Quote from rufus_4000:
Without going into the whole concept of market behavior modeling (which is a research top all by itself). let's talk about one "theoretical approach", beating VWAP by acting as a market maker. In a LMM (lead market maker) allocation matching algorithm, the benefits is obvious, the order will get filled at inside Bid (or inside Ask), therefore effectively beating the VWAP. In a non-LMM allocation matching algorithm, it just becomes a bit harder, hence the need to get out of undesired (aka above volume avg) positions, which can be done (in most cases except extreme volatility periods).
Of course, transaction costs aside, it is not a trivial matter to get up and running with a market making operation (and that's putting it mildly). But in terms of beating VWAP, that's one "theoretical" way.
Without using an agency model, look at the works by Neil Chriss and Robert Almgren on portfolio execution strategies, and ITG reports on market impact. These papers are by no means pragmatic, but they give you an glimpse into what the industry is working on.
As an aside, this is an extremely well studied problem in both research and practice for the past 7-9 years. why are you working on something like this now?
