What is Quarterly Theory?
Quarterly Theory divides a time period into four equal quarters and assigns each an expected role, typically consolidation in the first, an expansion or false move in the second, a reversal or continuation in the third, and a final move in the fourth. The division is applied recursively — to the year, the month, the week, the day and the session.
The framework overlaps heavily with the accumulation, manipulation and distribution model, adding a fourth segment and a fixed rule for where the boundaries fall. Because the boundaries are determined by the clock rather than by price, they are not subject to the hindsight labelling problem that affects most structural models.
That is genuinely a methodological advantage and worth naming. A framework whose segments are defined by time can be applied identically in real time and in retrospect, which is more than can be said for most of the vocabulary in this area.
What it shares with the rest of the family is that the behavioural claims attached to each quarter are stated loosely, and loose claims cannot fail.
How do you state Quarterly Theory as a testable claim?
State it as a conditional distribution claim: that a measurable quantity — range expansion, reversal frequency, directional persistence — differs between the quarters by more than sampling variation would produce.
This is the step that turns a framework into a hypothesis. Instead of asking whether the second quarter tends to produce a false move, ask what fraction of second quarters contain a move beyond the first quarter's range that is retraced inside the same period, and compare it against the same fraction for the other three quarters.
Framed this way the claim has a null. If the four quarters produce statistically indistinguishable rates, the framework adds nothing over dividing the day arbitrarily. If one quarter differs by a margin that survives an honest correction for the number of comparisons made, that is a finding.
The number of comparisons matters here more than usual, because the framework invites many. Four quarters, several period lengths, several metrics, and several instruments multiply quickly into hundreds of comparisons, and the best of hundreds looks impressive by construction. Any result has to be adjusted for how many were examined.
What does a fair test of a time-based framework need?
A fair test needs consistent clock handling, a null comparison against arbitrary divisions, an adjustment for the number of comparisons, and enough history that each quarter of each period still contains a usable sample.
Clock handling is the mechanical prerequisite. Quarters defined against a session require every bar timestamp converted to a stated market timezone, applied consistently across daylight-saving transitions. An inconsistency here shifts the boundaries relative to the market for part of the sample, which can manufacture or destroy an apparent effect.
The null comparison is the part most often skipped. Before concluding that the quarters matter, check what the same metrics look like when the period is divided at arbitrary points. If dividing the day into four random segments produces differences of a similar size, the framework is describing the shape of intraday activity in general rather than anything specific to quarters.
Sample sizing needs care because the recursion cuts it fast. Splitting the day into four quarters and then examining each quarter by session and by regime leaves each cell with a fraction of the original count, and cells with a few dozen observations cannot support a verdict.
Does time of day affect price behaviour at all?
Intraday activity does vary systematically with the clock, which is uncontroversial and follows from when regional participants are active — but a general variation in activity is not the same claim as a specific behavioural role per quarter.
The distinction is worth being precise about because it is where most confusion in this area lives. That volatility and volume concentrate around session opens and overlaps is well established and easily measured. That the second quarter of a period specifically produces a false move is a much narrower claim requiring its own evidence.
A framework can therefore be partly right in an uninteresting way. If the quarters happen to align with session boundaries, then any measured difference may be the session effect wearing a different name, and a test that does not control for it cannot distinguish the two.
Our own testing on session timing is published as a measured finding, including where the effect was smaller than the popular account claims. The same standard applies here: an effect that vanishes once sessions are controlled for is a session effect.
How would QuantParadox test a quarter-based rule?
QuantParadox grades time-conditional rules across a decade of minute-resolution history with session and quarter boundaries anchored to a stated clock, and applies out-of-sample splitting by default.
Minute resolution is what makes fine time divisions measurable at all. Dividing a session into quarters and asking what happened inside each one requires data at a resolution well below the quarter length, and a test built on hourly bars cannot answer questions about ninety-minute segments without interpolating.
The Reconciliation module is the relevant surface for this class of question, since it exists to identify which conditions carry a strategy and which bleed it rather than reporting a pooled figure. A quarter-based claim is precisely a conditional claim, and pooled results cannot address it either way.
The platform will not confirm the framework because it is widely taught. Where a time-conditional effect fails to survive out-of-sample grading or an adjustment for multiple comparisons, that is reported as a failure — which for frameworks inviting this many comparisons is an outcome worth expecting.