Methodology

How the system works.

The shape of each strategy, the live market-regime read we watch alongside them, the concentrated-money tracker, and the long elimination tournament behind the four sleeves. The architecture is open; the exact calibration values are not.

Mechanics

The shape of the system.

Every sleeve answers three questions in the same order: is the trend on, is the environment safe, and is the exit close?The questions are universal; the answers are tuned per asset. Specific thresholds and windows aren't published.

Sleeve
Trend
Risk filter
Exit
Apex Momentum
Long-term moving-average gate
Volatility throttle + rates overlay
Trailing stop · regime flip · shock guard
Bellwether Core
Long-term moving-average gate
Volatility throttle + tight rates overlay
Trailing stop · regime flip · shock guard
Block Catalyst
Long-term trend filter
Market-structure signal (non-price)
Trend break · drawdown rule · structure flip
Ether Surge
Long-term trend filter
Fast panic rule (outsized short-window drop)
Trend break · panic drop · cooldown re-entry

What these have in common. All four are trend-following with an explicit risk gate. None is mean-reverting. None uses options, shorts, or stacked leverage on top of the leveraged ETFs. Position size on each sleeve is fixed; there is no portfolio-level allocator voting between them. Each sleeve runs its own daily process, reads its own data, and writes its own state.

Independent sleeves

No allocator. No voting. Each sleeve has its own data feed and state.

Daily cron

Signals computed at scheduled times. Data failures degrade safely.

Monthly external review

Independent AI critics audit performance each month. Reports public.

Market regime · live

Read the weather before the signal.

A four-pillar read on trend, volatility, credit, and macro — rebuilt every day from public market and FRED data. It's a context dashboard, not a trading signal: no sleeve buys or sells off this number, so there's nothing here to overfit.

Context only · reads current conditions, does not forecastHow we read regime →
Second product · Smart Money

Track concentrated conviction.

A separate Quantsleeve tool that reads free SEC filings to follow concentrated hedge-fund managers — the ones who bet big on few names. Six live views, rebuilt from 13F and N-PORT data.

Strongest Buys

The highest-conviction new and added positions, ranked.

Hot Buys

Actionable

The freshest tradeable buying, surfaced first — the board worth watching.

Strongest Sells

Where conviction is being pulled — the largest exits and trims.

Private Unicorns

Private-company marks pulled from fund N-PORT filings. A window, not a trade.

What's New

Every fresh filing as it lands, newest first.

Sector Rotation

Where the concentrated money is moving, sector by sector.

Open Smart Money

Built from public filings on a ~45-day delay, and longs only. It shows what concentrated managers reported holding — a research lens, not advice or a real-time feed.

The research

Four strategies. Hundreds of variants discarded.

~60+
Apex Momentum versions tested
~370
Bellwether Core backtests swept
24×
Block Catalyst Calmar improvement
6
Ether Surge edges tested & rejected

The four sleeves on this page are the survivors of a long elimination tournament. None of them is the first version. None is the tenth.

The Apex Momentum sleeve began as a prototype that looked extraordinary until a triangulated review caught a look-ahead bug inflating its returns. The honest baseline was real but unlivable — drawdowns too deep to hold. Roughly sixty iterations later the locked version emerged: the same core idea, layered with regime gates, a volatility throttle, a rates overlay, and a structured recovery path. Bellwether Core is that same architecture retuned for the S&P through an autonomous sweep of hundreds of backtests; its locked configuration sits inside a wide plateau, where small parameter changes barely move the result.

Bitcoin walked through five labeled versions, each justified by an ablation study; the decisive gain came from a non-price market-structure signal that two independent reviewers proposed independently — convergence that suggests a real edge rather than a fit. Ethereum was the hardest. Reviewers agreed its edge was relative strength against Bitcoin; the backtest proved them wrong, so it was dropped, along with funding timing, trailing stops, momentum crossovers, channel breakouts, and stage analysis — each made the result worse out of sample. What survived was the proven Bitcoin engine with one rule removed: the highest-return and by far the roughest sleeve, and the page says so plainly.

Every strategy in production has been reviewed by at least two independent AI critics acting as CMTs, hedge-fund PMs, and risk managers. The reviews are saved in the project repo.

A note on what's public.The architecture is documented — the shape of each strategy, the categories of filter, the order of operations. The exact thresholds, position-sizing rules, and calibration values are not. The work is shared; the recipe isn't.

Scope

What this is not.

Honesty about the perimeter is the easier half of honesty.

Not a stock picker

The system never buys an individual company. It holds one of four instruments, or it holds T-bills.

Not intraday

Signals are computed on daily closes. A chandelier-breach check before the close exists only so a stop-out lands the same day rather than the next morning.

Not options or shorting

No premium-selling, no spreads, no put hedges. The only leverage is the leverage already baked into the leveraged Nasdaq and S&P 500 ETFs; Bitcoin and Ethereum are held spot.

Not a forecast

Nothing here predicts where any market is going. The system reacts to trend, volatility, rate moves, and funding pressure — and parks in cash when none favor a position.

It is a portfolio piece and a research log made public.

Read this part

What these numbers are, and what they aren't.

These figures are hypothetical — generated by running the strategy logic against historical data, net of estimated transaction costs and slippage, but without the friction of real execution. Live performance lands materially below backtest.

Backtests overstate. Survivorship bias, parameter drift, fitting to a sixteen-year window that included specific monetary regimes — all of it inflates results. Plan on drawdowns being deeper than the worst historical one.

This is a hobby quant project and a portfolio piece. It is not investment advice, a fund, or a service. The code, methodology, and live signal log are public so you can audit the work yourself.