ARMFA calculator of trade expectancy
The calculator of trade expectancy answers the question of whether the strategy has an edge. The simulation answers a second, no less important one: how different the results can be with one and the same edge. The five paths on the chart are built by identical rules — and they look nothing alike.
The expectancy simulator over distance
2,000 runs over the given distance, the seed is fixed. On the chart are the first five runs: identical rules, different paths. The outcome over the distance is not a return forecast: the model treats the win rate and the ratio as unchanged across all trades, and growth as compounding from current equity. In real trading both parameters change with the market, so the actual result is lower than the model one and the spread is wider.
How expectancy and the spread of outcomes are calculated
| Value | Formula or method |
|---|---|
| Expectancy per trade | win rate × (R/R − costs) − (1 − win rate) × (1 + costs) |
| A run | N trades, risk from current equity, outcome by the win rate |
| Median outcome | the middle of the distribution over 2,000 runs |
| Spread 10–90 % | the tenth and ninetieth percentiles of the outcomes |
| Runs in the red | the share of runs that finished below the start |
| Median drawdown | the middle of the distribution of maximum drawdowns |
Costs are counted twice and differently: they reduce the win and increase the loss. That matches reality — the spread is paid in both situations, not only on profitable trades.
The trade series simulator: why the spread matters more than the average
A positive expectancy does not mean that every stretch of two hundred trades will be profitable. The simulation shows the distribution: for a system with an expectancy of +0.3 R a noticeable share of runs ends in the red simply because of the order in which the wins and losses arrived.
What follows from this in practice
- A losing quarter does not mean a broken system
- It falls within the normal spread. The decision to change the rules is taken on expectancy over a long sample, not on the result of a stretch.
- A profitable quarter does not confirm the system
- Symmetrically: systems with zero expectancy also end up in the upper part of the spread. Only distance confirms anything.
- The median drawdown is the expected one, not the worst
- Half of the runs pass through a deeper drawdown. Planning should be done from the upper part of the distribution, not from the median.
How to use the calculation. Substitute your own win rate, ratio and costs from the journal rather than the planned values. Then look at the share of runs in the red: that is the probability that the next two hundred trades will disappoint you with a perfectly healthy system.
How many forex trades are needed for an edge to become visible
The spread narrows slowly — roughly as the square root of the number of trades. In practice this means that a conclusion about a system drawn from a quarter of trading is almost always premature.
| Distance | Share of runs in the red at an expectancy of +0.1 R | at +0.3 R | What can be claimed |
|---|---|---|---|
| 50 trades | about 29 % | about 7 % | Nothing: the spread covers the difference |
| 100 trades | about 25 % | about 2 % | The first cautious conclusions |
| 200 trades | about 19 % | less than 1 % | The edge is visible if it exists |
| 500 trades | about 9 % | less than 1 % | An estimate you can rely on |
The shares are given as an order of magnitude from simulations with the same assumptions as in the widget: a constant win rate and ratio, and the risk share taken from current equity. The exact values depend on the bet per trade and are checked with the sliders.
Related calculations. The depth of the drawdown along the way is shown by the losing streak calculator, the time to climb out of it by the drawdown calculation, and the upper bound of the bet for the edge you found by the Kelly share.
Frequently asked questions
Why are the runs so different under identical rules?
Because the outcome of a single trade is random. The order of wins and losses changes both the final result and the depth of the drawdowns along the way. That is exactly why one quarterly result says almost nothing about the quality of a system.
How many trades are needed for an edge to show?
It depends on the size of the expectancy: the smaller it is, the longer the distance has to be. At an expectancy of +0.3 R the spread becomes noticeably narrower after a few hundred trades — you can see it by increasing the distance with the slider.
Does the simulation account for a changing market?
No, and that is its main limit. The model assumes a constant win rate and ratio. A real system lives in changing conditions, so the result of a simulation is a lower bound of the uncertainty, not a full estimate of it.
Why is the outcome shown as a multiple rather than a percentage at high risk?
Because compounding over a long distance gives values like «+2,700,000 %», which read as a promise of returns. That is an artefact of the assumption of unchanged parameters, and a multiple shows the scale of the model result more honestly.
What does the 10–90 % spread mean?
The bounds between which eight runs out of ten ended up. The outer ten percent on each side are discarded as rare outcomes.
Why is the share of losing runs not zero with a good expectancy?
Because of the order of the trades. Even with an edge, part of the paths starts with a losing streak and does not have time to recover within the allotted number of trades.
How many runs are enough for a stable result?
Two thousand is enough for the median and the percentiles to stop changing noticeably on a repeat. Increasing it further refines the tails of the distribution but does not change the conclusions.
Which inputs should be taken from the journal?
The actual win rate, the average ratio on closed trades and the costs as a share of risk. Planned values from a tester give an optimistic picture.
Why is the generator seed fixed?
For reproducibility: identical inputs must give an identical answer. With a random seed the result would jitter after every touch of a slider, and there would be nothing left to compare two settings by.
How do you know the distance is not enough?
By the share of runs in the red: if it stays high at your expectancy, the result over that stretch is decided by chance and it is too early to draw conclusions about the system.
Read next
