The aims of this study were to evaluate Plusoptix performance in amblyopia prediction, to examine the best readings for that prediction and to determine if it is sufficient to have one measure or more. Data were collected in a two-year period in a whole-screening population of children aged three to four years; 1342 were included after applying exclusion criteria. This was a cross-section study including 54.4% public and 45.6% private schools; 51.2% were male, 50.7% were aged three years and 49.3% were aged four years. The authors applied seven models; three consecutive measures (P1, P2, P3), a mean of three consecutive measures (Pmean123), maximal (worst) of three consecutive measures (Pmax123), mean of first two measures (Pmean12) and maximal (worst of the first two measures (Pmax12). All models were significant for amblyopia prediction with hit rates of 97.8-98.2%. Astigmatism and difference in sphere between eyes were predictive for amblyopia in all models. Sphere and difference in cyls was not a predictor. Receiver operating characteristic (ROC) curves for all models were significant to predict amblyopia with no difference for models. This study is novel in considering amblyopia prediction instead of risk factors and evaluates internal consistency. In using Plusoptix to predict amblyopia they recommend any of seven models. They adjusted ROC curves to achieve 100% sensitivity for amblyopia diagnosis and found P1 and 2 measures to be better than P3. Maximal value between P1 and 2 was better than P1 alone. If using just one measure, they recommend P2. The authors conclude that two Plusoptix measures are better than P1 alone for amblyopia screening. Adding a third measure has no advantage. For highest sensitivity, Pmax12 was the best model in this study.