Search results

In Windows and Linux (I think), you also have the option of Powershell scripts. I use Powershell for simple scripts for downloading data, but for trading applications, C# is what I use.

Trying to explain markets movements as always due to cause and effect is surely a path to madness. I fell the better way is to have a probabalistic mindset toward market movements. The way I look at it is that price can't go up until it has gone down first. Buy low sell high is rational...

There you go. However one defines a trend, when it starts and when it ends are randomly distributed through time. Furthermore, one has to be able to distinguish between a pullback and a trend reversal in real time. I'm not saying it can't be done, just not by me. I only trade cycles and prefer...

Attempting to predict the future is a real thing using linear or non-linear methods, or even utilizing negative group delay. See attached.

Yes, and the reason for coin toss results with MAs is group delay (commonly called lag.) There are near zero lag MAs, but their filtering ability is very poor. That's physics at work, and not even Gauss, Euler, Fourier, Laplace, or von Neumann could invent a method to make it otherwise...

As you say, catching a trend is easy with an MA. What I have experienced is that the whipsaw periods eat up all the profits obtained during the trending periods. Since you don't provide any evidence that a 12 period SMA method has positive expectancy, I tend to doubt your claim. I don't use MAs...

If folks here insist on trend following with MAs, look at John Ehlers' Optimal Tracking Filter and Zero Lag MA.

Study up on topics such as oversampling and Nyquist sampling criteria.

Price is highly fractal in nature. 1 day, 1 hour, 1 minute, 1 millisecond. Of course there is a bottom limit given the speed of light and how fast a processor can process data.

Let me break it down for you. An SMA is a finite impulse response (FIR) lowpass digital filter. The group delay (ie lag) for an FIR filter is (n-1)/2 where n = length in days of your calculation window. Lag is the reason why moving averages don't work. Speaking of windows, an SMA is a boxcar...

The simple model that I like is that price is analogous to pink noise, until you get near the Nyquist frequency. Because financial data is sampled data, near the Nyquist frequency introduces a lot of aliasing noise. These articles by John Ehlers explains, and shows what to do about it.

Yes, timeless advice, but I need each of these precisely and mathematically defined. That way I can apply some data science and try to find positive expectancy before I spend lots of time and money validating a particular idea or philosophy (possibly years.)

Yes, one time frame is as good as another. It's all your personal preference as to how fast you want the action to be. When trading moving averages, you look to make a profit from the middle of the up swing and down swing. The smaller the time frame, the smaller those middles will be. Of course...

Look up the work of John Ehlers if you want to learn more about moving averages (lowpass digital filters.) Financial data is fractal in nature, so 4 hours does not necessarily mean anything more than 3 hours or 5 hours, or daily, or monthly. However, you should read Ehlers work on aliasing and...

Two-pole highpass filter. Look it up, no time to explain.

Imagine manually scanning all the charts in the SP100 every day looking for one of these patterns, or conversely watching a handful of stocks for weeks at a time hoping one of these patterns develops. No thanks, I need to be able to back test patterns on years worth of data in order to validate...

Price is random pink noise with stochastic amounts of skew and kurtosis. However, even that model is too simplistic. So, to somewhat answer your question, price prediction is a probabilistic problem. For the most simplistic and most flawed way of predicting price, see attached.

The thing is, these charts are all pattern recognition. Patterns are fine, but you have to be able to to define the patterns precisely enough to code into a computer algorithm. That way you can apply some data science and prove whether a given pattern has a positive expectancy. Otherwise, one...

Read his latest interview in S&C. https://www.mesasoftware.com/TechnicalArticles.htm He dismisses cycle measurement for market data. Look into his Zero Lag and Optimal Tracking Filter if you are looking for a trend indicator. If you like cyclical indicators, have a look at the "A Peek into...

What all this says is have a well diversified portfolio and a buy and hold strategy. Not what people want to see in a trading forum. It seems quite obvious that when bull and bear markets begin and end is randomly distributed in time. Perhaps buy on a down year, and keep adding if the next year...