Why do moving averages SUCK so much?
- Thread starter Saltynuts
- Start date
I haven't tested moving averages alone, but I have used moving average crossovers, and indeed they form a core part of my trading system. But the comments I'm about to make apply equally to almost any trading rule or indicator.
Based on my research, simple indicators like moving average crossovers have something like an average Sharpe Ratio of 0.25 for sensible ranges.
What does that actually mean? It means if you backtest a randomly chosen indicator, over a randomly chosen market, over a randomly chosen time period; then in expectation on average you will get a Sharpe Ratio of 0.25
That means some of the time you'll do a lot better. And some of the time you'll do worse. In fact, I get a negative Sharpe Ratio in something like a third of all instruments I trade on a randomly chosen moving average. That doesn't mean that a third of all the instruments I trade don't play well with moving averages, that's just what you'd expect to get by luck alone. None of the instruments has a performance that's significantly worse than zero (and I'm using the word 'significance' here in a statistical sense). Nor indeed, do any of them a performance that's significantly better than zero.
The point is you can't conclude from just one test anything meaningful about moving averages. If you get an exceptionally good result, well it's most likely you may just have lucked upon a flukey combination. If you get an exceptionally bad result, well it is most likely that you have just had bad luck.
(This means you should treat small sample tests done by other people dressed up as 'guru advice' with a strong dose of skepticism, since they've almost certainly cherry picked the instrument, moving average, and/or time period to get the result they want).
I trade moving averages, because when I pool data across multiple instruments (about 40 futures markets in my case, covering every sector) over long time periods (up to 40 years in some cases) I find that they work pretty well on average for a wide set of possible moving averages, and in aggregate across this entire dataset the performance is indeed statistically significant.
So it might be the case that SPY is uniformly awful when it comes to using a single moving average of any reasonable length; but it may be that the result is just bad luck and you would need to do a lot more testing to be sure.
GAT
%%
Same HERE. But a moving average that looked like one of the better ones[Tech Analysis of Stocks &Co\ magazine,a 233 ma............................................. I cant get much out of that on monthly charts.Same way with 55 month moving average.
Maybe its sdow+ not the ma??LOL
200 day ema helps
Show us.
Here is the real reason MA's can suck. MAs are lowpass filters. This immediately implies that price is a complex wave with component waves of many different frequencies, amplitudes and phases. This is the realm of digital signals processing (DSP) and I will treat it as such.
Lowpass filters work by filtering out the higher frequency component waves, and passing the lower frequency component waves in price. By filtering, I mean attenuating the energy of a component wave. All filters have a property called group delay, which is commonly called lag.
Lag is what typically makes MAs suck for most people. There are newer low-lag MAs (see the work of John Ehlers), but they still have some degree of lag. From physics, the lower the lag, the less the attenuation ability the filter has. Low attenuation means you are more likely to get whipsawed around by price.
Caveat, anything from the world of physics applied to finance is generally a misapplication of the physics. DSP is just a model of how price evolves, with all the associated inaccuracies typical of modeling financial time series.
Lowpass filters work by filtering out the higher frequency component waves, and passing the lower frequency component waves in price. By filtering, I mean attenuating the energy of a component wave. All filters have a property called group delay, which is commonly called lag.
Lag is what typically makes MAs suck for most people. There are newer low-lag MAs (see the work of John Ehlers), but they still have some degree of lag. From physics, the lower the lag, the less the attenuation ability the filter has. Low attenuation means you are more likely to get whipsawed around by price.
Caveat, anything from the world of physics applied to finance is generally a misapplication of the physics. DSP is just a model of how price evolves, with all the associated inaccuracies typical of modeling financial time series.
Last edited:
With an attitude like that, they will suck. I’ve never heard of a successful trader who only uses one piece of information. That would be a neat trick. There is zero context of anything in a 1 variable excel spreadsheet backtest. It lacks any sort of creativity or knowledge about the market.
how about a non secret trading method?
RSI Divergence Buy Signals in conjunction with daily 200ma Support for bottom fishing. > 70% success.
how about a non secret trading method?
RSI Divergence Buy Signals in conjunction with daily 200ma Support for bottom fishing. > 70% success.
Panzerman- great add. high pass filters are extremely valuable, and I for one can’t figure out what’s what without one. Not on any DSP I’ve ever programmed. It’s more than an analogy. It’s exactly what you do to come up with a system.
also, excel is a clunky backtest engine that isn’t very extensible without a ton of work. It seems less than ideal for creating a system.
also, excel is a clunky backtest engine that isn’t very extensible without a ton of work. It seems less than ideal for creating a system.
Pfft, so on his chart, you have to wait months and months for a set-up that MAY fail? Falling through the ice and drowning at the bottom. What a load of shit. How did it work out for him after 2009? How many more setups did he ever get on that DJIA?