You’re profitable on paper.
The question is how much you left on the table. Every journal shows you win rate and P&L. None of them tell you what produced that number.
Three things your journal knows
and doesn’t tell you
Not opinions, not self-reported feelings. Each one is a ratio computed from data your broker already exported, with a fixed threshold you can check.
How much longer you hold losers
Average loss duration against average win duration. Nobody feels this ratio. It only exists in the timestamps.
fires at 2.0xHow much of your edge fees take
Total fee and commission drag as a share of gross P&L. Negligible on any single trade, decisive across a year.
fires at 20%How you change after a loss
Win rate after a win against win rate after a loss, gated behind a consecutive-loss precondition so a single bad day can’t trigger it.
fires at 15% · 3 lossesThirteen patterns.
Zero configuration.
Most journals ask you to tag every trade. Two weeks in you stop, and the data that survives is the data you chose to remember — which is the data least likely to be true.
TradeYar reads timestamps, prices, quantities and fees. The part that can’t be forgotten and can’t be edited after the fact.
You enter none of these. They are computed from the same file you already download from your broker.
Prop firm rules, already configured
Pick your firm and the rule set loads with it — drawdown limits, profit target, minimum trading days, consistency percentage, phase names. Both trailing-drawdown methods are modelled separately, because end-of-day and intraday are not the same rule.
Firms that have shut down are flagged on any account still referencing them, with the date they closed.
Import once. Nothing typed twice.
CSV, XLSX and HTML statements. Execution logs are detected and rejected before they touch your history, because one fill per row is not one trade — and a journal that silently accepts them is wrong from the first import.
Duplicate detection matches on broker ticket first, fingerprint second. Re-importing the same file adds nothing.
Three states, not two
A number is either measured or it isn’t. When the data can’t settle a question, saying so is the only honest output — so the engine has a third state and uses it.
resolvedThe data answers it.
indeterminateBoth levels printed inside one candle. The candle doesn’t record which came first.
not_reachedPrice never got there.
// A rule in the engine: // fabricated data must never reach narration.
Every figure you see is computed before it is written. The engine calculates; the AI only puts it into a sentence. A number that wasn’t measured has no way to appear.
Start free. Upgrade when you’ve seen it.
Pattern detection is a Pro feature. The free tier imports, journals and tracks your accounts without limits — said here rather than after you’ve spent an evening setting it up.
Free
- Unlimited imports
- Dashboard and statistics
- Journal and playbook
- Account and rule management
- 3 AI trade reviews
Pro
$19.90/mo · $199/yr
- Everything in Free
- 13 behavioural patterns
- Mirror Score
- Reports and psychology
- Unlimited AI coach
Your trade history already contains the answer.
Import once. The engine reads what’s already there.
Import your trades