Definition
Round-trip analysis pairs every entry with its corresponding exit and treats the pair as a single transaction to be measured. For each round trip the analysis records the entry timestamp and price, the exit timestamp and price, the side (long or short), the size, the holding period, and the realised profit or loss net of costs. The result is a transaction-level dataset that supports a different family of diagnostic questions than a daily-returns dataset can answer.
The pairing logic is the crux of the technique. In simple strategies one entry is matched to one exit. In layered strategies that scale into and out of positions, fills must be matched using a convention — first-in-first-out, last-in-first-out, or weighted average — and the choice affects every downstream statistic. Once the pairing is established the analysis aggregates across trades: win rate, average winner versus average loser, holding-period distribution, profit per trade, and the slugging ratio that combines them.
Why it matters
How it works
The pipeline starts with a raw fill blotter — every executed order with timestamp, instrument, side, price, and quantity — and ends with a round-trip table. Building the table means walking the blotter chronologically per instrument, maintaining an open inventory, and emitting a closed round-trip every time the inventory returns to zero (or, for partial exits, every time the matching convention triggers a close). Each round trip records all the standard fields plus realised P&L, holding period, and any signal-attribution metadata that was attached to the originating order.
Once the table exists the diagnostic questions answer themselves. The histogram of trade P&Ls shows whether the strategy edge is broad or concentrated in tail outcomes. The win rate versus average win-to-loss ratio plot reveals whether the strategy depends on being right often or on letting winners run. The holding-period distribution exposes whether stop-losses, signal decay, or execution quirks are dictating exit timing. Aggregating by signal, by time of day, or by instrument turns the same table into an attribution surface that drives the next round of strategy iteration.