Evaluating index futures OHLC reversals with RSI and time stops

A three-layer exam

The evaluation question is not whether a chart shape looks like a reversal. It is whether an entry-and-exit rule can be examined as three published layers: an oscillator control, a same-session long trigger, and a calendar exit that ends every trade.

Editorial reading: the first layer makes the market’s mean-reverting bias explicit. The second layer turns candlestick-patterns into three falsifiable open-high-low-close tests. The third layer lets the holding period, not a price target or a trailing stop, decide when the hypothesis is over.

Publish the oscillator control

The relative-strength-index is a bounded momentum oscillator scaled from 0 to 100 and sampled over a fixed lookback. Threshold crosses supply the anti-trend control against which the pattern is compared.

The published control treated a drop below the lower threshold followed by a rise back through that threshold as a long, and a rise above the upper threshold followed by a fall back through that threshold as a short.

That oscillator rule was applied to a continuous equity-index futures series, with an initial stop and a per-trade allowance for slippage and commissions.

The same market was described as rarely persisting in one direction for many days and as a setting where anti-trend-model constructions were expected to fare better than trend-following ones.

Editorial reading: publishing this control first is what makes later pattern scores interpretable. Without it, a long reversal sits in an unnamed regime.

State the long trigger as three checks

The mechanical long trigger is built from the same session’s open, high, low and close plus a short lookback for an extreme low, so a repeatable chart shape becomes a testable entry.

The long-only reversal required three checks on the signal session. The close had to exceed the open by more than a stated point threshold. The high-low range had to clear a range-filter. A d-day-low had to sit inside the most recent several sessions.

When those three checks were met, the baseline entry was a buy on the next session’s close. Fills on the next open, a tick offset above the next high, or the signal-day close were listed as variants.

Pattern inputs were held fixed and labeled arbitrary rather than optimized. Both continuous and rollover contracts were used.

Editorial reading: the three checks are the whole long hypothesis. They do not add a confirmation oscillator at entry. The relative-strength-index stays in the control layer.

Retire every trade on a day-count

The reported pattern test used a time-stop as the only exit: flatten on the close after a chosen number of sessions, so exposure ends on the calendar rather than on a price target.

A holding-period-scan then swept calendar exits across a stated band of day-counts so the same entry could be scored under many fixed lives. The detailed long-only case used one pre-chosen session close.

Editorial reading: a time-stop keeps the exit from becoming a second, unstated price rule. The scan shows whether the same three-check entry changes character as the allowed life lengthens. It is not a search for a new exit.

A complete write-up therefore names the relative-strength-index control, the three same-session checks that arm candlestick-patterns, and the time-stop that ends the trade.