Every trade leaves traces — we make them visible

Most traders don't keep a journal. Of those who do, most quit after three months. The reason isn't lack of discipline — it's a design problem. Manual journaling is tedious, retrospectively biased, and screenshots age badly. Critical context — order book state, delta structure, regime — never gets captured because it feels obvious at the moment of the screenshot.

The OrderFlowAi Trading Journal solves this by turning journaling into a fully automated background process. Every trade — manual or executed through the CORTEX auto-trading strategy — gets recorded with its full market context. The trader doesn't decide what gets documented. The trader only decides when to look.

Why auto-capture instead of manual journaling

A good trade journal contains more than entry and exit. It contains the state of the market at the decision moment: the regime, the CORTEX direction, the iceberg signals, the session phase, the HTF trend, the delta sum of the last few seconds. You can't reliably capture that manually — you can only have it auto-logged at entry time.

The journal hooks into two events: OnOrderUpdate (NT8 strategy) and OnPositionFilled (bridge). At entry, a snapshot of the current market state is created. At exit, MFE (Maximum Favorable Excursion) and MAE (Maximum Adverse Excursion) are computed from the cached tick history — corresponding to the best point the position reached and the worst point where it sat closest to the stop.

MFE and MAE are the most important post-trade-review numbers. The PnL of a single trade is not what matters — what matters is the gap between actual exit and maximum MFE. That's where craft sits, or where the repeated mistake lives.

The six tabs in detail

MFE and MAE — the two numbers that say the most

MFE (Maximum Favorable Excursion) is the maximum gain a position reached during its lifetime. MAE (Maximum Adverse Excursion) is the maximum loss it had to sit through. Both are computed tick-accurately from the bar history and stored in ticks (not dollars) — so they're comparable across instruments.

The most important derived metric is MFE efficiency: the ratio between actual exit gain and MFE. A trader who consistently captures only 40% of their MFE is scaling out too early. A trader who captures 95% often runs back to stop — the trailing mechanism doesn't match market dynamics.

The AI Coach

The AI Coach is not a chatbot. It is a rule-based pattern detection module that scans the last 20–200 trades for statistically significant anomalies. The output is intentionally dry:

"78% of all losses over the last 30 days occur between 15:30 and 15:38 ET. Average loss in this window: -2.4 points. Recommendation: raise Early-Open guard to 8 minutes."

Exactly this kind of observation led to Strategy v7.4 in April 2026 with extended open-guard and NY soft-guard for aiDir=NEUTRAL and ATR spikes. The coach is not just analysis — it's a feedback loop: its insights become concrete code improvements and RL weight updates.

Privacy and locality

The Trading Journal stores data exclusively on the user's machine. No cloud sync, no automatic upload, no server-side storage. Data lives at %APPDATA%/OrderFlowAi/journal.json and is fully user-controlled.

If you voluntarily want to contribute anonymized trade statistics to improve the pretrained model, you can generate an RL export and upload it through the dashboard — strictly opt-in, teaches only aggregated signal combinations, no instrument-specific PnL.

Integration with CORTEX and the NT8 strategy

Every journal entry contains the full CORTEX signal snapshot at entry (aiDir, aiConf, aiAcc, aiRevPhase) and the rule-based score breakdown of the strategy (which signals contributed points). That makes it possible to verify retroactively why a trade fired — a diagnostic capability completely missing in manual execution.

For the auto-trading engine, the journal is also the learning data source: after every closed trade, the journal updates both internal signal weights and the CORTEX Neural Engine's RL table. Journal and neural engine form a closed feedback loop.

Frequently asked questions

Related articles