I wonder who decided to use a step function for the speed accuracy plot. They must have thought the convex hull would be wrong because you can't really make lin

I wonder who decided to use a step function for the speed accuracy plot. They must have thought the convex hull would be wrong because you can't really make linear combinations of algorithms (you could, but you'd have to use time not speed to make it linear). So I get why you would use step functions, but the step is the wrong way around. The current plot suggests accuracy doesn't drop if you need higher speeds