Research & tooling

Probabilistic Not predictable

Every position you hold has a range of outcomes, not an answer. Sanctuary Research publishes the shape of that range — scored by nine academic factor models, argued to consensus by four frontier AIs, and capped so no single name can decide your year.

Free tier: the daily briefing and one tracked portfolio. No card required.

The graphic below is an illustrative probability distribution: a dense centre of likely outcomes with thin tails either side. Its labelled points are:

the tail we size for p10 median p90
◀  worse outcomes p10medianp90the tail we size for Illustrative density — not a forecast better outcomes  ▶
01 The premise

You don't get to pick the outcome. You do get to pick the distribution.

That sentence is the whole company. Everything below it is plumbing.

Nobody knows what a single stock does next quarter. Anybody who tells you they do is selling you something, and the price of that something is usually your capital.

But the shape of the odds is knowable. Which names have historically sat in the better half of the spread. How wide the spread gets. How heavy the left tail becomes when leverage, accruals and distress signals stack up. That is a measurable thing, and it is measurable in advance.

So we don't publish a prediction. We publish a distribution, and then we go to work on its shape. Nine academic factor models to move the middle. Four frontier AI models arguing three rounds before anything reaches you. A hard 20% cap on any single name so the left tail can bruise you but not end you.

You will still only get one draw. The entire job is choosing a better bag to draw from — and then being honest with you about how much better, and how we measured it.

This is why there is no ticker of the day on this website, and no number at the top of it pretending to be a promise.

Shaping the densityillustrative
LEFT TAIL CENTRE RIGHT TAIL

Illustrative geometry, not a backtest. Screening and position caps do two separate jobs: the models shift where the centre sits, the caps decide how much of the left tail can ever reach your account.

02 How it works

Four stages. Each one narrows the distribution.

Score, then context, then debate, then monitor. Nothing skips a stage, and every stage leaves an audit trail you can open.

01

Score

Nine models, one composite

Nine academic factor models — growth, value, quality, momentum, distress and manipulation — run across the large-cap universe on each name's own earnings clock, not a calendar. Their outputs are distilled into a single composite weighted by walk-forward-validated predictive power: a model that only worked in the rear-view mirror gets a smaller vote.

Composite dashboardillustrative
NameFactor profileCompositeDecile
NAME 0188.4D1
NAME 0284.1D1
NAME 0379.6D2
NAME 0461.2D4
NAME 0522.7D9

Illustrative layout, anonymised names — not live scores, and not a view on any real company. Factor profile shows six of the nine model families; the composite is the walk-forward-weighted blend.

02

Context

Grounded, then challenged

A score with no weather report is half an answer. Macro forecasts are web-grounded and anchored to FRED prints and prediction-market odds rather than vibes. Sector stances are graded by a multi-model panel and then adversarially challenged — a second pass whose only job is to find stale evidence and unsourced claims and strike them.

Macro groundingillustrative
  • FRED CPIAUCSL · core m/m print vs consensus
  • FRED UNRATE · 3-month trend stable
  • FRED DGS10 · real-yield path rising
  • Odds next policy decision market-implied
  • Panel sector stance · industrials quorum met
  • Challenge citations past freshness window struck

Illustrative panel. Every claim in the live briefing carries a pinned source; anything that can't be pinned doesn't ship.

03

Debate

Four models, three rounds

Four frontier models propose, challenge and lock a 15-position book over three rounds. Quorum is required — a lone confident model doesn't get its way. A pre-trade vet hunts red flags, and a hard validation gate enforces the mandate: every ticker must resolve, every weight must sit inside the caps, or the book doesn't leave the room.

Debate transcriptillustrative

Round3 of 3 · quorum met

  • Model 01 · proposeOverweight quality-at-a-price. Fund it from the momentum sleeve, which is carrying the widest spread in the book.
  • Model 03 · challengeTwo of those names route the same end customer. Stated industry exposure understates the real correlation.
  • Model 02 · reviseTrim position 4. Add one defensive name to hold the industry sleeve under its limit.
  • Model 04 · concedeAccepted. Distress flag on the original candidate is thinner than I argued in round one.
  • Gate · validate15 / 15 tickers resolve. Max single weight 20%. Industry sleeve inside limit. Book locked.

Illustrative transcript. Model identities are shown as roles; the live record names each participant and preserves every round.

04

Monitor

Alerts you answer, not orders you receive

Drift, drawdown, concentration and thesis-break alerts run against your book, not a demo one. When something moves, you get a machine-argued rebalance proposal with its reasoning attached — and you approve or reject it. One click, full audit trail. Nothing trades on its own.

Alert feedillustrative
  • DriftPosition 07 has run above its target weight — rebalance proposed
  • ConcentrationSingle-name weight at its 20% ceiling — inside mandate
  • DrawdownBook inside its drawdown alert threshold — no action triggered
  • Thesis breakGuidance cut invalidates the margin case on position 11

Illustrative alerts. These describe operating limits and book mechanics — no performance figure is shown or implied.

03 Product

Four surfaces on one desk.

Also inside: the composite dashboard, multi-model deep research and risk proposals, per-ticker pre-trade research, side-by-side compare, and options & Greeks.

FreeBefore the open

Daily Briefing

Indices, futures, macro prints, market-moving headlines and a sourced market analysis and outlook. One screen before the open, with sources pinned to every claim.

Free + ProBuilt to outperform

Model Portfolios & Strategies

The Sanctuary model book and short-hold swing trades — every entry, target and stop published, scored in a ledger that models costs, keeps marks current, and holds every thesis answerable to its position.

ProFour models argue

AI Portfolio Debate

Frontier models argue three rounds to a consensus allocation — quorum required, every ticker and weight validated at the gate before it reaches you.

PremiumPass the challenge. Get funded.

Futures Strategies

Systematic intraday futures books — starting with the 8-Edge NQ book — engineered around the rules that actually end prop-firm evaluations: daily loss caps and trailing drawdown. Built for account challenges, tracked fully out-of-sample.

04 Pricing

Start on the free tier. Move up only if the desk earns it.

Monthly, no contract. Cancel from inside the app and the tier lapses at the end of the period — nothing to email anyone about.

Basic

Free

For reading the market before the open and watching one book.

  • Daily Briefing, every trading morning
  • One tracked model portfolio
  • Composite dashboard, limited view
  • AI Portfolio Debate
  • Futures Strategies
Start free

No card required.

Research

$14.99/ month

Every code repository from our research videos, plus the desk that produced them.

  • Everything in Basic
  • Every code repository from our research videos — full source, download any time
  • New repositories added as each video ships
  • Full stock rankings & screener
  • Unlimited stock detail pages, no daily cap
  • Quant ratings on every covered name
  • Options & Greeks, side-by-side compare
  • AI Portfolio Debate
  • Futures Strategies
Get Research

Card required. Cancel from inside the app.

Pro The full desk

$49.99/ month

For self-directed investors running their own book with real research behind it.

  • Everything in Research
  • AI Portfolio Debate — 4 models, 3 rounds
  • Model portfolios plus short-hold swing trades
  • Macro & sector theses, and a sector read on every stock page
  • Drift, drawdown, concentration & thesis-break alerts
  • Futures Strategies
Go Pro

Upgrade or cancel from inside the app.

Premium

$72.99/ month

For funded and prop-firm traders working an evaluation with a clock on it.

  • Everything in Pro
  • Futures Strategies, including the 8-Edge NQ book
  • Engineered around daily loss caps and trailing drawdown
  • Tracked fully out-of-sample, published as it runs
Go Premium

Research and tooling only. Never a broker, never a trade.

05 Start

Read one briefing. Decide from there.

The free tier is the daily briefing and one tracked portfolio, and it stays free. If the research is worth paying for you'll know inside a week — and if it isn't, we'd rather you found out before you paid us.

Or get the briefing by email

One email before the open. Unsubscribe from any of them.