Finclator reads a fixed roster of finance accounts on X, keeps only the explicit calls — Bitcoin to $85K by spring, not vibes — and scores each one against the price once its horizon has passed. Trust follows the record. The result is a single reading per asset and horizon.
| Short 0–3 mo | Medium 3–12 mo | Long 1–5 y | |
|---|---|---|---|
| BTC | — | — | — |
| Gold | — | — | — |
| S&P | — | — | — |
A language model reads every original post and keeps the ones that name an asset, a direction and a time frame — a price target if one is stated. Mood, news, past moves and questions are not calls. Every call carries the exact phrase that justified it, so any label can be checked against the tweet.
Short calls are judged after 3 months, medium after 12, long after 24. The move is compared against a flat band scaled to the asset's own volatility, so a 2 % wobble in Bitcoin doesn't count as a win. A stated target earns extra credit if the price actually got there.
Each account earns a score per asset and horizon from its matured calls, pulled toward the middle when the sample is small. Current calls are then aggregated with that score as the weight — a loud account with a poor record is outweighed, not banned.
A real row from the dataset. The highlighted phrase is the exact span the model used; entry and exit are daily closes on the instrument itself. Anyone with access can open the tweet and the price history and disagree.
Not a price forecast and not advice. It measures what a specific set of public commentators are saying and how reliable they have been on record — nothing about whether they will be right next time. Sentiment across finance Twitter has been bullish for most of the period covered, so a Buy reading is common; a flip to Neutral or Sell is the more informative event.
Finclator is a personal research project. The roster, every call, every outcome and every trust score are open to inspection in the panel — access is granted by hand to people I know. Ask for it and I'll approve you by email.