It is well known that the maximum likelihood (ML) detection for multiple-input-multiple-output (MIMO) systems requires high computational complexity. As a result, many suboptimal algorithms, achieving a tradeoff between computational complexity and performance, have been proposed. Among these algorithms, the sphere decoding (SD) detector is considered as a very efficient solution. However, the throughput of the SD is not constant and it greatly depends on for channel conditions. In general, the variation of the throughput is very large. As a result, its hardware implementation can be difficult or inefficient. In this paper, a low-complexity ML-like detection algorithm is proposed. While the proposed algorithm has a constant throughput, it can be computationally more efficient than the SD detection. The other advantage that the SD decoder doesn't have is that the proposed algorithm allows an easy tradeoff between complexity and performance.