[Purpose] To investigate clinical measures as valid predictors and discriminators of the level of community ambulation of hemiparetic stroke survivors. [Subjects and Methods] Seventy-five hemiparetic stroke patients were separated into a community ambulation group (>0.8 m/s) and a limited community ambulation group (0.4–0.8 m/s). The dorsiflexor strength of the affected side, Sit to Stand (STS) test, Timed Up & Go (TUG) test, Berg Balance Scale (BBS), Fugl-Meyer Assessment (FMA), Falls Efficacy Scale (FES), 2-min step test (2mST), and 6-min walk test (6mWT) were used for evaluation. The discriminative powers of the tests and measures were investigated using the receiver-operating characteristic (ROC) curve, and odd ratios were calculated to predict the level of community ambulation. [Results] The cutoff values for predicting the level of community ambulation (>0.8 m/s) were <14.77 s for TUG, <12.6 s for STS, a score >46.5 for BBS, a score >25.5 for FMA, a score <13.5 for FES, >7.5 kg for dorsiflexor strength, >30 times for 2mST, and >318 m for 6mWT. All clinical measures except FES had moderate accuracy according to the area under the curve of 0.76–0.88 (70–93%). [Conclusion] Clinical measures (except FES) have moderate validity in predicting the level of community ambulation of stroke survivors.