AbstractTwo sets of weather patterns describing variability in 850 hPa winds in Southeast Asia are presented and compared. Patterns are calculated using EOF/k‐means clustering with and without imposing a separation between planetary‐scale and regional‐scale circulation features. The former are labelled as tiered patterns while the latter are referred to as flat. The ability of the patterns to distinguish between known modes of tropical circulation variability is examined. This includes climate modes such as the seasonal monsoons, the El Niño Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) as well as sub‐seasonal modes including cold surges, phases of the MJO and Boreal summer intraseasonal oscillation (BSISO), tropical cyclones, Borneo vortices and equatorial waves. All these modes are well captured by the weather patterns except for the equatorial waves and the IOD. The tiered patterns are shown to better describe large‐scale modes of variability, while the flat patterns better describe the synoptic variability. Both sets of weather patterns are then used to study the likelihood of heavy precipitation depending on synoptic circulation by considering the regime‐conditioned probability of high‐percentile precipitation using the satellite‐derived Global Precipitation Measurement (GPM) dataset. It is shown that the pattern centroids explain up to 10% of the seasonally anomalous precipitation over land, and that a perfect weather pattern forecast would outperform a perfect MJO forecast. These weather patterns show promising potential in extending the useful forecast range for the risk of heavy precipitation, dependent on their forecastability.