Premium and Discount: Splitting a Range at Equilibrium, and What the Split Can Prove

8 min readQuantParadox research

Premium and discount is the oldest idea in trading, buy low and sell high, expressed with a ruler. Whether the ruler adds anything is a question with a clean experimental design, which is more than can be said for most concepts in this family.

The short answer

Premium and discount are the halves of a trading range above and below its midpoint, and a rule that only buys in discount or sells in premium is testable as a location filter by grading the same entries with and without it.

What are premium and discount?

Premium is the upper half of a defined price range and discount the lower half, with equilibrium as the midpoint between the range's high and low, and the concept holds that long entries belong in discount and short entries in premium.

The range is usually a swing-to-swing range on the timeframe being traded or one above it, sometimes called the dealing range. Its high and low are the most recent significant swing high and swing low; the midpoint is their average; and price's position relative to that midpoint is the only thing the concept reads.

The claim is a location claim. It does not say when to enter or what the trigger is. It says that, whatever the trigger, entries in the favourable half of the range have a better expectation than entries in the unfavourable half. That is a filter, and filters are the easiest kind of claim to test because the control is built in: the same entries without the filter.

Refinements exist, such as splitting each half again or requiring a return to a specific fraction of the range, and each refinement adds a parameter. The base version has only one, the definition of the range.

Which range defines the split?

The range that defines the split is a choice with no natural answer, and the honest test fixes it by rule: the most recent confirmed swing high and swing low under a stated swing definition, on a stated timeframe, as of the bar being evaluated.

The choice matters because ranges nest. At any moment there is a range on the five-minute chart, another on the hourly, another on the daily, and price can be in discount on one and premium on another simultaneously. A concept that lets the trader choose the range after seeing the trade will always find one in which the entry was in the right half.

Fixing the swing definition fixes the range. With a rule such as a swing high needing a stated number of lower highs on either side, the range at each bar is determined, and it updates only when a new swing is confirmed. That removes the choice, and it also means the range lags: a new high is not part of the range until the bars that confirm it have closed.

The timeframe is then the remaining parameter, and the sensible approach is to test more than one and report each rather than to pick the one that scores best.

Does entry location change the result?

Whether entry location changes the result is exactly the comparison a filter test makes: take every entry the underlying rule produces, split them by whether they occurred in the favourable half of the fixed range, and compare the two groups on the same stop, target and horizon.

The design is clean because the filter changes nothing about the entries themselves. The same trigger fires in both halves; the only difference is where in the range it fired. If the favourable-half group has a meaningfully better expectancy, and the difference holds on the out-of-sample segment, the location adds information. If the two groups perform alike, the filter is removing half the trades for no benefit, which is a cost rather than neutrality.

There is a plausible mechanism for the filter to help, which is that an entry in discount has more room to the range high before meeting supply, so a fixed target is reached more often. There is an equally plausible mechanism for it to hurt, which is that in a trending market the favourable half is the side the trend is leaving, and waiting for it means missing the move.

Both mechanisms are real and which dominates is a property of the instrument and the period, which is why the answer is measured rather than argued.

Where does the concept break down?

The concept breaks down in strong trends, where the range is redefined so often that discount keeps moving away, and in any test that allows the range to be chosen after the trade rather than before it.

In a persistent uptrend, each new swing high resets the range upward, the midpoint rises with it, and a trader waiting for discount is waiting for a pullback that may not come before the next leg. The filter then converts a trend-following rule into a rule that trades only the pullbacks, which is a different strategy with a different and usually smaller trade count. That is not a flaw in the concept so much as a description of what it does.

The after-the-fact range is the flaw. Any presentation of premium and discount that draws the range from the swing that turned out to matter has assumed the conclusion, and a rule tested that way cannot lose. The lagged, rule-defined range is less flattering and it is the only version that can be graded.

The refinement of splitting each half further, into quarters or finer, should be treated with suspicion for the usual reason: each extra threshold is another parameter to fit, and a rule that only enters in the lowest quarter of the range has been given a great deal of freedom to find a favourable history.

How does QuantParadox express premium and discount?

QuantParadox exposes price's position inside the current swing range as a single value from zero at the range low to one at the range high, computed as of the bar being decided, so a premium-or-discount filter is one comparison against a half.

The range behind that value is the platform's own: the most recent confirmed swing high and low under its swing rules, updated when a new swing confirms and not before. A trader whose hand-drawn dealing range uses a different swing definition, or a different timeframe, will find the platform's range differs, and the difference is worth checking on a chart before trusting the result.

Higher-timeframe ranges are available through the same as-of discipline, using the last completed bar of the higher timeframe rather than the one still forming, which is where hand-built versions of this filter most often leak.

The filter comparison, favourable half against unfavourable half on the same entries, is the kind of split the Reconciliation view reports directly, with the trade count for each side, which is the number that decides whether a difference between them means anything.

Questions people actually ask

What is equilibrium in trading?

Equilibrium is the midpoint of a defined price range, the average of its high and low, used as the dividing line between the premium half above it and the discount half below. It has no significance beyond being the halfway mark of whatever range was chosen, which is why the definition of the range matters more than the midpoint itself. A different swing rule or timeframe gives a different range and therefore a different equilibrium.

Should you only buy in discount?

Only if a test of the underlying rule shows that entries in the discount half outperform entries in the premium half by enough to justify skipping the premium ones. The filter has a cost, roughly half the trades, and in trending markets it can remove the trades that carry the result. Grade the same entries with and without the filter on the same history, compare the expectancy and the trade count, and apply the out-of-sample split to the comparison before adopting it.

Which timeframe should the dealing range come from?

There is no correct timeframe, and the choice is a parameter that should be stated and tested rather than picked to fit. A common arrangement uses a range from one timeframe above the entry timeframe, so that the location filter reflects a larger structure than the trigger does. Whatever is chosen, the range must be the confirmed one as of the bar being evaluated, using the last completed higher-timeframe bar and not the one in progress.

The only backtest that settles it is yours.

Build a strategy from a sentence, paste your own Python, or import your live trade history and have it graded. Five full backtests free, no card, and we'll tell you plainly when the result is indistinguishable from luck.

We publish research and tooling, not trading advice, and we make no claim about future returns. Everything above describes how to test an idea — not a reason to trade one.