Trade Duration Analysis for Reviewing Winners and Losers
An average holding time can hide a lot. One trade may have closed most of its position within minutes while a small remainder stayed open for hours. Another may appear as several completed trades because the report counts each partial exit separately.
Trade duration analysis starts by deciding what counts as one trade and when its clock starts and stops. Once those definitions are consistent, you can compare winners and losers, inspect duration groups and see which exits deserve a closer review.
This guide covers that measurement workflow. For the behavior behind keeping losing positions open, read Are You Holding Losers Longer Than Winners?.
Decide when the trade clock starts and ends
For a single-entry position closed in one exit:
Trade duration = exit timestamp − entry timestamp
A position entered at 09:40 and fully closed at 10:25 lasts 45 minutes. Use actual execution times rather than the time an unfilled order was submitted.
Store timezone-aware timestamps and calculate elapsed time on a consistent basis, such as UTC. A position held over a weekend includes those hours in elapsed duration. If you measure only market-open hours, label that as a separate measure.
Partial exits need an explicit convention. For this guide, position duration runs from the first fill that opens a position to the fill that closes its last remaining unit.
If 80% closes after ten minutes and the last 20% closes after four hours, position duration is four hours. For that single-entry position, quantity-weighted holding time provides a different view:
Weighted holding time = (0.80 × 10 minutes) + (0.20 × 240 minutes) = 56 minutes
The two figures answer different questions: how long the position existed, and how long its average unit was held. Keep them separately labeled.
For scale-ins, retain each fill’s quantity and execution time, and document how entries and exits are grouped. Average prices and a single pair of opening and closing dates cannot reconstruct every leg’s holding time.
Build a comparable sample
Choose one setup and strategy version. Keep intraday and swing trades apart. Separate markets, sessions or long and short positions when they have materially different trading rules.
Define the review period too. A closed-trade review can include positions closed during the month even if they opened earlier. Excluding those positions would remove some of the longest holds. State whether your date filter uses opening time, closing time or both.
Review open positions separately. Their final duration and result are unknown, and a long-running open loss will be absent from the completed-trade table.
Use consistent trade tags and flag missing timestamps, duplicate imports or uncertain grouping before calculating averages.
For this review, classify outcomes using net P&L after the costs you can reconcile:
- Winner: net P&L above zero
- Loser: net P&L below zero
- Break-even: net P&L equal to zero
Keep the same precision and cost treatment throughout. A trade that closes at its entry price may still be a net loser after commissions or financing charges. Check which costs are already included so you do not deduct them twice.
Compare duration groups with net outcomes
Here is a hypothetical sample of 60 completed positions from one intraday setup. Each position counts once. Results include costs and use R, where 1R is that trade’s original planned risk amount. Keep that denominator fixed; flag missing initial-risk records rather than guessing them.
| Completed position duration | Trades | Total net R | Average net R |
|---|---|---|---|
| Under 15 minutes | 18 | +6.3R | +0.35R |
| 15 through 60 minutes | 22 | +4.4R | +0.20R |
| Over 60 minutes | 20 | −10.0R | −0.50R |
The trade-level results sum to +0.7R. Account return also depends on the dollar risk and capital behind each trade. The longest-duration group deserves inspection, but the table cannot show what would happen if every position closed at 60 minutes. Closing earlier would use different prices and could change both winners and losers.
Choose duration bands that fit the setup’s intended behavior before looking for the best result. Keep the boundaries explicit: a trade lasting exactly 60 minutes belongs in the middle group here.
Within each duration group, record the number of winners, losers and break-even trades, plus average win and average loss after costs. A high win rate can coexist with a negative result when losses are much larger than gains.
Inspect the largest outcomes too. One unusually long loss can dominate a small group. How many trades are needed to evaluate a strategy explains why a trade count alone cannot validate a conclusion.
Check which trades drive the timing difference
Compare the average and median duration of winners and losers within the same setup. The median is the middle duration after sorting; when there are two middle values, average them.
Suppose five hypothetical losing positions lasted 10, 15, 20, 25 and 230 minutes. Their mean is 60 minutes, while their median is 20 minutes. The unusually long position explains much of the difference. Keep it in the record and inspect its fills, notes and planned exit.
Look for the specific reason a trade stayed open. Was it following a planned multi-hour condition, waiting for a partial fill, or being held beyond the original exit rule? Keep winners that broke the rule in the review as well. Outcome and rule adherence are separate classifications.
Test a time exit with its own price data
A duration table cannot calculate an alternative exit price. To test a 60-minute exit, specify the rule first: which positions qualify, what happens at minute 60, and which executable price the simulation uses.
For example, a hypothetical test might close a position at the first executable price after 60 minutes if it remains open and a predefined progress condition has not been met. Apply it to every eligible position, including those that eventually became winners. Positions already closed before the decision point retain their earlier exits.
Account for costs, partial exits and execution assumptions. Compare the original and candidate results across the same starting sample, and consider whether earlier exits would change later entries. Historical data may lack the bid, ask or liquidity needed to reproduce a realistic fill.
Keep track of the alternatives tested. Research on backtest overfitting explains why selecting the strongest result from many historical trials can produce a false discovery. Later, unused data can help challenge a candidate; one successful validation still does not establish future performance.
Bring the review into UltraTrader
UltraTrader’s Time Metrics view includes win rate by trade duration and average holding-time comparisons for winners versus losers and longs versus shorts. Use those views to identify the records worth examining.
For a custom calculation, check the underlying data before building a worksheet. UltraTrader’s published export field list includes Start Date, Close Date, Notes, Strategy and PNL. Check what your downloaded file actually contains and whether its timestamps, grouping and cost treatment match the review. A summary export may need supporting execution records for partial-exit calculations.
Retain the setup, entry and exit timestamps, initial risk, net result, planned time condition and actual exit reason. Calculate any additional measures separately and keep their definitions beside them.
At the next weekly trading review, choose one duration group and trace its results back to the trades. Finish with a specific observation or testable question, rather than a universal holding-time rule.
Educational content only. This is not investment advice. Trading involves risk of loss. Historical analysis and hypothetical examples do not guarantee future performance.