Back Testing Technical Startergies

There is only one technical indicator that I have seen that works. It is extremely simple to understand and apply. I don't understand (well I sort of do) why it works, but it probably works because "everyone" uses it.

I have never put this indicator to the scientific test, it is something that I just noticed that it worked far more often than it should have if it were random.

nitro
 
Hi all. Firstly MACD was tested and bought at the open when crossed up and was sold at the open when crossed down the following day after the signal was given since day data was used. Similar rules were used for other momentum or trending indicators described previously. Funny enough MACD was a really bad indicator when it came to results yet had you faded the indicator you could have been profitable.

All stocks were included, so even stocks like Enron which went bankrupt had to be included other wise the results would not be realistic. Volume was not controlled except in later tests in systems which were performing well and had potential. Tests were done on S & P stocks as well as the whole range of stocks.

Sure some of the strategies will show positive results by chance but we were looking for consistency and a good equity curve. Rather then strategies which had a few good years but were in deep draw down a lot of time eg 2-3 years at a time which would be extremely difficult to trade in reality.

NihabaAshi makes a good point the results were dependent on our methology. This means you may be able to use some of the indicators we tested which gave negative results in a profitable way. But this will have to be done through discretion rather then mechanical processes. Further more you may be able to write algorithms that combine some poor indicators with others that produce positive results.
 
Quote from Soros:

Funny enough MACD was a really bad indicator when it came to results yet had you faded the indicator you could have been profitable.
More...

I always joke about that. It seems like it might be true.

Also, messing around last night with stockfetcher.com's backtesting (free trial limited to two weeks of data), going long on "bullish MACD crossovers" always ended up with a huge negative ROI, and going short on those same "bullish MACD crossovers" was profitable. Hahaha. I cannot wait to do more backtesting.
 
Quote from Soros:

Their is a lot of literature debating weather technical analysis really works or not and if it does which indicators or theories are useful and which are just black magic? One of the best ways to address this question is to use today’s powerful computers together with specialized software and test the indicators, theories and strategies that have been developed and published over time. Over a period of one year together with a colleague we tested a number of renowned as well as some not so famous indicators and strategies.
More...

A) Why reproduce what others have already done? (as Wayne's post linked)

B) You cant cling on to the good performing indicators, because if you test a universe of indicators, that doesn't mean the good performers are any use.

Statistically, you are going to find that they range from the worst to the best performers. But your good performers could look good just by randomly being at the top of the heap. You need to apply significance testing to eliminate this possiblilty

Frankly, your good performers could be mostly garbage without this. And you should cross check your good performers with the texts Wayne showed, to see if it backs them up or throws cold water on them
 
Your way of optimizing is good. Finding the good parameter areas is important. You don't want to use a setting that if changed slightly affects performance greatly. 3D mapping of optimization can do that for you.
 
You must log in or register to reply here.
Back
Top