Natenberg had this to say; Over long periods of time the HV of the underlying contract will be the dominant factor affecting IV. Over short periods of time, however, other factors can also play a significant, perhaps even a dominant, role. He gives as examples economic data releases, meetings of economic ministers, earnings news, possibility of success or failure of new products, takeovers, etc, basically sentiment driven.
They only go up in the sense that the Black Scholes pricing model says they do. In this regard, this model may be, and is, imperfect. You could easily "fix" it to give implied volatilities that were more uniform, but then it would be less elegant mathematically. Or, intuitively, a lot of short term volatility averages out over the longer term.
Yes, Natenberg says both HV and IV are mean reverting and the graphed data provided supports that contention. Million dollar question: can one successfully trade the skew without access to historical IV?
I'm intend to simulate with data, but I think a lot of it is also going to be statistics: - am I making a small enough bet size in each individual trade, with the protective wings close enough in, to avoid ruin if I get an unusual string of adverse moves?
Ah, my comfort zone, trade management. I have been reading of an alternative approach whereby one cuts loss at a percentage of premium and similarly takes profit at a higher percentage of premium, thereby preserving the desired R:R ratio. This of course means more active trade management is required and leaves one open to the possibility of whipsaws. Unfortunately I have not been trading options long enough to offer an opinion as to which might be preferable under various circumstances.