Racetrack memory is a promising technology that offers high performance, low energy consumption, and high density. However, the current-driven movement causes the ports and data to be unaligned, leading to stop-in-middle errors. In this work, we find a previously unknown error when stop-in-middle happens, read-in-middle errors, that the in-middle read is sub-stable making the judgement of alignment unreliable. We further theoretically demonstrate that read-in-middle errors commonly occur within the middle 8 \(\%\) region of the domain wall, and this has been supported by experimental results. Since read-in-middle errors only occur during 0-to-1 transitions in the data, we propose a slothful data encoding method, called SEER, which reduces the proportion of 0-to-1 transitions in data, thereby mitigating the occurrence of read-in-middle errors. We specifically design a simple hardware encoding and decoding circuit for SEER, which supports low-power, high-concurrency encoding. Experimental results show that the tracks with our encoding improve 6 \(\%\) reliability for 8-bit data, while energy consumption is only 0.04pJ.