In all populations, as the time runs, crossovers break apart ancestor haplotypes, forming smaller blocks at each generation. Some blocks, and eventually all of them, become identical by descent because of the genetic drift. We have in this paper developed and benchmarked a theoretical prediction of the mean length of such blocks and used it to study a simple population model assuming panmixia, no selfing and drift as the only evolutionary pressure. Besides, we have on the one hand derived, for any user defined error threshold, the range of the parameters this prediction is reliable for, and on the other hand shown that the mean length remains constant over time in ideally large populations.