The case of the possible 11 yr periodicity in the historical light curve of OJ 287 over a century of observations is examined critically largely independently of possible models that may explain it. Although it is evident that there is at least a quasiperiodicity in the light curve because of the regular spacing of the nine observed outbursts, attempts to estimate an exact period have been unconvincing. We show that the principal reason is that the data sampling is too uneven to permit definitive conclusions to be obtained from standard data analysis methods such as Fourier transform analysis. However, it is also shown that this is at least in part because the five most recent light-curve events do not refer to the same outburst in the light curve as the previous ones. A simple statistical method is defined to calculate a best period from both the primary and secondary outbursts. The result is a best mean period of 11.844 yr, although there is some evidence that the period for the primary maxima may be significantly longer than this value. While the periodic model of the light curve has so far demonstrated considerable predictive power, final judgement should be reserved until after this predicted maximum in 2006.