This paper considers the problem of selecting good paths in a wireless mesh network. It is well-known that picking the path with the smallest number of hops between two nodes often leads to poor performance, because such paths tend to use links that could have marginal quality. As a result, quality-aware routing metrics are desired for networks that are built solely from wireless radios. Previous work has developed metrics (such as ETX) that work well when wireless channel conditions are relatively static (DeCouto , 2003), but typical wireless channels experience variations at many time-scales. For example, channels may have low average packet loss ratios, but with high variability, implying that metrics that use the mean loss ratio will perform poorly. In this paper, we describe two new metrics, called modified expected number of transmissions (mETX) and effective number of transmissions (ENT) that work well under a wide variety of channel conditions. In addition to analyzing and evaluating the performance of these metrics, we provide a unified geometric interpretation for wireless quality-aware routing metrics. Empirical observations of a real-world wireless mesh network suggest that mETX and ENT could achieve a 50% reduction in the average packet loss rate compared with ETX