In the example given, the author says that the odds of a given page having a typo is less than 3/20. Sure, but if we don't want a range, but an exact number? Th

In the example given, the author says that the odds of a given page having a typo is less than 3/20. Sure, but if we don't want a range, but an exact number? That sounds like a more interesting challenge to me. Formal statement: - You have observed N events, with 0 occurrences of X - Someone wants to make a bet with you about the likelihood of X happening - Once you've quoted a number, your counter-party then has the option of making an even bet about whether the actual likelihood is greater than or less than your prediction Eg: If you predict 3/N using a 95% confidence interval, then 95% of the time, the actual likelihood will be less than 3/N. Your counterparty will then win the bet 95% of the time, simply by predicting it to be lower. Your ideal strategy would be to quote a likelihood which is over/under 50% of the time, not 95%. Ie, you want to pick E such that 50% of the time, it matches the observation you've made (no occurrences), and 50% of the time it does not. E^N = 0.5 N log E = log 0.5 log E == log 0.5 / N E = 0.5^(1/N) For the example given, that comes out to 0.966. Ie, there's a 96.6% chance of no typos in a given page. Across 20 pages, this comes out to 0.966^20 => 50% chance of no typos. If your goal is to quote the single best estimate which can hold up well in a betting market, I believe this would be the ideal strategy