What this actually does that a spreadsheet doesn't
If you don't already have a CSV with these columns, setting up the core fields in Google Sheets first gives Claude cleaner data to work with than a raw, unformatted broker export.
Most brokers and prop firm platforms let you export your closed trade history as a CSV: symbol, entry time, entry price, exit time, exit price, size, P&L. You can build formulas in Google Sheets to calculate win rate and profit factor from that file, and you should. What a spreadsheet won't do on its own is look across hundreds of rows and notice that your position size tends to increase specifically after a losing trade, or that your holding time on winners has been shrinking over the last two months while losers get held longer. That kind of question, spotting a pattern across many rows, is exactly what Claude can be asked to do with the same file.
Setting it up
Most brokers and prop firm platforms export closed trades in a compatible format, similar to what a Notion trading journal would ask you to log manually.
- Export your trade history. From your broker or prop firm platform, export closed trades as a CSV with the standard columns: symbol, entry date/time, entry price, exit date/time, exit price, size, P&L.
- Upload the file directly in a conversation. Claude reads structured files like this and can run real calculations on the actual rows in front of it, not a generic estimate.
- Start with the standard numbers. Ask for total trades, win rate, profit factor, average win, average loss, and maximum drawdown, computed from the uploaded data specifically.
- Ask a specific behavioral question, not a vague one. "What do you notice" produces a generic answer. "Does my position size increase after a losing trade" or "is there a day of the week where my losses cluster" produces something you can actually check against your own memory.
- Verify anything surprising. If a pattern turns up, sizing increases on Fridays, holding time drops after two losses in a row, check it against what you remember about those specific sessions before treating it as settled fact.
The question worth actually asking
A generic prompt like "analyze my trades" tends to get a generic answer back: your win rate, your best month, a few surface observations. The more useful prompt names the exact behavioral question you actually suspect is true about yourself. If you already have a hunch that you oversize after a loss trying to make it back, ask that directly and let the data confirm or contradict it, rather than waiting for a general summary to happen to mention it.
The CSV can tell you what happened, row by row, with total accuracy. It cannot tell you why you did it. That part still needs a person, not a file.
Is it safe to upload trade history data?
Trade history CSVs typically contain trading data rather than personally identifying financial information, but it is worth removing account numbers or personal identifiers before uploading if your export includes them, as a general privacy practice with any AI tool. The core columns needed for this analysis, symbol, entry and exit prices, size, and P&L, do not require anything more identifying than that.
What the analysis can't close for you
Claude can confirm, with real numbers from your own history, that you double position size after a loss more often than not. It can't tell you what makes that specific moment, a red trade sitting on the screen, feel like the one exception where the rule doesn't apply. That's a different kind of question, and it's the one that actually determines whether seeing the pattern changes anything about what you do next time. It's the same gap covered in how to actually use your trading journal data to change behavior, not just observe it.
Frequently asked questions
What CSV columns does Claude need to analyze trade history?
At minimum: symbol, entry date and time, entry price, exit date and time, exit price, position size, and profit or loss. Most broker and prop firm platform exports already include these by default.
Is it safe to upload trade history data to an AI tool?
Trade history CSVs typically contain trading data, not personally identifying financial account information, but it is worth removing account numbers or personal identifiers before uploading if your export includes them.
Can Claude tell me why I make a specific trading mistake, not just that I make it?
It can identify the pattern reliably from the data. It cannot explain the underlying psychological reason that specific moment feels like an exception to your own rules, since that requires understanding a person's history and triggers, not just their trade data.
Funded prop trader at Apex Trader Funding ($100K) and Alpha Capital Group ($100K). MA Psychology candidate at Chandigarh University, with clinical training under Dr. Nitin Sethi and at Japneet Bakhshi Clinic. Cofounded TradeRoot with Khushi Narwal to work at the source of behavioral trading patterns rather than their symptoms.