I found something interesting during my back testing. Any advice is appreciated! I have an automated strategy that place ~200 buy orders every day. According to my back testing, the average PL% per trade is 0.32% and Sharpe Ratio is around 2 if I estimate the entry price as the 1-minute vwap during the 1 minute bar after the trade signal is generated. To clarify, the vwap I am using is calculated as: turnover during the minute / volume during the minute However, if I estimate the entry price as the average ask price of NBBO between the 4th second and 5th second after the trade signal is generated, the average PL% per trade is lowed to 0.24% and Sharpe Ratio is around 0.4. For example, if the trade signal is generated at 10:00:05, the entry price estimated by vwap is calculated as: turnover between 10:01:00 to 10:02:00 / volume from 10:01:00 to 10:02:00. The average ask price of NBBO is calculated as the average of hundreds of data points of NBBO ask price between 10:00:09 to 10:00:10. I am a bit surprised that the entry price estimated by vwap is lower than the average ask price of NBBO by 0.08%, which is a lot for this strategy. If what I found is correct, does it mean I should not place market orders or marketable limit orders to enter the position whose entry prices are the ask price? If yes, how can I get my entry price close to the 1 minute vwap calculated as above? I am using so-called zero-commission brokers at the moment which may be an inherent limitation for this type of strategy. If this strategy is so sensitive to slippage, do I have to use prime brokers for such strategies? Thanks!
Why would you be surprised that VWAP is lower than the average ask? VWAP includes trades happening at both the bid and the ask, as well as inbetween. With zero commission brokers there's basically nothing you can do. Backtesting against the ask, as you have done, is the correct thing to do here imho. With direct access you can do better in theory but that doesn't mean it's easy. You basically need to be HFT to do this yourself. Otherwise, look into smart order routers and algorithms. IBKR has pretty good price improvement, if you are trading liquid names then you might do pretty well using their adaptive algo.