Unless you’re a gang member or drug addict, driving is your top risk. But which risks can you handle and which can kill you? Since 96% of collisions are caused by human error and 75% are unconscious (the driver is not aware of the risk), we used one billion miles of real-life driving data to find out, feeding it into a model fusing 26 road context, driver action/attention and vehicle dynamics factors to predict collisions, near misses and safe driving. We found that drivers can handle most single risks, but multiple simultaneous risks can be deadly. Some combinations are 1,000 times more dangerous than regular driving. Our “SAFER” model can be fitted to any late-model vehicle. It can be trained with labeled data from historical trips or through reinforcement learning from observations of excellent and poor drivers. A time series matrix can help you see a few seconds into the future.