The only cost-effective way to control inbreeding in ex situ forest tree plantations is often to allocate trees in such a way that the possibility of close relatives mating is small and, consequently, inbreeding does not increase too much over time. The classical permutated neighbourhood methods look for the configuration in which no ramets of the same genet are planted in the surroundings (neighbourhoods) of a particular tree but deny the influence of more distant trees. Another limitation of these methods is that they cannot incorporate any other genetic (e.g. kinship) or ecological (e.g. phenology) information. We have developed a new method based on the minimisation of the global probability of generating inbred offspring for the whole population. Improvements of this method from the classical ones are: (i) it takes into account all the trees (whether near or far) and not only the neighbours; (ii) different pollen dispersal functions can be implemented, fitting the particularities of each species and population; (iii) it allows for the integration of all available information about the genetic relationship between trees; and (iv) it is flexible allowing for particular crosses to be banned or encouraged. The novel method showed a better performance than classical ones both for simulated data and a case study under a broad range of scenarios. Magnitude of the benefit depends on the actual and assumed parameters for the pollen dispersal function and the relationship between trees, but even in the simple case where only clone identity is considered some advantage can be obtained by implementing the new algorithm.