What the hell is he talking about? In the first graph, the trend is clearly a negative correlation between hours of sleep and accidents, even if you include the 5 hour data point.