There is one common thing among lotteries from all over the world: there is small number of winning tickets and considerably bigger number of losing tickets. Therefore, the probability that a ticket wins a lottery is quite low, usually so low that we think that it is almost sure the ticket loses. But, we would never say that we know that a ticket will lose, until we see results of the lottery in, for example, some newspapers. And the probability of newspapers making a mistake does not seem to affect our knowledge claims. But why is that, since newspapers could make a mistake more often than a ticket wins? This question presents trouble for fallibilism, which claim that S could know that p, even when the probability that p is less than 1. Contextualist theories give their typical brand of solution: we have a change of context between the two cases, and in one case standard for knowledge claims are higher than the standard in the other case. Because of that, one can know that S lost the lottery when she reads it in newspapers. In this paper, I will present analysis of the lottery paradox, and two types of epistemic contexutalism: simple conversational contextualism and inferential contextualism. I will also present two of the most popular solution based on simple conversational contextualism, made by Lewis and Cohen. Finally, I will introduce some problems for such solutions, and show that the problems could solved if we apply strategy and explanation of inferential contextualism, type of contextualism proposed by Michael Williams. [Projekat Ministarstva nauke Republike Srbije, br. 179041: Dinamički sistemi u prirodi i društvu: filozofski i empirijski aspekti]