I would be interested to see actual "real world" data vs "standard crash test" data. It's much easier to design something for excellent performance on a small

I would be interested to see actual "real world" data vs "standard crash test" data. It's much easier to design something for excellent performance on a small subset of rigorously defined tests, vs designing for any/every possible outcome.