Is trading gambling or skill? How to tell if your results are just luck
10 min read · Published 28 Oct 2026
Gambling is a game with a negative expected value, where you have no influence over the odds. Trading without an edge after costs is mathematically similar. Before costs it averages zero; after costs it loses, on average, what a trade costs. Trading with an edge after costs is not gambling in this sense. So the question is not “whether”, but “does your series differ from chance?” You can check that with numbers.
Abbreviations: R is a trade’s result in units of risk; −1 R means a loss equal to the distance to the stop. Expected value is the average result of one trade over a long series. An edge is a positive expected value after all costs. p is the share of draws in which chance alone produced a result at least as good.
The experiment: 1,000 traders with no edge
We simulated the results of 1,000 “traders”, none of whom has an edge. Each trade ends with a 1.5 R profit with a 40% chance, or with a 1 R loss. Before costs that is exactly zero: 0.4 × 1.5 − 0.6 × 1 = 0. We deducted a cost of 0.05 R from every trade, so the expected value is −0.05 R. It is a game as fair as a coin toss, only with a fee on every trade. Each “trader” first made 50 trades and then another 200.
After 50 trades, 347 of the 1,000 were in profit; 109 averaged at least +0.2 R per trade, and 14 at least +0.4 R. The best one won 31 of 50 trades, or 62%, and made 25 R, an average of +0.50 R.
Then we checked what the top 10% did next, that is, the 100 best after the first 50 trades:
| the 100 best after 50 trades | first 50 | next 200 |
|---|---|---|
| average per trade | +0.28 R | −0.06 R |
| win rate | 53% | 40% |
| in profit | 100 of 100 | 18 of 100 |
The other 900 averaged −0.05 R per trade over the next 200 trades. So the best had no edge over the rest, only a better start. After 250 trades, 237 of the 1,000 were in profit, almost one in four. Each paid 12.5 R in costs over that time, and the average result was −12 R.
What the calculator showed. We ran the best “trader’s” first 50 trades through the code of the “Luck or edge?” calculator, the same as on the site. Verdict: “Edge likely”, average +0.50 R, 90% range from +0.20 to +0.80 R. The calculator put the chance that the true average is positive at over 99%. Without the 3 best trades, +20.7 R was left. The “How many trades needed” box showed ≈ 25 and the label “Sample is large enough for this average”. Over the next 200 trades, the same “trader” averaged −0.04 R per trade and got the verdict “No edge in this data”.
The calculator’s arithmetic was correct. For a single series checked without picking, it is rarely wrong. After 50 trades, 45 of 1,000 traders with no edge got the verdict “Edge likely”; after 250, only 9 did. Over more draws, the share after 50 trades is about 3%. But if you pick the best of a thousand, you land on exactly one of those 45. One of them kept the verdict after 250 trades, because that trader got lucky again over the next 200.
Four questions that separate an edge from luck
1. How many trades do you have, and how wide is the range? An average of 50 trades is mostly noise. The “trader” in 100th place had +0.20 R, but a 90% range from −0.10 to +0.50 R, so it included zero. The calculator gave it the verdict “Not known yet” and showed how many trades you need to decide: ≈ 160. Over the next 200 trades, this “trader” lost 20 R.
2. Does the result survive without the best trades? One of the gold signal channels we scored made +0.068 R per signal over 89 signals. Without the five best signals, the average fell to −0.043 R (chapter 45). The calculator shows this in the “Without the best 3” box. This test catches a result that rests on a few lucky trades, but not luck spread evenly. Our best “trader” passed it easily.
3. Does the result beat random entries at the same cost? A placebo uses the same exits and the same cost, but random entry times. Indicator arrows that we tested on 1,293 entries made −0.30 R per trade, and random entries −0.25 R (chapter 5). So the signal was worse than chance, and the cost did the rest of the damage. The calculator cannot draw a placebo for you. But you can paste the results of random entries into it and compare them with yours.
4. How many versions were tested before this one was picked? We went through 1,344 variants of simple rules. Of the 1,273 variants with at least 100 trades, 268 ended in profit, or 21%. Meanwhile, the average trade across the whole family made −0.060 R. The average result of the 13 best shrank by about three quarters in the following years (chapter 7). That is the cost of picking the best of many, the core of overfitting. The calculator only sees the series you show it. If it is the best of many versions, its range is too optimistic, as it was for the best of our 1,000.
What typical guides don’t tell you
A high win rate is not an edge. A gold signal channel with over five years of history had a 47% win rate. Over 1,080 signals it made −0.012 R per signal, that is, zero (chapter 45). Our best “trader” won 62% of trades and had no edge.
The share of traders in profit does not tell you who has an edge. In the experiment with no edge, almost one in four was in profit after 250 trades. On top of that, results lists often show only those who survived. On a list of signals for copying, 11 of 1,920 entries met the basic quality conditions. The signals that had blown up were no longer on the list (chapter 44).
An edge can be real and still too small. We tested carry: buying currencies with a high interest rate against ones with a low rate. The rule picked the sides of pairs better than chance: placebo p = 0.004. A random choice did at least as well in only 4 out of 1,000 draws. But the strategy earned 2.9% a year, and its drawdown, the fall from a peak, reached 37.6%. After cutting the risk to a bearable level, 0.2% a year was left (chapter 38).
The same tests can also confirm a result. One signal for copying, over 978 entries and after our cost, had a sum of profits 1.88 times the sum of losses. A test that resampled whole days gave p = 0.0005 against a threshold of 0.05. So the result was real (chapter 44). An edge exists, but you have to calculate it rather than see it on a screenshot.
Check it yourself
The “Luck or edge?” calculator. Load a report from your trading platform, paste results in R in the “Paste results” tab, or choose the “From journal” tab. Read the result in this order:
- verdict and 90% range: “Edge likely” appears only when the whole range lies above zero; below 20 trades the verdict is “Too few trades”;
- “Without the best 3”: whether the total rests on a few trades;
- “How many trades needed”: how many trades it takes to tell your average from zero at your standard deviation;
- “Longest losing streak” and “Drawdown”: whether they are typical for your win rate and for random orders of the same trades.
Before you believe the verdict, answer question 4. A series picked after a good spell, or the best of several versions, overstates the result. In that case “Edge likely” only means it is worth collecting new trades, without changing the rules again.
The trading journal. If the journal on this device has trades with a result in R, the calculator loads them by itself instead of the example. In the “What to test” box you choose “Whole history” or a single strategy. The calculator skips trades without R and tells you how many. In the journal’s “Strategies” tab, record every change of rules as a new strategy. The journal itself reminds you that changing the rules midway means counting from zero. That way you know how many versions you have tested, and you judge each one separately.
Questions
Is trading gambling in the eyes of the law?
That is a legal question, and the answer depends on the country. Usually, games of chance and dealing in financial instruments fall under separate rules. We do not give legal or tax advice. This article is about the mathematics of results.
How many traders make money?
We have no reliable figure for all markets, and we do not repeat other people’s estimates. In the EU and the UK, CFD providers state in their risk warning the share of retail client accounts that lose money. That figure does not tell you who has an edge, though. In our experiment with no edge, almost one in four was in profit.
How many trades until I know I have an edge?
It depends on the size of the edge and how widely results scatter. The calculator uses the formula (2 × SD ÷ average)², where SD is the standard deviation of results. With an SD of 1.22 R, as in the experiment, +0.2 R needs about 150 trades and +0.1 R about 600. If the series is the best of many versions, only new trades, recorded after the choice, can settle it.
Does a high win rate mean an edge?
No. The win rate at which you end up at zero depends on the ratio of win to loss. With a 1.45 R win and a 1.05 R loss, it lies at a 42% win rate. With a 0.5 R win and a 1 R loss, you already need 67%. What counts is the average result in R after costs, not the share of winners.
Educational content, not investment or legal advice. Past results do not guarantee future results. The coin experiment is a simulation with an assumed distribution of results, not the history of real accounts. The numbers from our tests describe our reconstructions of strategies, on our data and at our costs.