The aims of this study were to examine the eVects of one self-selected and two enforced pacing strategies (constant and variable power output) on cycling performance during a time trial in which variable wind conditions were simulated. Seven male cyclists rode their own bicycles on a Computrainer cycle ergometer, which was programmed to simulate a 16.1 km time trial on a flat course with a 8.05 km h−1 headwind in the first half of the race and a 8.05 km h−1 tailwind in the second half of the race. Subjects rode an initial time trial (ITT) at a self-selected pace to the best of their ability. The mean power output from this trial was then used to calculate the pacing strategies in the subsequent two trials: Constant (C)—riders rode the whole time trial at this mean power output; and Variable (V)—riders rode the first headwind section at a power output 5% higher than the mean and then reduced the power output in the last 8.05 km so that the mean power output was the same as in the initial time trial and in trial C. Power output, heart rate and ratings of perceived exertion (RPE) were recorded every 1.61 km. Finish times, 8.05 km split times and blood lactate levels, pre- and post-exercise (to calculate Δ lactate), were also recorded in each trial. In the ITT, riders chose a mean ± SD power output of 267 ± 56 W in the first 1.61 km which was 14% higher than the overall race mean ± SD of 235 ± 41 W. Power outputs then dropped to below the race mean after the first few kilometres. Mean ± SD finish times in the C and V time trials were 1661 ± 130 and 1659 ± 135 s, respectively. These were significantly faster than the 1671 ± 131 s recorded in the initial time trial (p= 0.009), even though overall mean power outputs were similar (234 − 235 W) between all trials (p= 0.26). Overall mean RPE and Δ lactate were lowest in trial V (p < 0.05). Perceived exertion showed a pacing strategy by race split interaction (p <0.0001), but it was not increased significantly during the first 8.05 km of the V condition when power outputs were 5% higher than in condition C. Heart rate showed no main effect of pacing strategy (p= 0.80) and the interaction between strategy and race split did not reach statistical significance (p= 0.07). These results suggest that in a 16.1 km time trial with equal 8.05 km headwind and tailwind sections, riders habitually set off too fast in the first few kilometres and will benefit (10 s improvement) from a constant pacing strategy and, to a slightly greater degree (12 s improvement), from a variable (5% ± mean) pacing strategy in line with the variations in wind direction during the race. Riders should choose a constant power when external conditions are constant, but when there are hilly or variable wind sections in the race, a variable power strategy should be planned. This strategy would be best monitored with ‘power-measuring devices’ rather than heart rate or subjective feelings as the sensitivity of these variables to small but meaningful changes in power during a race is low.