"Trends are Back"

[dbphoenix]

There is no conflict between mean reversion and trending. Trend channels also have means.

 

[marketsurfer]

I d

Removing that long run drift from the total return series is a trivial matter (something like a symmetric moving average will do the trick, or kalman smooth if you're showing off). It takes about 10% of the total performance of trend following rules out. So perhaps 90% is due to stuff happening at a faster time scale.

s.

Now i am doubting you are who you claim to be------ there is no such thing as a symmetrical moving average.

Symetrical means taking data from both before and after to create an
average... but obviously with market data you don't know what the
future data points will be.

So WTF are you talking about?

peace
surf

 

[globalarbtrader]

I need to correct you--- there is no such thing as a symmetrical moving average.

Symetrical means taking data from both before and after to create an
average... but obviously with market data you don't know what the
future data points will be.

So WTF are you talking about?

peace
surf

There is such a thing as a symmetrical moving average - you just defined it. Formally, if it's time T in the past, and we want a 100 wide interval for our moving average, we take the average of the periods T-50 to T+50.

Obviously you can't use such a thing when building a trading system.... or when you backtest it will have forward data and look amazing.

However we're not building a trading strategy here. We're just applying a filter to past price data to adjust it, so we can then run our backtest on it.