Sample Size: How Many Bets Do You Need?
One of the most common mistakes in sports betting is judging a strategy based on too few bets. Variance means that short-term results can be wildly misleading. So how many bets do you actually need before you can trust your results?
The Rules of Thumb
As a practical starting point:
| Sample Size | Reliability Level |
|---|---|
| 30 bets | Minimum to see initial patterns, highly susceptible to variance |
| 100 bets | Enough to get a rough sense of whether your edge might be real |
| 300+ bets | Statistically reliable for evaluating your strategy |
| 1,000+ bets | High confidence in your true win rate |
Key Insight
The smaller your edge and the higher the variance, the more bets you need. If you're betting on coin-flip propositions with a 52% win rate, you'll need several hundred bets to distinguish your edge from random noise. If you're betting on heavy favorites with a 70% win rate, you'll see clearer results sooner.
Why Small Samples Are Dangerous
Imagine you go 6-4 over your first 10 bets. That's a 60% win rate—impressive!
But let's analyze this statistically:
- True edge assumed: 52% (you have a small edge)
- Expected wins: 10 × 0.52 = 5.2
- Standard deviation: √(10 × 0.52 × 0.48) = 1.58 wins
- Your result: 6 wins = 0.8 wins above expectation
- Z-score: 0.8 / 1.58 = 0.5 standard deviations
Going 6-4 is only half a standard deviation above expectation—this happens about 30% of the time by pure chance even with no edge at all!
Warning
A 60% win rate over 10 bets tells you almost nothing. You could be a genius or you could have zero edge—the sample is far too small to know.
The Math Behind Sample Size
Now imagine you continue and finish with 160 wins and 140 losses over 300 bets (53.3% win rate).
- Expected wins (at 52%): 300 × 0.52 = 156
- Standard deviation: √(300 × 0.52 × 0.48) = 8.65 wins
- Your result: 160 wins = 4 wins above expectation
- Z-score: 4 / 8.65 = 0.46 standard deviations
Even after 300 bets, a 53.3% result is less than half a standard deviation above a 52% expectation. This suggests your true win rate is likely close to your observed rate—but you still can't be highly confident.
The Formal Sample Size Formula
For rigorous analysis, use this formula to calculate exactly how many bets you need:
Required Sample Size
n = (Z × σ / E)²=(A1*B1/C1)^2Where:
- n = required sample size (number of bets)
- Z = Z-score for your desired confidence level
- σ (sigma) = standard deviation of a single bet outcome (≈1 for even-money bets)
- E = margin of error you'll accept
Common Z-Scores
| Confidence Level | Z-Score |
|---|---|
| 90% | 1.645 |
| 95% | 1.96 |
| 99% | 2.576 |
Example Calculation
You want to estimate your true edge within ±3 percentage points with 95% confidence:
n = (1.96 × 1 / 0.03)²
n = (65.33)²
n ≈ 4,268 bets
Note
Yes, you read that correctly. To be 95% confident your true win rate is within ±3% of your observed rate, you need over 4,000 bets. This is why professional bettors track thousands of wagers before drawing conclusions.
Practical Guidelines by Edge Size
The size of your expected edge affects how many bets you need to confirm it:
| Expected Edge | Sample Needed (95% confidence) | Practical Approach |
|---|---|---|
| 10%+ (obvious) | 100-200 bets | Clear signal emerges quickly |
| 5-10% (strong) | 300-500 bets | Reliable after a season |
| 2-5% (moderate) | 500-1,500 bets | Multi-season tracking |
| 1-2% (thin) | 2,000-5,000 bets | Professional-level volume |
Reconciling Theory with Practice
The formal sample size formula can seem discouraging—who has time for 4,000 bets? Here's how to reconcile theory with reality:
Rules of Thumb Are Starting Points
- 30 bets: Too early to conclude anything
- 100 bets: Preliminary signal—proceed with caution
- 300+ bets: Actionable but not conclusive
The Formula Is for Precision
- Use it when you need high confidence in your edge
- Essential for professional bankroll sizing decisions
- Explains why pros track everything over years
Progressive Confidence
As your sample grows, confidence increases even before reaching the "ideal" size. You don't need to wait for 4,000 bets to make any decisions—you just need to understand your confidence level at each stage.
Tip
Practical approach: Start with rules of thumb to build your dataset, then apply the formula as your sample grows to refine and validate your strategy. Don't be paralyzed by analysis—just be appropriately humble about small-sample results.
The Sample Size Reality Check
| Your Sample | What You Can Conclude |
|---|---|
| 10 bets | Almost nothing—pure noise |
| 50 bets | Directional hint only |
| 100 bets | Preliminary signal |
| 300 bets | Reasonable confidence |
| 500 bets | Strong indication |
| 1,000+ bets | High confidence |
📝 Exercise
Instructions
Evaluate whether this betting record provides sufficient evidence of an edge.
You've been tracking a new NFL prop betting strategy. After 50 bets, you've won 28 and lost 22 (56% win rate).
Questions:
- Calculate the standard deviation for 50 bets assuming NO edge (50% true rate)
- How many standard deviations above expected is your result?
- Is this result statistically significant?
- If you continue and finish 85-65 after 150 total bets, does this provide stronger evidence?
📝 Exercise
Instructions
Calculate the sample size needed for your specific situation.
You believe you have a 54% win rate on -110 bets (a 4% edge over true odds). To be 95% confident your actual edge is at least 1% (and not just luck), approximately how many bets do you need?
Key Takeaways
-
Don't judge on 20-30 bets—you need at least 100 for preliminary signals, 300+ for reliability
-
Small edges require large samples—a 2% edge needs thousands of bets to confirm
-
The formal formula n = (Z × σ / E)² gives exact requirements for any confidence level
-
Be appropriately humble about small-sample results—a 60% rate over 10 bets proves nothing
-
Progressive confidence—update your beliefs as samples grow, but don't overreact to early results
-
Track everything—you can't evaluate sample size if you don't have the data
In the final lesson of this chapter, we'll explore player performance variance—how to assess whether individual players are consistent or volatile for prop betting purposes.