What Is a Worked Example of Strategy Tagging?

What is strategy tagging?

Strategy tagging is a structured way to label each trade (or journal entry) with the strategy features that you want to compare later—such as entry logic type, time window, stop approach, and whether the trade was manual or rule-based. The core idea is to separate your decision process (the features you control and can describe) from outcomes that vary with market conditions.

A tag is a label from a predefined scheme. For strategy tagging to be useful, the scheme must be consistent: the same feature should produce the same tag across entries, even when the market changes.

How does a worked example of strategy tagging work?

Imagine you journal multiple trades and want to learn whether a “feature set” performs differently from another feature set. You will create tags and then compute simple summaries per tag.

Worked example setup (all assumptions stated):

• You review 10 closed trades in a single month.
• Each trade is assigned exactly one value for each of two tag categories:
  • Category A: Entry timing = {London session, Other}
  • Category B: Stop method = {Fixed-distance, Trailing}
• You will not use any real-time data. You already know the result in pips for each completed trade.
• You ignore compounding and treat pips as the only outcome metric.

Tagging rules (fixed):

1. If the entry time is within the defined London session hours, tag A = London session; otherwise A = Other.
2. If the stop distance at entry is a single fixed number of pips and does not move until exit, tag B = Fixed-distance.
3. If the stop is adjusted after entry (for example, moved stepwise as price moves), tag B = Trailing.

Data (numerical scenario): Assume the pips result for each tagged trade is:

• London session + Fixed-distance: trades = [8, 6, 5, 7] → total = 31 pips
• London session + Trailing: trades = [3, -2] → total = 1 pip
• Other + Fixed-distance: trades = [4, 2, -1] → total = 5 pips
• Other + Trailing: trades = [6, -3, 1] → total = 4 pips

Calculations (simple averages):

• London session + Fixed-distance: average = 31/4 = 7.75 pips
• London session + Trailing: average = 1/2 = 0.5 pips
• Other + Fixed-distance: average = 5/3 = 1.67 pips
• Other + Trailing: average = 4/3 = 1.33 pips

What you can and cannot conclude from this example:

• You can observe that, within this scenario, the tag combination London session + Fixed-distance has the highest average pips.
• You cannot claim this will predict future trades, because this is one small sample and market conditions may differ.

Limitations and risks (including failure modes)

1. Small samples and variance. Ten trades can produce patterns that disappear when more data is added.
2. Mis-tagging. If two journal entries that should share a tag are labeled differently, the comparisons become unreliable.
3. Overfitting to past conditions. A tag may appear effective only under a specific market “regime” (for example, when volatility or spread behavior changes).
4. Outcome metric mismatch. Using only pips ignores other effects like costs, slippage, or whether risk per trade is comparable.
5. Cost and execution dependence. The same strategy features can yield different net outcomes when execution quality or trading costs differ.

A concrete failure mode in this worked example: suppose you accidentally label a trade as Fixed-distance when your rules should have tagged it as Trailing. That single error can move pips totals and shift the averages, making the “best” tag combination look better or worse than it is.

Verification and next questions

To independently verify the logic, you should:

• Re-state the tagging rules in plain language.
• Apply them to each journal entry and confirm the assigned tags match the rules.
• Recompute the same summaries (counts and averages) per tag combination.
• Repeat the exercise on a different time window to see whether the relative comparisons remain similar.

If you want to go further, the next question is often: should you compare tags individually (A only vs B only) or jointly (A+B combinations like in the example)? Joint comparisons can reveal interactions but also increase the number of groups and reduce sample size per group.