AI trade review: what it finds in your trading journal

In short. AI is useful to a trader as a reviewer of their own history: it links facts from the journal (size after a loss, trades without a stop, losers held too long, the depth of floating drawdown) and tells you where money is at risk and which rule would fix it. The more stops, notes and tags the journal has, the sharper the conclusions; with fewer than 10 closed trades it is guesswork. AI gives no signals and no market forecasts, and its answer is not investment advice. In Davinchu, AI reviews an account, a single trade and a week using numbers the platform calculates itself.

What AI finds in a trade history

Any spreadsheet can count win rate and profit factor. AI's strength is elsewhere: it connects several facts and turns them into money and into the risk of losing the account.

A good conclusion reads like this: «after two losses in a row, size grew by half, and those trades lost more than the rest earned». It has numbers and a rule you can check in the journal.

What not to expect from AI

• Signals and forecasts. Your trade history does not tell where price will go.
• Investment advice. AI reviews your process: risk, discipline, behaviour. You make the decisions.
• Conclusions without data. If trades have no stops and no notes, AI cannot know whether there was a stop or why you entered. A good answer says what is missing.
• Unchecked arithmetic. A chatbot can miscount or «fill in» a number, so check the figures in its answer against the file.

What data it needs

The minimum is 10 closed trades in the period, and the longer the history, the more reliable the conclusions. A review gets its value from:

• a stop and a target on each trade — without them, risk to the stop can't be measured;
• entry and exit reasons and the lesson — AI checks them against what happened;
• setup, session, mistake, emotions and the «followed the plan» mark — to compare groups of trades;
• a strategy description and risk rules — then trades against your own rules stand out.

What to record for each trade — in «How to keep a trading journal in MT5: what to record».

AI review in Davinchu

Account review

In the «AI Analytics» menu, choose «Account to analyze» and «Period» («30 days», «90 days» or «All time») and press «AI account review». The threshold is 10 trades closed in the period: if there are fewer, the dashboard suggests a longer period and the request is not counted. AI receives a dossier calculated by the platform: trade results and behaviour, equity drawdown and how often it happens, open positions and their age, grid baskets, how far price has to move to loss thresholds, load by hour and by advisor, your Strategy and risk manager rules, deposits and withdrawals. Account numbers, logins, keys and names are not sent — this is described under «What is sent to the AI». The review runs in the background for up to a couple of minutes. It contains:

• «Key takeaway» — one or two sentences on risk and money;
• «Plan for the period» — 2–4 rules with a number, each with «How we'll check» and «Cost of the rule»: how many times it would have fired on this account's history and what would have followed;
• «Findings» — up to five, with severity from «Critical» to «Works»;
• «What works» and «What is missing for precise conclusions»;
• «Previous plan follow-up» — if at least a day has passed since the last review, the new one checks its rules against fresh numbers: «Done», «Partly», «Not done» or «Can't check».

For expert advisors, AI suggests limits that fit the strategy: a cap on positions and size per instrument, an emergency stop on equity.

Trade review

«Journal» → a closed trade → the «AI review» block in its card → «Review trade». AI receives the prices, stop and target, result with commission and swap, time and duration, your own notes (entry and exit reasons, lesson, «Execution (1–5)», «Followed the plan?», tags, emotions) and the context: how the account traded over 90 days overall and on this instrument, plus the five trades before this one. It judges the process, not the outcome: a verdict of «Good trade», «OK», «Room to improve» or «Process broken», then «Went well», «Could be better» and «Lesson:». The review is saved and reopens without a new request; «Refresh review» makes a new one.

Automatic Monday report

The «Automatic Monday report» toggle is on the «AI Analytics» tab. From the next Monday, at 9:00 in your time zone, AI prepares a report on all trades of the past week compared with the week before: «Strengths», «Areas to improve», «Recommendations», «Plan for the week». A «Weekly AI report» notification with the trade count, result and focus for the week arrives in the dashboard bell and in connected Telegram, email or Discord (the «AI reports and insights» switch in «Settings» → «Notifications»). A week without trades uses no request. Past reports open from the «Report history» block.

Tilt alert

Tilt is caught by rules right after a trade closes, with no AI request and nothing taken from the limit. Only manual trades count; expert advisor trades are left out. By default, «Tilt detected» arrives if 6 or more trades closed within the last 60 minutes (overtrading), or at least two of them lost after commission and swap and the last one's size is more than 1.4 times the average of the earlier trades in the window, up to three («revenge»); with three earlier trades:

last size > (V1 + V2 + V3) / 3 × 1.4

Example: trades of 0.10, 0.12 and 0.08 lots, two of them losers. Average size (0.10 + 0.12 + 0.08) / 3 = 0.10 lots, threshold 0.14. A 0.15-lot trade (+50%) triggers the alert, a 0.13-lot trade (+30%) does not. Each type arrives at most once an hour; the window and thresholds are changed in «Settings» → «Notifications» → «Tilt», and «Reset to defaults» restores them. Which rules stop tilt — in «Tilt, overtrading and revenge trading: how to stop».

AI requests by plan

One request is an account review, a trade review or a weekly report. If AI did not answer, the request returns to the limit, and a finished review reopens without a request. The limit resets on the 1st; what is left shows on the tab: «AI requests this month: N of M». On Plus, the automatic report takes up to 4–5 requests a month, one per Monday, leaving at least 10–11 for reviews. On Free, the two requests work well for a 90-day account review and one telling trade.

How to review your journal in a chatbot yourself

You can run your own review in ChatGPT or another chatbot that reads tables. Export: «Settings» → «Data» → «Data export» → «Trade journal» → «Download CSV». The file holds trades from all accounts: instrument, side, size, prices, stop and target, commission, swap, net result, open and close time in UTC, tags, entry and exit reasons, lesson. The id, account_id, ticket and account_name columns are not needed for a review — delete them before uploading.

Questions to start with:

1. «This is my trading journal. No signals or forecasts — review my process. First tell me how many rows the file has and what period it covers».
2. «Convert opened_at to UTC+3 and total net_pnl by hour and day of the week. Where am I consistently losing?»
3. «Compare trade size after a loser and after a winner. What did trades opened within 30 minutes of a loss bring?»
4. «How many trades have no stop (empty sl) and what is their result?»
5. «Which mistake tag cost the most? Suggest 2–3 rules with a number».

What to check in the answer:

• the row count and period match the file;
• you recalculated one figure yourself: win rate = winners / all trades × 100%, 23 winners out of 50 is 46%;
• each conclusion shows its calculation — ask for the formula or code;
• a rule with a number, not «be more disciplined»;
• no «buy» or «price target» — such an answer is no longer about your process.

The review in the dashboard needs no export: the platform builds the dossier and AI gets ready numbers. Which metrics to check yourself — in «How to evaluate a trading strategy: which metrics matter»; how the journal and statistics fill themselves — on the «Trading journal» and «Trading account analytics» pages.