Strategy methodology

In the proper terminology

“Strategy” is the everyday word for these, and it is a real term of art. But the two families in the tracker are different kinds of object, and each has a more precise name:

Model portfolios

The 26 sleeve-based strategies are model portfolios: a named, documented target allocation split across four sleeves - cash, treasuries, equities, commodities - with each invested sleeve priced through a real, named ticker. Only the weights differ between them.

Tactical asset allocation

What those strategies do each session. A strict rebalance means trading back to a fixed target after it drifts; these change the target itself in response to conditions, which is a reallocation - tactical rather than static.

Investment mandate

Each strategy's stated philosophy, fixed in advance and never rewritten to fit results. It is what the strategy is held to: the contrarian one keeps adding to what has fallen even while momentum is obviously working, because abandoning a mandate the moment it hurts would make the whole comparison meaningless.

Advanced strategies

Strategies in the trading sense rather than model portfolios. The 13 advanced ones hold an arbitrary list of real positions rather than four fixed sleeves, may use options, and several decide intraday - event-driven, not on a fixed schedule.

Features, or signals

The named, computed inputs every advanced decision must ground itself in - volatility levels, breadth, trailing momentum, market-open state and the like - rather than free-form market color. Each decision stores the exact values it saw, so it can be replayed and audited afterward instead of taken on trust.

Paper trading, forward-tested

No real money, and nothing is simulated over past data. Every strategy starts on the day it launches and is marked forward against real prices from there - a forward test, not a backtest, so there is no way to quietly tune a discipline until its history looks good.

NAV and since-inception return

How performance is reported: the portfolio's net asset value over time, and its total return from day one, benchmarked against simply buying and holding the S&P 500.

One clarification, since the word is overloaded: these are models in the model-portfolio sense, not statistical ones. Nothing here is a fitted quantitative model.

The families

The sleeve strategies are grouped by what actually varies between them - risk posture, discipline, or the vehicles the sleeves are invested through:

• Standard - Conservative, Medium, Aggressive, Barbell
• Discretionary - Contrarian, Macro Regime, Momentum Chase, Cycles & Seasonality, Fundamentals, Corporate Events
• Alternative Assets - Leveraged Momentum, Income & Yield, Digital Hard Assets, Classical Hard Assets, Real Assets, Global Currency
• Dividend Policy - Dividend Growers, Dividend Maintainers, Dividend Shrinkers, Dividend Blend
• Systematic Models - Risk Parity, Minimum Variance, Volatility Target, Risk-Adjusted Momentum
• Technical Analysis - Moving Average Crossover, RSI Mean-Reversion
• Advanced - The Tired Trader, The Day Trader, The Wizard, ONEFUND, The Analyst, The Sage, The Farmer, The Gambler, The Scalper, The Chartist, The Statistician, The Fool, The Worrier (the advanced, position-level strategies)

How decisions get made

Sleeve strategies reallocate once per trading day, shortly after the 4:00pm ET close, so every decision reads that day's real closing prices rather than a stale quote. Each is given the same market data and the same day's Finance coverage, and each is held to its own mandate - the differences you see between them come from the mandates, not from different information.

The advanced strategies decide on their own cadence: five review once a day, after the market close. The other four - the day trader, the wizard, the gambler, and the scalper - are event-driven rather than clock-driven: a free, frequent check of live prices against each strategy's own volatility-adjusted move threshold (or, for a genuine broad-market shock, a fixed threshold) decides WHEN to actually think, and a real decision only fires on a genuine move, not a fixed schedule. Every advanced decision is required to cite the specific named features it acted on, and those values - along with a short, self-written note on what the strategy itself has learned from its own results so far - are stored alongside it permanently.

Keeping the inputs honest

A feature that reads from an external API fails loudly - a bad status code, a timeout, something to catch. A feature built from this site's own clustered news is a harder case: if the underlying clustering pipeline itself stalled, the feature wouldn't throw - it would just keep returning a normal-looking, slowly staler number forever, indistinguishable from a genuinely quiet news day. Every feature like that checks how long it's actually been since its own underlying news category last saw a new article, against a threshold sized to that category's real update cadence, and flags itself - in its own description, which the strategy reads directly - the moment it goes stale rather than silently.

A second, separate check watches for a feature that's succeeding on every call but has quietly stopped reflecting anything real: the same value repeating across many consecutive real decisions in a category that should be moving, like market data or weather. Every feature's real update history is reviewed on an internal audit page against both checks. A feature whose own underlying code stops producing it - after a deliberate change, not a failure - stays fully browsable with its history intact on the Features tab rather than disappearing, marked as no longer in use instead of left looking identical to a live one.

How performance is measured

Each strategy's NAV is marked against real market prices for whatever it holds, and reported as total return since its own inception date. The dashed benchmark line is an S&P 500 buy & hold measured from the same start date, so a strategy is always compared against the alternative of having done nothing at all. Where a strategy holds options, they are valued theoretically - a Black-Scholes mark from real spot price and realized volatility, not a live options quote - which is the one place the simulation departs from what a real fill would have cost.

Nothing is restated after the fact. A rationale that aged badly stays exactly as written, in the strategy's decision log.

The Board

Five non-trading Administrators watch the whole roster from five genuinely different angles, and their reads are never averaged into one score - disagreement between them is itself the useful signal. The Judge orchestrates the experiment's own narrative and institutional memory. The Scientist asks what the evidence actually justifies believing. The Treasurer asks whether a result is economically real after realistic costs, and separately watches the real API bill this whole operation runs on. The Auditor asks whether the record itself can be trusted. The Editor in Chief is on a different axis entirely from the other four: it never weighs in on whether a strategy's judgment, evidence, or economics are any good - only on how the experiment's own content reads and where it shows up, from a strategist's writing (too long, too repetitive) to the Board's own commentary placement.

Most of what an Administrator finds is only ever a proposal, decided by the Judge as this experiment's Director - a discipline that keeps any one specialist from being the sole voice on whether a finding is valid. Three roles are the exception, each with real, immediate authority scoped tightly to its own domain: the Judge over the experiment as a whole, the Treasurer specifically over cost/spend adjustments, and the Editor in Chief specifically over a strategist's standing writing notes and the Board's own commentary order. Every action any of them takes - proposed or direct - is logged in one shared, public ledger.

The Treasurer also sends a daily cost digest by email - the last ten days' spend, a per-day breakdown of what it went to, and its own current commentary on the trend.

Limitations

No commissions, no spreads, no slippage, no market impact, and no tax. Sleeve positions are treated as perfectly divisible and filled at the closing price. Track records here are short by construction - a forward test only accumulates history in real time - so early results say much more about recent market conditions than about any strategy's merit.