Improvement Suggestions - Quant ShastrAI

The provided script is an excellent starting point, effectively visualizing a classic mean-reversion concept. It successfully identifies moments of statistical and momentum-based extremity. However, to evolve this from a discretionary dashboard into a professional-grade, automated trading system, we must systematically build a framework of rules that enhance its statistical expectancy and ensure its resilience across diverse market conditions.

This roadmap outlines three additive levels of enhancement, transforming the core concept into a robust, quantifiable strategy.

Level 1: Parameter Optimization & Dynamic Adaptability¶

The current script uses static, “hard-coded” values for its lookback periods and thresholds (e.g., RSI 14, Z-Score 20, RSI > 70). This is the primary source of curve-fitting risk. A strategy optimized for a specific historical period on one asset will likely fail when market volatility shifts or when applied to a different asset. Level 1 addresses this by introducing dynamic, volatility-normalized parameters.

Technical Logic & Implementation:

Convert to a Strategy: The first step is to transform the indicator into a strategy and define the core entry logic.

//@version=5
strategy("MR System v1: Dynamic Exits", overlay=true)

// ... [Inputs for RSI, Z-Score] ...
rsi_val = ta.rsi(close, 14)
z_score = (close - ta.sma(close, 20)) / ta.stdev(close, 20)

long_entry_condition = rsi_val < 30 and z_score < -2.0
short_entry_condition = rsi_val > 70 and z_score > 2.0

Implement ATR-Based Stop-Loss and Take-Profit: Instead of fixed-point or percentage-based exits, we will use the Average True Range (ATR) to define risk and reward. This ensures that our exits are proportional to the market’s recent volatility. In a volatile market, stops and targets are wider; in a quiet market, they are tighter.

// --- Dynamic Exit Inputs ---
atr_len = input.int(14, "ATR Length")
sl_multiplier = input.float(1.5, "Stop Loss ATR Multiplier")
tp_multiplier = input.float(3.0, "Take Profit ATR Multiplier")

// --- Exit Logic ---
atr_val = ta.atr(atr_len)

if (long_entry_condition)
    strategy.entry("Long", strategy.long)
    strategy.exit("Exit Long", "Long", stop = close - atr_val * sl_multiplier, limit = close + atr_val * tp_multiplier)

if (short_entry_condition)
    strategy.entry("Short", strategy.short)
    strategy.exit("Exit Short", "Short", stop = close + atr_val * sl_multiplier, limit = close - atr_val * tp_multiplier)

Quantitative Benefit:

This upgrade directly attacks curve-fitting and enhances adaptability. By normalizing risk parameters to volatility (ATR), the strategy’s risk-reward profile remains consistent across different assets (e.g., Forex vs. Crypto) and timeframes. This change is designed to reduce maximum drawdown and improve the Calmar Ratio (Annual Return / Max Drawdown). A static stop-loss might be too tight in a volatile market (leading to premature stop-outs) or too wide in a quiet one (leading to excessive risk). Dynamic ATR-based exits solve this, leading to a more stable equity curve and proving the strategy’s edge is not dependent on a specific, hard-coded risk value.

Level 2: Secondary Confluence & Noise Filtration¶

The Level 1 strategy will enter on every signal that meets the RSI and Z-Score criteria. Many of these will be “whipsaws” or “falling knives”—signals that occur in low-conviction environments or against an overwhelmingly powerful trend. Level 2 introduces secondary filters to improve the signal-to-noise ratio, ensuring we only take trades with a higher probability of success.

Technical Logic & Implementation:

Integrate a Higher-Timeframe (HTF) Directional Bias: A mean-reversion trade is most effective when it acts as a pullback within a larger, established trend. For example, buying a dip is safer if the weekly trend is bullish. We will use a higher-timeframe moving average to establish this macro bias.

// --- HTF Filter Input ---
htf = input.timeframe("D", "Higher Timeframe")
htf_ma_len = input.int(50, "HTF MA Length")

// --- HTF Logic ---
htf_ma = request.security(syminfo.tickerid, htf, ta.sma(close, htf_ma_len))

is_macro_bullish = close > htf_ma
is_macro_bearish = close < htf_ma

// --- Refined Entry Logic ---
long_entry_condition_l2 = long_entry_condition and is_macro_bullish
short_entry_condition_l2 = short_entry_condition and is_macro_bearish

Add a Volume Spike Confirmation: A true exhaustion or capitulation event is typically accompanied by a significant spike in volume. Filtering for this ensures we are trading genuine market turning points, not just low-volume drift.

// --- Volume Filter Input ---
vol_ma_len = input.int(50, "Volume MA Length")
vol_multiplier = input.float(1.75, "Volume Spike Multiplier")

// --- Volume Logic ---
vol_ma = ta.sma(volume, vol_ma_len)
is_volume_spike = volume > vol_ma * vol_multiplier

// --- Final Level 2 Entry Logic ---
long_entry_condition_l2 = long_entry_condition and is_macro_bullish and is_volume_spike
short_entry_condition_l2 = short_entry_condition and is_macro_bearish and is_volume_spike

// ... [Strategy entry/exit calls using L2 conditions] ...

Quantitative Benefit:

These filters are designed to surgically remove low-expectancy trades. By avoiding counter-trend entries against a strong macro trend and filtering out low-volume noise, we significantly increase the strategy’s Win Rate. While the total number of trades will decrease, the quality of each trade increases dramatically. This has a direct and positive impact on the Profit Factor (Gross Profit / Gross Loss), as we are systematically eliminating a cohort of losing trades from the system’s performance history.

Level 3: Structural Architecture & Regime Detection¶

A professional-grade system must recognize that no single market philosophy (like mean reversion) works all the time. Markets cycle between trending and ranging phases. A mean-reversion strategy will perform well in a choppy, range-bound market but will be systematically destroyed in a strong, persistent trend. Level 3 rebuilds the strategy’s core architecture to be “regime-aware.”

Technical Logic & Implementation:

Implement a Market Regime Filter: We will integrate a mechanism to classify the current market environment. A simple and effective way to do this is by using the ADX (Average Directional Index). A high ADX value indicates a strong trend (hostile to mean reversion), while a low ADX indicates a range-bound or non-trending market (ideal for mean reversion).

// --- Regime Filter Inputs ---
adx_len = input.int(14, "ADX Length")
adx_threshold = input.int(25, "ADX Trend Threshold")

// --- Regime Logic ---
[di_plus, di_minus, adx_val] = ta.dmi(adx_len, adx_len)

// The strategy is only active if the market is NOT in a strong trend
is_mean_reversion_regime = adx_val < adx_threshold

// --- Final Level 3 Entry Logic ---
long_entry_condition_l3 = long_entry_condition_l2 and is_mean_reversion_regime
short_entry_condition_l3 = short_entry_condition_l2 and is_mean_reversion_regime

if (is_mean_reversion_regime)
    if (long_entry_condition_l2) // Note: we use L2 conditions here
        strategy.entry("Long", strategy.long)
        // ... exit logic
    
    if (short_entry_condition_l2)
        strategy.entry("Short", strategy.short)
        // ... exit logic

Note: The logic is structured to check the regime first. If the regime is not suitable, no other calculations or trade checks are performed, making the system computationally efficient and logically sound.

Quantitative Benefit:

This structural change is the ultimate defense against “Black Swan” events and prolonged, unfavorable market cycles. By dynamically toggling the entire strategy off during strongly trending periods, we achieve a significant improvement in Robustness. The primary quantitative benefit is a drastic reduction in the depth and duration of drawdowns. This prevents the catastrophic losses that occur when a mean-reversion strategy repeatedly tries to fade a powerful, sustained trend. This enhances the strategy’s long-term viability and survivability, directly improving its Sortino Ratio by minimizing severe downside volatility. The system is no longer a one-trick pony; it is an intelligent agent that knows when its core competency is applicable and, more importantly, when to stand aside.