How to Separate Manual Trades and EA Trades
Run an Expert Advisor and place the occasional manual trade on the same MetaTrader account, and the history records both as one undifferentiated stream. Every statistic computed on that stream — win rate, average trade, drawdown — then describes a trader who does not exist: half algorithm, half human. Separating the two is mostly a matter of conventions decided before the trades happen, and there are five workable ways to do it, each trading cleanliness against effort.
Key takeaways
- Blended account statistics average an EA and a human together — a profitable EA can hide losing manual trades, and the reverse, for months.
- Magic numbers are the most reliable separator for EA trades, but manual trades and EAs left at a default of 0 land in the same bucket.
- Order comments are free text the broker can append to, rewrite or truncate — useful as notes, too fragile to be the only separator.
- Symbol partitioning and separate accounts trade convenience for cleanliness: one costs flexibility, the other adds administration.
- Tagging trades in a journal after import works even on history that was never tagged, at the cost of ongoing manual effort.
- A written register — which magic number and comment prefix means what — is what keeps the separation intact over years.
Why blended statistics mislead
An account that mixes automated and discretionary trading produces one history, and every aggregate computed on it — win rate, average trade, longest losing run — describes the mixture, not either of its parts. The mixture can look unremarkable while the parts are anything but.
Consider a 12,500 account over four months: 160 closed trades, net +450. Flat-ish, slightly positive, nothing to act on. Split the same history by origin and a different account appears:
| Metric | Blended view | EA only (magic 4101) | Manual only (magic 0) |
|---|---|---|---|
| Closed trades | 160 | 110 | 50 |
| Net result | +450 | +1,340 | −890 |
| Win rate | 49% | 56% | 34% |
| Average trade | +2.81 | +12.18 | −17.80 |
| Longest losing run | 5 | 4 | 7 |
Blended, this account suggests a mediocre strategy. Separated, it is a working EA paying for a manual habit that loses on average and loses in streaks. The reverse pattern is just as common — a careful discretionary trader concluding they “can’t trade” while an EA quietly bleeds in the background. Either way, the blended number points the review at the wrong target, and it can do so for months before the gap becomes obvious.
Five ways to draw the line
Separation is a labelling problem, and MetaTrader offers several places to put the label — plus two options that avoid the platform entirely. Each has a distinct failure mode:
Magic numbers
The integer an EA attaches to its orders, kept by the server for the life of the trade. Reliable and machine-readable — but manual trades carry magic 0, and so does any EA whose magic input was never changed from a default of 0.
Order comments
Free text attached when the order is sent, readable at a glance. Brokers may append to, rewrite or truncate it — an MT4 partial close typically replaces it — so it cannot be the only separator.
Symbol partitioning
The EA trades EURUSD only; manual trades happen on other symbols. No setup and it works on past history — but it collapses the moment you want to trade the EA's symbol by hand.
Separate accounts
Manual trading on one account, EAs on another. The cleanest split every report will ever show — at the price of extra logins, split funding and margin that is no longer shared.
Journal tagging
Assign trades to strategies after importing the history into a journal. Works on any history, even untagged years — but someone has to keep tagging, indefinitely.
In practice the methods stack rather than compete: magic numbers identify the EAs, comments annotate the manual trades, and the journal is where both labels become groups. How the magic number itself works — and why it beats the comment for identity — is covered in the magic number guide.
A convention that still reads cleanly next year
The labels only help if they are assigned consistently, which means deciding the scheme once — before the next trade, not after the next confusion. A workable hybrid convention needs three parts: numbers for the EAs, comment prefixes for the manual side, and a register that records what means what.
A hybrid labelling convention
- EAs: one hundred-block per strategy — 41xx for the Asian-range EA, 42xx for the swing EA — with the last two digits as the version (4101, 4102 …).
- A revision that changes entries or exits gets a new number; cosmetic changes do not.
- Manual trades: magic stays 0, and a short comment is typed at entry — MAN-BO for breakouts, MAN-NF for news fades, MAN-X for anything unplanned.
- Journal mapping: each magic number to a strategy name, each comment prefix to a manual setup.
- The register — which number and prefix means what — lives in the journal notes, not in memory.
The MAN-Xprefix earns its place: it gives unplanned trades somewhere honest to go, which is the difference between a category you can review and a habit you can deny. And because brokers can rewrite comments on partial closes and stop-outs, the prefix is a convenience layer on top of magic 0, never the foundation — mangled comment fields are one of the data traps examined in the trade data quality guide.
What changes once each strategy reports its own numbers
The point of the split is not tidier reports; it is that different questions become answerable. Return to the account above. The blended view could only ask “is this account working?” The separated view asks two sharper questions — and the manual one changes character entirely once timing is added:
From split to decision
- Manual bucket: −890 over four months — but 28 of the 50 trades, and −610 of the loss, fall inside the two weeks the EA spent in drawdown.
- Reading: most of the manual damage is intervention trading around the EA, not a failing manual strategy.
- EA bucket: +1,340 across 110 trades with a worst run of 4 losses — consistent behaviour worth evaluating on its own terms.
- The sizing question for the EA and the discipline question for the manual trading are now separate decisions. Blended, neither could even be asked.
Per-group figures also stop strategies from borrowing each other’s credibility: a drawdown belongs to the strategy that produced it, and a strong month no longer launders a weak one. Which figures deserve attention per group — and at what sample sizes they start to mean something — is the subject of the journal metrics guide.
Habits that keep the separation clean
Separation decays without maintenance — not dramatically, but one unlabelled trade at a time. The habits that prevent it are small and mostly monthly:
- Assign before the first trade.A new EA gets its magic number from the register before it ever runs on the live account — retro-labelling is the expensive version.
- Audit the magic-0 bucket monthly. Every trade in it should be one you remember placing. Anything else means an EA or panel is trading untagged.
- Never reuse a retired number. A number recycled onto a new strategy splices two unrelated histories into one statistic.
- Re-check inputs after updates.Reinstalling or updating an EA can reset its inputs to defaults — including the magic number.
- Note interventions on EA trades.A manually closed EA position keeps the EA’s magic, so only a journal note preserves the fact that the exit was human.
- Spot-check comments after partial closes. If the broker rewrote them, the magic number and the journal tags are what still hold the line.
None of this requires special software — a spreadsheet and discipline will do. A journal that imports your own MetaTrader history and groups it by magic number simply automates the grouping half, leaving the register, the prefixes and the monthly audit as the human half. Ten minutes a month is a fair price for statistics that describe the trader who actually produced them.
Frequently asked
Can I separate trades that are already in my history with no tags?
Not inside MetaTrader — the magic number and comment are fixed when the order is sent and cannot be edited afterwards. The practical route is retroactive tagging in a journal after import: group the trades that already carry magic numbers, then assign the rest by symbol, session and size patterns. It is slower than tagging at order time, which is why a convention matters going forward.
Do partial closes break the separation?
They can break comment-based separation: in MT4 a partial close creates a new ticket whose comment typically references the original order, replacing whatever was typed at entry. The magic number survives partial closes on both platforms, which is the main reason to treat comments as notes rather than identity.
What happens to the statistics when I manually close an EA's trade?
The trade keeps the EA's magic number, so the result still lands in the EA's statistics even though the exit was yours. The per-strategy numbers then measure a hybrid of the EA's logic and your interventions. Noting manual interventions in the journal is the only way to tell the two apart later.
When is a separate account for manual trading worth the admin?
Usually when manual trading is a continuing activity rather than the odd trade. Two accounts give perfectly clean reports with no conventions to maintain, but margin is no longer shared, funding has to be split, and there are two histories to review. For occasional manual trades, magic numbers plus a comment convention achieve a similar analytical separation with less overhead.
Related guides
What Is a Magic Number in MetaTrader?
The integer ID that lets an EA recognize its own trades — and how it splits one account into separate strategies.
Why Data Quality Matters in Trade Analytics
The record-keeping faults that corrupt trade analytics — missing rows, doubled imports, shifted days — and the fix for each.
Which Metrics Matter in a Trading Journal?
The six core journal metrics, what each one hides on its own, and how to read them in pairs — with a worked 20-trade sample.
Related free tools
Free, no login required.
Related NuvoraSync features
Sources & further reading
- MQL4 Documentation — OrderSend() — the magic argument — the identifier MT4 stores with each order it opens.
- MQL5 Documentation — MqlTradeRequest structure — the magic and comment fields carried by MT5 trade requests into positions and deals.
Want to analyze your own MetaTrader account data automatically?
NuvoraSync is a read-only MetaTrader journal and analytics workspace. Connect MT4 or MT5 once and your trades, drawdown and performance update on their own — no manual entry, no signals, just your own data.
This article is for educational purposes only. It does not provide trading signals, investment advice, financial recommendations, broker recommendations or trade execution.