I know that the need for a portfolio/strategy to be self-financing (the purchase of a new asset needs to be funded by selling of an older one/ones) is very helpful when attempting to price derivatives due being able to create replications of the derivative's payoffs, etc.
However, do we gain anything (such as a "wrong" but more useful/parsimonious model) from requiring this condition of a portfolio/strategy outside of valuation, specifically, when we're also not working under the risk-neutral measure? e.g. analyzing the profitability of a trading/investment strategy in practice.
Or does this assumption just move our model further away from how our portfolio/strategy would actually function due to market frictions (slippage, transaction fees, etc.) without providing a greater upside to model correctness/performance?