Lolium perenne (perennial ryegrass) breeding programs are primarily based on recurrent selection to develop improved populations. For accurate phenotyping of many target traits in these programs, measurements need to be performed on populations planted in swards, as opposed to single plants. Resolving the allele frequencies within these populations at multiple SNP locations may allow us to associate the allele SNP frequencies with the phenotype. In order to test this approach, a strategy is required to accurately determine allele frequencies in perennial ryegrass populations. We have used a genotyping by sequencing (GBS) approach that involved genome complexity reduction with restriction enzymes, barcoding fragments to allow multiplexing, and pooling samples for sequencing on the Illumina GAII and HiSeq 2000. Different enzymes were tested to generate a range in the extent of genome complexity reduction. We have applied this approach to eight perennial ryegrass cultivars to get an insight into expected SNP harvest and the coverage required to accurately determine allele frequencies. The developed strategies will ultimately be used to characterise genetic variation in a training population being used to develop models for genomic selection.