Every trader gets told to journal. Almost nobody keeps it up.

The problem isn't discipline. It's that manual journaling asks you to do data entry right after a loss — the exact moment you least want to. So you skip a day, then a week. Now your spreadsheet covers your good stretch and none of your bad ones.

That's worse than nothing, because you'll draw conclusions from it anyway.

So I wrote a Python script. One command, and every closed trade lands in a spreadsheet with the context already attached.

The thing that broke my first version

I assumed one trade equals one row.

It doesn't. MT5 doesn't store trades — it stores deals, meaning individual fills. Scale into a position with two orders, take profit in two pieces, and that one trade is four deals in the database. They share a position_id, but nothing marks them as one idea.

My first version treated each deal as a trade. It reported roughly double my trade count, and a badly wrong win rate — the scale-in fills all showed zero profit, so they got counted as scratches.

The fix is grouping by position_id, then splitting on the entry field: 0 opens or adds, 1 closes or reduces.

for position_id, group in deals.groupby("position_id"):
    opens = group[group["entry"] == 0]
    closes = group[group["entry"] == 1]

Once you're grouping, entry price stops being obvious too. Two fills at 2310.5 and 2311.5 means your entry is 2311.0 — not 2310.5, which is what you get reading the first fill. So prices are volume-weighted.

Open positions will wreck your stats

If you pull the last 30 days, some positions in that window are still open. They have an opening deal and no closing deal.

Group them naively and you get a trade with a real entry, no exit, and zero P&L — which lands in the sheet as a flat loss. Every open position quietly drags your win rate down.

if opens.empty or closes.empty:
    continue

Two lines. Same check catches the reverse: positions you closed inside the window but opened before it.

The automation isn't the point

Pulling the data is plumbing. The labels are the product.

For each trade the script derives session, hold-time bucket, asset class, and outcome. That's what turns a flat list into something you can pivot on — and a pivot table is a mirror:

• Am I actually better in London, or does it just feel that way?
• Do my scalps make money, or do a few swings carry them?
• Is gold my edge or my ego?

None of those are answerable from a list of trades. All of them are answerable from a list of labelled trades.

My Take

The script removes the excuse. It doesn't do the work.

It still only sees what I did, never why — no note on the setup, no record of what I was thinking. Those are the fields that make a journal actually useful, and they're the ones that can't be automated.

Which is probably the honest lesson. Automate the bookkeeping so you have no reason left to skip the thinking.