Only repositories designated LIVE, LIVE PATH, or VALIDATION enter the lab. Each surface translates active research into an explorable model without replacing the underlying methodology or source.
1 live spine5 live paths3 live drivers11 validation repositories20 active repositories
01
Eligibility rule
ONLY ACTIVE WORK ENTERS THE LAB.
LABS is the interactive surface of the current BLAQUE BAUX research program. The governed spine, LIVE PATH, LIVE DRIVER, and VALIDATION repositories are visible here; Research, Concept, and Null work remains in the Corpus rather than being dressed up as an active model.
Current organization state / Aug 22 2026
02
Featured interactive studies
The research, manipulable.
Each study begins with the repository's actual thesis and reported finding, then exposes the idea through a restrained interactive surface. Reported metrics remain fixed to the research record; sliders and regime controls are used to explore the mechanism, not to manufacture new backtest claims.
HOW TO READ THE LABS
Finding is what the repository currently reports. Interaction lets you interrogate the mechanism behind that finding. Source remains authoritative for methodology, assumptions, code, and validation.
STATUS MATTERS
LIVE means the work has graduated into the governed spine, live path, or live driver. VALIDATION means the idea has survived enough research to be studied here, but has not earned the production bar.
LAB 001 / LIVE
BASE
The governed spine / risk structure
Repository finding: the production spine combines an inverse-vol / equal-risk-contribution base with 12-month time-series trend, volatility-targets each sleeve, blends them, then applies a drawdown regime brake. Reported production profile: ~0.97 Sharpe, ~1.30 Sortino, ~−11% max drawdown.
Research record / locked
The reported profile above does not change when you move the controls. It remains the repository's documented result.
Scenario explorer / user controlled
Supply hypothetical sleeve outcomes and a brake setting. The display performs transparent blend arithmetic only. It is not a new backtest or forecast.
red / hypothetical lossgreen / hypothetical gainrecord remains fixed
LAB 002 / LIVE PATH
BRICS
Gulf-pocket diversifier / dollar regime
Repository finding: the tradable BRICS core is the Gulf pocket—Saudi Arabia, UAE, and Qatar. The dollar-regime overlay improves reported Sharpe from +0.48 to +0.51 while reducing max drawdown from −35% to −20%. The governed driver remains paper / dry-run.
Repository finding: the bond–equity relationship is useful as a governed regime control rather than a static correlation claim. The live driver publishes a regime signal; repository validation reports a 41% reduction in full-cycle equity drawdown with the overlay active.
Repository finding: the leverage wrapper was rejected; the keeper is managed QQQ exposure using trend and volatility targeting. The governed driver consumes the BONDS regime overlay as a marginal, secondary control and remains paper / dry-run.
Repository finding: the keeper is a mega-cap blue-chip momentum tilt with volatility targeting and per-name caps. The BONDS overlay is wired but remains off because full-cycle validation showed no benefit. The driver remains paper / dry-run.
mega-cap momentumvolatility targetregime overlay off
LAB 006 / LIVE PATH
BENCHMARK
Market internals / risk-regime gate
Repository finding: market internals coincide with risk rather than lead it, but the composite still improves a cost-adjusted SPY gate. Repository validation reports Sharpe improving from +0.87 to +1.11. The emitted regime is honestly classified as volatility timing, not breadth alpha.
Repository finding: a naive equal-weight QQQ / VUG / XLK growth book becomes more ownable when the validated rate-regime overlay de-risks during rising rates. Reported Sharpe improves from +0.96 to +1.04, max drawdown contracts from −35% to −30%, and roughly 90% of the return is retained.
Repository finding: the replicable All-Weather book is a low-volatility risk-parity compounder, but it closely resembles the family spine with reported correlation of 0.96. Its bonds-regime overlay targets the right 2022 weakness yet fails the full-cycle bar because the book already self-diversifies, so the overlay remains off.
static All-Seasons2022 weakness mappedoverlay fails full-cycle bar
LAB 009 / LIVE DRIVER
BLACKSTONE
Listed private equity / market-regime gate
Repository finding: listed private-equity managers behave as a levered pro-cyclical market and credit factor rather than a diversifier. BLACKSTONE is the first consumer sleeve to earn BENCHMARK's market-regime overlay: reported Sharpe improves from +0.76 to +0.95 and max drawdown contracts from −45% to −31%.
BX / KKR / APO / CG / ARESpro-cyclical factormarket regime on
LAB 010 / VALIDATION
BRASH
Growth vs ruin / governed aggression
Repository finding: aggression is a distribution choice, not a Sharpe choice. The first pass reports that strong edge can tolerate 2–3× leverage, while weak or negative edge can move rapidly toward ruin; the keeper is fractional Kelly plus a volatility target.
Repository finding: fast reversal dies on turnover. The keeper is slow 10-day cross-sectional reversal, gated to choppy regimes; the live driver uses SPY efficiency ratio < 0.35 and is dollar-neutral.
Repository finding: the static Brown−Blue spread is not alpha; the keeper is rotation. The first pass reports spread trend at +0.73 Sharpe and camp rotation at +0.85 versus +0.72 hold.
Blue / growthRegime: growthBrown / inflation-value
LAB 013 / VALIDATION
BITDOLLAR
Governed crypto trend
Repository finding: the dollar-axis thesis was rejected, but trend plus volatility targeting survived. Real BTC/USD + ETH/USD validation reports +0.72 Sharpe / −12% max drawdown versus buy-and-hold +0.29 / −76%.
Repository finding: Brown and Blue are distinct but the static spread is a disguised value/rate bet, not alpha. The keeper is rotation: trend the spread or hold the stronger 126-day camp. The research records a −35% spread move in the 2020–21 growth regime and +40% in the 2022 inflation regime.
Repository finding: the cap-ladder thesis was rejected. Large-cap drawdown bounce survives better after cost: +0.52 net versus +0.45 for small caps. The smaller-cap ladder also shows lower standalone Sharpe and deeper drawdowns; the keeper moved back up to quality large caps.
Repository finding: apparent beta-neutral momentum still carried about +0.20 residual beta. The hedged prototype reduces beta to −0.04 and leaves a more modest but genuine stream: Sharpe +0.34 and alpha +3.5%/yr. True neutrality costs the beta tailwind.
unhedged +0.20alpha +3.5% / yr at full hedgetarget ≈ 0
LAB 017 / VALIDATION
BLUNT
Crude → refiner / one-day follow-through
Repository finding: five of six tactical ideas failed or were only marginal. The survivor is crude leading refiner equities by one day: crude is the signal, CRAK is the traded asset, volatility-targeted to 10%. The prototype reports net Sharpe ~1.06 at 3 bp/side.
Repository finding: no single hedge covers both deflation and inflation crises. The keeper is a diversified basket: 40% TLT / 30% GLD / 20% DBC / 10% VIXY. Validation passes on insurance criteria: on SPY's worst-5% days the sleeve returns +79% while SPY returns −82%, with corr-SPY −0.29.
Repository finding: the cross-section flips from reversal to momentum around 20 days. Reversal dominates at ≤10d; momentum appears at ≥20d. The keeper blends 20/40/60-day growth, beta-neutral, with reported net Sharpe +0.50; acceleration itself was rejected as a reversal signal.
Repository finding: directional APAC is mostly US beta and contains only ~2.5 effective bets across 12 ETFs. But APAC is the one region with a positive country-rotation pulse (+0.30 long-short). The research also records severe FX drag in Japan: EWJ +138% versus currency-hedged DXJ +367%.
country dispersionrelative strength, not betaFX matters
Prototype note: repository names, LIVE / VALIDATION eligibility, thesis language, and quoted research metrics are taken from the BLAQUE BAUX GitHub corpus. Reported metrics are labeled as repository findings. Interactive response curves and regime controls are explanatory UI demonstrations unless explicitly tied to a reported scorecard value. Production LABS should execute repository code, a verified browser-safe port, or an embedded notebook before presenting newly computed results.
Human judgment / machine scale
One research desk. Three kinds of intelligence.
The analyst frames the question. Market data supplies the evidence. Local AI compute accelerates the search. Governance decides what survives.
