Inexperienced helmsmen often oversteer because of the lag that occurs between changing the rudder angle and a change in the vessel's heading. Predictive displays are a common way of mitigating the effect of lag on human control. Accordingly we developed a predictive display to show the position and heading of a vessel a short time in the future. With this display, the helmsman's task becomes that of keeping the predictor on the planned path. In effect, the predictor is steered, not the vessel. Our predictive model was statistical and based on data gathered from a 40 foot survey vessel carrying out a variety of maneuvers while the position, heading, speed through the water and rudder angle were continuously recorded. The advantage of such a predictor is that it can, in principle, be generated automatically, without any need for a model of hull shape or vessel dynamics. We evaluated the predictor by having both experienced and inexperienced helmsmen steer a pre-defined figure-of-eight course. The results showed substantial reduction in cross track error for inexperienced participants to the point that their performance was indistinguishable from those that were more experienced.