Yes, but once you know a bit of the math and engineering around digital filters, it allows you to design filters in ways that further reduce (but not eliminate) lag and increase attenuation. John Ehlers is the man for DSP in finance if you want to learn more. However as Real Money pointed out, price processes are not digital signals. That is a somewhat philosophical debate you can have as to whether moving averages should be used at all, or any indicator for that matter, and just go with price action like you indicated.
Just a personal anecdote. If you consider something like an arbitrage spread, eg. 10*SPY - SPX then you can find many uses for higher level analysis techniques. This is similar to a price process, but very different in certain characteristics. I have found that arbitrage spreads are very useful for gaining information about order flow. There is information embedded in this signal.... In fact, DSP seems like something that I might want to add to my current set of proprietary scripts and algo's for this reason.
Been there, done that. Below is some code that I made for a butterworth filter that I played around with about a decade ago. I have also implemented some of Ehler's filters...and backtested that too. Did lots of stuff with some of his filters and genetic algorithms involving other combinations of common technical indicators and filters. Then around that time, I became interested in ProfLogic's work, but was not able to replicate his results. After that, I have pretty much given up on moving average based systems. Only exception might be the RSI and Keltner bands -- indicators which are not intended to indicate trend, but rather indicate how far a move has gone.
Thanks, but the pictures are just copied and pasted from my lecture slides, and my latest book. So not as much effort as you might expect GAT
Back years ago when I used MA's: 10 days (2 weeks) - likewise 21 days (month) - not that it matters using 20 instead 13 weeks (1/4 year) 12 months (for obvious reasons) and short time frame 60 minutes and multiples (for obvious reasons). Also using odd lengths (like fib numbers) so as to NOT follow the crowd never showed much of an edge to me. Fibs themselves though are invaluable to me used with other time and price techniques.
A moving average model is not necessarily bad or good, it is just a model. One could argue that how you use a model is more significant than the model itself. One thing I realized is that a simple moving average (from TA world) is useful to beginners since it's a very simple model for neophytes to intuitively grasp and apply. You can get into more sophisticated models, like say cointegrated vectors (as applied to spread comment above). They can be very difficult to grasp without the right background, but again, they are just models and suffer from similar issues, on top of being possibly less intuitive. In both cases, similar issues arise, like 1) how to properly model (SMA with one series is pretty simple) 2) how to parameterize (what parameters, how many, which ones, what is optimal, etc..) This can get very complex, and SMA again, is very simple and easy to grasp, visually. 3) What features/factors to use? SMA again, makes this simple, it's just lagged values of a series with simple weighting schemes. 4) Optimality constraints can get very complex. And you really have to ask what is optimal? Sharpe, Expectation, Terminal wealth, profit factor, etc... a lot of that is subjective. So until, you start to define this, you don't even really know what you are trying to achieve. 5) multivariate analysis and applications, IMO, are a big step up (trader evolves beyond one series models, this can take years to realize). Here a whole new world opens up. 6) then you have validation methodologies, like lookback window optimization, walk-forward, cross-validation, etc.. etc.. more potential overfitting that has to be dealt with. .. .. Point being that moving averages aren't necessarily evil or bad, but most (like myself) tend to have disdain for the simplicity and lack of rigor of TA as it is peddled to the masses over and over. Someone could have a model based only on simple moving averages, that I would argue could beat super sophisticated deep learning models, based upon how each is applied. I once used a sophisticated evolutionary algorithm to find that the optimal crossing period parameters for EOD S&P500 index were 220/221. Is that the best? Maybe, maybe not. But I used a very simple objective criteria for that. One person's answer may be completely different than another's, without divulging all the criteria behind it. Because it works, doesn't cut it. Using statistical rigor (t-tests etc) as mentioned earlier is good (much better), but not necessarily a panacea, either. I digressed a bit from topic, but that's some 2c on moving averages. PS. (esp @ OP)... If you want to look at some of this kind of stuff in slightly more detail (not too math heavy), Emilio Tomasini , Urban Jaekle, just came out with a new (2nd)version of "Trading Systems: A new approach to system development and portfolio optimisation " I have the pdf version, and looking forward to reading it. Printed version not released yet, AFAIK
Some researchers have suggested varying the length of the moving average based on market realized or implied volatility, a current example being Adaptive VIX Moving Average with Ehlers Alpha Formula by David Varadi December 4, 2019 I wonder if traders have found such modifications to be useful.
simple question. There is absolutely no significance between various periods. Because in the first place, you shouldn't be using indicators. Why? because price moves in various manners, fashions, patterns, speed !!! So if you want to reach financial freedom, remove all indicators !
A moving average doesn't have to be an indicator, it's a statistical measure and fundamental to traditional trend following but understand your point. IMO, over the past decade or so moving average crossovers have become less effective for the following reasons: Markets have become more efficient. Markets are using more sophisticated information. Inferior methods tend to be exploited by other market participants. Politics play a bigger part in markets than it did previously, creating more periods of uncertainty. Investors have become more risk-averse.