Why It Matters

Poker is a game with extreme short-term variance. Even if you are technically superior to your opponents, you can lose over hundreds of hands. Understanding variance and standard deviation helps you:

• Distinguish between short-term luck and long-term skill
• Set reasonable profit expectations
• Manage your bankroll and mental resilience

Basic Concepts

Variance: A statistical measure of how much results are spread out. In poker, the more your profit per hand fluctuates, the higher the variance.

Standard Deviation (SD): The square root of variance, in the same unit as profit (e.g., big blinds per 100 hands). Common ranges:

• Cash games: approximately 70–100 big blinds per 100 hands
• Tournaments: higher, due to sharp changes in payout structures

Sample Size: Number of hands played. The larger the sample, the closer your estimated win rate is to the true value.

Step-by-Step Process

1. Collect Data

• Use tracking software (e.g., PokerTracker, Hold'em Manager) to export profit per hand (in big blinds or chips).
• You need at least several thousand hands for the data to be meaningful.

2. Calculate Sample Mean and Standard Deviation

Let x_i be the profit per hand, N the number of hands, and mean μ = (Σx_i)/N.

Sample standard deviation formula: SD = sqrt[ Σ(x_i - μ)² / (N-1) ]

Most software automatically outputs standard deviation per hour or per 100 hands.

3. Estimate the Range of Your True Win Rate

Confidence interval (95% confidence level) for the true win rate μ_true: μ ± 1.96 × SD / sqrt(N)

Example:

• Sample win rate: 5 BB/100 hands
• Standard deviation per 100 hands: 80 BB
• Number of hands: 10,000 (i.e., 100 blocks of 100 hands)
• Standard error = 80 / sqrt(100) = 8 BB/100 hands
• 95% confidence interval: 5 ± 1.96×8 = [-10.68, 20.68] BB/100 hands This means your true win rate could be negative! You need more hands.

4. Calculate Required Sample Size

Given a desired margin of error E (e.g., you want your win rate estimate to be within ±2 BB/100 hands), the number of 100-hand blocks needed: n = (1.96 × SD / E)²

Using the same example, if SD=80 and E=2: n = (1.96×80/2)² ≈ (78.4)² ≈ 6146 blocks of 100 hands, i.e., 614,600 hands.

Common Mistakes

• Jumping to conclusions too early: Win rates from a few thousand hands are unreliable.
• Ignoring differences in standard deviation: Different game types (6-max vs. full ring) have different standard deviations.
• Not considering bankroll management: High variance requires enough buy-ins.

Advanced Tips

1. Use Bayesian methods: Combine prior information (e.g., the average win rate at that stake) to update your estimate.
2. Segment your calculations: Compute standard deviation by month or session to check the stability of your game.
3. Consider risk measures: For example, downside deviation (Sortino ratio) focuses more on loss volatility.

Summary

Variance and standard deviation are the foundation of scientific poker management. By using statistical methods, players can avoid emotional decisions and rationally evaluate their progress. Remember: Short-term results do not define your true skill level; only a large sample size is convincing.