Estimation of fitness is a key step in experimental evolution studies. However, no established methods currently exist to specifically estimate how successful new alleles are in invading populations. The main reason is that most assays do not accurately reflect the randomness associated with the first stages of the invasion, when invaders are rare and extinctions are frequent. In this protocol, I describe how such experiments can be done in an effective way. By using the nematode model, Caenorhabditis elegans, a large number of invasion experiments are set up, whereby invading individuals carrying a visual marker are introduced into populations in very low numbers. The number of invaders counted in consecutive generations, together with the number of extinctions, is then used in the context of individual-based computer simulations to provide likelihood (Lk) estimates for fitness. This protocol can take up to five generations of experimental invasions and a few hours of computer processing time.