The Track Record
The paper book's time-weighted return since inception, measured against the benchmark the strategy actually answers to — and published with everything needed to recompute it.
Cumulative TWR
−3.79%
since Apr 29
CAGR
−7.2%
annualized from 188d — extrapolated
Volatility
23.2%
annualized, daily basis
Sharpe
-0.21
rf = 0%
Max drawdown
−13.5%
current −10.8%
Cash
24.3%
of NAV today
Turnover
7.7%
trailing 12m, one-sided
- Flagship (TWR)
- Civilizational-Builder Basket (benchmark)
- SPY (reference index)
- QQQ (reference index)
- Bootstrap book, never traded (counterfactual)
The benchmark is the civilizational-builder basket defined in the strategy document — not the S&P 500. SPY and QQQ are plotted for context only and are labeled accordingly: tracking error against broad indices is expected and accepted.
Monthly returns
| Month | Portfolio | Benchmark | SPY | QQQ |
|---|---|---|---|---|
| Apr 2026 * | +2.30% | +2.68% | +0.99% | +0.93% |
| May 2026 | +3.30% | +0.74% | +5.26% | +10.57% |
| Jun 2026 | −0.88% | −1.82% | −1.28% | −0.26% |
| Jul 2026 | −8.50% | +0.59% | +0.03% | −6.57% |
| Aug 2026 | +6.14% | +3.90% | +2.68% | +4.18% |
| Sep 2026 | −2.81% | −2.43% | −0.58% | +3.21% |
| Oct 2026 | −3.27% | +2.90% | −0.91% | −6.66% |
| Nov 2026 * | +0.60% | +1.26% | −0.20% | −0.70% |
* Partial month — measured from inception or through the latest priced day, not a full calendar month. Exact measurement windows are in the JSON export.
What if the book had never traded?
The dashed line on the chart is the Apr 29 bootstrap book bought once and never touched — its cash earning the same T-bill yield the live book's cash earns, the whole thing marked at the same daily closes. It is a counterfactual, not a benchmark: it was knowable only after the fact. The difference between the live line and it is what every trade since inception added, or cost. The tables below replay the public trade tickets with the live series' own rules to say which decisions and which names carried that difference.
Live book
−3.26%
time-weighted, since inception
Bootstrap, never traded
+7.31%
same closes, same cash yield
Activity cost
−10.57pp
live minus never-traded
| Decision source | Trades | P&L of what it opened | What its sales did after | Whipsaw on its buys |
|---|---|---|---|---|
| Bootstrap (Apr 29 seed) | 15 | +4.55pp | 0.00pp | 0.00pp |
| Quarterly Audit | 8 | −0.08pp | +0.42pp | 0.00pp |
| Investment Committee | 6 | +0.23pp | −0.12pp | 0.00pp |
| Bulletin | 54 | −1.22pp | −3.23pp | −0.18pp |
| Weekly Brief | 106 | −5.42pp | −2.17pp | −0.23pp |
| Monthly Brief | 19 | −1.64pp | −0.40pp | 0.00pp |
| Cap-breach cure (mechanical) | 16 | 0.00pp | −0.71pp | 0.00pp |
"What its sales did after": positive means the sold slice rose after the sale (the exit cost money); negative means it fell. Whipsaw: buying a name back above the price it was last sold at. All figures in percentage points of the $100k reference NAV.
- KTOS · held−3.95pp
- MP · closed−1.69pp
- CCJ · closed−1.11pp
- PSN · closed−0.87pp
- LUNR · closed−0.75pp
- LLY · held+3.79pp
- ASML · held+1.07pp
- VRTX · held+0.83pp
- NVO · closed+0.57pp
- CME · held+0.33pp
Through 2026-11-03 · price basis, like every line on this page · read the since-inception review
The civilizational-builder basket
Equal weight across the four pillars (25% each), equal weight across names within a pillar, rebalanced on the first trading day of each calendar quarter. Computed on the same price-return basis as the portfolio. The list is versioned and append-only — currently v2026.1.
Disclosed: Initial list, constructed June 2026 retroactive to the portfolio's April 2026 inception. Disclosed as such on the Track Record page: the names were chosen with knowledge of April–June price action, so the benchmark's pre-June history carries hindsight risk. Versioned in git from here forward.
- NVDA · NVIDIA
AI accelerator monopoly with the CUDA software moat.
- MSFT · Microsoft
Hyperscale compute and the broadest enterprise AI distribution.
- AVGO · Broadcom
Custom AI silicon and the networking fabric of the data center.
- TSM · TSMC
The leading-edge foundry every AI chip ultimately passes through.
- VRT · Vertiv
Power and cooling infrastructure for the data-center build-out.
- CEG · Constellation Energy
Largest US nuclear fleet, prime beneficiary of compute power demand.
- VST · Vistra
Merchant power generation leveraged to data-center electricity demand.
- CCJ · Cameco
Western uranium supply for the nuclear restart.
- ETN · Eaton
Electrical equipment across the grid build-out and electrification.
- GEV · GE Vernova
Gas turbines, grid equipment, and nuclear services for new generation capacity.
- LMT · Lockheed Martin
The flagship US prime: aircraft, missiles, and missile defense.
- RTX · RTX
Munitions, air defense, and engines with deep replenishment backlogs.
- NOC · Northrop Grumman
Strategic deterrent programs: B-21 and the nuclear triad.
- GD · General Dynamics
Submarines and land systems at the center of naval recapitalization.
- PLTR · Palantir
Software-defined defense and the AI targeting layer.
- LLY · Eli Lilly
GLP-1 franchise leader in the metabolic-disease wave.
- NVO · Novo Nordisk
The other GLP-1 duopolist, with global obesity-care scale.
- ISRG · Intuitive Surgical
Robotic surgery platform with a compounding installed base.
- TMO · Thermo Fisher
Life-science tools: the picks and shovels of biology research.
- REGN · Regeneron
Antibody discovery engine with a genetics-driven pipeline.
Rebalanced: 2026-04-29 (2026-Q2) · 2026-07-01 (2026-Q3) · 2026-10-01 (2026-Q4)
Did de-risking beat doing nothing?
When the crash protocol acts, the page keeps a shadow book: the executed book exactly as it stood at the protocol's first action in the episode, held unchanged from there and marked at the same daily closes as the real series. Both lines are computed on a price basis. The shadow's cash earns no T-bill yield, which tilts the comparison slightly against the protocol — a bias accepted rather than corrected, so the protocol never gets the benefit of the doubt.
Every Tier 2 episode, and every Tier 1 episode in which the protocol actually traded, is scored this way; an episode still open shows as provisional. Lose to the shadow in two consecutive episodes and the protocol is flagged for the operator's review. Nothing about it changes automatically — the flag is a question for a human, not a switch.
| Episode | Trades | Protocol | Shadow | Difference | Verdict |
|---|---|---|---|---|---|
Nov 2, 2026 → open Tier 2 · index drawdown + 2 AI-unwind tripwires | 0 | — | — | — | Unscorable the NAV series does not yet cover the window |
Scored through 2026-11-02 · 0 completed episodes · mode live · Tier 2 now
How these numbers are made
The money is notional. The book runs against a fixed $100,000 reference. Positions are weights of that notional, and the daily value is rebuilt deterministically from the public trade record — the same trades listed on the Trades tab. Nothing on this page comes from a private ledger.
Returns are time-weighted. The headline series is a time-weighted return (TWR) index: daily returns geometrically linked, with any external cash flow neutralized by subperiod linking. This book has no external flows — capital is fixed at $100k — so the TWR equals the simple NAV chain, but the math is implemented (and tested) in its general form. You can recompute it row-by-row from the CSV below.
Prices come from Polygon.io. Daily official closes for every name, the benchmark constituents, SPY and QQQ. Scheduled rebalance trades fill at that day's close; intraday autopilot trades fill at their stamped market-hours fill price.
Dividends are excluded — everywhere. Every line on this page (portfolio, custom benchmark, SPY, QQQ) is computed on a price-return basis. That makes the comparison consistent, but it means each line slightly understates its total return — particularly for dividend payers in the basket. Total-return accounting is a documented future upgrade; until then, no line gets the benefit the others don't.
Fees are zero. No commissions, management fees, borrowing costs, or taxes are modeled. A real implementation would pay all of these.
Paper fills are generous. Simulated trades assume unlimited liquidity at the reference price: no bid-ask spread, no slippage, no partial fills, no market impact. This flatters the results relative to real execution, and is the single biggest reason these numbers should be read as a record of decisions, not of achievable returns.
Risk statistics. Volatility is the sample standard deviation of daily TWR returns, annualized by √252. Sharpe uses a 0% risk-free rate (stated, not hidden — at today's rates this flatters the ratio). CAGR compounds over calendar time and is flagged on the page whenever the track is shorter than a year, because annualizing a short record extrapolates it. Turnover is the trailing-12-month sum of absolute traded dollars, halved (so replacing the entire book once equals 100%), over average NAV — the opening bootstrap trades excluded.
Verify it yourself
Everything needed to recompute this page from $100,000 is in two files: the daily series and the full trade history (bootstrap opens included). The decision record behind the trades is sealed in an append-only hash chain — each journal entry's hash commits to everything before it, so history can't be quietly rewritten.