Deploying a high-frequency (HF) radar on one anchored floating platform for the maneuverability appears to be a good alternative. In this case, because the yaw rotation during one coherent integration time may be rather large, an effective method to get precise direction-of-arrival (DOA) estimations is important for reliable ocean surface current mapping. In this article, we propose an adaptive data channel conversion method to compensate for this kind of rotation. In this method, radar data can be converted from rotating antenna channels into pseudo-fixed beam channels by following the minimum-mean-square-error rule between the wanted pseudo-fixed beam and a reference beam. Simulations with a uniform circular array and one kind of yaw rotation, the range of which is 90 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> , are conducted. After compensation, the DOA estimation of one simulated object becomes unbiased and performs nearly as good as that on a fixed platform. Besides, the high root-mean-square errors of radial ocean currents, caused by the yaw rotation, also decrease. Furthermore, we introduce a field experiment of one anchored floating HF radar, during which the DOA estimations of two onshore radars' direct waves improved drastically. The field ocean currents, deduced from this floating HF radar, also indicate the feasibility of the proposed method. This method offers a new idea to compensate for a floating platform's yaw rotation and can help introduce the traditional onshore HF radar to a much more flexible offshore scenario.