I dont know how aggressive liquidity consuming hft does it. In doing market making and trying to make spread or narrow it, so i can provide liquidity to market order submitters and try to balance my inventory. It should work as long as average spread is greater than the volatility.. if volatility is too big adverse selection eats the profits
Since you probably are one of the guys posting quotes. Is NBBO quote data useful for backtesting if considering all bid and ask quotes as valid price points?
I'm not sure, I do forward testing with stochastic simulation calibrated from historical data. I'm just now going live so its still a bit of an experiment for me. I need to be coding some unit tests right now actually cause I still have a bug where its reacting to its own orders
For best results you should probably use the full order book data, using the BBO only the results would be not as accurate, there are some papers on the topic, but depending on what your doing BBO data can be just fine