I realize that this question may be verging on asking for the proprietary/"secret", so if suggestion of a general approach that doesn't divulge details isn't really possible, I understand.
From what I've seen of the literature, the usual approach to profiting off of modeling options seems to be of a sell-side/market-making nature.
To my knowledge, the approach usually goes like this: Some sort of potentially-modified model (e.g. BSM, Heston, etc.) is chosen, the model is then fit to quoted vanilla options prices on said underlying, then this resulting model is either used to price more exotic options (in the sell-side context) or to calibrate greeks so as to market-make, hedge, and, as a result, profit off of the spread (in the market-making context).
Let us assume that, henceforth, "options" will generally refer to European or American options, specifically, calls and/or puts.
I realize that there are ways to leverage these pricing models to make some sort of "speculative" profit by, for example: using the model fitted on something other than market prices to identify mispriced (in the context of the model) options, buying and/or selling the options accordingly, and then delta-hedging our position until (hopeful) convergence to capture the disparity.
I'm wanting to know if there is a general approach that is used in practice to predict the future value of an option or set of options, compare it to it's current price, and determine if a profit can be made through either going long or short the option or set of options without the need to dynamically (or even statically?) hedge (since this may not be feasible/possible) or invoke discounting expectations under the real-world measure (due to the numerous ways that this may complicate things).
Does this just come down to doing something like this post suggests and just choosing/creating a pricing model, forecasting the parameters over some future time horizon (say until expiration), and then plugging these forecasts into the pricing model to get the prediction of the future value of the option over said time horizon?
Alternatively, would there be anything inherently "wrong" with treating the option prices acoss strikes for the same expiration (although, this could be extended to multiple expirations as well) as a multivariate time series/sequence and then applying the various methods designed to model those to the option prices? Or do the complications inherent in non-European options (early exercise, etc.) undermine this approach (for prediction of American Options)?
Any input/references would be greatly appreciated, thanks!