Looking for a trend using the Moving Average

The standard types of averaging include: 

• Simple averaging (SMA)

• Exponential averaging (EMA)

• Smoothed averaging (SMMA)

• Linearly-weighted average (LWMA)

To the advanced ones:

• Adaptive JMA of smoothing (JJMA)

• Ultralinear JRSX smoothings (JurX)

• Parabolic approximation (ParMA)

• Wilson's algorithm (T3)

• Variable Index Dynamic Average (VIDYA)

• Adaptive Moving Average (EMA)

To the rare:

• Digital filters

• Various kinds of complex approximations

• Options where the weights for each bar are determined by a neural network

When working with this indicator, you need to be very careful, because in fact, only time is included in its variability (period setting parameter). This means that when we select the best parameters of this indicator, we try to find a pattern in time cycles. Many trading gurus of the past decades have persistently discouraged attempts to use time cycles in predicting future stock prices. In his book, Larry Williams wrote: using time cycles to predict prices is a sure way to get into a madhouse.

In modern terms, trading systems with time cycle length selection do not pass the forward test, since there are no logically justified patterns in the length of the time cycle. Or simply put, they do not pass the test of the future.

The moving average itself does not indicate the direction of the trend. Therefore, we will have to use interpretation. We looked at what they write on the Internet about this, and the most common is "the slope of the moving average perfectly shows the trend direction, use the MovingAverage direction to determine the trend." Well, to be honest, this is a pretty bold statement, and we'll check it first. So, Gold (XAUUSD), the H1 timeframe and 13 years of real history - let's go.

Let's select the parameters on the site for optimization and find IMP_opti.

The best parameters were: Period = 85, Mode = Simple, Price = Median Price. The results are disgusting even in the optimization area. IMP_opti = 0.81. Looking at the chart, I really don't want to participate in this with my money.

As expected, the test of the obtained parameters on the entire site looks very bad. The remnants of the tiny profits were finally wiped out. And on the reverse forward test section (where we have a very strong flat), the balance chart plunged into the abyss. In addition, the PC indicator (Prediction Coefficient) indicates that no real patterns have been found.

IMP_all = 0,06 

PC = 7%

RBTI = 0,43

In the previous test, we saw extremely poor performance. Let's try to increase it. We cannot change or improve the indicator, as it is the purpose of our test. This means that we will improve the way we interpret the Moving Average.

Let's add options for determining the trend: 

1) The price crosses the indicator line.

2) A shock bar, the entire price range of which is higher (or lower in the case of a sell transaction) than the indicator line.

3) The shadow touches the indicator line.

We will also give you the opportunity to use different trend interpretation methods for long and short positions.

Already at the stage of parameter selection, we see a significant increase in the results - IMP_opti = 2.63. This is three times better than the previous result. The best indicator settings were: Period = 181, Mode = Simple, Price = Close, Shift = 2, Open BY Strategy = color change, Open SELL Strategy = ghost cross MA.

The flat section in the reverse forward test is still bad. But overall, the overall result was positive.

IMP_all = 0,55 

PC = 20%

RBTI = 1,59

Congratulations to the Moving Average indicator for assigning it an RBTI rating: 1.56