Poker Variance Calculation: Win Rate Standard Deviation and Sample Size Guide

Why It Matters

Variance in poker is the main source of short-term result fluctuations. Even with a positive win rate, you can experience prolonged downswings. Understanding the relationship between standard deviation and sample size allows you to reasonably assess your true skill level and avoid overconfidence or discouragement due to short-term luck.

Basic Concepts

Win Rate

Usually expressed as profit per 100 hands (or per hour) in bb/100. For example, 5bb/100 means an average profit of 5 big blinds per 100 hands.

Standard Deviation

Measures the dispersion of results around the average per 100 hands. Typical values range from 70 to 120 bb/100, depending on game type and style. Higher standard deviation means greater fluctuation in individual results.

Variance

The square of the standard deviation. Used mathematically for calculations, but standard deviation is more common in practical discussions.

Sample Size

The total number of hands (e.g., 10,000 hands). For calculations, it is often converted into “per hundred hands” units to keep the standard deviation consistent.

Step-by-Step Procedure

Step 1: Collect Data

Export your hand history using database software (e.g., Holdem Manager, Poker Tracker). Include the number of hands, profit/loss, and total hands for each session.

Step 2: Calculate Overall Win Rate and Standard Deviation

The software usually provides your overall win rate (bb/100) and standard deviation (bb/100) directly. If calculating manually, first compute the weighted average win rate per session, then calculate the standard deviation.

Step 3: Calculate Standard Error

The standard error measures the variability of the sample mean:

$$SE = \frac{SD}{\sqrt{N_{100}}}$$

where (N_{100}) is the number of 100-hand samples. For example, with 10,000 total hands, (N_{100}=100). If SD = 80 bb/100, then SE = 80 / √100 = 8 bb/100.

Step 4: Construct a Confidence Interval

95% confidence interval = win rate ± 1.96 × SE.

Example: win rate 5 bb/100, SE = 8 bb/100 → interval = [5 – 15.68, 5 + 15.68] = [-10.68, 20.68]. This means that after 10,000 hands, there is a 95% chance your true win rate falls within this wide range.

Step 5: Determine Required Sample Size

Suppose you want the margin of error for your true win rate estimate to be within ±2 bb/100 at a 95% confidence level. The required number of 100-hand samples is:

$$N_{100} = \left(\frac{1.96 \times SD}{\text{margin of error}}\right)^2$$

Substitute SD = 80, margin of error = 2 → (N_{100} \approx (78.4)^2 \approx 6147), i.e., 614,700 hands. This shows that poker requires extremely large samples for accurate evaluation.

Common Mistakes

• Confusing swing with variance: A downswing is a concrete manifestation of variance, but variance itself is a constant statistical parameter.
• Drawing conclusions from small samples: Even after 10,000 hands, the confidence interval may still be wide, so you cannot determine whether your true win rate is positive.
• Ignoring downswing probability: Standard deviation can be used to calculate the probability of a downswing (e.g., the chance of losing a certain number of big blinds). The formula requires specialized knowledge, but software can generate charts directly.

Advanced Tips

• Use tools: PokerTracker and Holdem Manager have built-in reports that automatically show standard deviation and confidence intervals. There are also online calculators (e.g., Primedope’s variance calculator).
• Risk management: Use standard deviation to determine bankroll requirements. Suppose you play NL100, with a standard deviation of 80 bb/100 and a win rate of 5 bb/100. A 20 buy-in downswing could last several months. A general guideline is to have at least 20 times the standard deviation in your bankroll.
• Compare different games: Cash games typically have lower variance than tournaments due to their more stable structure. Tournament play, especially with ICM considerations, greatly increases variance.

Summary

Variance is an inseparable part of poker. By calculating standard deviation and using an appropriate sample size, you can scientifically evaluate your skill level and make more rational decisions. Remember: winning or losing 100 buy-ins does not prove your ability; only 100,000 hands of data begin to offer meaningful reference.