Search results

Trend Following: An EMA in forward-reverse mode (makes a true zero lag) Parameters: alpha = 0.5 Long Condition: 1. price[0] < price[1] 2. Reverse MA[0] > Reverse MA[1] 3. Reverse MA[1] > Reverse MA[2] 4. Reverse MA[2] < Reverse MA[3] Note: Reverse MA[0] is today's value, not the...

One article I've found very helpful is "Whiter is Brighter", by John Ehlers. Essentially modeling price as pink noise, and then whitening the price data to make all "frequencies" (i.e. trend periods) have the same spectral energy density, just choose your favorite time period! Longer trending...

Try these two methods with your favorite EMA: 1) Forward-reverse filtering 2) Twicing (also called DEMA), followed by sharpening.

Way too much work! Just predict one day into the future and use an 1-pole EMA in forward-reverse mode, you can look that method up yourself. It is true zero lag, and doubles the attenuation of the filter, but it also non-causal.

Either way, it is still only one sample per day. Read about Nyquist sampling theory.

Moving Averages are lowpass digital filters. SMAs are finite impulse response (FIR), and EMAs are infinite impulse response (IIR). John Ehlers is the man whose work you want to read. Technical Papers (mesasoftware.com)

You may want also pose these issues over at Wilmott.

My strategy is to not use indicators with these amounts of lag. I am currently testing a zero lag (or near zero lag) strategy.

Lag in an EMA, which is a one-pole infinite impulse response (IIR) lowpass filter, is (N-1)/2. So: 4 days lag 9.5 days lag 24.5 days lag 49.5 days lag Try using these filters in forward-reverse mode. You'll have to look that one up. Perhaps try using these with a Kalman style gain factor to...

The problem with defining support and resistance classically it is always a backward looking calculation. There are a million ways to define a trend as well as high and low. You have to be able to define both in real time imo. Here is how I do it currently in real time. First, define a trend...

90% winning strategy right here! Please do not actually do this, but it is a good learning exercise.

An EMA is a 1-pole infinite impulse response (IIR) lowpass filter. The are hundreds of ways to make them adaptable to market conditions. John Ehlers is the man for DSP in finance. However, even if you are a Ph.D. Mathematician from MIT, you can't overcome the fact that less lag means less...

Buy low, sell high is rational behavior, but the problem is that low and high are relative and not absolute. You need a method that distinguishes low and high, a pullback from a trend reversal. IMO, that way to do that is with an oscillator indicator with zero lag. I've attached an article in a...

Market data is non-stationary, and can be modeled as pink noise. In the pink noise model, the power in the cycle content of the data increases at 6dB per octave. This increase in power spectral density is what John Ehlers calls spectral dilation, and it distorts many traditional TA indicators...

This is Arnold Schwarzenegger's perspective when people refer to him as a self-made man.

Ideally, yes and some recommend using 252 trading days instead of calendar days.

Solves for future price X: X = exp(sigma*t*x)*S where S = current price, sigma = volatility, t=sqrt(days/365), x = std dev. Solves for standard deviation x: x = log(X/S)/(sigma*t) Skew and kurtosis? Who needs them. Stochastic volatility? No problem.

And how do you propose to identify when these up moves begin and when they end with anything other than 50% precision over a statistically relevant amount of trades?

Yes, and how I do it is turn pink noise (i.e. price) into white noise with a 5-bar highpass filter: 0.0909*x[0]+0.4545*x[1]-0.4545*x[3]-0.0909*x[4] Now, the group delay (i.e.lag) for the filter is 2 days, so I run the Voss predictor algorithm to reduce the lag. Next, I normalize the data with...

If market prices are not random, then make them random, if that suits your trading style. 1. Model price data as pink noise 2. Whiten the data 3. Peak limit the data 4. Perform a Fisher Transform on the whitened and peak limited data John Ehlers' work explains all this in detail.