LEARN · EN · easyquanttrading.com

The anatomy of a trading strategy: the five parts most people forget

Ask someone to describe their strategy and you will usually hear an entry condition: buy when this crosses that. That is one part in five. The other four decide whether the strategy makes money, and they are the ones people leave implicit — which means they never get tested.

Part one: the universe

The universe is what you are allowed to trade. It sounds trivial and it is not. A rule that works on major currency pairs may be hopeless on a thinly traded commodity, and a rule that works on one index may fail on another because of differences in how the underlying behaves.

Worse, the universe is where survivorship bias enters. If your list of instruments is the ones that exist today, you have quietly excluded everything that failed. For a strategy that trades a single instrument this does not matter much; for anything that selects across a list, it is the first thing to check.

Part two: the entry

This is the part everyone can describe and the part that gets optimised to death. An entry rule is a condition on data that is available at the moment of the decision. Everything else in this article exists to make sure that condition means something.

The most common defect in an entry rule has nothing to do with the logic. It is ambiguity. 'Buy when momentum turns up' cannot be tested, because two people will code it two ways and get different results. 'Buy when the 14-period RSI crosses above 50 on a closed bar' can be tested, and disagreed with, which is better.

Part three: the exit

Exits come in two flavours and most strategies combine them badly. A signal exit closes the position when a condition is met — an opposite signal, a trailing stop, a time limit. A target exit closes it at a predetermined price.

Which one you choose changes the distribution of your trades more than most entry tweaks do. Target exits produce high win rates with small wins and occasional large losses. Signal exits produce lower win rates with larger average wins. Both can be profitable; mixing them without knowing which does what usually produces the worst of both.

Exit typeTypical win rateTypical shapeMain risk
Profit targetHigherMany small wins, few large lossesOne big loss erases many wins
Signal exitLowerFewer, larger winsLong flat periods with no progress
Time-basedDependsMechanical, easy to testIgnores whether the idea is playing out
Trailing stopMidLocks in gains, cuts trends shortWhipsaw in choppy markets

The exit determines the trade distribution. Choosing it deliberately is most of the work.

Part four: the sizing

Sizing converts a signal into a position. The two common approaches are fixed size and risk-based size. Fixed size is easier to backtest and easier to blow up, because it does not shrink after losses. Risk-based size keeps the loss per trade roughly constant and makes the equity curve smoother, at the cost of complexity.

The part people forget is rounding. If your calculated size falls below the broker's minimum, you cannot take the trade at the intended risk — you either take it larger or you skip it. On a small account this happens constantly, and it means the live strategy is not the strategy that was backtested.

Part five: the limits

Limits are the rules about when you stop. Maximum positions open at once. Maximum daily loss. A circuit breaker after a set number of consecutive losses. Minimum time between entries.

These are not risk-management decoration. On many strategies, a single limit — such as a maximum of one position per instrument — changes the drawdown materially while barely touching the return. They are also the part most often omitted, because they feel like constraints on a system that is supposed to be finding opportunities.

Why writing it down matters

A strategy that cannot be written down completely cannot be tested, and a strategy that cannot be tested cannot be trusted. The five parts above are the checklist: if any of them is 'I decide at the time', then that part is being executed by judgement, and the backtest you ran describes a different system.

This does not mean judgement is bad. It means judgement should be a deliberate override with its own record, not an invisible component of a strategy you believe you have measured.

FAQ

Can a strategy be profitable with only an entry rule?
In a backtest, sometimes — usually because the tester closes positions at the end of the test and the implied exit happens to be favourable. Live, you need an exit and a size, so the question answers itself.
Which part matters most?
For whether the strategy makes money, the entry and the exit together. For whether you can actually run it, the sizing and the limits. The failures that end accounts are almost always in the second group.
How specific does an entry rule need to be?
Specific enough that a programmer who dislikes you would implement it the same way. If there is a judgement call in the definition, it will be made differently under pressure than it was in the backtest.

More guides

Not investment advice. Historical results do not guarantee future performance. EasyQuant is a research factory — you execute on accounts you control.