Patterns & edge

Beyond raw P&L, Journal has three layers of pattern discovery: Edge Finder, Patterns, and Trend Alignment. Each takes your trade history as input and tries to surface something actionable.

Edge Finder

Location: /app/journal/edge

Per-symbol statistics computed from your round trips:

• Win rate — pct of round trips that were profitable
• Expectancy — average P&L per trade (can be negative)
• Average winner vs loser — the asymmetry is what matters, not the raw win rate
• Risk-reward — derived from winner/loser sizes
• Trade count — ignores symbols with too few trades (< 3) as noise

The Edge Finder groups symbols into four buckets:

• Strong edge — high win rate, positive expectancy, enough trades to be significant
• Moderate edge — positive expectancy but with warnings (small sample, high variance)
• Negative edge — you've lost money on this symbol over enough trades to matter
• Too few trades — < 3 trades, can't conclude anything

Patterns

Location: /app/journal/patterns

Patterns goes further than symbol-level stats. It groups your trades across multiple dimensions and asks: what setups consistently worked for you, and what setups consistently bled?

The current pattern dimensions:

1. Trend alignment — were you trading with the larger trend, against it, or neutral?
2. Volatility regime — calm, normal, spike?
3. RSI zone — oversold (< 30), neutral (30-70), overbought (> 70)?
4. Holding period — 0-3d, 4-7d, 8-14d, 15-21d, 22-30d, > 30d
5. Position size — small, medium, large as percentage of capital
6. Sector — IT, Banks, Pharma, etc.
7. Trade-of-day — first trade, middle of day, end of day
8. Day-of-week — Monday through Friday
9. Post-loss behaviour — did you revenge-trade after a losing trade?
10. Scale-in/out — did you add to winners, losers, or neither? 11-13. Cross-dimension — Trend × Volatility, etc.

For each dimension, the engine computes: win rate, expectancy, total P&L, trade count. High-conviction findings become rules you can backtest.

Rules

Rules formalise the findings from Patterns into backtestable hypotheses. Example:

"On days when RSI(14) > 70 and you're buying (entry = long), your win rate drops from 38% (baseline) to 22%. Avoid buying overbought symbols."

The rule engine turns this into a backtestable filter, runs a mini-backtest on your own history (applying the rule retroactively to see what your P&L would have been), and reports the counterfactual.

If the mini-backtest shows the rule would have saved you money, you can "Publish" it — which exports it as a DSL strategy filter you can apply to future deployed portfolios.

Trend Alignment

Location: /app/journal/trend

A focused view that splits your trades into "with trend" vs "against trend" based on SMA(200) direction on entry. Three panels:

• With-trend performance — win rate, expectancy, total P&L for trades taken in the direction of the larger trend
• Against-trend performance — same metrics for counter-trend trades
• Per-symbol breakdown — which symbols favour with-trend vs against-trend for you specifically

The insight: most retail traders think they're good at counter-trend trades but the data usually says otherwise. If your with-trend win rate is 50% and against-trend is 30%, you know what to do.

Prerequisites

All three surfaces need:

• Tradebook data (see Tradebook import)
• At least a few months of trading history — patterns are noisy below ~50 trades

The Patterns engine specifically needs daily candles for each symbol you've traded to compute the dimensions (trend, vol regime, RSI zone at entry). If candles are missing for a symbol, its trades are marked "N/A" and excluded from pattern analysis.

Related

• P&L analysis — the raw P&L this all builds on
• Benchmark — performance vs market baselines
• Strategy DSL — the rule language Patterns exports to