The efficiency and selection gain of yield trial experiments is affect
ed by experimental design. A larger number of genotypes, G, increases
the chances of including markedly superior genotypes, and a larger num
ber of replications, R, reduces errors and thereby increases the chanc
es of correctly identifying truly superior material. But assuming that
cost rises approximately as the number of plots GR, containing resear
ch cost imposes a tradeoff between G and R. This paper gives tables sp
ecifying the number of replications that optimizes selection of genoty
pes with the highest true yields. It also quantifies the impact of sub
optimal choices. Even with an optimal level of replication, the value
of a selection experiment rises logarithmically with the number of plo
ts and hence the cost. Since benefits from more data follow this sever
e law of diminishing returns, statistical strategies for decreasing er
rors and improving selections are quite important.