allows bounding signals by London/NY sessions. Given the reliance on structural trend, restricting signals to high-volume hours reduces the likelihood of false crossovers typical in the Asian session.
• Pending Verification: With an E2 grade and only 72 signals in the evaluated EURUSD 1h slice, the out-of-sample edge is not yet proven. Do not deploy these defaults live without expanding the backtest sample size on your target asset.

Risk Disclosure

Not financial advice. The performance metrics presented here are derived from historical backtesting and automated signal evaluation, which inherently suffer from hindsight bias. Backtested performance does not predict future results. The algorithms and logic discussed are for educational and research purposes only. Real-world trading involves slippage, commissions, and execution delays that can significantly impact results. Always independently verify code logic and conduct walk-forward, out-of-sample testing before allocating capital to any algorithmic strategy.