I think the 'nonlinearity' they are referring to is the difference between expected value and probability. It's possible that A beats B most of the time, and yet the average amount by which A beats B is negative, because when A does lose it loses big.
That’s mis-understanding the effect. It has nothing to do with “losing big but rarely” you could put 20s instead of 6s and it wouldn’t change a thing. It’s just that when you roll A against B it will win more than half, but B against C will have B winning more than half, and finally C vs A will have C winning more than half.
