results. Therefore, nonadjusted data are best for day traders with very short trading
horizons, who more often than not close out all their trades at the end of the day. To
overcome the distortions on the historical system’s testing results induced by the
nonadjustment method, the point-based back-adjusted contract was invented.
If the new front contract is trading at a premium compared to the old con-
tract, the entire historical time series leading up to the roll will be adjusted upward
with that distance. A similar but opposite adjustment takes place each time the new
contract is trading at a discount. Figure 6.7 shows what the October 1987 market
action looks like using a point-based back-adjusted contract, with the latest roll
made in September 1999. In this chart, the high on October 2 is at 567.35, and the
low on October 20 is at 415.35, for a total difference of 152 points. However,
because of the transition upward, the original percentage difference of 45.6 per-
cent now has decreased to 26.8 percent (152 / 567.35). For each roll, all these val-
ues continue to change slightly, presumably upward, resulting in an
ever-decreasing percentage difference.CHAPTER 6 Quality Data 73
FIGURE 6.7
October 1987 using point-based back-adjusted contracts.