Background and purpose Preoperative posterior tilt in undisplaced (Garden I–II) femoral neck fractures is thought to influence rates of reoperation. However, an exact method for its measurement has not yet been presented. We designed a new measurement for posterior tilt on preoperative lateral radiographs and investigated its association with later reoperation. Patients and methods A consecutive series of 113 patients, ≥ 60 years of age with undisplaced (Garden I–II) femoral neck fractures treated with two parallel implants, was assessed regarding patient characteristics, radiographs, and rate of reoperation within the first year. In a subgroup of 50 randomly selected patients, reliability tests for measurement of posterior tilt were performed. Results Intra-and interclass coefficients for the new measurement were ≥ 0.94. 23% (26/113) of patients were reoperated and increased posterior tilt was an accurate predictor of failure (p = 0.002). 14/25 of posteriorly tilted fractures ≥ 20° were reoperated, as compared to 12/88 of fractures with less tilt (p < 0.001). In multiple logistic regression analysis including sex, age, ASA score, cognitive function, new mobility score, time from admission to operation, surgeon's expertise, postoperative reduction, and implant positioning, a preoperative posterior tilt of ≥ 20° was the only significant predictor of reoperation (p < 0.001). Interpretation The new measurement for posterior tilt appears to be reliable and able to predict reoperation in patients with undisplaced (Garden I–II) femoral neck fractures.