To assess the wind power resources of a large continental shelf area, we analyze the 18-year hourly wind records from meteorological stations in the US Middle Atlantic Bight (MAB), comparing areas of coast, estuary, and open shelf. We calculate winds at turbine hub height for the sea breeze compared with synoptic winds and, for each type of site, we compare the seasonal and daily phase match to electrical load. To improve large-scale ocean power resource calculations, we derive an iterative algorithm to determine the surface roughness coefficient (z0). Our method calculates z0 for specific times and locations over the ocean, rather than the prior practice of using a generic z0 that is constant across time and space. Due to lower surface roughness of the ocean, wind speeds are notably higher at hub height, so that in the MAB we find that a representative open shelf site has three times the power content of a nearby land site. Regarding phase match to daily electric load, we find the sea breeze adjacent to the coast is a very good match to this region's electric power load profile. However, the open shelf wind speeds are so much higher (10.9 m s–1 versus 5.7 m s–1 for the comparison period) that the near-coast phase advantage is obviated. We also find more consistent wind power production offshore, with single sites producing at least some power 88 to 92% of the time. By modeling electrically interconnected sites, power production improves to 96.3% with as few as three interconnected wind sites and to 99.3% with 5 interconnected sites.