For the new cultivar release in Brazil it is necessary to carry experiments of Value for Cultivation and Use (VCU). In the VCU the experimental plot size and design is an important piece of information. Since this detail is not available for tobacco in Brazil, the purpose of this work was to establish the ideal size and design of the tobacco plot for experimental evaluation. Data was used from the final evaluation trials from the tobacco breeding programme from Souza Cruz S.A., from the variety groups flue-cured Virginia and Burley. The number of lines in the evaluation was 10, allocated in plots formed by 6 rows of 7 plants. The replication number was four per trial. From these trials data collected for individual plants included: yield (weight of cured leaves per plant), total sugar (%), and total alkaloids (%). Using the bootstrap technique and the curvature of the Variation Coefficient, the ideal size and design of the plots were defined.

From the data analysis it was possible to conclude:

• The performance of the plants changed with the position on the column and plot, although no interaction lines × position happened.
• The experimental accuracy increased with the number of plants in the plot.
• It was suggested that the ideal plots to evaluate tobacco lines were 2 rows of 10 plants without borders.