How to Review Your Trades Each Week Without Overreacting

A weekly review is one of the more useful habits in trading and one of the easiest to get wrong. Done well, it is a short piece of bookkeeping: what happened, where it came from, and whether the trades matched the plan. Done badly, it is a verdict on a strategy reached from five days of data, which is far too little to support one. What follows is one calm way of reading a week of your own trades, not a view on what anyone ought to trade.

Why one week says so little

Start with a hypothetical setup that genuinely has an edge. It wins 55% of the time, and winners and losers are both 1R. Its expectancy is 0.55 × 1 minus 0.45 × 1, which is +0.1R per trade, so a trader taking 15 of these trades a week expects to finish the week about 1.5R up.

The spread around that expectation is wide. The standard deviation of a single trade is just under 1R, so over 15 trades it is about 3.9R, more than twice the expected result. Worked out exactly, this setup finishes a 15-trade week in the red about 35% of the time, roughly one week in three. It finishes 5R down or worse about 8% of the time, roughly one week in thirteen. Even across four such weeks, 60 trades, it still ends the month negative about 18% of the time.

None of those weeks would mean the edge had gone. They are what a real but modest edge looks like through a narrow window. The same logic runs the other way: a strong week from a setup with no edge at all is just as ordinary. This is the background every weekly review sits against, and the reason how many trades it takes to validate a strategy is measured in hundreds rather than dozens.

Start with counts, not the result

Before the result, the count. How many trades were taken, compared with a typical week, and how many per setup and per session? A week with twice the usual number of trades is information in its own right, and whether those extra trades looked like the first ones or drifted away from them is the question taken up in overtrading: what the data usually shows.

Counts also set the honest resolution for everything that follows. A figure computed from four trades can be noted, but not interpreted.

Results by condition

The next step is splitting the week by condition. In NoxarQuant every trade is tagged with its setup, session (Asia, London, New York AM or New York PM), volatility bucket, RSI bucket and trend, and trades sharing all five tags form a cluster. A weekly review in those terms asks a simple question: which clusters did this week's trades land in, and how do their running totals look now that this week has been added?

That framing matters. A cluster with three trades this week has not been tested this week; it has had three trades added to its history. The classifier is deliberately slow to move for this reason. A cluster with fewer than 12 trades gets no verdict at all, and strong requires at least 20 trades and an SQN of at least 2.5, as set out in what SQN measures. One week rarely moves a cluster across a threshold, and when it does, the cumulative count is the thing to read, not the week.

The calendar view lays out each day's result with a weekly total and trade count, and can be filtered to strong groups only, promising groups only, or everything except the groups that lost money. That makes one useful split quick to see: whether a losing week came from trades inside established clusters, where losses are an expected part of the distribution, or from trades scattered outside them.

Rule breaks belong in their own column

Results describe outcomes; rule breaks describe process. Sorting the week's trades into four groups shows why the two are worth keeping apart: followed the plan and won, followed the plan and lost, broke a rule and lost, broke a rule and won.

The first two groups are the strategy doing what it does. The third is a mistake that was at least visible. The fourth is the one that does the most damage over time, because a rule break that pays off feels like good judgement and tends to be repeated. Counting rule breaks regardless of whether they won or lost, and totalling what they cost or earned, stops that fourth group from hiding inside a good week.

What counts as a rule break is the trader's own definition: a moved stop, a trade outside the planned sessions, size above the plan, an entry with no setup behind it. It only needs to be recorded the same way every time, as a field rather than a comment, which fits the argument in what a trading journal should track that a journal is evidence with a schema.

Read the biggest losers one at a time

Averages smooth over the trades that usually explain a bad week, so reading the three largest losses individually often says more than any summary statistic. For each, a few factual questions:

• Was the loss roughly the size the plan allowed, or larger?
• Was the position the planned size?
• Did the trade fall into a cluster with a meaningful history, or one too thin for a verdict?
• How far did it go against the position before it closed?

A planned 1R loss in a well-sampled cluster is the cost of trading that cluster, and nothing about it needs explaining. A 3R loss from a stop that was widened during the trade is a different event entirely, even though both appear as one red line in the journal. Maximum adverse excursion shows whether losses are ending where they were planned to end.

Separating process from variance

Put together, those steps let a bad week be split into two parts. Suppose, hypothetically, a week ends 5R down across 15 trades. Thirteen trades followed the plan, winning six and losing seven, for a net of minus 1R. The other two were not in the plan at all: an entry with no setup behind it, taken at double size, lost 2R, and a second unplanned trade at double size, taken straight after it, lost another 2R. Together those two come to minus 4R.

For the example setup, 13 trades have an expected result of +1.3R and a standard deviation of about 3.6R, so minus 1R is well within one standard deviation of normal. That part of the week is variance. The other 4R is process, and it is four fifths of the loss. The headline figure of minus 5R describes neither part well. The split describes both.

The reverse case is just as common and easier to miss. A week that finishes 4R up can contain +6R from two oversized trades and minus 2R from everything else. The total looks fine. The structure underneath does not.

What one week cannot show

Some conclusions a single week does not support, whichever way it went:

• That a setup has stopped working. A losing week from a positive expectancy setup is a routine event, as the numbers above show.
• That a setup has started working. A winning week from a cluster with eight trades is eight trades.
• That a change made on Monday caused Friday's result.
• That a run of wins or losses says anything about the next trade. Streaks of the length a week contains are normal in independent sequences, as covered in how long a losing streak can last.

What a single week can show reliably is process: how many trades were taken, whether they matched the plan, and whether the losses were the size they were meant to be. Those facts are fully observable after five days. Edge is not.

The point

A weekly review is bookkeeping for a sample that takes months to mean anything. Its job is to keep the record clean and the process visible, so that when there are finally enough trades to say something about edge, the data underneath is worth reading. Looking at your own weeks this way, with the trade count next to every number, is the kind of measurement our case study is built around.