The inheritance of genes in sexual reproduction has a major chance component. During meiosis homologous chromosomes separate and migrate to different poles. The haploid daughter cells each have one allele for each gene, but which allele they have is random. Essentially meiosis is like flipping thousands of coins and getting either a head (one allele) or a tail (the other allele) for each one. At a single locus each parent passes on either a head or a tail with a 50% probability of each. Over an entire population each allele should be passed on proportionately to its frequency. In other words if 10% of the alleles in a population are A then allele A should show up in 10% of the gametes.
However, results do not always follow probability, especially with small sample sizes. The results of one coin flip are independent of the results of the next coin flip. If a coin is flipped 2 times it is not unlikely that the results will be 2 heads or 2 tails. If a coin is flipped 10 times it is very unlikely that the result will be 10 heads or 10 tails. And if the coin is flipped 100 times it is so unlikely that all the flips will be tails (or heads) that for practical purposes it can be regarded as impossible. Similarly in a small population, random chance can significant change the frequency of alleles in a short time. In a large population, genetic drift has only very small effects in any given generation.
http://glencoe.mcgraw-hill.com/sites/9834092339/student_view0/chapter20/simulation_of_genetic_drift.html