A Wi-Fi fingerprint, measured as a vector of received signal strength (RSS) values of signals from access points (APs), has been used in Wi-Fi indoor positioning systems. However, a fingerprint collected in a large indoor venue with a huge number of APs would suffer from high contention in the wireless channel, which leads to defective fingerprints with missing RSS values from a certain APs and thus inaccurate estimation of user's location. This paper proposes Selective AP Probing (SAP) scheme, which reduces the contention and thus solves the missing AP problem. Through the performance evaluation using actual data consisting of 1734 APs, it is shown that the proposed scheme reduces the positioning error by up to 55%. In addition, SAP reduces control and management traffic for indoor positioning and thus avoids the throughput degradation problem during the probing stage.