What if you flipped a coin ten times and used the results to predict the ratio of heads to tails? You could get the answer drastically wrong.
Even if you flipped the coin 50 times, you could have the same result, as shown below. But as you flip the coin more, you wash out the randomness and get estimates that are much closer to 50/50. You might miss by a percent or two, but never ten.
In a very basic way, this illustrates two things: 1) Small samples are dangerous. They produce unreliable results and 2) Big samples cure the problem and provide estimates that can be relied on.
But far too often, attorneys rely on small samples as their only source of information. This either means they can’t take the win rate, average damages, fault allocation and more seriously – or worse, they take it seriously and rely on flawed data.