Attachments forums

List of attachments posted on this forum.

All files on forums: 77279


amdudus, Fri May 18, 2018 12:17 am

Crash test indicator Moving Average
I think almost everyone knows about it, and many also use it in their trade. But have you ever wondered how effective it is, and what is the probability of a successful transaction?
Here I will give a shortened electronic translation from a Russian site (publication of 2015).
"Preparing for research In order to more accurately assess the effectiveness of moving averages, it is necessary to leave only a fluctuation in the price (of the schedule) from all factors affecting the performance of the indicator. In other words, we remove the spread, commissions, swap, slippage, which in our study act as extra noise. Thus, we created ideal conditions, which in nature can not be, however, it is such testing that will reveal the potential of the indicator. Naturally, quotes should also be qualitative. In the following experiments we use tick quotes, which allows us to achieve 99.90% quality modeling.
Purpose: To determine the effectiveness of the indicator Moving Average. At the same time, an indicator that, when equal to SL, TP makes profitable trades with a probability of at least 53-55%, can be considered effective. Otherwise, if you include commissions, spreads, swaps, it will automatically become unprofitable. Although I'm more than confident that in the end we will get a deal success of about 50%. For tests from the set of trading strategies for MA, I chose the three most popular ones, and we will test them using trading robots.
Study number 1 - A simple intersection of moving averages.
The first and most common trading strategy for moving averages is the intersection of moving averages. A signal to enter the market is formed when a fast MA crosses a slow MA. Accordingly, we buy when the fast MA is from above, and we sell when the fast MA is from the bottom. In this case, the exit from the market for this strategy can be of two types: fixed SL / TP, or when there is a reverse intersection. Of course, it would be more correct to test the input and output on the intersection, but the version with fixed SL / TP is more revealing for us. In addition, if the entrance to the market does not show any positive results, then the output will not affect anything. Based on the strategy described above, an advisor with three variables was created.
For them we will select the optimal values in the range:
1. SL = TP in the range from 10 to 100 points;
2. Fast MA with a period of 10 to 60;
3. Slow MA with a period from 40 to 100;
So, after carrying out the optimization, it turned out that the most stable parameters are: SL, TP = 40 pip; FastMA = 40; SlowMA = 80. The probability of a successful transaction under strategy No. 1 was 50%. As I expected, optimization is just fitting parameters for a certain period of time. Using optimized parameters for forward test, we got zero results.
My opinion is that a profitable pattern does not need optimization. In addition, other tests showed that the type of moving average and timeframe did not have a significant effect on the result.
Study number 2- Crossing of two middle ones with a filter, from a long-term moving average.
Add an additional filter to the simple intersection, which should exclude counter-trend inputs into the market. The signal to the input is formed when the fast MA crosses the slow one, and all three MAs are directed in one direction. As we have already defined above, optimizing the parameters does not give anything in principle, so I see no reason to run several-hour optimization, we will immediately proceed to testing on standard parameters:
1. SL = TP = 30 points;
2. Fast MA = period 20;
3. Slow MA = period 40;
4. Long-term MA = period 150;
We are testing this strategy: The probability of a successful transaction under strategy No. 2 was 51.72%.
Wow! The percentage of profitable trades exceeded the mark of 50 by as much as 1.7% ... but ... in real conditions with a spread, commission and swap, this percentage easily turns into a negative one. Let's move on to testing the latest strategy
Study number 3 - 200 MA, as support / resistance.
The oldest classical strategy. A moving average with a period of 200, which "symbolizes" 200 working days a year.It is used as a very strong support, or resistance. There is another version of this strategy, when the breakdown of this line is traded. In any case, it is enough for us to conduct only one test, since these strategies are mutually exclusive. The probability of a successful transaction under strategy No. 3 was 51.70%. Again to no avail. Another study proves that, no matter how you twist the Mashki, and whatever you do with them, it will always be nothing more than an "eagle" or "tails".
Summary table with the results Let's collect some more statistics, and put all the results into the table.
As can be seen from the table, the average probability of a successful transaction was almost 51%. Yes, and this 1%, I would easily write off the error. It will not be superfluous to show what will happen if the real spread and commission are used in these studies. For example, the second strategy for the yen showed the best result. By results of researches it is established, that at the indicator Moving Average there are no any positive (profitable) regularities. And the result of its application will not be much different from the usual coin toss. Probably, someone will say that these tests are "about nothing", because it was necessary to add a filter that will determine the trend, tk. Trade with MA is possible only when the market trend. With the latter I agree! However, if you have such a unique filter that determines the trend, then why do you need a random MA? Just open a deal on a trend, and that's it. "
All files in topic