Just to directly address this - the key thing wrong in this formula is that it's ^2*state, not ^state. The state transition table operates both on the current state and the input you read from the tape, so for a binary turing machine with 5 states, you have a 10-entry transition table.
Ah, yes, of course the read bit should move into the exponent, since it's an input, not an output of the function. But the key point I was making is that there exists a formula. (I don't really care what the formula is.) The part I was not understanding was the complexity of "essentially different".