Correcting Construction Drift After a Cycle

Date: 2026-05-25 Author: OLTA Research Abstract: Many crypto baskets in 2026 still carry constituents whose original thesis was written for the 2021-2022 cycle. The constituents drew down 80% or more from cycle highs, never recovered, and continue to dilute basket performance against the assets that did survive. This paper documents the institutional principle behind a post-cycle constituent refresh, the framework OLTA applied across the May 2026 catalogue refresh, and the trade-offs an allocator should understand when a basket is reconstituted rather than rebalanced.

1. The problem

A basket whose constituents were chosen at the top of a cycle and never reviewed afterwards is carrying dead weight whether or not it acknowledges it.

The pattern is consistent. In 2021 and 2022 a wave of altcoin majors traded at multiples that implied the projects would dominate their sectors within five years. File-storage tokens were priced as if on-chain file storage would replace centralised cloud storage. Layer-1 alternatives to Ethereum were priced as if three of them would each command tens of billions in market capitalisation. Metaverse, gaming, and creator-economy tokens were priced as if those use cases would settle on chain within a single product cycle. Several DeFi blue chips were priced as if their early protocol revenue would compound through a multi-year fee-capture window.

By mid-2026 the empirical record is clear. Many of these constituents sit 80% to 95% below their 2021 peak with no constructive base, no resurgent narrative, and no incremental holder demand. They trade with low correlation to the broader market not because they decoupled from beta but because they failed. The pattern is concentrated in specific sectors: alternative L1s outside the top three by ecosystem activity, storage tokens with no enterprise adoption, meme tokens whose 2022 launch narrative collapsed into illiquid tail, legacy payment-rail tokens overtaken by stablecoin-native settlement, and a tail of DeFi tokens whose fee capture never crystallised into token value.

A basket constructed against the 2021 thesis and not refreshed since is structurally exposed. The drawdown reflects the failed constituents more than the broader market. The correlation profile is distorted because dead-cat-bounce tail returns are not the same statistical object as live-market returns. The headline Sharpe degrades even when the broader market is positive, because the basket is not actually expressing the broader market.

2. The institutional principle

A basket is not a museum piece. A basket is a thesis expressed through a constituent list. When the thesis is no longer supported by the empirical record, the basket should refresh its constituents, retire if the underlying thesis was wholly falsified, or be explicitly relabeled as a historical artefact.

The principle is not novel. Every institutional index methodology specifies a constituent review cadence and the criteria under which a constituent is replaced. MSCI specifies float-adjusted market cap thresholds and explicit deletion rules. The S&P U.S. indices specify a minimum continued listing requirement and exclusion criteria for companies whose business has materially declined. The Bloomberg fixed-income indices specify maturity thresholds and credit-rating exclusions that remove deteriorated bonds regardless of how they arrived. The principle in each case is the same: an index that carries its constituents indefinitely is an artefact, not an investable.

Crypto baskets have generally not enforced this discipline. The dominant pattern is to set constituents at launch, rebalance the published weights on a regular cadence, and treat the constituent list as fixed unless a project is delisted. This made sense in 2017-2020 when each cycle was understood as additive. After the 2022-2024 cycle the pattern is no longer defensible. The asset class now has its own pile of failed projects, and a methodology that cannot remove them publishes a number that overstates structural beta.

The institutional answer is a periodic catalogue review with explicit dragger criteria. The May 2026 OLTA refresh was that review.

3. Scope of the review

The review pass covered every basket in the active OLTA catalogue and identified more than 60 constituent positions across roughly 20 baskets that fit the dragger profile: structurally large drawdown from a 2021-2022 cycle high, no constructive base, no narrative resurgence, low to negative incremental holder demand. The exact list is internal to the methodology brief; the framework that identifies it is published here.

The dragger framework reads from four signals:

• Drawdown from prior cycle high. A constituent sitting 80%+ below its 2021-2022 peak with no constructive base over the subsequent two years carries a high weight on the dragger profile.
• Realised beta to the broader crypto market over the recent twelve months. A token whose realised beta has collapsed (the price has stopped tracking the broader market) is signalling either detachment from the asset class or liquidity thin enough that price discovery is no longer informative.
• Holder concentration and on-chain activity. A token whose holder count is net flat or declining and whose on-chain activity (transactions, active addresses, protocol revenue) has not grown is signalling that the original thesis has not been validated by demand.
• The narrative thesis itself. A constituent whose sector thesis has been overtaken by an empirical replacement (file storage by enterprise cloud, generic L1 by ecosystem consolidation, generic metaverse by mature gaming platforms) is dragging on the basket whether or not its price has caught up.

The four signals are scored together. A constituent that scores high on three or four is a removal candidate. A constituent that scores high on two is reviewed but typically retained at a reduced weight. The signal weighting and removal threshold are documented in the methodology brief.

4. The OLTA response

The refresh applied surgical changes across roughly 17 baskets. The pattern varied: in some cases a single failed constituent was removed and replaced with a structurally healthier token; in others several constituents were cut and the freed weight was redistributed to existing holdings whose thesis remained intact; in a small number of cases the basket was reweighted toward post-cycle replacements that were unavailable at the original launch. The methodology brief documents the per-basket changes.

One basket was retired outright. Its thesis was that tokens which had drawn down severely in 2022 would mean-revert through the subsequent cycle. The empirical record falsified that thesis across the basket's own backtest history: the constituents did not mean-revert; the survivors held their depressed valuations and the failures fell further. Removing or replacing constituents would not have fixed the basket because the thesis itself was wrong. The retirement preserves the entry as a historical artefact while removing it from active rotation. This is the institutional pattern: a falsified methodology should be retired, not patched.

Benchmark baskets were intentionally not refactored. An equal-weight benchmark of the top-100 crypto names exists to serve as a passive control group against the actively constructed baskets. Refactoring it would defeat the purpose: the benchmark is informative because it carries whatever the top-100 set holds at the moment, draggers included. The cycle-tested benchmark basket is also untouched; its constituents are fixed at six majors with continuous deep history, and that constraint is the basket's defining feature.

5. The trade-off

The refresh is not free. Removing a dragger from a basket has three consequences.

First, the basket loses optionality on dead-cat-bounce. A failed constituent that bounces 5x off its base from a 95% drawdown still ends 75% below its prior cycle high. That bounce, multiplied across several constituents, can add tens of percent to a basket's recent-window return. A basket that has removed those constituents cannot capture that return if it materialises. This is a real cost, sized against the alternative cost of carrying the constituents through their multi-year base.

The math of the trade-off is straightforward. A constituent at weight w and volatility v contributes roughly w*v in vol terms and roughly w*r in return terms over a window where it returns r. A dragger at 50% realised volatility but close to zero return contributes w*0.50 to vol and roughly zero to return. The basket's Sharpe is dragged down mechanically. Removing the constituent recovers the Sharpe. The question is whether the option value (the probability-weighted return from a bounce) exceeds the carrying cost. For a constituent with 80%+ drawdown, multi-year base, and no narrative resurgence, the math typically favours removal.

Second, the refreshed basket has a backtest discontinuity. The published backtest before the refresh reflects the old constituents; the backtest after reflects the new. A naive concatenation would overstate the track record. OLTA does not concatenate; the refreshed basket's published backtest is computed against the new constituent set across the available common window, and the prior-version backtest is preserved internally for traceability. The refreshed track record starts from the refresh date.

Third, the refresh signals that OLTA holds an opinion about which constituents are alive and which are draggers. That signal is part of the value the catalogue offers and the kind of signal an allocator should challenge. An allocator who disagrees with a specific removal can express that view by holding the underlying token directly.

6. What is not disclosed

The published framework documents the dragger profile, the signal set, the basket-by-basket scope, the retirement principle, and the trade-offs. The methodology brief documents the exact constituent removals per basket, the exact replacement weights, the precise signal thresholds that triggered each removal, the per-basket forward Sharpe estimates that justified each refactor, and the timeline of when each refactor became effective. The brief is available to institutional counterparties and grant reviewers under standard research-distribution terms.

The institutional pattern is to publish the framework and reserve the per-basket recipe. A desk that wants to apply the framework to its own catalogue can do so from the published material. A desk that wants to lift the OLTA per-basket refresh verbatim cannot, because the per-basket inputs are not disclosed.

7. Practical reading for an allocator

An allocator looking at the OLTA catalogue in mid-2026 should read the May 2026 refresh as the catalogue's first cycle-tested constituent review. The published headline numbers reflect the post-refresh constituents. Prior-version backtests are preserved internally but are not the published track record.

Three concrete reads follow.

First, the catalogue is now structurally less exposed to dead-2021-narrative tokens. The basket-versus-BTC profile of the active set has improved structurally rather than because of a specific market event.

Second, the catalogue's reported Sharpe figures should be understood as forward-looking conditional on the refresh holding. If the broader market produces a regime where the removed draggers do bounce, the refreshed catalogue will underperform a hypothetical unrefreshed catalogue over that window. The refresh expresses a view, and the view is that the carrying cost of the draggers exceeded their option value. That view is held by the OLTA desk; allocators are free to hold a different view.

Third, the next constituent review will run on a documented cadence. The intent is for catalogue refresh to be the institutional norm rather than a one-time event. The methodology brief documents the review cadence and the criteria under which an off-cycle refresh would trigger.

8. Caveats

• The dragger framework is one set of judgments about which constituents are structurally drag and which are temporarily dislocated. Reasonable analysts can disagree about specific cases.
• Removing a constituent crystallises a view that the constituent will not recover. If it does recover, the basket misses the bounce. This is an unavoidable property of any constituent review.
• The published track record of a refreshed basket starts at the refresh date for the post-refresh construction. Prior performance is preserved internally but is not the headline number.
• The retired basket is no longer in any active rotation. Its entry remains in the catalogue as a historical artefact and is documented in the methodology brief.
• Equal-weight and cycle-tested benchmark baskets are intentionally not refreshed. Their role is to serve as control groups against the actively constructed baskets.

The methodology brief documents the per-basket parameters and is available on request.

OLTA Research Desk · 2026-05-25