This paper proposes a novel Scale and Location Sensitive (SLS) loss and a simple Multi-Scale Head (MSHNet) for infrared small target detection (IRSTD). The SLS loss addresses the limitations of existing loss functions by incorporating scale and location sensitivity. Specifically, it introduces a weight for the IoU loss based on target scales and a penalty term based on the center points of targets to improve localization accuracy. The MSHNet is a simple modification of the plain U-Net with a multi-scale head that produces multi-scale predictions. By applying the SLS loss to each scale of the predictions, MSHNet outperforms existing state-of-the-art methods. The SLS loss is also effective when applied to existing detectors, demonstrating its generalization capability. The proposed method achieves better detection performance, lower inference time, and fewer floating point operations compared to existing methods. Experiments on two datasets show that MSHNet achieves the best results in terms of IoU, detection probability, and false alarm rate. The SLS loss is shown to be effective in distinguishing targets of different scales and locations, leading to improved detection performance. The method is simple and efficient, making it suitable for real-time applications.This paper proposes a novel Scale and Location Sensitive (SLS) loss and a simple Multi-Scale Head (MSHNet) for infrared small target detection (IRSTD). The SLS loss addresses the limitations of existing loss functions by incorporating scale and location sensitivity. Specifically, it introduces a weight for the IoU loss based on target scales and a penalty term based on the center points of targets to improve localization accuracy. The MSHNet is a simple modification of the plain U-Net with a multi-scale head that produces multi-scale predictions. By applying the SLS loss to each scale of the predictions, MSHNet outperforms existing state-of-the-art methods. The SLS loss is also effective when applied to existing detectors, demonstrating its generalization capability. The proposed method achieves better detection performance, lower inference time, and fewer floating point operations compared to existing methods. Experiments on two datasets show that MSHNet achieves the best results in terms of IoU, detection probability, and false alarm rate. The SLS loss is shown to be effective in distinguishing targets of different scales and locations, leading to improved detection performance. The method is simple and efficient, making it suitable for real-time applications.