Private. Backtested. Yours. Never advice.
Technicals × news × social × fear
WebSocket prices · smart alerts
Your data never leaves
Action scores · risk · correlation · leverage
Installable · works offline
Optional Claude reviews
We tested our own signal across five time horizons — and found it barely works at a week, is inverted at 30 days, and only pays over months. We changed the product to match, and published the scripts so you can check us.
| Signal said… | +7d | +30d | +60d | +90d | up% @60d | n |
|---|---|---|---|---|---|---|
| Bullish (≥60) | +0.03% | +5.31% | +14.75% | +12.96% | 39.2 | 646 |
| Neutral | +0.01% | −2.46% | −2.73% | −0.29% | 35.3 | 184 |
| Bearish (<45) | −0.23% | −0.82% | −4.73% | −5.20% | 33.1 | 1096 |
| Whole market | −0.12% | +1.08% | +2.00% | +1.36% | 35.4 | 1926 |
Run 21 July 2026 — a rolling window, so these drift. Edge over the market by horizon: 7d +0.2 · 30d +4.2 · 60d +12.8 · 90d +11.6 pts. Note the "up%" column: even at its best the signal is right LESS than half the time — the edge is the size of the winners, not how often it's right. Separately, ranking coins against each other, our signal is mildly INVERTED at 30 days, so the app uses a contrarian signal there and ships NO 7-day signal at all. 1,926 = overlapping windows every 3 days (correlated, not 1,926 independent bets); confirmed on 198 non-overlapping ones. Top-20 actively-listed coins only (survivorship bias). Excludes fees and slippage. Scripts: /research in the repo.
No — never. Scolta is an educational research tool. It shows data and its own track record; it never tells you to buy or sell, and never touches your exchange accounts.
On your device. Portfolio, alerts, watchlists — in your browser, on your machine. Nothing personal is uploaded, ever.
One fair one-time price at launch. Waitlist members get the early-bird price.