The OLTA Rebalancing Calendar: Effective Dates, Announcements, and Drift Monitoring

1. Institutional convention

The reference practice is well established and, across three of the largest index families in the world, structurally identical. S&P Dow Jones Indices rebalances the S&P 500 and Nasdaq-100 quarterly, effective after the close of the third Friday of March, June, September, and December, with the change taking hold at Monday's open; the change is announced roughly five trading days in advance, with pro-forma weights published alongside the announcement so that index-tracking capital can pre-position. MSCI runs quarterly and semi-annual index reviews, effective the first of the month following each review. FTSE Russell reviews quarterly, on the same March/June/September/December cadence.

Underneath the differences in exact timing, the pattern is one pattern: a fixed calendar known years in advance, an advance announcement that carries provisional weights rather than a bare notice, and a mechanical effective date with no discretion inserted between the announcement and the event. The value of the pattern is not the specific dates — it is what the pattern removes. A fixed public calendar removes information asymmetry: every market participant sees the same pro-forma weights on the same day, rather than some counterparties learning of a change ahead of others. A mechanical effective date removes discretion risk: the committee cannot quietly reconsider between the announcement and the event, because the event is not a decision, it is the execution of one already made and disclosed.

The pro-forma weights matter as much as the date. An announcement that says only "a rebalance occurs on this date" tells a reader nothing they can act on; an announcement that publishes the actual weights that will take effect lets index-tracking capital pre-position ahead of the mechanical event, which is precisely what keeps the effective date itself from being a liquidity event. This is a second reason the grid on which those weights are published is not a cosmetic detail — a companion paper in this collection (Weight Precision in Index Construction: When the Hundredth of a Percent Matters) addresses the publication precision of computed targets directly; the calendar policy here assumes whatever grid that paper's recommendation settles on, since the pro-forma vector an announcement carries is the same vector the effective date later fixes into share counts.

2. The 24/7 adaptation

"After the close of the third Friday" presupposes a close. OLTA's products do not have one — crypto constituents trade continuously, and the diversified and cross-asset baskets inherit that continuity from their crypto legs. The institutional anchor has no referent in this product set, and the desk faced a choice between two ways to fill the gap: adopt an anchor unrelated to anything else in the stack, or anchor to the boundary the engine already uses.

The engine was not built with this policy question in mind, but it settled the question anyway. The backtest simulation already fires rebalance events on UTC calendar-month and calendar-quarter crossings — the boundary the engine calls internally is a straightforward date comparison against the first of the month or quarter, in UTC. Every Sharpe ratio, drawdown figure, and turnover statistic published elsewhere in this research collection was computed against that exact boundary. Had the calendar policy adopted a different anchor — an approximation of the third-Friday convention, for instance — every published figure in the collection would become basis-misaligned: simulated on one calendar, governed on another, until the entire engine was re-run on the new grid. Adopting the identical boundary as policy collapses that gap. What is published is what is promised, on the same rulebook. The rebalancing ledger already tracks this property per index, labeling each as basis-aligned or not.

This is why the desk's core recommendation is deliberately unglamorous:

Effective dates are calendar boundaries — the first of the month, the first of the quarter, both at 00:00 UTC. Layered on top is an announcement discipline: each scheduled rebalance is announced five calendar days before the effective date, carrying pro-forma weights (computed-weight indices publish the headline vector from the latest compute run; committee-set indices republish their standing targets). Targets then freeze one calendar day before the effective date — any data arriving after the freeze rolls forward to the next cycle rather than reopening a target already announced.

A Phase-2 option exists, not yet proposed, for the cohort of tokenized-equity indices settled through Dinari and Backed Finance: migrating that cohort specifically to the US third-Friday convention, so that effective dates fall on trading days of the underlying venue rather than an arbitrary UTC date. The trade-off is stated plainly rather than glossed: this would be an engine boundary change, and every published equities statistic in this collection would go basis-misaligned until re-simulated on the new grid. It becomes a live question once equities settlement in the product is a real mechanism rather than a simulated one — not before.

The announcement content differs by weighting scheme, and the calendar policy is explicit about which. Committee-set indices republish their standing targets at each announcement — a re-statement, not a change, absent a separate governance decision to move a weight. Computed-weight indices publish the actual output of the latest pipeline run, whatever that run produced, as the pro-forma vector; the announcement is not an opportunity to smooth or override a computed result before it goes public. This distinction is the same one that separates a policy weight from a computed weight throughout this research collection, and the calendar's announcement discipline is built to preserve it rather than blur it at the one moment — publication — where the two might otherwise look alike.

3. The proposed calendar

Two calendar-driven cohorts are proposed, each with its own cadence:

A fourth, non-calendar cohort is covered separately in Section 4.

The calendar's first live test doubled, awkwardly, as a sequencing question. An earlier draft of this proposal bundled three separate matters into the 2026-08-01 date: the calendar policy's own adoption, a set of overdue label corrections, and three pending computed-weight adoptions — treating the adoptions as a single "off-cycle corrective" event sitting outside any index's own cadence. That bundling does not survive scrutiny against the calendar's own logic. Two of the three adoption candidates depend on a comparison gate that does not yet exist as running code; announcing pro-forma weights against a gate that has not been built, let alone run, would violate the same evidentiary discipline the calendar exists to enforce. And treating any rebalance as "off-cycle" sits uneasily beside the contrast drawn in Sections 1 and 2 — institutional rebalances are not exceptions carved out of the calendar, they are the calendar.

The revised sequencing drops the off-cycle concept entirely and instead lands each pending adoption on its own natural boundary:

2026-08-01 carries the calendar policy's own adoption plus a set of label corrections that were overdue independent of any adoption question — two indices whose names currently claim a hand-picked or balanced construction they no longer accurately describe are relabeled back to descriptions that match their actual weighting, and a third's exit label awaits an index-committee decision on naming. None of this is a computed-weight adoption; it is disclosure hygiene that should not wait on a gate that governs a separate decision.

2026-09-01 is the natural monthly boundary for the two computed-weight candidates furthest along in their evidence review — a hierarchical-risk-parity scheme and an equal-risk-contribution scheme, both evaluated under the pre-registered walk-forward protocol of From Labels to Computation: HRP, ERC and Inverse-Volatility Weights in Practice. Their adoption is conditional, not scheduled outright: the pre-registered comparison gate that adjudicates each has to exist as running code, execute against both, and file a verdict before the T-5 announcement carries pro-forma weights. Publishing pro-forma weights ahead of that verdict would be exactly the credibility failure Section 1 describes institutional practice as designed to prevent.

2026-10-01 is the natural quarterly boundary for the third candidate, an inverse-volatility scheme, whose own adoption question was reopened earlier this month by a delisted-constituent replacement and now carries an open concentration question on its largest resulting position. The index committee has been asked to settle that concentration question as a governance decision ahead of the announcement, not as a post-hoc adjustment once pro-forma weights are already public — adjusting a computed rule's output after the fact, in response to what the rule produced, would defeat the purpose of computing it at all.

Under this sequencing, every adoption falls on a date that would exist on the calendar regardless of any single index's readiness — the closest analogue a 24/7, continuously-computed book has to the discipline a fixed third-Friday date provides a traditional index. No index adopts ahead of its own natural cadence, and none is held back past it once its evidentiary gate clears. The cost of this discipline is stated plainly rather than absorbed silently: one index carrying a computed-weight-claiming label continues to run on hand-set weights for an additional month beyond the originally floated date, and the announcement accompanying the calendar's launch discloses that gap explicitly rather than leaving it to be discovered.

4. Drift monitoring

A small cohort of four indices is declared drift-triggered rather than calendar-triggered: they rebalance when their weights wander materially from target, not on a fixed date. The mechanism is a daily check, run at 00:00 UTC, comparing each index's live weights against its target vector, against a published trigger threshold of 25 percentage points of drift on any single constituent.

Current levels, from the most recent ledger regeneration:

None of the four sits within half the published threshold, and no drift event has fired in any reconstructed window to date for any of the cohort. Publishing the current levels alongside the threshold, rather than the threshold alone, is a deliberate choice: a rule with no visible reading invites the question of how close the book actually runs to its own limit, and the honest answer at present is not close.

Drift monitoring is a distinct instrument from a different turnover question the calendar makes visible elsewhere in the book. One index outside the drift-triggered cohort — a three-name basket rebalanced on the quarterly calendar — turns over roughly 17 percent of its NAV at every event and roughly 76 percent annualized, the highest turnover figure in the tradable book. That basket is calendar-driven, not drift-driven, and its churn is a function of its construction (a small basket, re-ranked every quarter, will always move a large fraction of itself) rather than a signal that anything is malfunctioning. Whether its rebalance band or cadence should change is a governance question the calendar surfaces by publishing the number, not one its mechanics resolve on their own; it is flagged for review at the next quarterly cycle rather than acted on here.

Together, the calendar and the drift trigger cover the entire tradable book under one of two disclosed rules: a fixed date known well in advance, or a fixed threshold checked every day and published alongside its current reading. Under this proposal, nothing in the book rebalances on a basis a counterparty could not have looked up beforehand. The accountability this creates runs in both directions — it binds the desk to a rule it published, and it gives anyone reading this paper the means to check, on any given day, whether the desk is following it.

Status: desk proposal, pending index-committee ratification. No element described in this paper — effective dates, the announcement and freeze cadence, the drift threshold, or the phased adoption sequencing in Section 3 — is yet in effect. Figures cited reflect the 2026-07-23 ledger regeneration (22:13 UTC).