Although device-free human activity recognition (HAR) has been among the commonly investigated HAR methods in the past few years, a neglected topic in this field is the sensor placement in real environments. To address this, we investigate the effect of sensor placement on the recognition performance of fine-grained human activities in a real cooking environment. We used a WiFi-based system that recognizes cooking activities from the Channel State Information (CSI) and the Received Signal Strength Indicator (RSSI) reflected by the user. We experimented with three sensor placement strategies, each having the user and the performing area in line-of-sight (LOS) and non-line-of-sight (NLOS), leading to 6 different placements. The results show a statistically significant difference between the distribution of CSI and RSSI features for different fine-grained activities when the user and the performing area are in the line of sight.