Every trade you take is a bet with uncertain outcomes. Even if your analysis is perfect, you're still dealing with probabilities, not certainties.
Let's say you have a 60% win rate. That sounds good-you win more often than you lose. But what does it actually mean in practice?
It means that out of every 100 trades, you expect to win about 60. But that's an average. In any given 10-trade sequence, you might win 4, you might win 8, you might win 6. All of these are consistent with a true 60% win rate.
Probability distributions tell us the range of possible outcomes and how likely each one is.
For a system with 60% win rate over 10 trades:
| Wins |
Probability |
How Often This Happens |
| 0-3 wins |
5.5% |
Rare but possible |
| 4 wins |
11.1% |
Common |
| 5 wins |
20.1% |
Common |
| 6 wins |
25.1% |
Most likely outcome |
| 7 wins |
21.5% |
Common |
| 8-10 wins |
16.7% |
Less common |
Notice that even with a true 60% win rate, you'll see only 4 wins out of 10 trades about 11% of the time. That's roughly once every 9 sets of 10 trades. It's not your system breaking-it's just probability doing what probability does.
One of the most dangerous probability errors is believing that past results influence future independent events.
"I've lost 5 trades in a row, so I'm due for a win."
This is wrong. Each trade is independent. The market doesn't know or care about your recent results. After 5 losses, your next trade still has the same probability of winning as always.
The opposite error is also common: "I've won 5 in a row, I'm on fire-let me size up." This overconfidence leads to exactly the kind of oversized losses that destroy accounts.
Variance is a statistical measure of how spread out your results are from the average. In trading terms, it tells you how much your trade outcomes fluctuate around your expected value.
Variance = Average of (Each Result - Mean Result)²
A simpler way to think about it: variance measures the bumpiness of your equity curve.
- Low variance: Results cluster tightly around the average. Steady, predictable equity curve.
- High variance: Results swing wildly. Roller coaster equity curve with big wins and big losses.
It determines your experience. Two systems can have identical expected value but completely different variance. System A might make $10,000 per year with smooth monthly gains. System B might make $10,000 per year but with months of +$15,000 followed by months of -$10,000. Same expectation, vastly different lived experience.
It affects position sizing. High variance systems require smaller position sizes to avoid ruin. The Kelly Criterion directly accounts for variance-more variance means you should bet smaller.
It impacts psychological sustainability. Humans don't optimize for expected value-we experience the journey. High variance systems are psychologically grueling, even when profitable. Many traders abandon winning systems because they can't handle the variance.
It determines drawdown risk. The higher your variance, the larger your potential drawdowns. A high variance system will experience deeper troughs, which you must survive to capture the long-term edge.
In basic statistics, we often assume returns follow a normal distribution (bell curve). This simplifies the math and works reasonably well for many applications.
Under normal distribution assumptions:
- 68% of outcomes fall within 1 standard deviation of the mean
- 95% of outcomes fall within 2 standard deviations
- 99.7% of outcomes fall within 3 standard deviations
For trading, this would mean that extreme outliers-events more than 3 standard deviations from normal-should happen only 0.3% of the time.
Crypto market returns absolutely do not follow a normal distribution. They exhibit what statisticians call "fat tails" or "leptokurtosis"-extreme events happen far more often than a normal distribution predicts.
Under a normal distribution, a 5-standard-deviation event should happen once every 4,776 years. In crypto, we see moves of this magnitude multiple times per year.
| Event Size |
Normal Distribution Prediction |
Crypto Reality |
| 2σ move (~5%) |
Every 22 trading days |
Multiple times per month |
| 3σ move (~7-8%) |
Every 1.4 years |
Multiple times per year |
| 4σ move (~10%) |
Every 126 years |
Yearly occurrence |
| 5σ move (~15%) |
Every 4,776 years |
Every few years |
The March 2020 crash, the May 2021 crash, the November 2022 FTX collapse, multiple flash crashes-these are all events that "shouldn't" happen under normal distribution assumptions but absolutely do happen in crypto.
Fat tails mean that extreme outcomes are more probable than standard models suggest. This has massive implications for how you manage risk.
Nassim Taleb popularized the concept of "black swans"-rare, unpredictable events with massive impact. In crypto, black swans are actually gray swans-they're not as rare as people think.
If you build your risk management around normal distribution assumptions, you're dramatically underestimating tail risk. You might think a 30% drawdown is nearly impossible based on your historical variance, then experience exactly that in a single day.
Position sizing must be conservative. Because extreme events are more likely than normal distributions suggest, you need larger safety margins. What looks like a "safe" position size under normal assumptions might be dangerous in fat-tailed markets.
Stops don't always work. In fast-moving markets, your stop loss might not execute at your intended price. Slippage during extreme events can be enormous. A stop at -5% might fill at -15%.
Correlations spike during crises. Assets that normally move independently tend to crash together during extreme events. Diversification provides less protection than you might expect when it matters most.
Leverage amplifies tail risk. A position that seems safe with 3x leverage under normal assumptions can be wiped out by a fat-tail event. This is why crypto leverage tends to destroy traders eventually.
Rather than assuming normal distributions, experienced traders often:
- Use historical worst-case scenarios, not standard deviation, for risk planning
- Assume correlations will be 1.0 during crises
- Size positions so that even a 30-50% gap move is survivable
- Use options for tail risk hedging when available
- Keep significant capital in stablecoins or off-exchange
Let's get practical. Here's how to calculate variance from your actual trade data.
First, find your average result across all trades:
Mean = Sum of All R-Multiples ÷ Number of Trades
Example (10 trades): +2R, -1R, +0.5R, -1R, +3R, -1R, +1.5R, -1R, +2R, +1R
Sum = +6R
Mean = 6R ÷ 10 = +0.6R per trade
For each trade, calculate how far it deviated from the mean, then square it:
| Trade |
Result |
Deviation from Mean |
Squared Deviation |
| 1 |
+2R |
+1.4R |
1.96 |
| 2 |
-1R |
-1.6R |
2.56 |
| 3 |
+0.5R |
-0.1R |
0.01 |
| 4 |
-1R |
-1.6R |
2.56 |
| 5 |
+3R |
+2.4R |
5.76 |
| 6 |
-1R |
-1.6R |
2.56 |
| 7 |
+1.5R |
+0.9R |
0.81 |
| 8 |
-1R |
-1.6R |
2.56 |
| 9 |
+2R |
+1.4R |
1.96 |
| 10 |
+1R |
+0.4R |
0.16 |
Variance = Sum of Squared Deviations ÷ (Number of Trades - 1)
Sum of squared deviations = 20.90
Variance = 20.90 ÷ 9 = 2.32
- Standard deviation is simply the square root of variance: Standard Deviation = √Variance = √2.32 = 1.52R
Your system has a mean of +0.6R per trade with a standard deviation of 1.52R.
This means:
- About 68% of your trades will fall between -0.92R and +2.12R (mean ± 1 SD)
- About 95% of your trades will fall between -2.44R and +3.64R (mean ± 2 SD)
A standard deviation of 1.52R relative to a mean of 0.6R tells you this is a relatively high-variance system. Your individual trades swing quite a bit around your average result.
The Sharpe ratio-a key metric for risk-adjusted returns-directly incorporates variance:
Sharpe Ratio = Mean Return ÷ Standard Deviation
From our example: 0.6R ÷ 1.52R = 0.39
A Sharpe ratio of 0.39 is modest. Generally:
- Below 0.5: Low risk-adjusted returns
- 0.5 - 1.0: Acceptable
- 1.0 - 2.0: Good
- Above 2.0: Excellent (and rare)
Understanding variance intellectually is one thing. Experiencing it emotionally is another.
Humans are terrible at intuitively understanding variance. We see patterns where randomness exists. We draw conclusions from samples too small to be meaningful.
After 5 losing trades, you feel like your system is broken. After 5 winning trades, you feel invincible. Neither feeling reflects statistical reality.
- Antidote: Before reacting to recent results, calculate whether they're within expected variance. If you have a 55% win rate, losing 5 of 7 trades is annoying but statistically unremarkable. Don't change anything.
High variance systems inevitably experience extended drawdowns. During drawdowns, traders often:
- Start questioning their system
- Second-guess entries and take fewer trades
- Cut winners short to "lock in" any profit
- Size down dramatically after losses
- Eventually abandon the system at the worst possible time
This behavioral decay transforms a winning system into a losing one-not because the edge disappeared, but because the trader couldn't handle normal variance.
Before you start trading any system, calculate what drawdowns you should expect. Not "might possibly experience once in a blue moon"-what you should expect as a normal part of operation.
For a system with:
- 50% win rate
- Average win of 2R
- Average loss of 1R (expectancy = 0.5R per trade)
You should expect:
- Losing streaks of 5+ trades multiple times per year
- Monthly drawdowns of 10-15% regularly
- Quarterly drawdowns of 20-25% occasionally
If these numbers make you uncomfortable, either reduce position size until they're acceptable or find a lower-variance system. Don't start trading assuming you'll never face normal variance.
A confidence interval tells you the range within which your true edge likely falls, given your observed results.
Your observed results are just a sample. If you've taken 100 trades and averaged 0.3R per trade, your true edge isn't necessarily 0.3R-it's somewhere near 0.3R, but you're not certain exactly where.
Confidence intervals quantify this uncertainty.
For a 95% confidence interval around your mean:
CI = Mean ± (1.96 × Standard Error)
Where Standard Error = Standard Deviation ÷ √(Number of Trades)
Example:
- Mean: 0.3R per trade
- Standard deviation: 1.8R
- Number of trades: 100
Standard Error = 1.8 ÷ √100 = 0.18R
95% CI = 0.3 ± (1.96 × 0.18) = 0.3 ± 0.35
Your true edge is somewhere between -0.05R and +0.65R with 95% confidence.
If your confidence interval includes zero (like the example above), you cannot be confident your edge is actually positive. Your observed results might be entirely due to luck.
To narrow your confidence interval (increase certainty), you need either:
- More trades (more data)
- Lower variance (more consistent results)
- A larger edge (more signal relative to noise)
| Trades |
Standard Error (if SD = 1.5R) |
95% CI Width |
| 25 |
0.30R |
±0.59R |
| 50 |
0.21R |
±0.42R |
| 100 |
0.15R |
±0.29R |
| 200 |
0.11R |
±0.21R |
| 500 |
0.07R |
±0.13R |
This is why you need so many trades to validate a system. With only 25 trades, even a system with a true edge of 0.3R would have a confidence interval including zero.
Every trading decision is made with incomplete information. Probability and variance don't tell you what will happen-they tell you what might happen and how likely each outcome is.
When facing any trading decision, calculate the expected value:
EV = (Probability of Win × Amount Won) - (Probability of Loss × Amount Lost)
Only take trades where EV is positive. This seems obvious but many traders chase low-probability, high-reward trades that have negative EV because of the allure of big payoffs.
Example:
- Trade has 30% chance of hitting 3R target
- Trade has 70% chance of stopping out at -1R
EV = (0.30 × 3R) - (0.70 × 1R) = 0.9R - 0.7R = +0.2R
This trade has positive EV even though you'll lose most of the time. Over many similar trades, you make money.
How much should you risk on a positive-EV opportunity? The Kelly Criterion provides the mathematically optimal answer:
Kelly % = (Win Rate × Average Win / Average Loss - Loss Rate) / (Average Win / Average Loss)
For a system with 45% win rate, 2:1 reward/risk:
Kelly % = (0.45 × 2 - 0.55) / 2 = (0.90 - 0.55) / 2 = 0.175 or 17.5%
Most traders use "fractional Kelly" (usually 25-50% of full Kelly) to reduce variance at the cost of some expected return. Full Kelly is mathematically optimal but produces wild swings.
For complex decisions with multiple possible outcomes, decision trees help visualize expected values:
Trade Entry
├── 40% → Win at 2R target → +2R
├── 25% → Small win at 1R (trail stop) → +1R
├── 20% → Breakeven exit → 0R
└── 15% → Stop loss hit → -1R
EV = (0.40 × 2) + (0.25 × 1) + (0.20 × 0) + (0.15 × -1)
EV = 0.80 + 0.25 + 0 - 0.15 = +0.90R
Understanding variance is useful. Managing it is essential.
Your position size should be calibrated to your system's variance, not arbitrary percentages.
High variance systems need smaller positions. Low variance systems can handle larger positions.
Use this framework:
- Calculate your maximum historical drawdown (in R)
- Set position size so that experiencing that drawdown 1.5x wouldn't ruin you
- If your max drawdown was 20R and you want to survive 30R, size so that 30R = 30% of account max
Running multiple uncorrelated strategies reduces overall portfolio variance while maintaining expected return.
If Strategy A has variance of 2.0 and Strategy B has variance of 2.0, but they're uncorrelated, running both (properly sized) produces lower combined variance than either alone.
This is the core principle of portfolio theory-diversification reduces variance without necessarily reducing return.
High variance can work in your favor on winning trades. Trailing stops allow you to capture larger moves when variance pushes in your direction, while still protecting against reversals.
Rather than fixed targets (which cap upside), trail your stop as price moves favorably. You'll give back some profit on most trades but occasionally capture outsized winners.
Set rules for what happens when variance exceeds normal bounds:
- If you hit X consecutive losses: Stop trading for the day
- If drawdown reaches Y%: Reduce position size by 50%
- If drawdown reaches Z%: Stop trading entirely and review
These rules prevent you from compounding losses during high-variance periods when your judgment is most compromised.
Judge your trading decisions by process, not outcome. A good trade can lose money. A bad trade can make money. Variance makes individual outcomes unreliable feedback.
Ask: "Did I follow my system?" not "Did I make money on this trade?"
Over time, good process produces good outcomes. Over individual trades, the connection is much weaker.
You can reduce variance by: taking more selective trades (smaller sample with higher probability), using tighter stop losses (caps downside but may reduce win rate), trading less volatile assets, using smaller position sizes, or diversifying across uncorrelated strategies. There's usually a tradeoff-reducing variance often reduces expected return.
For a 50% win rate, losing streaks of 5-6 trades are common (happen every few months). Losing streaks of 8-10 trades are rare but expected yearly. For a 40% win rate, even longer streaks are normal. Calculate expected losing streaks for your actual win rate before panicking.
Not necessarily. High-variance strategies can be extremely profitable-they just require appropriate position sizing and psychological resilience. The key is understanding the variance before you start and sizing accordingly. If you can't handle the variance psychologically, choose a different strategy.
Compare your recent performance (last 30-50 trades) to your long-term performance. If the difference is within one standard deviation, it's probably variance. If it's beyond two standard deviations and persists, it might be edge decay. More data helps distinguish signal from noise.
Above 1.0 is solid for crypto given its inherent volatility. Above 2.0 is excellent. Below 0.5 is concerning-your risk-adjusted returns are poor. Most successful retail crypto traders operate in the 0.7-1.5 range.
Leverage directly multiplies variance. If your unleveraged variance is 1.0, 3x leverage gives you variance of 9.0 (3²). This is why leverage destroys so many traders-not because their edge is negative, but because the amplified variance produces drawdowns they can't survive.
Trading is fundamentally about making decisions under uncertainty. You will never have certainty. The best analysis in the world gives you probabilities, not guarantees.
The traders who succeed long-term are the ones who:
- Understand that individual trade outcomes don't define system quality
- Size positions so that normal variance doesn't threaten survival
- Make decisions based on expected value, not hopes or fears
- Don't confuse noise for signal or signal for noise
- Accept that even the best systems have bad stretches
Variance is not your enemy-it's the price of admission to markets that offer profit opportunities. Without variance, there would be no edge to capture.
Your job isn't to eliminate variance. It's to manage it so that you're still standing when the long-term edge manifests in your favor.
The market doesn't owe you smooth sailing. It owes you nothing. But if you understand probability and manage variance appropriately, you can extract value from the chaos.
Calculating variance, standard deviation, confidence intervals, and Sharpe ratios manually is tedious. Most traders never do it because the friction is too high. Thrive eliminates that friction.
✅ Automatic Variance Calculation - Import your trades and Thrive instantly calculates standard deviation, variance, and statistical significance of your results.
✅ Rolling Sharpe Ratio - Track your risk-adjusted returns over time. See when your Sharpe ratio is improving or degrading.
✅ Confidence Intervals - Understand the uncertainty around your edge. Know whether your results are statistically meaningful or potentially just luck.
✅ Drawdown Analysis - See your historical drawdowns in context of expected variance. Distinguish normal fluctuation from actual system problems.
✅ Weekly AI Coach - Get personalized insights about your variance patterns and recommendations for managing psychological challenges during high-variance periods.
✅ Outlier Detection - Thrive flags when your results are outside expected ranges, helping you catch both problems and opportunities faster.
Understanding probability and variance is what separates professionals from amateurs. But you don't need a statistics degree to benefit-you just need the right tools.
Make better decisions under uncertainty.
→ Start Understanding Your Trading Variance
