Your strategy as a statistical experiment
Any rule-based trading strategy can be described by a few simple statistics:
- the probability of a winning trade,
- the probability of a losing trade,
- the average size of a win,
- the average size of a loss.
From these numbers we can compute the expectancy of the system – its average profit or loss per trade over a very large number of trades. For example, a system that wins 60% of the time with 1R profits and loses 40% of the time with 1R losses has a positive statistical edge. On paper, over thousands of trades, it should make money.
But in the real world we never see “infinite trades”. We see limited samples: 10 trades this month, 40 trades over the last quarter, maybe 100 trades in a full year. Inside those samples, randomness can hide or even invert the true edge of the system. That is exactly what I wanted to show with the simulation behind these slides.
The experiment: random systems vs a system with an edge
I built a small program to simulate equity curves under two conditions:
- Random systems: equivalent to flipping a fair coin. Each trade has 50% chance to win 1R and 50% chance to lose 1R.
- System with real edge: the same structure and risk, but with a 60/40 advantage – 60% wins, 40% losses. This is the blue equity curve in the charts.
For each simulation, the program generates around 30 equity curves starting from the same initial capital. Grey lines represent purely random systems. The blue line represents the system with a real mathematical edge.
In a perfect world, the blue line would always be above the grey ones. In reality, when we look at small samples, the picture is very different.
Looking at only 25 trades: randomness dominates
In the first chart you see about 25 trades per curve. The blue equity line – our “good” system with 60/40 odds – does not clearly dominate. In fact, some of the random systems end the sample above it. If you didn’t know which line had the edge, you might even conclude that the blue system is “worse” than some random ones.
This is exactly what happens when traders judge a strategy after 10, 15, 20 trades. A perfectly valid system can look terrible, while a random or overfitted system can look brilliant. The sample is simply too small to separate skill from luck.
This is why system-hopping is so dangerous. If you abandon a strategy every time you hit a short losing streak, you will never stay long enough to let the edge express itself.
At 50 trades: the edge begins to appear (but still isn’t obvious)
In the second chart, we extend the simulation to 50 trades. Now the blue curve starts to show some improvement. On average, it tends to drift higher than many random curves, but there are still random systems that finish above it. The picture is better, but still messy.
This is similar to what you might see after a few months of consistent trading. Patterns begin to emerge: you sense that your method “probably” has something, but the noise is still strong. If you look only at equity, you might not be fully convinced yet.
At this stage, good traders rely on their process metrics (expectancy, R-multiple distribution, risk per trade) and on solid backtesting, not just on how the equity curve “looks”.
At 100 trades: the law of large numbers does its job
In the third chart, we go out to 100 trades. Now the magic of large numbers starts to become visible. Most of the random equity curves hover around break-even or drift slowly downward. The blue curve – the system with a real 60/40 edge – increasingly stands out, rising clearly above the pack.
It’s still not a straight line; there are drawdowns and noisy sections. But the direction is clear: as the number of trades grows, the realised results of the edge system converge towards its theoretical expectancy. Random systems, on the other hand, tend to oscillate around zero or gradually erode capital.
This is what the Law of Large Numbers tells us: as the number of observations increases, the average outcome tends to the true expected value. In trading terms: the more quality trades you execute with the same rules, the more your actual performance will resemble the mathematical edge of your system.
What this means for your trading
There are several practical lessons hidden in these charts.
1. Don’t judge a strategy on 10–20 trades
Ten trades is nothing. Even twenty-five is almost nothing. You can flip a coin ten times and easily get seven heads or seven tails in a row. In the same way, a strong trading system can start with a losing streak, and a weak or random system can start “on fire”.
If your process is sound and your backtesting is rigorous, give the system enough room – often at least 50–100 trades – before making big conclusions. That does not mean you ignore obvious structural problems, but it does mean you stop overreacting to normal variance.
2. Large samples protect you from illusions
Many “edges” on social media are just statistical noise: they look incredible on 20 or 30 historical trades, but collapse when you test them over hundreds of examples or across different market regimes. Forcing yourself to evaluate systems on larger samples is a way to defend yourself from these illusions.
In my own work, both with manual strategies and with bots, I never trust a pattern until I have tested it on a sufficiently large dataset and seen how it behaves in different years, different assets and different volatility environments.
3. Risk management still matters
In the slides, we ignored commissions, spread and slippage to keep the example simple. In real trading, these frictions reduce your edge, especially on high-frequency or very short-term strategies. A 60/40 system with 1R wins and 1R losses may become closer to 55/45 once you include trading costs.
This makes it even more important to think in large numbers. A small edge applied over thousands of trades can still produce meaningful results, but only if your risk per trade is controlled and your position sizing is consistent.
4. Process over outcome
Once you accept the role of large numbers, you naturally shift focus from “Did I win this trade?” to “Did I execute my plan correctly?”. A single trade is almost pure noise; one hundred well-executed trades are a signal. Your job is not to make every trade win, but to give your statistical edge as many fair chances as possible.
Conclusion: trust the math, not the noise
The equity charts from this experiment tell a simple story:
- Over very few trades, randomness completely dominates. A losing system can look brilliant and a winning system can look broken.
- As the number of trades grows, a real edge separates from random behaviour.
- To trade like an engineer, you must design your strategy based on data, and then judge it over large numbers, not isolated streaks.
In other words: if you have a verified edge, your best friend is not the next trade – it is the next hundred trades. Your mission is to survive long enough, with consistent risk and discipline, for the law of large numbers to work in your favour.
