dozu888,
its impossible and unhealthy to put everything under scrutiny. Nobody has time for that. We as traders must be logical and simple in approach day to day.
Any Math Whiz Here??
- Thread starter schizo
- Start date
its impossible and unhealthy to put everything under scrutiny. Nobody has time for that. We as traders must be logical and simple in approach day to day.
All you wanted to know about moving averages (plus even more) can be found here: https://alphaarchitect.com/2017/07/...leriy-zakamulin-moving-average-basics-part-1/
The above chart shows that the crossover for the EMA line took place 2 bars prior to the SMA line. This got me thinking whether it's possible to hack the formula to make the lines respond faster, eg. faster crossover.
Don't forget to also read part 2.
true...
although in this case it's easy.. e.g. say you back test buy/sell based on ema alone, and prove that it basically provides zero edge... then you can just remove it from your system and it won't make any difference
say if a purely random environment you are right 50% of the time.
if you become a good PA reader.. what does that give you.. 55% perhaps... not bad.. if you are really disciplined you can make a living out of that.
but once you look at the first derivative of the price, e.g. ema, that 55% is now down to 52%, if you are lucky, or 2nd derivative, like macd, now you are down to 50.5%, basically random again.
now if the market has been trending and the trend follow indicators will work for a while, only to get the method crowded enough to be vulnerable for the counter trades.
this is why most traders don't make any money... they basically go yo yo from 1 randomness driven 'system' to another, armed with some 'back testing results'.
but it gets worse... if they are truly randomness traders, you'd think perhaps half of them make money... or perhaps just below half due to commission slippage... but reality is 95% lose money.. why?
because they get manipulated and played by my pro boys, who are experts in human behavior...
so the key is to get on the right side... the side of the pro boys.
you can do that with price action... it's a crude way.. you get 55% of the time correct, 60% if you are really really good.
but there is a better way - read my thread 'trading is easy'... instead of price action you do situation analysis, narrative analysis... that's how my boys play the crowd... now your success rate gets a boost... ignoring the narrative is like pretending you are trading in a vacuum and just staring at the screen with the price candle going up and down... you are leaving a useful on the table - the narrative.
and this method will not be crowded, because YOU are simply better than the rest... every story every narrative is never the same but you see thru the truth never the less.
The above chart shows that the crossover for the EMA line took place 2 bars prior to the SMA line. This got me thinking whether it's possible to hack the formula to make the lines respond faster, eg. faster crossover.
You're overfitting. But if you care, reducing the timeframe for the EMA will give you faster crossovers.
If you really care, you can program this in about 20 minutes using pinescript and can see it for yourself. The rule of TA alchemy is for each time you fit to a signal closer, other signals become less readable. The goal is to find balance.
Come on dozu, MACD isn't a derivative. It's a signal attenuator. An IIR filter. First year DSP stuff. It is not a ROC indicator.
The best systems I've ever dreamed up had solid backtests run to statistical significance, testing via walk forward and ruin analysis, and then traded with small money until they've worked for a bit. Of all the systems I've used, zero of them have come from a book.
I will however agree that by themselves moving averages are useless. They need context to work. Gone are the days of MACD trendfollowing getting you a spot in the next Market Wizards.
Last edited:
Yes it is possible (I'm a math whiz). A few fast adapting averages.....
Wilder's Smoothed Moving Average (just reduce the period)
Doubly Exponential Moving Average
Hull Moving Average
Polynomial Regression Averages
You can use all kinds of them. The thing is though, unless you are nice with stochastic processes then you won't know what you are doing with these things.
You also need to know which averages are good to measure fast adapting averages against. In other words, you need to know what averages are slow to adapt or resist certain types of price movement.
TL;DR You are talking about the rate of convergence of estimators/indicators. You have to be a math whiz to get anywhere with this kind of thing. Back testing is overrated.
Wilder's Smoothed Moving Average (just reduce the period)
Doubly Exponential Moving Average
Hull Moving Average
Polynomial Regression Averages
You can use all kinds of them. The thing is though, unless you are nice with stochastic processes then you won't know what you are doing with these things.
You also need to know which averages are good to measure fast adapting averages against. In other words, you need to know what averages are slow to adapt or resist certain types of price movement.
TL;DR You are talking about the rate of convergence of estimators/indicators. You have to be a math whiz to get anywhere with this kind of thing. Back testing is overrated.
Thanks, but seriously, it doesn't work that way. If you reduce the period for one indicator then you must reduce it for the other. Moreover, such a garden-variety trick that everyone already knows is not very useful. IMO one can no longer rely on publicly available indicators. It needs to be customized and, more importantly, different.
You brought up a good point. In the future, you not only need a back- and forward-testing, but a system that will learn to detect and adapt on an ongoing basis--hopefully, in real time.
The truth is you need a real leading indicator. You need to do analysis on related markets and then use that to trade with an edge.
You can try using Kaufman's adaptive average (KAMA).
Usually, it's based on a 10 day period, but can be adapted a bit.
MIDDLE:=(HIGH+LOW+CLOSE)/3;
SPD :=REF (MIDDLE, 1)-REF (MIDDLE, 10);
VLT :=REF(SUM (ABS (CLOSE- REF ( CLOSE,1 )),10),1);
EFF :=ABS ( SPD / VLT );
FASTC:=2/5;
SLOWC:=2/21;
CONST:=POW ((EFF*( FASTC-SLOWC )+SLOWC), 2);
KAMA:=SMA(MIDDLE,CONST);
basically it's a smoothed moving average where the constant changes depending on how "disturbed" is the signal. If the trend is very clean and the price goes smoothly from A to B, then the EFF value approaches 1 and the constant approaches the FASTC value. If the signal is choppy EFF approaches 0 and the constant approaches the SLOWC value. So depending on the choppiness of the signal you have different responsiveness of the SMA.
Usually, it's based on a 10 day period, but can be adapted a bit.
MIDDLE:=(HIGH+LOW+CLOSE)/3;
SPD :=REF (MIDDLE, 1)-REF (MIDDLE, 10);
VLT :=REF(SUM (ABS (CLOSE- REF ( CLOSE,1 )),10),1);
EFF :=ABS ( SPD / VLT );
FASTC:=2/5;
SLOWC:=2/21;
CONST:=POW ((EFF*( FASTC-SLOWC )+SLOWC), 2);
KAMA:=SMA(MIDDLE,CONST);
basically it's a smoothed moving average where the constant changes depending on how "disturbed" is the signal. If the trend is very clean and the price goes smoothly from A to B, then the EFF value approaches 1 and the constant approaches the FASTC value. If the signal is choppy EFF approaches 0 and the constant approaches the SLOWC value. So depending on the choppiness of the signal you have different responsiveness of the SMA.