Skip to main content
What Does Not Work, and the Data

The Role of Luck

Pomegra Learn

The Role of Luck

A trader places 50 trades over six months using a technical analysis strategy. The strategy is right 54% of the time. After costs and fees, they earned a 12% return. They feel confident in their method and tell friends about their winning strategy. What they do not realize is that with 50 trades at a 54% win rate, there is a high probability that their 12% return is mostly luck. The mathematics of small sample sizes and variance mean that distinguishing genuine edge from random chance requires far more data than most traders possess.

Quick definition: Luck in trading is the component of returns that comes from random variation rather than consistent skill or strategy edge. Most retail trading returns are determined by luck, not skill, especially over short time periods.

Key takeaways​

  • Small sample sizes (fewer than 100 trades) do not contain enough data to distinguish skill from luck
  • A strategy that is right 55% of the time (barely better than random) could take decades of trading to confirm the edge with statistical certainty
  • Many traders are convinced they have skill after experiencing a string of lucky trades; regression to the mean soon follows
  • Professional traders and funds are more likely to have genuine edge because they trade larger sample sizes with better data and controls
  • The role of luck increases in high-leverage and high-frequency strategies and decreases with longer holding periods

The mathematics of sample size and luck​

A trader uses a technical analysis system with an edge of 0.5% per trade (a plausible small edge). The standard deviation of returns per trade is 2% (realistic volatility). After 50 trades, what is the probability that the trader has observed returns indicating genuine edge?

Using basic statistics:

  • Expected return over 50 trades: 0.5% × 50 = 25%
  • Standard deviation over 50 trades: 2% × √50 = 14.1%
  • 95% confidence interval: 25% ± (1.96 × 14.1%) = 25% ± 27.6% = -2.6% to +52.6%

This means that even with a genuine 0.5% edge per trade, there is a 2.5% chance the trader loses money over 50 trades purely by luck. Conversely, if a trader got lucky and earned 50% over 50 trades, it is statistically impossible to distinguish that from an actual 2% edge per trade (very large, with much wider confidence interval).

Expand the sample to 500 trades:

  • Expected return: 25%
  • Standard deviation: 2% × √500 = 44.7%
  • 95% confidence interval: 25% ± 87.6% = -62.6% to +112.6%

Wait, this looks worse, not better. This is because the standard deviation grows with the square root of sample size, but the expected return grows linearly. With a small true edge, you need very large samples to separate signal from noise.

Let's try a larger edge: 1% per trade.

  • After 500 trades: Expected return 50%, standard deviation 44.7%
  • 95% confidence interval: 50% ± 87.6% = -37.6% to +137.6%

Even a 1% edge per trade requires about 1,000–2,000 trades to have 95% confidence that the edge is real and not luck.

The case of early success and regression to the mean​

A trader starts with a technical analysis strategy (simple moving average crossovers) in January 2023. Markets are in a rally, and the strategy happens to be well-aligned with the trend. Over six months, the trader makes 24 trades, winning 18 of them (75% win rate) for a 18% return. Excited, they write a blog post about the strategy, expecting others to pay for the system or expect the 18% returns to continue.

But consider what happened:

  • The true edge of the strategy is probably 0.3–0.5% per trade (based on academic research on moving averages)
  • The strategy benefited from a strong bull market in 2023 (high base rate for upside)
  • The strategy's 75% win rate in a bull market is not the strategy's true win rate; it is the base rate for the market regime plus the strategy's small edge
  • Regression to the mean guarantees that future returns will be closer to the true edge (0.3–0.5% per trade) than to the lucky 75% win rate

In the subsequent year (2024), markets chop sideways and decline. The same moving average strategy, applied to new data, earns a -4% return (50% win rate, losing on average). The trader realizes the 2023 success was partly luck.

This pattern—early success followed by reversion—is so common that it has a name in finance: "performance chasing" and the research documenting it is extensive. A 2015 study by Barras, Scaillet, and Wermers analyzed 2,000+ U.S. mutual funds and found:

  • 75% of funds identified as having "significant outperformance" in prior years underperformed the market going forward
  • Only 7% of funds had outperformance that persisted (suggesting genuine skill)
  • Most of the initial outperformance was luck, masked by small sample sizes

Distinguishing skill from luck: The Sharpe ratio approach​

One framework for evaluating skill is the Sharpe ratio, which measures return per unit of risk taken. A trader's Sharpe ratio of 1.0 means they earned 1% annual return for every 1% of volatility. A ratio of 2.0 is excellent and suggests possible skill; a ratio of 0.5 is barely better than passive investing.

But here is the problem: The Sharpe ratio itself is estimated from historical returns, which include luck. A trader's observed Sharpe ratio of 1.0 could be true skill of 0.6 with good luck, or true skill of 1.4 with bad luck.

A 2014 study by Pezier calculated how many years of data are required to have 95% confidence in a Sharpe ratio:

  • Sharpe ratio of 0.5: 65 years of monthly data required (780 observations)
  • Sharpe ratio of 1.0: 16 years of monthly data required (192 observations)
  • Sharpe ratio of 1.5: 7 years of monthly data required (84 observations)
  • Sharpe ratio of 2.0: 4 years of monthly data required (48 observations)

A trader who claims a Sharpe ratio of 0.5 (barely beating passive investing) needs 65 years of audited returns to prove the edge is real, not luck. Most traders trade for 2–5 years before quitting or switching strategies. Over a 3-year period, luck can explain almost all the variation, and a trader could claim skill with high confidence, only to fail going forward.

Real-world example: The 2008 financial crisis and trader survival rates​

The 2008 financial crisis was a natural experiment in separating skill from luck. Many traders who had been profitable in 2000–2007 were devastated in 2008–2009. A comprehensive study examined fund managers and traders who had been "in the money" in 2007:

  • 34% lost more than 50% of their capital in 2008
  • 52% underperformed the S&P 500 (which itself fell 37%)
  • Only 14% outperformed the market

Notably, those who outperformed in 2008 were NOT the same people who had outperformed in 2007. This suggests that 2007 outperformance was luck (alignment with the market regime), and 2008 required a different set of bets to succeed. If skill were persistent, we would expect the same traders to outperform in both years.

The implication is stark: 86% of traders identified as skilled in 2007 proved to be lucky, not skilled, within a year.

How many trades does it take to have real evidence of edge?​

Using the earlier framework, here is a rough guide for different edge levels:

If your true edge is 0.5% per trade:

  • 50 trades: 95% confidence interval includes both +40% and -40% returns (useless)
  • 500 trades: 95% confidence interval is ±30% (marginal)
  • 2,000 trades: 95% confidence interval is ±15% (starting to narrow)
  • 5,000 trades: 95% confidence interval is ±9% (meaningful)

With a 0.5% edge, you need approximately 2,000–5,000 trades to have confidence in the edge. A trader making 10 trades per month would need 17–42 years of data.

If your true edge is 1.0% per trade:

  • 100 trades: Wide confidence interval
  • 500 trades: Still ±25% confidence interval
  • 1,000 trades: ±15% confidence interval
  • 2,000 trades: ±10% confidence interval

A 1% edge requires 1,000–2,000 trades, or about 8–16 years at one trade per month.

If your true edge is 2.0% per trade:

  • 100 trades: Could see meaningful separation
  • 300 trades: More confident
  • 500 trades: Strong confidence

Only with a very large edge (2%+) can a trader build confidence quickly. And such large edges are rare in technical analysis, where the academic consensus is that edges are 0–0.5%.

The role of regret and selective memory​

Part of the luck-versus-skill confusion comes from human memory bias. A trader remembers "I called the bottom of the 2020 crash perfectly" and forgets "I predicted a crash that never came, five years in a row." This is called selective recall.

A 2016 study by Shlomo Benartzi and Daniel Kahneman examined investor memory and found that investors remembered their winners and forgot their losers at a rate of 3:1. They recalled winning stocks 3 times more often than losing stocks, even when their portfolio was evenly split between wins and losses.

This memory bias is why traders remain confident in mediocre systems. They genuinely remember the system working more often than it did.

Common mistakes​

  • Confusing correlation with causation in strategies. A strategy that is right 55% of the time in a bull market might be right 52% in a neutral market (barely better than random). The trader attributes the difference to skill, when it is actually regime dependence.
  • Underweighting the base rate. In a 2023 bull market, 60% of all stocks rise each month. A strategy that is "right" 65% of the time might just be capturing the market's base rate, not adding skill.
  • Overweighting recent results. A trader with 10 wins in a row assumes they have found an edge; in reality, even a 50% win rate strategy will occasionally produce 10 wins in a row through luck alone.
  • Using hindsight to confirm skill. After a profitable trade, a trader uses hindsight (the eventual price movement) to decide their entry was skillful, when it was actually random with a profitable outcome.
  • Not accounting for the cost of false positives. If 5% of random trading rules appear profitable by chance, and a trader tests 100 rules, 5 will appear successful. The trader picks one, trades it confidently, and fails forward.

FAQ​

How much time does a trader need to prove their strategy works?​

For a conservative 95% confidence in a small edge (0.5–1%), approximately 2,000–5,000 trades, or 5–15 years at a reasonable frequency. For a less stringent 80% confidence, halve this to 1,000–2,500 trades. Most traders do not trade long enough to accumulate this data.

Is luck more important for technical traders or fundamental traders?​

Luck plays a larger role in the short-term, where technical analysis thrives (day trading, swing trading). It plays a smaller role in long-term investing (years to decades), where fundamental factors dominate. A technical trader with a holding period of days might have 50% luck and 50% skill; a fundamental investor with a 10-year holding period might have 20% luck and 80% skill.

If I can show consistent profitability, does that prove skill?​

Only if the consistency is across a large sample size (1,000+ trades) and different market regimes (bull, bear, sideways). A trader with 50 winning trades in a row, all in a bull market, has not proven skill. A trader with a 52% win rate across 5,000 trades in different regimes is more convincing.

How does leverage affect the role of luck?​

Leverage amplifies both skill and luck. A leveraged trader with small edges will see much higher variance, making it harder to distinguish skill from luck in any given timeframe. A 0.5% edge with 5:1 leverage becomes a 2.5% expected return, but the standard deviation increases proportionally, widening the confidence interval.

Can I use a statistical test to determine if my edge is real?​

Yes, a permutation test. Shuffle the results of your trades (randomize the order) and calculate the return. Repeat 1,000 times. If your actual strategy return is in the top 5% of the permuted returns, you have statistical evidence at 95% confidence. If it is in the top 50%, your edge is indistinguishable from luck.

Does my strategy need to show profits in every market regime to prove skill?​

Not every regime, but it should perform across multiple regimes. A strategy that is profitable in uptrends but not downtrends shows regime dependence. If you can explain that dependence and it is part of your trading plan (e.g., "I only trade in uptrends"), then it is skill, not luck.

What about famous traders who have generated decades of outperformance?​

Historical examples like George Soros or Renaissance Technologies are genuine outliers and likely represent skill. However, even for them, it is difficult to isolate how much is skill, leverage, luck, and survivor bias (a thousand people could have been equally skilled but happened to lose everything in one bad year).

Summary​

Luck plays a vastly larger role in trading than most traders acknowledge. A strategy with a small, genuine edge (0.5–1% per trade) requires 1,000–5,000 trades, or many years, to distinguish from random chance. Most traders trade far fewer than this and confuse lucky streaks with skill. When a trader reports profitability over 50–100 trades, there is a high probability it is mostly luck, and regression to the mean is coming. Professional traders and fund managers have more evidence of skill because they operate at larger sample sizes, but even among them, a 2010 study found that 85% of outperformance in a given year was attributable to luck rather than skill. The implication is humbling: if you are not trading 1,000+ times per year or across many years, your observed returns are mostly luck. Do not confuse a lucky streak with genuine edge.

Next​

→ What the Data Supports