Compare

Chart Library vs Danelfin — an AI score versus calibrated base rates.

Danelfin distills a stock into a single AI Score (1–10) — its model’s probability that the name beats the market over the next few months. It is a prediction, packaged as a rating.

Chart Library deliberately refuses to do that. It returns no score and no buy/sell call. For any (symbol, date, timeframe) it retrieves the real historical analogs of that exact setup and reports what they actually did next — a calibrated forward-return distribution with an audited coverage receipt, over API/MCP for an agent to reason on. The difference is philosophical: a rating tells you what to think; base rates tell you what happened, and how often the band held.

Prediction vs calibration

  • Danelfin = a directional score. One number per stock; you trust it or you don’t. The hard part — how often a given score is actually right — is opaque.
  • Chart Library = a calibrated distribution. The cohort’s p10/median/p90 and up-rate per horizon, plus the receipt: the nominal 80% band held 80.8% across 300K+ audited cases, PIT-flat. You can see exactly how much to trust it.
  • Single-name prediction hits a noise floor. The forward-return signal for one stock is close to irreducibly noisy — which is why we don’t sell a point forecast; the durable edge is honest, calibrated base rates, not a sharper guess.
  • Built for agents. Danelfin is a website you read; Chart Library is an endpoint your agent calls (MCP, REST, SDKs) and cites in its reasoning.

Frequently asked questions

Does Chart Library give a stock score or rating like Danelfin?
No, by design. It returns the calibrated forward-return distribution of a setup's historical analogs plus a coverage receipt — not a 1-10 score or a buy/sell call. It surfaces base rates; the decision stays with you or your agent.
Is it a Danelfin alternative?
Only if you're looking for honest historical base rates instead of a predictive rating. They're different philosophies: Danelfin predicts (AI Score); Chart Library calibrates (what analogs did, with audited coverage). Some users consult a rating and the base rates together.
Why not just trust an AI score?
Because a score with no published, audited calibration is a confidence you can't verify. Chart Library's value is the receipt — measured predicted-vs-realized coverage across 300K+ cases — not a number to take on faith.
Try it

Run a cohort_analyze call.

Free Sandbox tier — 1,000 calls/day, no authentication. MCP install for Claude or Cursor takes 30 seconds.

Related