Long-short attribution separates a backtest into its long and short trades so you can see whether the result is carried by one direction. For a prop trader, that matters because a strategy that looks balanced in its final equity curve may be relying on a market condition that only favoured one side.
The aggregate return cannot answer that question. It nets together trades that may have had different entry logic, holding periods, costs, and drawdown behaviour. A direction split does not prove that either side has an edge. It tells you where to look before you call the whole system robust.
What long-short attribution actually measures
Long-short attribution is a reporting split, not a new trading signal. Each completed trade is assigned to the direction opened, then the same measurements are calculated for the long book and the short book separately.
The split is most useful when it preserves the original test rules. Do not give the long side a different cost model, omit open equity from one side, or quietly exclude a difficult period. Those choices turn a diagnostic into another optimisation exercise.
| Review item | What it reveals | What it cannot prove |
|---|---|---|
| Net result by direction | Whether either side carries the aggregate outcome | Whether the outcome will repeat |
| Trade count and holding pattern | Whether the comparison rests on similar opportunity sets | That more trades mean better evidence |
| Drawdown and floating equity | Which direction put the account under pressure | That final closed profit captured the worst path |
| Results across market conditions | Whether a direction depended on a particular environment | A permanent regime forecast |
| Costs by direction | Whether execution changes the apparent advantage | That the same costs will apply at every broker |
Keep the direction label from the order itself. A short trade closed for a loss does not become a long observation because the market later rose; attribution describes the decision made at entry, not a hindsight opinion about the market.
The combined curve can hide a one-sided dependency
A final curve is a summary, not an explanation. A strategy may earn most of its result from long trades while short trades contribute little or repeatedly deepen the drawdown. The reverse can happen in prolonged risk-off or declining markets. Neither pattern is automatically wrong. It becomes a problem when the trader has not noticed it.
That distinction matters in a prop evaluation because the account experiences the sequence, not the final chart. A weak side can create a concentrated run of losses or floating pressure even when the other side eventually lifts the total result. The question is not whether a short book is permitted to lose. It is whether the portfolio and its account constraints were tested with that weakness visible.
Losing streaks in prop backtesting help examine the sequence of adverse trades. Market-regime backtesting adds the environment in which the direction split occurred. Put those reports beside the attribution table rather than treating any single view as a verdict.
Make the comparison fair before drawing a conclusion
The cleanest comparison uses the same ledger for each side. That means the same date range, symbol universe, entry and exit rules, position-sizing rule, spread, commission, swap, slippage assumption, and treatment of overlapping positions.
Inspect the account path, not just closed trades
A prop firm may measure a constraint on equity rather than on a tidy list of closed positions. If a directional book carries open losses while another book is profitable, the combined closed-trade total can be reassuring at exactly the wrong moment. Preserve floating equity, concurrent exposure, and the timestamps of any account-rule breach in the directional review.
The funding model is useful for keeping the equity path beside the account constraint. The prop firm remains the authority for its current rules and measurement conventions, so record the rule source used for any specific evaluation rather than reconstructing it from memory.
Do not make a losing side disappear after the fact
Removing shorts because they impaired a historical curve is a strategy change, not a reporting improvement. It may be a valid hypothesis to test, but it must be stated before the new result is judged and checked on data that was not used to choose the change.
The same warning applies to filters that happen to remove mostly one direction. A trend filter, a session filter, or a volatility threshold can alter directional exposure. Report that effect plainly. Otherwise a supposedly neutral filter can conceal a decision to trade only the historically flattering half of the sample.
A practical attribution ledger
Before deciding whether to keep, change, or remove a directional rule, make a ledger another trader could audit. It should include the following fields for the combined strategy, the long side, and the short side:
- the test window and the data source;
- entry, exit, sizing, and cost assumptions;
- each trade's direction, timestamps, realised outcome, and exit reason;
- the open-equity path and concurrent exposure;
- results by declared market condition; and
- the untouched sample used to challenge any proposed change.
This is deliberately less exciting than a headline return. It is also more useful. It distinguishes a system that has a lopsided historical result from a system whose directional dependence has been tested and accepted as a stated limitation.
realbacktesting is a trading-software studio for cTrader built around inspectable, reproducible tests. Its published methodology makes the data, execution, cost, and validation assumptions visible. That does not convert a backtest into a live track record. It gives a trader something more defensible than a screenshot: a result whose directional dependence can be challenged.
Frequently asked
Should long and short trades always be reported separately?
They should be separated whenever direction could plausibly change the strategy's behaviour or account risk. The split is a diagnostic; it does not require the two sides to have identical results.
Does a profitable long side prove that it has an edge?
No. A profitable historical segment is evidence to investigate, not proof of a repeatable advantage. It still needs a clear rule set, realistic costs, and a check on untouched data.
Can removing the weaker direction improve a strategy?
It can change the historical result, but that is a new hypothesis rather than proof of improvement. Define the change before judging it and test it outside the sample that suggested it.
The stubborn takeaway is simple: a combined equity curve can tell you that a strategy made money; long-short attribution tells you which side you are actually trusting.