Sizing and Rebalancing the Two Legs — Decision Rules for the Barbell

This is Part 2 of two. Part 1 laid out the architecture: two structurally orthogonal poles, zero in the middle. Part 2 below covers the sizing math per leg, the rebalancing cadence per pole, and the five failure modes that cause well-designed barbells to quietly collapse back into balanced portfolios over time.

Why sizing matters more than architecture

The architecture in Part 1 is recoverable from first principles in about ten minutes of reading. The sizing discipline takes ten years to internalise. Most real-world barbell implementations don't fail because the architecture was wrong — they fail because the sizing drifted, the rebalancing got tactical when it was supposed to be passive, the tactical pole bloated past its budget after a strong year, or a single-name story-stock allocation crept past the position-size discipline the framework was supposed to enforce. The architecture is the easy part. The sizing is where the framework actually lives or dies in a real portfolio.

This piece works through the math one leg at a time, then the rebalancing cadence per pole, then the five failure modes the Closelooknet reference portfolios were designed to defend against.

Sizing the strategic pole (66 to 75% of total)

The strategic pole splits between two cap-weighted vehicles. QQQ for the US growth-and-tech tail, VEU for the global rest. The internal split is the first sizing decision.

Closelooknet's reading: roughly 60% QQQ / 40% VEU within the strategic-pole budget. Three reasons:

• The Bessembinder evidence shows the US has produced disproportionate skewness relative to its share of global market cap. The top five global wealth creators (Apple, Microsoft, Amazon, Alphabet, Tencent — four US, one Chinese) are 0.008% of all firms but account for 10.3% of all global net wealth created over thirty years. Tilting strategic exposure toward the US growth-and-tech tail reflects this.
• QQQ's filter is structurally clean for skewness capture — it excludes NYSE listings and most of the slower, smaller, structurally weaker US equity universe, leaving a basket dominated by platform-scale, network-effect, software-and-semis businesses where the wealth has actually compounded. From 2019 to 2024 the largest 30 names in the NDX contributed an average of 88% of the index's total return; in 2024 alone, 98%. The filter is the point.
• VEU's role is structural ballast — Tencent, ASML, TSMC, Novo Nordisk, LVMH, Saudi Aramco are the non-US wealth-creators the US-only QQQ misses. The 40% weight reflects the non-US share of global GDP and equity-market capitalisation roughly, with the explicit acknowledgment that non-US skewness is even tighter (1.41% of firms, per the Bessembinder 2023 sample) and the cap-weighted index will mechanically tilt toward the few names that are doing the work.

Investors with stronger conviction in the US-growth thesis can run 70/30 QQQ/VEU. Investors with stronger conviction in EM and the European sovereignty thesis can run 50/50. The 60/40 internal split is the Closelooknet default — neutral, defensible, and survives Bessembinder either way.

The strategic-pole-as-a-percent-of-total decision is largely set by the inverse of the tactical pole. If conviction in the current cycle's tactical opportunities is high (and the Money Temperature reads Dawn or Early-Ramp), the strategic pole sits at 66%; if conviction is moderate or the cycle is mid-to-late, it sits at 75%. The strategic pole's job is not to be optimised — it's to be present.

Sizing the tactical pole (25 to 34% of total)

The tactical pole holds the four Closelooknet reference indices plus optional single-name exposure where the operator-authority signal is identifiable. The internal weighting is where most of the decision work actually happens.

Closelooknet's default tactical-pole allocation, with the current AI-infrastructure cycle running and the agentic stack early-maturing:

Sleeve	Weight in tactical pole	Weight in total portfolio (at 30% tactical)	Cycle phase
Rubin Build-Out 100	40%	12%	Early-Ramp (AI infra)
HALO Growth 100	20%	6%	Dawn (physical-world)
Euro-AI Sovereign 50	15%	4.5%	Dawn (EU sovereignty)
AW25	10%	3%	Dawn (agentic stack)
Pattern 03 cohort	15%	4.5%	Active (operator authority)

Three points the table embeds:

• Rubin gets the largest weight because the AI-infrastructure cycle is among the most mature and clearly readable of the four tactical sleeves. The sector breadth (eighteen sub-sectors from optical to memory to power) is itself a diversification across the cycle. The 40% weight is the structural read, not an enthusiasm.
• Pattern 03 sits inside the tactical pole's budget, not on top of it. The 15% Pattern 03 weight is part of the 30% tactical-pole envelope. Single-name story-stock exposure does not get its own additional budget — it competes with the index sleeves for the same capital. This is the Closelooknet discipline against the "just one more story stock" failure mode covered below.
• Total adds to 100% of the tactical pole, not 100% of the portfolio. Multiply each by the strategic-vs-tactical split to get the position size in total-portfolio terms. The right-hand column shows the math at a 30% tactical pole.

When the AW25 cycle matures (the agentic stack moves from Dawn to Early-Ramp), the AW25 weight rises and one of the maturing sleeves' weight falls. When a new structural wave emerges that doesn't fit any existing Closelooknet index — and the Money Temperature + the Pattern Scanner jointly confirm it — a new sleeve gets added and the existing weights rebalance. The framework is not static; it is signal-driven.

Rebalancing cadence per pole

The two poles have structurally different rebalancing rules. Conflating them is a common implementation error.

Strategic pole — drift-tolerant, annual review only, not tactical

The strategic pole is held with a 20-year horizon. The QQQ/VEU split will drift over time as QQQ tends to outperform VEU during US-growth cycles and underperform during global-rest catch-up cycles. Closelooknet's reading: let it drift. An annual review confirms whether the drift has pushed the QQQ/VEU split outside a 50/50 — 70/30 envelope. Inside the envelope, no action. Outside the envelope, a once-a-year cash-flow rebalance (using new contributions to top up the underweight side, not by selling the overweight side) brings it back toward 60/40.

Three things explicitly excluded from strategic-pole rebalancing:

• No defensive rebalance into bonds on VIX spikes.
• No "the market is overvalued" tactical reduction.
• No 60/40-style rebound-to-balance after QQQ runs hard (a common failure mode — see below).

The strategic pole is allowed to do its job, which is to remain present. Letting winners run inside the strategic pole is the point of holding the strategic pole.

Tactical pole — signal-driven, exit on wave maturity

The tactical pole has the opposite cadence. Quarterly review minimum. The trigger for action is not calendar; it is signal. Three structural signals matter:

• Wave maturity. When a sleeve's underlying wave moves from Early-Ramp into Sunset under the Money Temperature framework, the position is exited. This is not a price-based call — it is a regime call. Sunset is when the wave has done its structural work and the marginal capital is no longer flowing to its sectors. Rubin Build-Out 100 will eventually hit Sunset; when it does, the position is exited regardless of how strong the recent return looks.
• New wave emergence. When a new structural wave emerges that the existing four sleeves don't capture — and the Pattern Scanner + Cointegration Lab jointly confirm the regime shift — the framework adds a new sleeve and rebalances the existing weights to make room. This is rare. The four current sleeves cover most of the structurally readable cycles for the next 2-3 years on the Closelooknet read.
• Pattern 03 cohort changes. The Pattern 03 cohort rebalances continuously per the published rule (admit on Tier-1 capital or Tier-3 architecture admission; exit on 13F disappearance). The 15% tactical-pole budget for Pattern 03 is reallocated within the cohort on each admission or exit; the budget itself doesn't change unless the operator-authority signal degrades structurally.

The tactical pole is allowed — and required — to act on signal. The 1-to-3-year holding period is a floor, not a ceiling. A wave that has clearly Sunset gets exited even if the floor has not yet been reached. A wave still in Early-Ramp after three years stays held.

Five failure modes

Most real-world barbell implementations fail at one of these five places. The framework is designed to defend against each.

1. Tactical drift past 34%

When the tactical pole has a strong year — Rubin Build-Out 100 was +87.56% YTD at the time of the most recent Heresy III print — the tactical pole's share of total portfolio mechanically grows. If the tactical pole was 30% at the start of the year and Rubin alone returned +80% while the strategic pole did +15%, the tactical share could easily drift to 38-42%. Past 34% the framework no longer reads as a barbell — it reads as a tactical-heavy growth portfolio with strategic exposure as the smaller ballast. The defense: an annual rebalance back into the strategic pole, using gains from the tactical pole to top up the strategic pole. This is the one place where active rebalancing is required.

2. Premature exit on tactical drawdowns

The opposite failure: a tactical sleeve drops -25% in a sharp correction, and the temptation is to exit before the wave has matured. The 1-to-3-year holding period is a contract with the framework, not a guideline. The Rubin Build-Out 100 experienced a -38% drawdown across mid-2024 before resuming its compounding cycle. Any investor who exited at the drawdown's worst point converted a structural-wave bet into a momentum-chasing one. The defense: the exit signal is wave maturity, not price. If Money Temperature still reads the underlying cycle as Early-Ramp, the drawdown is noise — hold.

3. "One more story stock" bloat

Single-name story-stock exposure inside the tactical pole has a budget — 15% of the tactical pole, 4.5% of the total portfolio at a 30% tactical weight. The temptation, after a Pattern 03 winner like INTC at +437%, is to add a second methodology — "Pattern 04 — Elon Musk Strategic Authority", "Pattern 05 — Mark Zuckerberg Capex Cycle" — and treat each as if it deserves its own budget. The framework does not budget for ad-hoc story-stock methodologies. If a new operator-authority signal emerges and survives the same diligence that produced Pattern 03 (replicable rule, public capital allocation, demonstrably >54% directional accuracy across multiple cycles, published exit discipline), it competes with Pattern 03 for the same 15% budget. The defense: published methodology requirements, not anecdotal conviction.

4. The 60/40-rebound-to-balance temptation

After a year where the strategic pole's QQQ has done +25% and VEU has done +6%, the temptation is to rebalance the QQQ position down — to "lock in gains" and bring the QQQ/VEU split back toward 60/40. This is the failure mode that has cost the asset-management industry trillions of dollars across the last two decades. The Bessembinder evidence says the strategic pole's job is to ride the winners — when QQQ outperforms, that's QQQ doing its job, and selling it down breaks the very mechanism the strategic pole is supposed to capture. The defense: rebalancing inside the strategic pole is cash-flow only (use new contributions to top up the underweight side), not sales-based. Inside the strategic pole, avoid selling the winner.

5. Single-name concentration without exit discipline

The Pattern 03 cohort's discipline is the exit rule — when NVIDIA's 13F shows the position has been closed, the cohort closes it. APLD was admitted in September 2024 on a $160 million NVIDIA equity placement; it appeared in three consecutive 13Fs, then disappeared. Pattern 03 closed APLD on 2026-02-17, the filing date of the 13F that showed no position. APLD traded materially lower in the months that followed. The rule does not assume NVIDIA's conviction is permanent. The failure mode is to hold a single-name position past its operator-authority exit signal because the personal narrative has become attached. A scanner without an exit rule is a graveyard. The defense: the exit signal is mechanical (13F disappearance, divestment announcement, displacement by a contradicting partnership) — not discretionary.

What the framework looks like in practice

The Closelooknet reference portfolios are the running implementation of all of the above. Methodology published, constituents published, rebalance dates published, base date fixed, live performance visible — including the drawdowns, including the rebalancing moves, including the structural wave calls. Skin in the game. The pieces work as a cross-link block:

• Heresy I + Heresy II establish the Bessembinder evidence base.
• Heresy III derives the architecture from the evidence.
• Part 1 of this strategy explainer is the operational distillation of the architecture.
• Part 2 (this piece) is the sizing and rebalancing discipline that keeps the architecture functioning over time.
• Pattern 03 is the worked example of the single-name story-stock methodology inside the tactical pole.
• The four Closelooknet indices — Rubin, HALO, Euro-AI, AW25 — are the tactical-pole sleeves expressed as tradable indices.
• The Money Temperature + Pattern Scanner + Cointegration Lab are the signal infrastructure the tactical pole's rebalancing cadence depends on.

What this isn't

Two things to be explicit about, since the diary framing matters:

• This is published as a market diary, not investment advice. Whether, when and how to adjust an existing portfolio depends on tax position, cost-basis, holding-period optimisation, and personal risk tolerance — all of which call for a licensed advisor.
• This is not a claim that the specific weights (60/40 strategic-pole internal split; 40/20/15/10/15 tactical-pole internal split at 30% tactical share) are the only set of defensible weights one could choose. They are the Closelooknet defaults — the read of the current cycle, the current Money Temperature regime, and the current operator-authority signals. They will evolve as the underlying cycles evolve. The framework is the sizing discipline, not any specific weight set.

One question worth sitting with

The five failure modes above are not academic. Each of them has cost specific investors and specific institutions significant capital over the last decade. Failure mode #4 — the 60/40-rebound-to-balance temptation — is alone responsible for a meaningful share of the active-management underperformance documented in the SPIVA scorecards. The Bessembinder evidence says that letting strategic-pole winners run is the mechanism by which the strategic pole works. The industry's response, almost universally, has been to keep selling the winners. A reader of this piece who is honest with themselves about which of the five failure modes has appeared in their own portfolio over the last five years has done most of the diagnostic work the framework requires. The sizing math is the easy part. The discipline to actually follow it through a cycle is the entire game.