Testing the Accumulation Signal on the Top 20 ASX Stocks across 10 Years
There are many ways of trying to get value out of a particular signal, particularly when that signal may be only one component in a more complex trading system. For example, a signal might be valuable in a trading system as a trend detector, an entry signal, a filter, an exit signal, a money management tool and the list goes on.
Therefore, although a tool may not be useful in one role, nonetheless, that same tool may be quite valuable if used in a different manner. We see this quite commonly with traders often using the same tool for very different tasks.
However, one way of assessing whether to use a tool as an entry technique is to assess how much value that tool is providing relative to a completely random signal.
Now in the top 20 ASX stocks across the last 10 years or so, as a rough rule of thumb, closing prices on any given day are about 53% likely to close higher than the closing price of the prior day.
So, in principle, a signal is doing better than random if it selects stocks that are higher in closing price the day after the signal more than 53% of the time
If any signal is doing better than random any trader should want to know about it.
Now if we refer to table 1, which shows Closing Price versus the Accumulation signal on ASX top 20 stocks for the last 10 years, we see some very interesting data.
Number of Days After the Accumulation Signal
| Annual Strips |
No of signals |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
9 |
10 |
15 |
20 |
30 |
| 1 |
30 |
53.3 |
63.3 |
63.3 |
60 |
60 |
66.7 |
70 |
76.7 |
73.3 |
76.7 |
76.7 |
70 |
79.3 |
| 2 |
44 |
61.4 |
70.5 |
65.9 |
56.8 |
56.8 |
56.8 |
59.1 |
63.6 |
65.9 |
65.9 |
59.1 |
56.8 |
63.6 |
| 3 |
26 |
50 |
76.9 |
56 |
66.7 |
66.7 |
62.5 |
43.5 |
59.1 |
70 |
65 |
70 |
55 |
75 |
| 4 |
25 |
84 |
76 |
64 |
60 |
48 |
60 |
72 |
76 |
64 |
56 |
56.5 |
61.9 |
57.1 |
| 5 |
29 |
58.6 |
62.1 |
62.1 |
69 |
62.1 |
69 |
72.4 |
72.4 |
75.9 |
82.8 |
79.3 |
79.3 |
79.3 |
| 6 |
13 |
69.2 |
69.2 |
69.2 |
69.2 |
76.9 |
69.2 |
69.2 |
76.9 |
69.2 |
76.9 |
61.5 |
69.2 |
61.5 |
| 7 |
14 |
71.4 |
50 |
64.3 |
64.3 |
64.3 |
78.6 |
85.7 |
92.9 |
92.9 |
92.9 |
42.9 |
21.4 |
28.6 |
| 8 |
8 |
50 |
42.9 |
46.4 |
50 |
50 |
42.9 |
50 |
53.6 |
53.6 |
53.6 |
60.7 |
57.1 |
76.9 |
| 9 |
22 |
63.6 |
77.3 |
72.7 |
68.2 |
68.2 |
68.2 |
63.6 |
72.7 |
86.4 |
90.9 |
77.3 |
77.8 |
85.7 |
| 10 |
38 |
65.8 |
63.2 |
63.2 |
73.76 |
63.2 |
63.2 |
60.5 |
60.5 |
60.5 |
63.2 |
65.8 |
76.3 |
73 |
| Total signals |
269 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| Average % |
|
62.7 |
65.1 |
62.7 |
63.79 |
61.6 |
63.7 |
64.6 |
70.4 |
71.1 |
72.3 |
64.9 |
62.4 |
68 |
Table 1: Closing Price Versus the Accumulation Signal on the Top 20 ASX Stocks Over the Last 10 Years.
Table 1 was derived as follows:
- The data on the top twenty ASX stocks over the last 10 years was divided into 10 one- year periods. So in the table, Annual Strip 1 represents from July 2003 to July 2004, while Annual Strip 2 represents 1 year of price data from July 2002 to June 2003.
- In each strip the Accumulation signal was generated using the default settings, (obtainable by going to the Wizard Indicator settings module within the system).
- Then the closing price of the day on which the signal was generated was checked against the closing price from 1 day after the signal, all the way through to 10 days after the signal, and then 15, 20 and 30 days after the signal.
- Then the percentage of the time that a subsequent close on each of the indicated days was higher than the close of the signal day was calculated.
Perusal of Table 1 indicates:
- The average percentage correctness across all of the time strips and across all of the time periods after the Accumulation signal varies from 61.62% to 72.39%, with an overall average across all periods of 65.67%.
- If we assume that randomness can be represented by the benchmark of 53% then the signal beats randomness in this data 114 times out of 130 data points, being about 87% of the time.
- However, we should note that in the 8th time slice the performance in the first 10 days was only about random and also in the 7th time slice the performance after 10 days was very poor. We should always bear in mind that any signal can go bad without warning, and for a good period of time. For traders this is a reminder that money management is always a priority.
We also see by looking at Chart 1 that the pattern of the percentage closes has an interesting overall time characteristic. Generally the accuracy of the indicator seems to increase somewhat from about 7 to 10 days after the signal is given. This, of course may be simply a quirk of the data, but it is data like this that gives traders pause for thought.
Using classical statistics the standard deviation of all of the percentages in table 1 is about 11.7%. This means that the true value of the percentage closes on the Accumulation with about a 66% probability is between 77.43 and 53.92 (assuming a normal distribution). Or stated another way, there is about an 83% chance on the basis of this data that the indicator is doing better than random.
It is rare to find indicators that perform this well using classical statistics. Normally any edge is much more subtle and requires analysis with sophisticated back testing techniques.
If you check out the Market Analyser you will see that MDSnews gives all traders the opportunity to do their own back testing in an easy and understandable manner.
Good Luck!
Chart 1: Accumulation Average Percentage Closes for the 13 Time Periods