Abstract: Adding a trace amount of polymer into turbulent flow can significantly reduce wall friction. This drag reduction technique has been widely applied in fields such as fire-fighting, pipeline transportation, and biomedicine. Polyethylene oxide (PEO) is an efficient drag-reducing polymer, whose performance is affected by multiple parameters. In this study, a gravity-driven circulating pipe-flow system is employed to experimentally investigate the drag reduction characteristics of PEO solution injection in turbulent pipe flow. The effects of Reynolds number, injection angle (seven angles), injection rate and relative molecular mass (total of 7 kinds) on the drag reduction rate (R_D) are systematically examined. A normalized polymer flux K_p, which is suitable for pipe flow, is proposed to collapse the experimental data. Results show that R_D initially increases roughly linearly with lg K_p and then approaches a saturation level. This trend is analogous to the previously reported K-scaling law for polymer injection in turbulent boundary layers. Moreover, the dependence of R_D on molecular weight exhibits an S-shaped trend. By fitting the data with a sigmoidal function, the optimal molecular weight range for maximum drag reduction can be predicted. These findings provide useful guidance for the optimization and prediction of polymer injection parameters in drag-reduced turbulent pipe flows.