Edge or Luck? — The Statistical Foundation of Profitable Trading
"It ain't what you don't know that gets you into trouble. It's what you know for sure that just ain't so."
— Mark Twain
The Question Nobody Asks
You've been trading for three months. Forty-two trades. 57% win rate. Average winner 2.3R, average loser 1R. Expectancy +0.72R per trade. You're profitable.
Here's the question that should keep you up at night: Is this real, or did you get lucky?
At forty-two trades, the honest statistical answer is: you don't know yet. Your observed 57% might be your true rate. It might be 47% and you happened to flip a few extra heads. The difference — one profitable long-term, one slowly destructive — is invisible in forty-two trades.
This chapter gives you the statistical tools to answer the most important question in your trading career: do I have a real edge, or have I been lucky?
Sample Size — Why Forty Trades Tells You Almost Nothing
A fair coin flipped 40 times. You expect 20 heads. But getting 24 (60%) or 16 (40%) is completely normal. The range of "normal" results from NO edge is enormous in small samples. A 60% win rate over 40 trades could easily be pure chance.
| Trades Taken | What You Can Conclude |
|---|---|
| 10-20 | Almost nothing. Too few to matter. Don't make any sizing decisions. |
| 30-50 | Basic patterns visible, but confidence is LOW. Observed stats could be wildly different from true stats. |
| 50-100 | Getting useful. If expectancy is clearly positive, you probably have something. But could still be noise. |
| 100-200 | Meaningful. Probability of pure luck drops significantly. Start trusting the numbers. |
| 200-500 | Robust. Observed stats are likely close to true stats. Confidence is high. |
| 500+ | Statistical bedrock. You can trust these numbers. Refine based on them. |
Rule of thumb: Don't increase risk% until 100+ trades of evidence. Don't consider system "proven" until 200+.
The t-Test for Traders — Your Edge Detector
The t-test answers: "Is my average profit significantly different from zero?" In plain English: am I really making money, or is the apparent profit within random chance?
Calculate your t-statistic monthly. A rising t-stat means edge strengthening or sample growing. Declining = edge eroding. If it drops below 2.0 after being above, investigate.
Monte Carlo Simulation — Your Range of Possible Futures
A backtest shows ONE path. But that sequence was only one of millions of possible orderings. What if your biggest winner came first? What if your worst losing streak was twice as long?
Monte Carlo shuffles your actual trade results 10,000 times, producing 10,000 equity curves — all from your REAL edge, just in different sequences.
- 50th percentile (median): Your expected outcome. The "normal" path.
- 95th percentile: Best case. Don't plan for this.
- 5th percentile: WORST case. THIS is what you must survive.
Your backtest showed 20% max drawdown. Monte Carlo 5th percentile shows 38%. Plan for the 38%. If you can't handle that drawdown, reduce position size until you can.
Tools: Excel (RAND() shuffling), Python (~15 lines with numpy), QuantConnect/AmiBroker (built-in). If you only run ONE additional analysis beyond basic metrics, make it Monte Carlo.
MAE and MFE — Your Data Tells You Where to Stop and Target
The most actionable statistical tools you'll ever use. They turn YOUR trading data into empirically optimized stops and targets.
Maximum Adverse Excursion = how far each trade went AGAINST you before resolving. Winners tend to have small MAE. Losers tend to have large MAE.
If 90% of winners never went more than 1.2R against you but your stop is at 1.5R — tighten to 1.3R. Same winners, smaller average loss. Expectancy improves automatically.
Maximum Favorable Excursion = how far each trade went IN YOUR FAVOR before ending. Shows where most winners peak before pulling back.
If 80% of winners reach 2.5R before pulling back to 2R — take 50% at 2.5R, trail the rest. Average realized R improves.
"Let your data tell you where to stop and where to take profit. Don't guess. Measure." After 50 trades: run MAE. After 100: run MFE. After 200: run combined analysis. Every system has an empirically optimal stop and target. MAE/MFE finds them. Everything else is guessing.
How Long Until I Know If My System Works?
The honest mathematical answer nobody provides:
| Win Rate | Trades for 95% Confidence | At 2-3 trades/week |
|---|---|---|
| 55% (good trend-following) | ~385 trades | 2.5-3.5 years |
| 60% (excellent) | ~96 trades | 8-12 months |
| 45% with 3:1 R:R (asymmetric) | 80-100 trades (t-test) | 6-10 months |
The asymmetric approach partially addresses this. Because winners are larger than losers, positive expectancy shows up in the t-test faster. 3:1 R:R doesn't just improve profitability — it makes your edge DETECTABLE in fewer trades. The signal is stronger.
Realistic Return Expectations
"The path to millions is boring. That's why so few walk it." It's not one amazing trade. It's not a secret strategy. It's positive expectancy, executed consistently, compounded patiently, for years and years. The math is certain. The discipline is everything. The time is the price of admission. Pay it willingly.
Part III: Executing the Trade is now complete. You have entries, exits, risk management, backtesting, and statistical validation. Part IV confronts the final boss — your own psychology.
Ready to Apply These Principles?
Get 18 free institutional-grade indicators on TradingView and join the Nexural trading community.