The P300 of the event-related potential is typically obtained for infrequent target stimuli that are embedded in a sequence of frequent irrelevant stimuli. The P300 has been suggested as a marker of high-level cognitive processing and might be useful in ophthalmology to confirm the diagnosis of a functional disorder. However, typical P300 measurements require relatively lengthy recording sessions. It would therefore be desirable to minimize the required time and to maximize the signal-to-noise ratio by finding the optimal balance between parameters such as stimulus probability and the number of target trials or the recording time. This is different from previous studies, which assessed the amplitude only. We recorded event-related potentials to visual stimuli using standard oddball paradigms with various target frequencies ranging from 2:1 (target majority) to 1:16 (massive non-target majority). We compared the signal-to-noise ratios for a fixed number of target trials as well as for a fixed total recording time and assessed effects of the immediate stimulus history. As expected, P300 amplitudes depend strongly on target infrequency. This did not reach saturation within the range tested. For a given number of target trials, the signal-to-noise ratio also increases with target infrequency. For a given recording duration, the signal-to-noise ratio is optimal around 1:8. In the 1:4 condition, the signal-to-noise ratio can be improved by excluding trials that were preceded by a target trial.