Three after-close filters on a two-basket pairs procedure

Three filters after the close

The archive describes a mechanical trading system whose rule-based entry is a pairs-trading sequence on two index baskets. A trade is allowed only after the close when the broader basket finishes higher, its close-to-close percentage price change exceeds that of the narrower basket, and its closing-volume percentage change also exceeds that of the narrower basket.

Price change is the close-to-close percentage move. Volume change compares the session's volume with an eight-day simple moving average of volume.

That three-filter closing procedure was eligible on 29 of 334 sessions from January 1, 2002 through April 30, 2003, or 8.68 percent of days, about one playing day every 11.52 sessions.

Orders after the regular close

The procedure places orders between 16:00 and 20:00 ET. Any premarket placement must occur before the next session moves away from the prior close, so discretionary regular-hours timing is not required.

Background volume, spreads, and range

From January 2, 2002 in the sample, average daily volume was 85 million shares for the narrower basket and 34 million for the broader basket, with regular-session spreads described as generally less than a penny.

In that background window the narrower basket's average high-low range was 0.975, or 3.5 percent of a 28.22 average close, while the broader basket's average range was 1.97, or 2.0 percent of a 96.77 average close.

Tabulations on the 29 signal days

On the 29 signal days, a 100-share overnight short of the narrower basket was tabulated at a 32.62 average, a 43 median, and a 62.07 percent positive-day share, with the seven highest scores summing to 872 against a 458 cumulative figure for the seven lowest.

A same-day 100-share long in the broader basket paired with a 300-share short in the narrower basket was tabulated at 784, described as a 0.015 percentage figure, using a close-price ratio near 3.39 to 1.

A later window and a rising path

Eligible-day frequency rose from 20 of 289 sessions, or 6.92 percent, through February 20, 2003 to 9 of 48 sessions, or 18.75 percent, from February 21 through April 30, 2003, with the later window's tabulated success rate at 66.67 percent versus 55.55 percent earlier.

While the narrower basket rose 12.56 percent from 24.37 on December 31, 2002 to 27.43 on April 22, 2003, the short-the-narrower-basket rule was still tabulated at a 2 percent figure. The archive used that window as a case that the same abstention-plus-signal sequence can be examined when the underlying path is higher.