Search results

Nobody can give you absolute positive returns under all market regimes. When the market is down big, I'll bet even Jim Simmons is down big. There are plenty of managers who have positive returns, even market beating returns, year after year.

The person whose done a lot of work in measuring market periodicity is John Ehlers, with his MESA and autocorrelation periodogram algorithms. I have not done much investigation with either method, so I can't speak directly to performance. They are interesting ideas, however I suspect that they...

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...

Remember, frequency = 1/period. So a lowpass filter with a cutoff frequency of 0.1 will have a 10 day period. The filter's cutoff period controls the group delay (lag) and attenuation. How much lag you can stand in your MA, and the smoothness of the curve will be determined by the number of...

Moving averages are lowpass filters. The most important parameter for trading is the filter's group delay, otherwise known as lag. To be useful for trading the group delay needs to be zero or almost zero. Also, to be useful, filters need to be causal (you can look up that definition.) Next in...

That is a mighty big IF. Most people never get to the point of trading profitably. Instead of spending years bouncing from one trading idea to another and losing lots of money looking for profitability, why not apply a bit of science to the trading hypothesis first? Conduct proper backtesting...

expiated, Just curious if you trade the slope of the ITL directly, or wait for a bar to change color and only then enter in the direction of the slope? Example: wait to go long on a red bar when the slope of the ITL is still positive. How about your exit rules?

The thing about H-A is that even if you base your indicators off of its' closing value, you are still sampling once per day. The H-A close is the arithmetic mean of the OHLC, but is still just one sample per day. I'm not getting into Shannon information or Nyquist theory in this post, but...

Real world distributions have skew and kurtosis. Add some positive skew and assume some leptokurtosis, and then do the calculation again.

I also recommend you look into the Fisher Transform by John Ehlers. The transform will "Gaussianize" your data over a given lookback window.

An random walk isn't a perfect model, but is the single best model of market movement. Besides, any number of studies doing R/S analysis have shown values barely above 0.5 for daily data, which is pretty damn close to random walk. However, once you go to monthly data, long range memory starts to...

First of all, if prices evolve as a random walk, then all of technical analysis is charlatanism. The degree of randomness is perhaps a philosophical debate that you need to have with yourself. As far as adaptive indicators, there are plenty of methods to dynamically adjust and tune parameters...

The point myself and others are making is that markets switch regimes, often violently, between random and nonrandom. Saying that autocorrelation is the best market model is wrong. However, that doesn't mean you can't money under the autocorrelation assumption, you just have to use it in the...

My issue with the differencing method is that anytime you take a difference, you add 6dB of noise at the Nyquist frequency. You can see this in the differenced data versus the original data. The differenced data may be stationary, but is also much noisier.

Financial markets can be modeled under any number random variable models and their assumed distributions (ie normal, lognormal, pareto etc.) The problem is that none of these models explains reality significantly better than any of the others. The reason is that the underlying processes are more...

Ultimately yes, but Gaussian statistics do a poor job of describing price action. However, traders have long used simple heuristics to deal with probability, such as buying just below obvious support to get a perceived low risk entry.

When you realize that moving averages are just lowpass filters in the digital signals processing world, then group delay comes out of the math(the negative derivative of phase with respect to frequency.) Once you view MAs as digital filters, you are implying that price is a complex signal plus...

I should have added that it is traditional moving averages used in the traditional way (ie crossovers) that don't work. Certainly, some of the newer low-lag MAs used in unconventional ways might work, but even then, it's a low probability proposition that they would. All of TA is a low...

For those of you using EMAs, try using this formula as an alternative calculation to reduce the group delay of the filter: alpha = 2/(N+1), where N is your lookback period EMA = alpha*(price[0]+k*(price[0]-EMA[1]))+(1-alpha)*EMA[1] where k is a gain factor you choose. I suggest values in the...

I don't have to. MAs don't work, and you are not some 1:10,000,000 person who has figured out how to make them work.