Advanced macd question | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Important Question for amateur and advanced traders 15 replies

Advanced question on iCustom function! 16 replies

Traditional MACD histograms vs MT4's MACD Coding 8 replies

Advanced Markets question 0 replies

Advanced MT4 question about Time 0 replies

Advanced macd question

  • #1
  • First Post: Dec 13, 2017 11:57am Dec 13, 2017 11:57am
So, I have a system that uses mad to confirm stochastic crossovers under the bought and sold levels. My question is this. The macd fluctuation levels change with price. When eur jpy was 130 in the 1990s, 1.5 or greater macd at oversold stochastic levels signalled a strong reversal consistently. In 2000, when eur jpy was near 110, the macd level of 1.5 was never reached at all. It fluctuated under 1. What equation can I use to based off current price to determine adequate macd levels to validate a trade.
  • #2
  • Dec 14, 2017 3:53am Dec 14, 2017 3:53am
If you understand what Fast EMA and Slow EMA are, then you know the answer. They are the same EMA's you can place on your chart and see how they look in comparison to price movement. When the Fast crosses the Slow, you have a 0 line cross on the MACD. If the EMA's are flat, you need smaller values to see any movement, or lower your time frame so you can see more range between high and low. If you don't want to alter your settings, lowering the time frame is probably the best idea.

Personally, I think MACD and Stochastic are among the worst indicators a person could use, so this is where my advice must end. Except that, whatever happened in the 90's, 2000's, last week, or yesterday, really has nothing to do with what happens today. Each new trade is comprised of new traders, new lot sizes, and new reasons for entering/exiting. Trading a dead trade is like showing up to play a baseball game a week late. No one is there now. It's over. Good Luck.
You cannot be extraordinary by being normal
  • #3
  • Dec 14, 2017 11:58am Dec 14, 2017 11:58am
MACD is the difference between two moving averages. You can show that the histogram is an estimator of the trend up to a multiplicative factor with lag and distortion. You can remove the distortion by using a pair of SMA instead of EMA. The price to pay is an increase of the lag.

The MACD estimates the trend in the unit of the chart ($ per bar). This means it uses absolute values. When the price increases 1% from 130 it reaches 131.3. When it increases 1% from 110 it reaches 111.1. In absolute value this is 15% less. Because MACD estimates this trend its value will also be 15% less now than it used to be when the price was 130.

You have at least two choices to "fix" this. The first is to normalize the price by taking a point in time as a reference. That's not easy. The second is to normalize the MACD. That's what I do for my (unbounbded, zero mean) oscillators. The bonus is that the oscillator looks the same on any chart (asset+TF). I personally use Z-score (the value divided by the standard deviation of the value over a long history of at least 300 bars).

HTH
No greed. No fear. Just maths.
  • #4
  • Dec 14, 2017 4:02pm Dec 14, 2017 4:02pm
Quoting dkrock
If you understand what Fast EMA and Slow EMA are, then you know the answer. They are the same EMA's you can place on your chart and see how they look in comparison to price movement. When the Fast crosses the Slow, you have a 0 line cross on the MACD. If the EMA's are flat, you need smaller values to see any movement, or lower your time frame so you can see more range between high and low. If you don't want to alter your settings, lowering the time frame is probably the best idea. Personally, I think MACD and Stochastic are among the worst indicators...
I'm not actually using the macd for crossover confirmation. I'm using it to confirm a stochastic crossover. Stochastic give lots of false crossovers. The macd value helps validate the stochastic levels. So if macd value is below -2 with a stochastic signal and a -2.5through 2.5range , its statisticly more likely to go up than down according to my sampled backtesting. The problem of course is prices change. So if the eur jpy goes down 30%, the macd range will shrink, so 2.5 is no longer a useful setting. Hence the folowung comment referring to standard deviation.
  • #5
  • Dec 14, 2017 4:07pm Dec 14, 2017 4:07pm
Quoting PipMeUp
MACD is the difference between two moving averages. You can show that the histogram is an estimator of the trend up to a multiplicative factor with lag and distortion. You can remove the distortion by using a pair of SMA instead of EMA. The price to pay is an increase of the lag. The MACD estimates the trend in the unit of the chart ($ per bar). This means it uses absolute values. When the price increases 1% from 130 it reaches 131.3. When it increases 1% from 110 it reaches 111.1. In absolute value this is 15% less. Because MACD estimates this...
Interesting. Like dk said, trends change randomly. What if the fluctuation range changes quickly in a 300 bar sample. Would that not create false signals? I'll have to expirement with the best tradebar count. Too much data will allow for more changes, which could create more false signals.
  • #6
  • Dec 14, 2017 4:10pm Dec 14, 2017 4:10pm
Quoting Davidhoyle
{quote} I'm not actually using the macd for crossover confirmation. I'm using it to confirm a stochastic crossover. Stochastic give lots of false crossovers. The macd value helps validate the stochastic levels. So if macd value is below -2 with a stochastic signal and a -2.5through 2.5range , its statisticly more likely to go up than down according to my sampled backtesting. The problem of course is prices change. So if the eur jpy goes down 30%, the macd range will shrink, so 2.5 is no longer a useful setting. Hence the folowung comment referring...
Basically, I'm trading way before the macd emas cross over or under 0. I'm trading at the peak of the fluctuation of macd values. So if the range is 2- -2 I'm trading at 1.7 - -1.7 plus a stoch cross from overbought or sold levels.
  • #7
  • Dec 15, 2017 1:24pm Dec 15, 2017 1:24pm
Quoting Davidhoyle
{quote} Interesting. Like dk said, trends change randomly. What if the fluctuation range changes quickly in a 300 bar sample. Would that not create false signals? I'll have to expirement with the best tradebar count. Too much data will allow for more changes, which could create more false signals.
The 300 bars aren't to estimate the trend but the variance of the histogram values.
No greed. No fear. Just maths.
  • #8
  • Last Post: Dec 16, 2017 3:35pm Dec 16, 2017 3:35pm
Quoting Davidhoyle
{quote} I'm not actually using the macd for crossover confirmation. I'm using it to confirm a stochastic crossover. Stochastic give lots of false crossovers. The macd value helps validate the stochastic levels. So if macd value is below -2 with a stochastic signal and a -2.5through 2.5range , its statisticly more likely to go up than down according to my sampled backtesting. The problem of course is prices change. So if the eur jpy goes down 30%, the macd range will shrink, so 2.5 is no longer a useful setting. Hence the...
Okay. Let's think logically for a second. I have money in the market. I want to make a profit. I am going to use an indicator that gives me lots of false crossovers. Does that make sense to you? If it don't work, don't use it. Improve your signals by fixing the stoch values, deleting the stoch, or replacing the stoch. How can you trade head faking yourself??
You cannot be extraordinary by being normal
0 traders viewing now