Fortress Model Portfolios — how the management works
← Performance dashboard

How we manage the models

One idea drives everything: keep each client's familiar risk level, then earn a better risk-adjusted return through smarter, evidence-based implementation — never by taking more risk.

The standard we hold ourselves to over the July 1 – December 31, 2026 evaluation: a higher risk-adjusted return (Sharpe) than the firm's portfolio manager, at matched-or-lower risk. Performance is measured against the PM, not asserted.

01 The model, in three parts

Fixed

The methodology

The construction rules are academic and stable. They're optimized for risk-adjusted return and do not chase the market. This is the foundation and it rarely changes.

Continuous

The data

We refresh the inputs as often as the data allows — prices daily, fundamentals after filings — so the methodology always acts on a current picture.

Responsive

The agents

Software watches the book continuously and drafts the prescribed response for a human to approve. It makes a fixed methodology timely — it never invents strategy and never trades on its own.

02 How the portfolios are built

Evidence, not opinion, sets the weights. Each model is built from a documented stack:

Equity is held at or slightly below the PM's level for each profile; the small reduction funds the diversifier sleeve so overall volatility stays matched-or-lower.

ProfileEquityFixed incomeDiversifierJob
Conservative Income~30%~60%~5–8%Preserve capital, pay income
Moderate~45%~45%~8–10%Income with measured growth
Balanced~58%~30%~10%Growth and stability in balance
Growth~73%~12%~10–15%Long-term growth, managed ride
Strategic Growth~82%~0%~12–18%Maximize long-term growth

03 The 20 models

Five risk profiles, each implemented four ways, so every client gets the same strategy in the vehicle that fits their account: ETF (low-cost, near set-and-hold), Mutual Fund (the PM's approved menu plus a managed-futures sleeve), Individual names — Taxable (single stocks for tax-loss harvesting, municipal-bond ladder), and Individual names — Non-Taxable (same equity book, Treasury/IG-corporate ladder, trades freely).

04 Choosing and removing holdings

Every position has to earn its place. A name clears a screen (quality, value, low-risk, liquidity, diversification fit), then an add-gate: it must clear a research report with acceptable risk/reward, fit the model (sector caps, ≤ ~5% per name, right profile and account type), and beat the holding it would displace. New positions start at roughly half size and scale up on confirmation.

Names also carry an eligibility tier so the equity character matches the profile — durable-yield, low-beta names in the conservative models; higher-growth names in the aggressive ones; a core held everywhere. This lifts income where clients want it and places winners where there's risk capacity, without raising total risk.

A holding is trimmed or sold when any pre-defined trigger fires — thesis break, factor deterioration, valuation stretch, a limit breach, a better opportunity, or a governance/risk event. The triggers are written down up front so exits aren't emotional, and every change gets a documented memo with a client-ready explanation. Two living lists run alongside the book: a watchlist of researched candidates and a removal flag on anything approaching an exit trigger.

05 When we trade — and why we trade rarely

We monitor continuously but trade on rules, never on a calendar. A trade only fires when a holding drifts outside its tolerance band:

Tax efficiency is built into the order of operations for taxable accounts: first route incoming cash, dividends, and coupons to underweight sleeves (no sale); then harvest losses into a not-substantially-identical replacement, respecting the 30-day wash-sale window; and only realize gains last, within a budget, using the highest-cost lots. Tax-deferred accounts (IRAs, 401ks) rebalance freely since there's no tax cost, and the highest-turnover sleeves are held there where possible.

06 The agents that keep it fresh

Two kinds of software assistant support the process. Both propose; a human disposes. Neither ever places a trade or changes the strategy.

07 Oversight