Analyzing your trades is not about staring at monthly P&L and deciding whether you are good or bad. It is a repeatable loop: start with clean data, classify setups, measure risk in R, separate process from outcome, then isolate one recurring mistake and fix it on a large enough sample. The seven-step method below fits into a 45 to 60-minute weekly review. It works with a spreadsheet or an automated journal, as long as you use the same format every time.
What is a trade review?
A trade review is a structured reading of a group of trades, not a verdict on the latest P&L. It compares groups (setup, session, instrument), checks that the data is clean and chooses one action for the week. A useful review produces a written decision, not just a chart.
Five key takeaways
- ▸Analysis starts with clean data: fees included, duplicates removed, time zone consistent.
- ▸R-multiples let you compare trades of different sizes; dollars alone say nothing about risk.
- ▸A win outside the plan is not a good trade, and a loss inside the plan is not a failure.
- ▸Review groups first (setup, session, instrument), not isolated trades.
- ▸One written rule tested over 30 trades beats ten changes after a bad week.
Why most trade reviews fail
The classic review looks like this: open the journal on Sunday night, look at the total, see a loss, find the worst trade, then change a rule. By Monday the new rule is forgotten. This loop fails for three reasons. It judges a process on one outcome. It mixes trades from several accounts, instruments or sessions. And it changes too many things at once, so you cannot tell what worked.
A good review answers precise questions: where is the edge, where is the leakage, and what concrete action fixes it this week. It does not try to predict the next trade. It tries to make next month less random than last month.
Step 1: clean the data before drawing conclusions
A wrong statistic is worse than no statistic. Before calculating anything, check that each row is a complete trade and that net P&L includes every cost. A manually filled journal often forgets commissions, swaps or partial closes. A CSV import can duplicate rows if you import the same period twice.
- ▸P&L is net of commissions, fees and financing, not just entry minus exit.
- ▸Every trade has an initial stop and a risk amount in account currency, even if the stop was never hit.
- ▸Partial closes are handled the same way for every trade: one row per exit or one aggregated row.
- ▸The journal time zone matches your broker, and sessions use the same convention.
- ▸Setup names come from a closed list, not from improvisation at entry time.
- ▸Importing the same file twice does not create duplicates.
- ▸The net P&L of three random trades matches the broker statement.
If your journal cannot show the initial stop and risk, you cannot calculate reliable R. Fix that before analyzing anything else.
Step 2: classify setups with a fixed taxonomy
A setup is a market configuration you can describe in one sentence, with a trigger and an invalidation. "Breakout" is not enough: too many different trades fit that label. "London open range break with a close above the first 30-minute high" is more useful because you can verify whether the condition was actually met.
Limit yourself to two or three main setups for a quarter. Add a secondary tag for context (trend, range, news) instead of creating a new category for every variation. The goal is at least 30 trades per category before comparing results.
| Setup | Context | Trigger | Invalidation |
|---|---|---|---|
| Range break | Clear 30-minute range | Close outside the range | Back inside the range |
| Pullback | Established intraday trend | Rejection at a value area | Close below the swing |
| Post-news continuation | After a major release | First 5-minute close outside the pre-news range | Back inside the pre-news range |
These examples are illustrations, not strategy recommendations. What matters is that each setup has a testable definition and always uses the same name.
Step 3: measure R-multiples, not just dollars
R is your unit of risk. If you risk $100 on a trade and make $200, the result is +2R. If you lose $100, it is -1R. This unit lets you compare a trade on a $5,000 account with one on a $50,000 account, or two position sizes in the same week.
The formula is simple: net result divided by initial risk. Initial risk is the distance from entry to stop, multiplied by point or lot value and position size. It must not be recalculated after you move the stop, otherwise you erase the very information you are looking for. For the link between stop, target and expectancy, read our risk/reward ratio guide.
| Trade | Initial risk | Net result | R |
|---|---|---|---|
| EUR/USD long | $75 | $150 | +2.0R |
| MES short | $60 | -$60 | -1.0R |
| BTC perp long | $120 | $30 | +0.25R |
A 40-trade example: 16 winners averaging +2.04R, 24 losers at -1R. Total P&L is +8.6R and expectancy is +0.215R per trade. A 40% win rate looks weak, but the strategy is positive because average wins are twice average losses. That is exactly the conclusion a dollar total hides.
Step 4: separate process from outcome
A winning trade can be a bad decision, and a losing trade can be a good decision. If you judge only the outcome, you learn to repeat dangerous behavior because it paid once. The matrix below is the heart of the review: it crosses result with plan adherence.
| Result | Plan followed | Reading | Action |
|---|---|---|---|
| Win | Yes | Good trade | Repeat, do not overinterpret |
| Win | No | Good result, bad trade | Identify the broken rule |
| Loss | Yes | Good loss | Accept the cost of risk |
| Loss | No | Bad loss | Fix the trigger or the stop |
In the 40-trade example, 28 trades follow the plan and make +12.8R. The 12 off-plan trades lose -4.2R. The overall result is still positive, but the leak is visible: without off-plan trades, performance would be better. That is more useful than saying "the month was average".
Step 5: review by session, day and instrument
Global aggregates hide sharp differences. A trader can make money on EUR/USD in London and lose on the same instrument late in the US session. To see it, filter by session, weekday and instrument, but keep at least 30 trades per group. Below that, you are looking at noise, not a trend.
| Session | Trades | Net P&L | Win rate | Expectancy |
|---|---|---|---|---|
| London | 18 | +6.1R | 44% | +0.34R |
| New York | 14 | +3.9R | 43% | +0.28R |
| Asia | 8 | -1.4R | 25% | -0.18R |
| Total | 40 | +8.6R | 40% | +0.215R |
The Asian session has only eight trades: too few to conclude it is a loser. But the fact that it is the only negative group justifies reducing size or pausing the setups traded there until the sample is larger. That is a risk decision, not statistical certainty.
Step 6: find mistake clusters
Do not look for the perfect mistake in an isolated trade. Look for groups. If five of your twelve off-plan trades are revenge trades after two losses, the problem is not the setup: it is the rule that follows a losing streak. The fix belongs to that sequence, not to the chart.
| Mistake tag | Frequency | P&L | Possible fix |
|---|---|---|---|
| Revenge trade | 5 | -2.6R | 30-minute pause after two losses |
| Moved stop | 4 | -1.1R | Fixed stop, no intervention |
| Early entry | 3 | -0.5R | Wait for the signal close |
Rank mistakes by cost, not frequency. A rare but expensive mistake may deserve priority. Here, moving the stop costs less than revenge trading, but it is easier to fix with a mechanical rule. Start with the correction you can apply every day.
Step 7: write one rule and verify it on a sample
A useful rule is specific, observable and tied to a trigger. "Be more disciplined" is not a rule. "After two consecutive losses, stop entering for 30 minutes and resume only with an A setup" is one. Write it as: if [condition], then [action], because [observed data].
Test it on the next 30 comparable trades. Record before and after: mistake frequency, group P&L, rule adherence. Change nothing else during the test. If the rule reduces leakage without damaging the whole, keep it. If not, remove it. An untested rule is just another belief.
One rule per week. Add three and you will not know which one caused the change, then you will abandon the review within a month.
A full example: what a 40-trade review says
Take the same 40 trades. Overall P&L is +8.6R, with 40% winners and +0.215R expectancy. The "post-news continuation" setup loses -2.2R over eight trades, while the range break makes +7.2R over twenty and the pullback +3.6R over twelve. The twelve off-plan trades cost -4.2R. The Asian session is negative on a sample that is too small.
| Setup | Trades | Winners | Net P&L | Expectancy |
|---|---|---|---|---|
| Range break | 20 | 8 | +7.2R | +0.36R |
| Pullback | 12 | 6 | +3.6R | +0.30R |
| Post-news continuation | 8 | 2 | -2.2R | -0.275R |
The decision is not to delete the news setup after eight trades. It is to write a stricter rule and test it on the next 30 comparable trades. For example: wait for the first 5-minute candle close outside the pre-news range instead of entering in the first minute. If the rule shows no improvement over 30 trades, remove the setup from the plan.
The weekly review agenda
An effective review has a fixed order and a time limit. Do it at the same time every week, with the journal already imported. The goal is not to read everything, but to leave with one decision.
| Time | Task | Expected output |
|---|---|---|
| 5 min | Check the import and missing trades | Complete data |
| 10 min | Read the scorecard and curves | Three gaps identified |
| 10 min | Tag mistakes and count clusters | One priority mistake |
| 10 min | Review three good and three bad trades | Context and triggers |
| 10 min | Write one rule and its test | One action for the week |
If time is short, do the first three rows. The scorecard and mistake counts often identify the action for the week. Our weekly review method covers a five-step variant, and our journal template lists the columns you need to run it.
The scorecard and common mistakes
A scorecard is not a report card. It is a small table that forces you to look at the same numbers every week. Adapt the thresholds to your strategy: they are guardrails, not universal truths.
| Metric | Example | Guardrail | Action if outside |
|---|---|---|---|
| Expectancy | +0.215R | Positive over 100 trades | Cut the worst setup |
| Win rate | 40% | Read with average win | Do not optimise it alone |
| Average win / loss | 2.04 / 1.0 | Above 1.5 | Review exits |
| Profit factor | 1.36 | Above 1.3 | Check costs |
| Plan adherence | 70% | Above 90% | One process rule |
| Off-plan losses | -4.2R | Under 10% of gross losses | Remove the trigger |
Most mistakes are methodological rather than technical:
- ▸Changing strategy after three losses, before an exploitable sample exists.
- ▸Judging a setup on fewer than 30 trades and confusing variance with edge.
- ▸Ignoring commissions, spread and slippage in net P&L.
- ▸Mixing several accounts, instruments or sessions in one average.
- ▸Reviewing only losing trades and ignoring off-plan wins.
- ▸Adding tags endlessly instead of keeping a short, stable list.
- ▸Confusing a high win rate with profitability.
To connect these numbers to risk management, use the position size calculator before every trade. A journal cannot fix badly sized risk after the fact. Our performance analysis guide also explains how to read drawdown and profit factor without telling yourself stories.
FAQ
What is a trade review? It is a structured reading of a group of trades, not a verdict on the latest P&L. It compares groups, checks data quality and chooses one concrete action for the week.
How many trades do I need to analyze a strategy? At least 30 per group for a first read, and around 100 for a stronger conclusion. Below that, you can fix a process, not judge an edge.
Should I analyze every trade? Log every trade for data quality, but focus the deep review on three representative wins and three losses. The rest is analyzed in groups.
What exactly is R? Net result divided by initial risk. If you risk $100 and make $200, you have +2R. Moving the stop does not change the initial risk used in the calculation.
How long does a useful review take? About 45 to 60 minutes a week once the data is imported. A five-minute daily note is enough to capture context while it is fresh.
What if I break a rule? Log it, measure its cost and choose one correction for the week. If the broken rule concerns daily risk, stop for the day and resume tomorrow.
Sources
For planning and risk management: CME Group, CFTC and Van Tharp Institute. For MetaTrader 5 history export: MetaTrader 5 Help. The numbers in this article are teaching examples, not real trading results.