What the Model Does
Aether does not predict prices. It assesses whether the weight of evidence on a stock is unusually strong or weak relative to what the current price already assumes.
The Four Layers
Every assessment runs through four layers, in this order, and the output states where they agree and where they disagree. Disagreement is information, not noise to average away.
1Business Quality
- —Moat type and moat direction.
- —Management capital allocation.
- —Whether the business is explicable in two sentences.
2Valuation
- —Absolute multiples vs. the company's own history and its peers.
- —A reverse read: what growth does the current price already assume, and is that bar conservative or heroic?
3Technical / Timing
- —Primary trend, relative strength vs. sector, volume confirmation.
- —Never overrides Layers 1–2. Affects when to act, not whether.
4Macro / Catalyst
- —Sector cycle position.
- —A named, dateable catalyst — or an explicit statement that none exists.
The Funnel
How picks are selected.
Universe
A curated universe of covered stocks across the sectors we follow.
Quality
Positive or credibly-turning free cash flow; sector-appropriate leverage; no forensic red flags; an identifiable moat.
Valuation
Attractive on at least two of three valuation lenses.
Survivor counts are published at each stage so the filter is visible. See them live on Coverage and Picks.
The Compass Score
Every covered stock gets an Aether Compass score from 0 to 100 — a fixed, fundamentals-only formula, computed the same way on every page:
Core = 0.35 × Valuation + 0.30 × Growth + 0.35 × Financial Health
- ·Momentum is deliberately excluded — the score reflects what a business is worth and how it is growing, not where its price has been. Momentum is still shown on each stock page, in a separate spot clearly labeled "Not part of the score".
- ·Aether's own prediction-accuracy history is not part of the score either; matured calls are graded separately on this page.
- ·Scores map to labels at fixed thresholds: below 40 Weak, 40–59 Mixed, 60 and above Strong.
- ·The same formula feeds the funnel and the research list — same data in, same score out, no discretion.
What the output contains
- ·A conviction score, 0–100
- ·A signal: BUY / HOLD / SELL
- ·A transparent scenario calculator — arithmetic you control, not a forecast
- ·Stated invalidation criteria — what would prove this wrong
The Scenario Calculator
For forward-looking questions we don't publish a forecast. Every stock page has a scenario calculator instead — transparent arithmetic you control, over a 5-year horizon (switchable to 10):
Price in N years = Revenue today × (1 + growth)N × net margin ÷ shares outstanding × P/E
- ·Inputs default to the company's own history: revenue growth and net margin from the 5-year median, P/E from the 5-year median. Shares outstanding is the current value, held constant.
- ·Bear, base and bull presets: growth and margin use the lowest, median and highest 3-year rolling averages of the past 5 years; P/E uses the 25th, 50th and 75th percentiles. Rolling averages and medians keep a single freak growth year from dominating the defaults.
- ·Guardrails on the presets: growth is capped at 25%/yr for Base and 35%/yr for Bull (20% and 25% on the 10-year horizon) and Bear is floored at -20%/yr; every P/E preset is clamped to 8–40×. When a cap or floor changes a value, the calculator says so under that input. The sliders still go to 50% growth and 60× P/E, so you can test more aggressive assumptions yourself.
- ·A "what's priced in" line shows the yearly revenue growth needed to justify today's price at the base margin and P/E.
- ·The calculation is shown step by step — revenue → profit → profit per share → price — so you can check the math yourself.
- ·Loss-making companies and those with under 3 years of data see "Not enough data for a meaningful scenario" instead of numbers.
- ·It's a calculator, not a forecast. Results depend entirely on the assumptions.
What We Won't Claim
- —We don't publish a headline accuracy number and won't fabricate one.
- —Every prediction is logged and graded against what actually happened at the stated horizon.
- —Those numbers are published on this page — including the bad ones — once the sample is meaningful.
- —If the model doesn't beat a passive index fund over time, we'll say so.
Known Limitations
- —The model is backward-looking in its inputs. It has no special ability to anticipate novel events.
- —Small and micro-caps are harder to assess — thinner data.
- —Early-stage, pre-revenue and loss-making businesses get no scenario at all — the calculator says "Not enough data" rather than guessing.
- —The world changes faster than any model updates.
Track Record
Every prediction is logged and graded against what actually happened at the stated horizon. Those numbers are published here — including the bad ones — once the sample is meaningful.
Calibration data is still accumulating — 0 resolved predictions on record. Per-band hit rates will appear here once each confidence band has at least 5 resolved calls, and confidence language across the platform will then be anchored to those demonstrated rates.
Until that data is available: treat our assessments as one informed input among several — not as a standalone signal.
Source Credibility — One Bar for Everything
We score credibility the same way for external news and data sources and for our own predictions. A user comparing our prediction to a news headline should be able to assume both were held to the same visible bar.
- •Data freshness: How old is the underlying number — a live quote vs. a delayed feed vs. a stale filing.
- •Source reliability: Is the origin a primary filing, a reputable aggregator, or an unverified secondary report?
- •Corroboration: Is the claim backed by more than one independent source?
- •Falsifiability: Is the claim specific enough that you could tell if it were wrong?