This paper presents simulation model that estimates average text entry time and average number of user inputs per character for an unambiguous virtual keyboard with different character predictors. The simulation is focused on VK operated by binary user inputs (ON,OFF), that could come from physical buttons, wink detection systems, etc. The model allows us to use different predictors (k-grams, n-grams or any combination of them), prediction list scanning methods, lengths of prediction list and prediction turn-on delays. The simulation give us optimal list length and prediction delay for different predictors and scanning methods.