it's just programmed to run 40% faster since the target duration of the trades is only 8 days or less.
Here's a smoothed and refined version of yesterday's screen shot showing the momentum in both TLT (A) and TBT (B) as well as the % change dynamics for both A & B...which is actually a chart of comparative standard deviation decay. Note the intersection (zero line crosses) of A & B. These are the times when the skew favors either TLT (red) or TBT (blue). Its complete serendipity that the length of the trough between the Skew crosses is almost consistently 8 days.
These opportunities can also be traced using the actual trades signaled by the pairs algorithm (Excel spreadsheet below). Based on the screen shot run at 7 AM PST today (Friday) the 30 Day Skew chart is displaying a cross of TLT and TBT SD decay and the open TBT Long position was closed. A run of the same screen at the end of the day today shows this was a smart move as TBT deteriorated steadily for the remainder of the day.
In most case below we see a sequential reversal of S-A (short A) versus L-B (long B), although there are several instances over the 6 month look back where the trend is so strong that back to back long,long or back to back short. short signals fire.
What the spreadsheet DOES NOT show is the effect of using an 8 day time stop.
As a result we see loser trades like #5, which the program allowed to run for 30 days. Had the 8 day fixed time stop been employed the net result of that trades would have been a 3.13% gain.....not a 12.46% loss. Using that time stop would have upped total returns to almost 70% over the 6 month study period.
Here's how QQQ versus QID looks over the past 6 months...still using an 8 day time stop.
