For odds of b to 1 with win probability p: f = (bp - q)/b, where q = 1 - p.
The asymmetry that matters. Half Kelly gives about 75% of the optimal growth rate. Twice Kelly gives a growth rate of exactly zero - you have a real edge and make nothing. Beyond that, a winning strategy becomes a losing one.
That is why practitioners use fractional Kelly. If you overestimate your edge by a factor of two, full Kelly on your estimate is twice Kelly on reality.
Also: uncertainty about your edge argues for a smaller bet, not a larger one.