|
The other factor to consider when modeling intraday entries with daily bars is the order in which fills would have occurred. Say your system identifies 20 or more candidates each day of stocks with a "bull flag" and you want to test buying the first 5 that break above yesterday’s high. By the end of the day, 10 of them broke above the high. Which 5 should the model buy? (In practice, running such a system would require either a realtime scan or a mechanism that places the initial 20 orders and then quickly cancels the remaining 15 after the first 5 are filled.)
Some systematic traders worry obsessively about this question and even refuse to trade a system not modeled with perfect fidelity. I would again point out the law of large numbers and the degree of randomness in the market here. If you just look at the next 10 trades, it might matter a lot which 5 of the 10 you assume were filled. But if you look at the next 1000, it makes less difference which 500 are selected.
RealTest of course supports modeling this either way, depending on your preference. The key is to understand what is going on under the hood, be sure it makes sense to you, and be sure your live trading matches the model. Taking some time to study the trade list from each backtest (or even the Test Details Log if needed) is the best way to achieve this.
An excellent way to objectively measure the relevance of intraday trade entry order in your system is to use the Random function in your EntryScore formula, run the same test 100 times, and compare the results.
|
