- Data for testing
- Point A
- Failed Point A
- Point C
- Pivot range
- W-pivot range -new
- Points A and C through the pivot
- Closing method -new
- Dynamics of Points A and C
- Significant time frames
- Entry and exit
In this chapter I will try to answer the following questions related to the ACD system.
First, I will briefly describe the testing strategy. As already mentioned, the historical data of 1 minute bars for 10 years, starting in January 2009, was used. The program calculated trading indicators, such as the number of trades, the percentage of profitable trades, profit, p/l, the average profitable and loosing trade for different opening range time frames and breakout levels.
Understanding how the program calculates point A is best done using an example. Suppose the opening range is 20 minutes. Then, if the market breaks the upper border of the opening range, it goes further, say, 7 ticks up, and the low of 10 consecutive 1-minute bars is above the upper border of the opening range plus 7 ticks, then A up, equivalent to 7 ticks, is established. The trade closes at the last minute of the trading day. Such a closure is chosen simply to compare results for different parameters, in our case it is the opening range time frame and breakout level. Although in real trading, it hardly makes sense to close at the last minute of the trading day. Only one Point A is allowed during a trading day. Point A can be established to the penultimate minute of the trading day so that closing can be done at the last minute. This is also not applicable in practice, but formally does not contradict to the ACD theory. If Point A is established and the market breaks through the opposite side of the opening range by 1 tick, the program closes the trade, simulating stop loss at Point B. Hereinafter, transaction costs are not taken into account. The program sequentially iterates over the breakout level values for potential Point A from 0 to 100 ticks. Point A corresponds to the breakout level that brings maximum profit for the whole history.
The test results are presented in Table A.2 in the Appendix for SBER stocks issued by Sberbank, the main bank of Russia, as an example. These stocks are the most liquid on the Moscow exchange and quite volatile, which is a necessary condition for the applicability of the ACD system. The duration of the opening range varies from 4 to 40 minutes, and the breakout levels above or below the opening range for potential Point A are from 0 to 100 ticks. The breakout levels are grouped at intervals, such as 0-10, 5-15 ticks, etc., to smooth out random deviations of Point A and not overload you with a large amount of data (imagine a table with 100 rows for levels from 0 to 100 ticks, multiplied by 5 timeframes of the opening range of 10 pages!). The last row in the table for each opening range time frame averages the results for all breakout levels. The first column of the table shows the average size of the opening range in ticks after the slash. Obviously, it increases with the duration of the opening range.
From Table A.2 obtained for SBER stocks one can make the following observations:
I made similar calculations for different stocks and futures. The values of Point A and its range in terms of profit for different opening range time frames and securities are presented in Table A.10 in the Appendix. The 1st and 2nd row for each security contain A (C) value and its range, respectively (as already mentioned, it is better to consider the range of A values, such as 0-10, 5-15, etc., to avoid random deviations of Point A). As can be seen from the table, Point A and its range are not the same for different securities. Point A does not always lie in the range due to random deviations. Not only A value, but also its range varies quite unpredictably depending on the duration of the opening range, although they generally decrease with increasing the opening range time frame.
The value of Point A does not depend on the average size of the opening range in ticks. For example, the A value for an MGNT stocks with an average open range of 166 ticks is 7 ticks, while for GMKN stocks with an average open range of 70 ticks, it is 64 ticks, with the same 20-minute open range for both.
The average profit and p/l for breakout levels from 0 to 100 ticks for different opening range time frames and securities are presented in Table A.11 in the Appendix (see the 1st row for each security).
The total accumulated profit for about 10 years and profit per trade from breakout levels from 0 to 100 ticks for different opening range time frames and securities is shown in Fig. 2.1 and Fig. 2.2 respectively.
As can be seen in Fig. 2.1 and Fig. 2.2, the maximum accumulated profit and profit per trade usually occur for a higher opening range time frames and breakout levels. This is especially true for profit per trade, because the number of trades decreases sharply. For different stocks and futures, the optimal duration of the opening range in terms of profit varies, the 4-minute opening range is usually the worst in this regard. The maxima of the curves in Fig. 2.1 corresponding to point A are highly blurred. The accumulated profit by days from Point A is displayed in Fig 4.1 in chapter 4.
As shown in chapter 9, a fairly long history of at least 1000 days should be used to calculate Point A. Let's compare the accumulated profit from point A, calculated over 1000 days, with a profit from two breakout levels of 0 and 15 ticks for different stocks and futures. As you remember, Point A corresponds to the breakout level that brings maximum profit during the period used to calculate it, in our case 1000 days. Profit from Point A is calculated on the day after the 1000-day period. Breakout level of 15 ticks was chosen, because in The Logical Trader you can most often find low values of Point A.
The accumulated profit by days from Point A, recalculated daily and two breakout levels is shown in Fig. 2.3. The number of trades "n" and the average profit per trade "p" in ticks are indicated for each line. The accumulated profit from Point A is greater or less than from breakout levels, depending on the security. All lines are similar in shape.
As indicated on mbfcc.com, a good approximation of the current A value for any stock will be to take the 30-day average range H-L and use 20-25% of this value. Let's compare the accumulated profit from Point A based on the maximum profit for 1000 days and from Point A as 20% of H-L for 30 days, for different stocks. The results are shown in Fig. 2.4. It’s hard to declare a winner which Point A is better. By the way, both Points A are not correlated, as shown in chapter 9. For Point A, based on volatility, it should be noted that it varies over a wide range; its maximum value often exceeds 100 ticks. In practice, it hardly makes sense to use such high A values to open a trade and recalculate it daily. While Point A, based on a maximum profit for 1000 days, can be relaculated once every 6 months - see chapter 9.
Fig. 2.3 and Fig. 2.4 show losses for some securities. This does not mean that the ACD method does not work for these securities. It works in trending markets and can not be used constantly, while backtesting is carried out not depending on market behavior. To make a profit, you need to exit not at the end of the trading day, but in a different way. To enter a position, it is advisable to use additional conditions, for example, pivot. This is discussed in chapter 7 and chapter 8.