Methods/Basis for prediction Y

[HNick]

New to the site but figured this would be the best place to ask.

So the general idea is to be able to predict the future of an asset price right? Statistical models require, for the most part, that a "Y" be specified. A common Y for asset prices is the close price.

So my basic question is: How do we know what "Y" to use? Surely the close price can't be the best because I would certainly like to know more then just the close, like how much risk was involved getting to that close price (how much did the asset move in price before settling on the close).

I understand that trading models can be easily evaluated without having a specific Y other then the general "return" because you have specific exit and entry points but lets say I was trying to improve a trading strategy (who isn't?) and want to find another variable to help decrease the mistake of the first model.

My ideas where to look at slope (linear or expo) over the trading period or maybe something like a sharpe ratio?

Am I missing the point? Is this something everyone decides on their own?

Thanks in advance.

 

[Ash Singh]

Close is fine for very liquid instruments.
If you are analyzing daily time-series of an illiquid (or you trade a siginificant percent of the daily volume) security, VWAP is better than close.