Tagging Your Trades: How to Find Your Most Profitable Setup

Tags turn a list of trades into reviewable segments. Only then can you answer which of your setups carries the result and which consumes it. What matters is not how many tags you use but how they are structured: five fixed dimensions with clear pick lists say considerably more than twenty freely typed labels.

Why is a plain trade list not enough?

A list of entry, exit and result answers exactly one question: how much did you make or lose? It does not answer the one that moves you forward, which is why.

A strategy's headline number is almost always a blend. One setup finances another's losses, good market phases cover weak ones, and rule breaking trades hide inside the total. Without segmentation you only see the result of that blending.

This is where tags come in. They are not extra information but the precondition for any metric to produce an action. Which metrics suit that purpose is covered in the overview of the metrics that matter most.

Which five dimensions actually pay off?

These five carry most of the insight in my assessment, and together they cover idea, environment, circumstances and execution.

• Setup type. Which entry signal was present, breakout or pullback? The single most important dimension, because it points straight at your rules.
• Market regime. Trending or ranging, measured by a fixed filter rather than by feel.
• Time of day. Particularly revealing in continuously running markets, since both liquidity and your own attention vary across the day.
• State. One word about your condition at entry, chosen from a fixed list.
• Rule compliance. Yes or no, independent of outcome. The only dimension that separates a system problem from an execution problem.

Deliberately absent: tags for the outcome. Win or loss already sits in your numbers, and as a tag it invites assigning convenient categories after the fact.

Why do twenty tags say less than five?

Because every additional dimension splits your sample. That is arithmetic rather than preference.

An example: with 100 trades and three setup types, roughly 30 trades fall to each setup, which supports a first conclusion. Combine setup with two market regimes and you are down to about 16 per combination. Add three times of day and roughly five remain. Five trades say nothing.

The practical rule follows: analyse one dimension at a time rather than all at once. And expand your tag set only once the existing dimensions each hold enough cases.

How many trades do you need per tag?

The number that matters is not your total trade count but the count per segment. Thirty to fifty trades under identical rules is a reasonable lower bound, and that bound applies per value, not in total.

In practice: trading three setups means you need correspondingly more trades for a solid conclusion about all three than someone trading only one. Anyone experimenting with many setups simultaneously unintentionally extends the time to their first usable insight considerably.

That is a strong argument for trading few setups cleanly at the start rather than many in parallel. The review then arrives months earlier.

How do you tell whether a dimension discriminates at all?

Not every plausible sounding dimension actually says something about your trading. That can be checked, and checking saves a lot of unnecessary recording effort.

Compare your chosen metric across the values of one dimension and look at the gap between the best and the worst. If all values sit close together, that dimension does not discriminate for you and can be dropped. If they diverge clearly, you have found a lever.

Sequence matters here: gather enough cases per value first, then compare. Comparing after ten trades per segment measures randomness and may lead you to drop the most useful dimension because it happened to look unremarkable.

Dropping a dimension is progress rather than a step back. Every dimension you remove returns sample size to the remaining ones and makes their analysis reliable sooner.

Why do tags have to be fixed pick lists?

Because free text cannot be grouped. Writing "breakout" once, "break out" another time and "breakout after consolidation" a third time produces three categories with a third of the cases each instead of one holding all of them.

So define your pick lists in advance and keep them short. Three to five values per dimension is almost always enough. If a category seems missing, that is a deliberate decision at month end rather than a spontaneous entry during a trading day.

In a spreadsheet you enforce this through data validation, in journal software through predefined fields. Why spreadsheets are additionally error prone here is covered in trading journal spreadsheet.

When do you assign which tag?

Separating tags by timing matters more than choosing the tags themselves.

Before entry belong setup type, market regime and state. Those three describe your decision, and they cannot be reconstructed later, only remembered. Memory adjusts itself to the outcome.

After the exit belong rule compliance and reason for exit. Both can be judged objectively once the trade is closed, and both require knowing how it actually went.

Time of day derives automatically from the timestamp and never needs assigning at all. It is the one tag an exchange connection handles for you.

Why must tags never be edited afterwards?

Because doing so fits the past to the outcome, usually without you noticing. A trade planned as a breakout that ended in a loss easily becomes a messy breakout or a special case on review. That erases precisely the information you needed.

The exception is a genuine recording error, such as an obviously mis clicked category. Correct that, but flag the correction. Everything else stands, even if you would classify it differently today.

If you find a category was systematically unsuitable, change the list going forward and leave the past untouched. A visible break in the series is more honest than one smoothed over afterwards.

How do you analyse by tag?

Take one dimension and compare a single metric across its values. Profit factor works well here because it is dimensionless and makes segments of different sizes comparable, see calculating profit factor.

Three analyses tend to produce the most value:

1. By setup type. Does one setup carry the result while another consumes it?
2. By rule compliance. If results drop sharply on the deviations, your problem is execution rather than the system.
3. By market regime. In which phase is your strategy structurally weak?

This analysis belongs in the monthly slot rather than in a live trading day. How to keep the two apart is covered in trade review. A structured trading journal supplies the fields.

When should you expand your tag set?

Only once two conditions hold. The existing dimensions each contain enough cases to support a conclusion, and you have a concrete question the current tags cannot answer.

The second condition is the more important one. A dimension with no question behind it creates effort without return and spreads your sample across even more segments. When in doubt, fewer is right.

Conclusion

Tags are the bridge between recording and insight. Five dimensions with fixed pick lists suffice: setup, market regime, time of day, state and rule compliance. Assign them at the right moment, never edit them afterwards, and analyse one dimension at a time. The decisive number is not your total trade count but the count per segment.

Disclaimer: this article is for informational purposes only and does not constitute investment advice. The example figures are illustrative. Trading securities and crypto assets carries the risk of loss, up to and including total loss of capital.