For independent Bernoulli trials with success probability , the sample proportion approaches in probability. Its variance is , and the probability of any fixed positive error tends to zero as tends to infinity. Neither a finite-sample tail probability nor an individual path must improve at every step.独立重复进行成功概率为 的伯努利试验,样本中的成功比例会依概率趋近 。其方差为 ;当 趋于无穷时,偏离任意固定正幅度的概率趋于零。有限样本的尾部概率和单条路径,都不必逐步改善。
Three seeded simulations reveal the running proportions through 256 trials; the gold line is . The same finite paths replay as an illustration, not a proof. This concerns concentration of the average; the central limit theorem instead describes its scaled fluctuations.三条固定种子的模拟路径逐步显示前 256 次试验的成功比例,金线表示 。动画重复的是同一组有限样本,只作示意,不构成证明。这里关注平均值的集中;中心极限定理关注经过缩放后的波动形状。