Seasonality filters

I'm still without full internet, but I shall spare you the details. In short; Virgin Media have excelled themselves in indeptitude.

However I thought I'd share something i've been playing with. Well, more adapting really. Market Sci has been putting out some great work on seasonality effects and I have to say that I'm coming round to them as a useful trading input.

I spent a good few hours trawling through the library of studies available at http://www.cxoadvisory.com/blog/ and found there is some significant evidence that seasonality effects are meaningful and persistent.

The evidence is strongest for the Turn of the month effect and the Halloween effect (AKA sell in May effect).

But how to employ these?

An increasingly popular method is to use seasonality to modulate position size. Market Sci has been suggesting this for a while and it does appear to have merit.

Another method I have played with is to use seasonality as part of a system score board.

So each day you gather a number of inputs each with a different positive or negative score between -1 and +1 and add up the score for that day. A basic idea would then be to go long above 0 and short below 0 and out of the market if the score is exactly 0.

Here are the inputs at the score for each day:

DVIB Combo Buy signal: +1
A Williams Vix Fix buy signal (WVFIX in top 95% of values for last 30 days) +0.5
Turn of the month (excl last day):* +1 each.
Months November to April: +0.25 for every day.

DVIB Sell signal: -1

*Turn of the month = 1st, 2nd, 3rd, 2nd to last and 3rd to last day of the month. Not last day.

You add up the score for each day for your daily total. You'll note that giving each day a score of 0.25 points from November to May will lend the method a positive bias. The theory is that this is not a bad idea given the positive seasonal bias during this period.

Returns in 2010

This method is very profitable when back tested further, but I didnt want to major on that as I thought it better to demonstrate that the seasonality studies can have relevance today.

DIVB is the DVIB Combo method
S&P 500 is the S&P 500 index
DVIB+S is the daily scoring method outlined above.

Not only has the method out performed the market in 2010, it has also out performed the already excellent DVIB Combo method in 2010. These results take no account of friction/ fees etc.

I'm not saying this is a perfect method, but based on my research, for example the first trading day of the month appears to have as much validity as a trading input as other quant based analysis. Perhaps I over play its importance with a score of 1, but this could be changed depending on your confidence in it. If it confirms a buy signal great, if there is a sell signal on the first day of the month, I'm wary and this score card method would make you flat for the day unless there were other inputs.

Other seasonality inputs to consider include:

Fed Day: +0.5
Day before thanksgiving: +0.5
Day after thanksgiving: +0.5
2 days after thanksgiving: -0.5
Day after options expiry: -0.25

Next I'll see if using this scoreboard to adjust position size improves things. Food for thought anyway.

Comments

[Aly S.]

Nice work. You may want to also consider the days leading up to and following Christmas Day as I believe there's a bullish bias somewhere in there as well (traditionally the last 5 TDs of Dec. and first 2 TDs of Jan. have a seasonal edge, as I understand it).

I'm surprised more aren't commenting here. Your blog is really great!

[Aristotle]

Nice work! I've missed your posts.