Quantitative trading which we will often use this one step "back-test", is to use past historical data to test the profit and loss of a policy, and then use this to estimate the profit and loss profit and loss when the firm. However, this back-tested way there are a lot of pitfalls.
There are many online now known as the "Victory Act" strategy, winning high when backtesting profit retracement extremely low. But if you really take these strategies do real trading, it is likely to lose the final to go bankrupt. The reason behind these winning method mostly used the "next function" that trap.
The so-called future functions can be divided into two categories.
Case
One is the obvious next function is a reference to the historical data "part of the future." Such as policy based on historical know t + 3 is a huge candle, and therefore issued a buy order in t day. This transaction can not be issued on the spot in real intraday trading signals, because the program does not know the ups and downs of the next three days the situation. But only when they do come to t + 3 rose when we come in t Day "show" out of the buy and sell signals, which is naturally Wan try Souls up.
another situation
Another situation is relatively hidden future functions. This next function is often used to represent a current value of future values, but in fact the value of the future is constantly changing, and therefore according to the current value calculated strategy trading point is unstable.
Here is a simple example, many people like to use the "closing price the day" to calculate the point of sale. If history is back-tested, so no problem, because history's closing price is unchanged. However, when you know how trading day closing price of it? Most programs can only be a default of the current price is the latest closing price for the day. The problem is, when the policy of the transaction, the latest price is not equal to the final closing price! This function exists in the future, will lead to the sale of the signal point frequently appear and disappear. If a sale is unstable, then it shows that it does not comply with the core logic strategy. At that time point of view may be reasonable, the results point of sale appeared. But with the back of the data, this sale point-of-policy, the result of the sale point disappeared. But once the transaction is not like trading point that even if the revocation. If the final sale point proved to be wrong, then the transaction is also a great opportunity loss.
How to solve?
To solve the problem of future functions, there are several ways.
One strategy to get the source code to identify whether there is the next function codes manually or by a program of analysis.
Another simple way is to first simulated trading for some time, to see if it is the case the transaction is unstable existence.
Of course, if both can not get the source code, there is no way to simulate the disk test, the most simple identification method is to look at winning trading strategy. If a winning trading strategy unusually high, it is almost certain that the policy contains a future function. Otherwise, other people make money, how would such a strategy made public yet?
Source: quantitative investment club
Further Reading:
1. a quantitative strategist Confessions (Good text strongly recommended)
2. classic quantitative trading strategies available in the market are here! (Source)
3. futures / stock data Daquan query (History / real-time / Tick / finance, etc.)
5. From the high-frequency trading to quantify, can not read five books
6. HFT four factions Big Secret
