Skip to content

Discretionary → Systematic

For traders who have built up a discretionary playbook over years and want to codify those rules as declarative JSON and validate them with backtests, rather than leaving them at the level of "gut feel" or "market sense."

Three Problems Discretionary Trading Carries

Problem What Happens in Discretionary Trading How AlphaForge Solves It
Reproducibility "Market sense" and "intuition" are hard to verbalize, so the source of edge is opaque Express the strategy as JSON so anyone can reproduce the exact same signals
Emotional Bias Aggressive on winning days, timid on losing days — decisions swing with P&L Apply one fixed ruleset across the full history to quantify the bias
Hindsight Bias "I should have cut here" is always obvious after the fact Walk-forward validation evaluates using only information available before the trade

Translate Discretionary Rules Into Declarative JSON

If your discretionary rule is "buy when the 20-day MA crosses up and RSI(14) < 30, stop out at 2× ATR," you can declare this directly in an AlphaForge strategy JSON.

# Start from a template
forge strategy create my_playbook --template ma_rsi_atr

# Edit the strategy JSON to match your discretionary rules:
#   - Entry conditions (MA period, RSI threshold)
#   - Exit conditions (ATR multiple, take-profit width)
#   - Sizing (fraction of account balance)

For logic beyond what plain JSON expresses (composite conditions, regime detection), see the HMM and multi-timeframe examples in Strategy Templates.

Quantify Discretionary Skill vs. Emotional Drag

Apply the same rule mechanically across 2018–2025 and the gap between this systematic result and your actual discretionary P&L is the size of your emotional bias.

# 1. Backtest the codified rule
forge backtest run QQQ --strategy my_playbook

# 2. Walk-forward to evaluate only with information available beforehand
forge optimize walk-forward QQQ --strategy my_playbook --folds 5

# 3. Compare against your actual discretionary record
#    - Systematic backtest Sharpe: 1.10
#    - Personal discretionary Sharpe : 0.40
#    → Emotional bias was eroding 0.70 Sharpe points

The point isn't to disown discretion — it's to quantify the edge you took discretionarily, and carve out the reproducible parts.

Bridge to Hybrid Execution

If you're not ready to commit to full automation, AlphaForge supports a hybrid intermediate stage: mechanize the signal, keep execution discretionary.

Backtest → Optimize → Generate Pine Script
                                  ↓
                       TradingView alerts (notification only)
                                  ↓
                       Confirm and execute manually

You can let AlphaForge raise alerts for the patterns you "always miss," while keeping the execution decision in your own hands. There's no full-automation prerequisite, so the psychological migration cost stays low.

First Steps

# 1. Start from a simple MA + RSI template
forge strategy create my_playbook --template ma_rsi_atr

# 2. Fetch 5 years of historical data
forge data fetch QQQ --start 2020-01-01

# 3. Backtest the codified ruleset
forge backtest run QQQ --strategy my_playbook --json

# 4. Capture the baseline for emotional-bias comparison
forge journal record my_playbook --note "Baseline vs. discretionary record"