Or we have to construct/synthesize the data ourselves? (i.e. write our own "naive" VWAP algorithms, and of course accordingly, also write our own "naive" VWAP execution algorithm for order execution? Thanks a lot!
Because in back test I use closing price, but I won't be able to hit closing price exactly in real trading, instead of attribute all deviations into slippage, maybe VWAP will give me something better...
If you are concerned about accurately trading at the settle price, why would you not also be concerned about accurately trading at VWAP?
We have found that using VWAP (or TWAP) when creating systems helps us remove some of the tendencies to overfitting to data. The argument behind this is that, provided that you believe that the price generating process includes a random component, the VWAP price gives you a more stable sample of a days price. This means that when you evaluate how good the signals from the system are, you will not rely upon using just one sample - the close price.