Tuesday, November 23, 2004

Turkey Day Effect

Performed a fun system test on the major indicies today. Tested for the Thanksgiving Holiday effect on the stock market. If you buy on the Wednesday before Thanksgiving at the market's open and sell at the market's open on the Monday of the following week...you get some impressive results:

Nasdaq Composite
Time Period: 1990 - 2003
Total Trades: 14
Total Win%: 92.86%
Total Loss%: -7.14%
Average Profit: 1.72%
Largest Profit: 5.37%
Only one loss and it was less than half a percent. Not bad.

S&P 500
Time Period: 1962 - 2003
Total Trades: 42
Total Win%: 83.33%
Total Loss%: 16.67%
Average Profit: 0.93%
Average Loss: -0.62%
Largest Profit: 2.64%
Largest Loss: -2.46%

DJ-30
Time Period: 1951 - 2003
Total Trades: 51
Total Win%: 58.49%
Total Loss%: 41.51%
Average Profit: 1.11%
Average Loss: -0.92%
Largest Profit: 3.89%
Largest Loss: -5.23%

Take special note of the length of time frame tested here. The Nasdaq Composite enjoys the best record but with the shortest time frame tested. The Dow Jones is the worst performer but with the longest time frame tested.

I'm afraid the longer we extend the test period...the worse the results will become. But, continuing on this fun factor...let's see what happens if we use the same time period for all 3 indicies in the test:

Nasdaq Composite
Time Period: 1990 - 2003 (same period as previous test)
Total Trades: 14
Total Win%: 92.86%
Total Loss%: -7.14%
Average Profit: 1.72%
Largest Profit: 5.37%
Only one loss and it was less than half a percent. Not bad.

S&P 500
Time Period: 1990 - 2003
Total Trades: 14
Total Win%: 71.43%
Total Loss%: 28.57%
Average Profit: 0.79%
Average Loss: -0.30%
Largest Profit: 2.64%
Largest Loss: -0.72%

DJ-30
Time Period: 1990 - 2003
Total Trades: 12
Total Win%: 64.29%
Total Loss%: 35.71%
Average Profit: 0.83%
Average Loss: -0.23%
Largest Profit: 2.58%
Largest Loss: -0.79%

As expected, the DJ-30 results improved as the tested time period was shortened by almost 40 years. A little odd is the S&P 500 results performing worse under the shorter time period. This throws a minor wrench in my theory of results improving with shorter time periods. Especially since this shortened time period was the roaring bull market 90's.

There might be something to the Taylor Turkey system. A little tweak here, a trend filter there...might improve the results even more. Then again, that could leave you with even less data points in your test. A system with few tested data points has a tendency to forward test rather poorly. Maybe this test should remain as it started...just a fun project.

Please note, the above system results were judged merely by the Total Win% variable. I'm fully aware there are many more important factors to consider when evaluating a system. But, in order to keep the post short, simple, and fun...decided to go with a ratio most investors understand.

No comments: