Methodology
See-Market is built on one promise: every read is transparent and every call is recorded. No black box, no cherry-picking. Here's the gist: a transparent quant baseline gives each market a Strength score, the AI Oracle adds its own directional call — which can disagree with that score — and both are graded in public.
Strength — the quant baseline (0–100)
The quant baseline gives each market a Strength reading from 0 to 100: roughly where today's price sits within its own recent range, blended across short, medium and longer timeframes so no single window dominates. 50 is the neutral midline — above 50 leans bullish, below 50 leans bearish, and the further from 50, the stronger the read. Readings hugging the middle (about 45–55) are a no-call and aren't graded. Strength is a position-in-range gauge, not a probability of going up.
The strength bar
The bar under each read shows how far Strength sits from the neutral 50 — a stretched, one-sided reading fills more of it; a balanced one barely moves it. It's a quick visual of how decisive the reading is, not a probability.
The AI Oracle's call
On top of the quant Strength, the AI Oracle makes its OWN directional call — bullish, bearish or neutral — and it can ride the Strength reading or fade it (e.g. call a pullback on an overstretched tape), explaining itself in plain language and weighing the day's notable news when there's something real to cite. So yes: Strength can read high while the Oracle calls bearish — that gap is the whole point. We track the Oracle's calls separately from the quant baseline, so you can see whether its judgement actually beats the raw model. Not investment advice.
The accountable ledger
Every read is appended to a public ledger with its date, price, the Strength reading, both calls (quant baseline + AI Oracle) and a fixed 5-trading-day horizon. Five trading days later we compare the close to the close on the read day: a bullish call is a hit if price rose, a bearish call if it fell (no-calls aren't graded). Each result — hit or miss — is locked in and never edited.
Hit rate = hits ÷ graded calls. We publish two, side by side and honestly:
- Quant baseline — the transparent range-position model. Its backtest spans a bullish window, so it's a sanity check, not proof of edge.
- The AI Oracle — the AI's own calls, recorded live going forward. This is the one that shows whether the AI adds value; still accumulating.
When each read is recorded
Reads are recorded once per trading day at 22:00 UTC — after Asia, Europe and the US have all closed — so every stock and commodity uses that day's closing price, never a mid-session snapshot. Scoring and grading are close-to-close on each instrument's own trading calendar, so markets that close at different local times stay comparable. Crypto trades 24/7 and has no close, so Bitcoin and Ethereum use the 22:00 UTC snapshot.
Data & limits
Market data comes from Yahoo Finance, so intraday prices can lag. This is a research project and a transparent quant + AI market read — not investment advice. Markets can and do move against any read.