Mechanism: what “Average Win Loss” measures
Average Win Loss (often shortened to AWL) describes the average size of positive outcomes versus negative outcomes in a set of trades. The exact meaning depends on the rules used to label outcomes.
A trade is typically counted as a win if its realized result is above zero (or above a chosen threshold) and a loss if it is below zero (or below a chosen threshold). The “average win” is the mean size of wins; the “average loss” is the mean size of losses (sometimes reported as a negative number, sometimes as an absolute magnitude).
Because you must define the win/loss label and how trades are closed, the timeframe becomes central. If you observe outcomes after different holding periods—minutes versus days, or within different calendar windows—you are changing both:
- the holding period used to realize results (when you decide the trade outcome), and
- the period in which price paths are sampled (how much movement and variation a trade has time to experience).
How timeframe changes win/loss statistics
Timeframe affects AWL through several general, mechanically stable channels.
-
More time allows more divergence With longer holding periods, a position has more opportunity to move away from entry. This can increase both the chance of larger wins and the chance of larger losses. In other words, the distribution of outcomes can widen over time, even if the average long-run drift is unchanged.
-
Outcome timing changes what you call “win” If you shorten the measurement window, some trades that would eventually turn into wins may be counted as losses if your rules stop them early. Conversely, delaying measurement can reclassify outcomes when a trade is allowed to recover.
-
Costs and microstructure matter more on short horizons On short timeframes, small frictions (like trading costs) can represent a larger share of realized results. That can shift average win and average loss, especially when many trades have small raw price movement.
-
Regime changes and rare events matter more on longer horizons On longer timeframes, markets can enter different regimes (for example, higher volatility versus lower volatility). This can make AWL reflect a mixture of conditions rather than a single stable environment. Rare but impactful events can also change the averages.
Example scenario with explicit assumptions
Assume a simplified process where trades are held for either 1 hour or 5 days, and each trade’s outcome is recorded at the end of its holding period. Also assume the same entry method is used in both cases, but the exit time differs.
- Over 1 hour, many trades may close after limited movement. Some trades will end slightly positive and some slightly negative, so average win and average loss may both be relatively small.
- Over 5 days, the same trades may experience broader movement. Some may reach much larger positive results; others may move farther negative. So the averages can spread apart.
This illustrates the key point: changing holding period changes what portion of the price path becomes realized, which changes the computed AWL.
Limitations, failure modes, and what you can verify
Timeframe can make AWL look better or worse without any change in underlying skill, because measurement rules change.
Material limitations / failure modes
- Selection bias from the measurement rule: If win/loss labeling depends on holding time, AWL comparisons across timeframes may be comparing different “experiments.”
- Cost sensitivity: If costs are not accounted for consistently, short-horizon AWL can be misleading.
- Non-stationarity: Historical relationships between timeframe and AWL do not guarantee future behavior.
How to verify using only general facts A reader can verify the timeframe effect by checking that all of the following are consistent when comparing AWL across holding periods:
- the definition of a win and a loss (thresholds and zero-point),
- the exit rule that finalizes the realized result,
- the cost treatment used in the dataset (at least conceptually, even if exact figures are not available), and
- the sample coverage (for example, whether both timeframes cover similar market conditions).
Control point (important question) Ask: *Are you comparing AWL across timeframes using the same classification and exit logic, only changing the holding period?