The website is updated each weekend. For each indicator, new data is used to generate a long SPY or cash position for the next week. For the momentum example, SPY’s dividend-adjusted close as of Friday is the main input. Using this, I calculate the 12-month total return. For each indicator on this site (except for the macro data), I take a 4-week average of the main indicator input. So for this example, I’m taking the 4-week average of 12-month total return momentum. Why four weeks? To reduce false positives and whipsaws when an indicator is bouncing slightly above or below its filter rule. There’s nothing special about a 4-week average. You could use two or eight weeks and reach similar results.
Data is compiled as of Friday’s close. Buying or selling decisions occur on Monday’s close. I do this, as opposed to making trades at Monday’s open, simply because I had a more reliable data source for dividend-adjusted close data. It’s also important to reflect realistic transaction costs. Each historical performance graph factors in a $10 trade commission and a 0.02% spread on SPY for each buy or sell. Commissions and spreads are lower now, but considering SPY started in 1993 I chose to use these above-average numbers.
Each graph shows data as of each Monday. So for example, look at June 30, 2014 in the momentum graph. As of that Monday’s close the Momentum Filter portfolio had a value of $925,560 and the B&H portfolio had a value of $600,836. The 23.1% momentum metric you see is that weekend’s value for the upcoming week. So at the close on Monday June 30th you would have stayed invested for the following week. This is because the filter rule for momentum is to be invested if the 4-week average of SPY’s 12-month total return is above 0%.