Evidence Coach

The Evidence Coach reads your trading record and says what it found: the one pattern costing you the most, the one paying you best, and the exact trades behind each. Every string it produces has the same shape — your data says X, here are the trades — because every finding is arithmetic over your own record, checkable trade by trade. There is no AI involved, and nothing to take on faith.

It runs on every plan, on every Home board, and it speaks for the first time at the end of your first import — before you've journaled a single day.

What it looks for

Seven detectors, each judging your record against itself (your own median loss, your usual pace, your hours):

• Oversized losses: single losses far beyond your typical one, charged only for the excess over your own median.
• Revenge re-entries: entries hurried after a loss, measured against your own usual gap between trades, reported only when the hurry costs money.
• Hour of day: the windows, in your timezone, where your results say you shouldn't be trading, and the ones that pay you.
• Unreviewed trades: whether the trades you never review perform worse than the ones you do.
• Broken rules: what the trades where you broke your own setup rules cost, next to the ones where you followed them.
• Unplanned days: how days traded without a sealed plan compare against the days you planned.
• Setup fading: a setup whose recent trades have turned against its own older record.

The first three read pure trade history, so an import is enough. The other four read your journaling (reviews, plans, rule marks) and speak once that history exists.

The floor, and the silence

Every detector has a minimum sample: twenty trades before the coach says anything at all, and its own floor per finding. A pattern claimed from four trades isn't evidence, it's an anecdote. Below the floor, the coach tells you exactly how many trades it still needs. When nothing crosses a threshold, it says that nothing did, rather than inventing a finding to fill the space. And a flagged pile that still made money is not called a leak: the coach reports what costs, not what merely looks undisciplined.

The evidence, opened

Each finding shows its money impact and links the trades behind this: a drawer with the full sample, the baseline the finding was judged against, the direction your record was already moving, and every cited trade (day, symbol, result) linked to the trade itself. One press opens the whole citation as a filtered view of the trade log.

Disagree? Not a leak dismisses that reading for thirty days. The record may still be saying it next month, so the dismissal is remembered, not permanent, and the next-costliest finding steps up in its place.

From evidence to commitment

When the costliest leak is single oversized losses, the daily and weekly Plan offer the coach's cap as a max single loss target: the detector's own threshold, so accepting means "no more of exactly what was flagged". The target is marked as coming from your evidence and the Debrief scores it like any other; declining is remembered too. See Reviews for how targets are scored.

The coach never tells you what you should do. It tells you what your record says, shows the trades, and lets the numbers argue.