What Average Win Loss Means in Forex
Average Win Loss is a simple descriptive measure that compares the typical size of winning trades to the typical size of losing trades. In forex, it usually answers a question like: “When trades are winners, how large are the winners on average compared with losers on average?”
To keep the concept checkable, you need to separate the measurement from any interpretation. The measurement is mechanical: it uses recorded trade results. The interpretation is conditional: it depends on how you define wins and losses, how you measure the trade outcome, and what period and dataset you use.
A key point is that “average win loss” is not the same as predicting future results. It summarizes what already happened in the chosen sample. Historical relationships can weaken when market conditions, execution quality, or trading costs change.
A Simple Model for the Calculation
A practical way to define the metric is to compute two averages from the trade list:
- Average Win (AW): the mean outcome among all winning trades.
- Average Loss (AL): the mean outcome among all losing trades.
Then Average Win Loss is often expressed as one of the following ratios (different writers use different formats):
- AW/|AL|: the average win divided by the absolute value of the average loss.
- AL/AW or AW−AL: less common, but sometimes used depending on whether the goal is comparison or difference.
Because definitions vary, the most important “input” is your exact rule for calculating the win and loss amounts.
Define what “outcome” means
For each trade, you must decide what number you call the outcome. Common choices include:
- Net profit/loss after costs (preferred for realism when you have reliable cost data).
- Gross profit/loss before costs (can overstate results when spreads or commissions matter).
- Pips movement vs. account currency value (pips are not always comparable across position sizes).
Whatever you choose, use the same basis for every trade in the sample.
Use a consistent sign convention
Winning trades are typically recorded as positive outcomes, and losing trades as negative outcomes. When you compute AL, you often take the average of the negative numbers and then use the absolute value later (|AL|) so that a ratio like AW/|AL| stays interpretable.
Assumptions for examples
To make any example verifiable, you must state assumptions such as:
- Whether costs are included.
- Whether partial closes are treated as separate trades or as part of one outcome.
- Whether you exclude break-even trades (zero outcome) or keep them in the win/loss sets.
If you do not state these, two people can compute “average win loss” and get different answers from the same raw trading log.
Inputs, Outputs, and the Sequence to Compute It
Here is a checkable sequence that does not assume any specific platform or indicator.
Inputs
You need a dataset of completed trades, where each trade has at least:
- Trade outcome value (in consistent units).
- A label for win vs. loss based on your rule (e.g., outcome > 0 is a win; outcome < 0 is a loss).
- Optional but helpful: trade size or the basis you used to compute profit/loss.
Step-by-step output generation
- Select the time window and include only completed trades from that window.
- Split trades into two groups:
- Wins: outcomes above zero (according to your definition).
- Losses: outcomes below zero.
- Compute AW as the arithmetic mean of win outcomes.
- Compute AL as the arithmetic mean of loss outcomes.
- Compute Average Win Loss as a chosen comparison, commonly AW/|AL|.
What the number represents
If AW/|AL| is greater than 1, it indicates that average winning trades are larger than average losing trades in magnitude (under your measurement rule). If it is less than 1, average losses are larger in magnitude than average wins.
This does not indicate “good” or “bad” by itself. A strategy can have smaller average wins than average losses and still produce positive net results if the win rate is high enough, or if losses are controlled elsewhere. Average win loss is only one part of a larger picture.
Evidence Through a Worked Example (With Explicit Assumptions)
Assume the dataset contains 10 completed trades, and you measure each trade’s outcome as net profit/loss in the account currency, after spreads and commissions. Also assume:
- Wins are trades with outcome > 0.
- Losses are trades with outcome < 0.
- Zero outcomes (if any) are excluded from both averages.
Suppose the five winning outcomes are: 6, 4, 8, 5, 7 (in account currency).
- AW = (6+4+8+5+7)/5 = 30/5 = 6.
Suppose the five losing outcomes are: -3, -4, -2, -5, -4.
- AL = (-3-4-2-5-4)/5 = -18/5 = -3.6.
Then AW/|AL| = 6 / 3.6 = 1.67.
This value means: on average, wins are about 1.67 times as large as losses are in magnitude, using the stated net-outcome definition. A second person who calculates the same metric with different assumptions—like gross outcomes before costs, different inclusion of break-even trades, or pip-based outcomes with varying position sizes—may reach a different result even from the same underlying trades.
That is why the calculation is “mechanics,” while interpretation depends on assumptions.
Limitations and Common Failure Modes
Average win loss is useful for describing the shape of historical outcomes, but it has material limitations.
1) Sample size and instability
With few trades, AW and AL can move a lot due to random variation. The metric can look stable in one period and change significantly in another period.
2) Changing costs and execution
If spreads, commissions, or execution quality vary across time, and you use net or gross outcomes inconsistently, the metric can reflect cost differences rather than trade behavior.
3) Selection bias in what you include
If you exclude certain trades (for example, the ones executed under unusual conditions) or if you only log trades that “fit” a rule, the averages may not represent all outcomes.
4) Mixed risk models
If position sizing or risk rules vary by trade, then “average” outcomes may mix different exposure levels. The metric can then reflect changes in sizing more than changes in decision quality.
5) “Average” can hide distribution
Two datasets can share the same AW/|AL| ratio but have very different outcome distributions (for example, occasional large losses vs.