Cost-efficient and low-power sensing nodes enable to monitor various physical environments. Some of these impose extreme operating conditions, subjecting the nodes to excessive heat or rainfall or motion. Rough operating conditions affect the stability of the wireless links the nodes establish and cause a significant amount of packet loss. Adaptive transmission power control (ATPC) enables nodes to adapt to extreme conditions and maintain stable wireless links and often rely on knowledge of the received power as a closed-feedback system to adjust the power of outgoing packets. However, in the presence of a significant packet loss, this knowledge may not reflect the current state of the receiver. In this article, we propose a lightweight n-step predictor that enables transmitters to adapt transmission power in the presence of lost packets. Through extensive practical deployments and testing, we demonstrate that the predictor avoids expensive computation and still achieves an average prediction accuracy exceeding 90% with a low-power radio that supports a transmission rate of 250 kb/s (CC2538) and 85% with a low-power radio that supports 50 kb/s (CC1200).