See in 5 seconds whether a stock looks strong or weak. We run the same academic factor research that built modern quantitative finance — across 1,000+ companies — and give you one honest 0–100 score per stock. No AI hype. No jargon. Just the maths that has worked for 50 years.
Free for everyone while we test it on real markets. Research signal, not buy / sell tips — use it as a second opinion on your own homework.
Framler is a multi-factor stock research engine covering 1,000+ US equities. Thirteen peer-reviewed factor families compress each name into one honest 0–100 score with a conformal prediction interval — built on public data, methodology disclosed, no AI in the scoring path.
Free during early access. Research signal, not investment advice — a transparent factor frame for your own analysis.
NVIDIA Corporation
Strong Signal
Framler is built in-house from peer-reviewed quant finance research. No AI guesswork. No black boxes. Every output traces back to its inputs.
Open any score and see which factors helped, which hurt, and by how much. No black-box numbers — every reading traces back to its drivers.
How well past scores predicted real returns. Updated daily. Honest about what we don't yet have enough data to claim.
Rerun the engine through 2008, 2020 and 2022. We publish the misses too — including where the model failed.
Every score comes with a confidence range. Wider range = less sure. Tighter = more sure. We don't hide it.
When you score 1,000+ stocks daily, some "look strong" just by chance. We use a published statistical filter (FDR) to flag what's real.
24 academic patterns where multi-factor configurations historically associate with directional moves — VALUE TRAP, SHORT SQUEEZE, CONFIRMED BEAT and more.
● LiveScores, verdicts and price changes pulled from today's universe — click any ticker for the full breakdown.
● Top forecasts today
Highest Framler-scored tickers as of the latest scoring run. Each card links to the full forecast — factor decomposition, regime context, and the 90% prediction interval per horizon.
We don't ask you to take our word for it. The architecture is published, the tests are live, the invariants run continuously. Read the code, run the diag, replay the backtests yourself.
Unit tests covering the factor pipeline, copula math, conformal calibration, and every public formula. Pass on every commit.
See test list →9 public structural invariants on /coherence + 59 internal engine self-checks on /diag/engine. Run continuously against production data. If any one breaks, the page goes amber and we get paged.
View invariants →Each factor module cites the academic paper it implements. The methodology page maps every block to its source.
Open methodology →No in-sample fits, no look-ahead. Every weight change replays the 2022 bear market before deploy; GFC 2008, COVID 2020 and 2024 vol cluster are coded and pending historical-bar backfill.
See backtests →Framler is a free multi-factor stock research engine that scores 1,000+ US equities daily using 13 peer-reviewed academic factor families — quality (Novy-Marx 2013), value (Fama-French 1992), momentum (Jegadeesh-Titman 1993), insider flow (Seyhun 1998), post-earnings drift (Bernard-Thomas 1989), NLP tone (Loughran-McDonald 2011), options flow, and more. Each score includes a Mondrian conformal prediction interval and is regime-conditioned via Bayesian Online Changepoint Detection. The methodology is fully public — every factor cites a paper, every weight is documented, and 9 structural invariants run live at /coherence.
No AI in scoring. Framler is a deterministic probability engine — given the same inputs, it always produces the same score. There is no LLM in the scoring pipeline, no neural net trained on opaque data, and no opaque inference. The architecture is grounded in 25+ peer-reviewed quant-finance papers; the pipeline (Bayesian posterior, conformal intervals, walk-forward CV) is documented end-to-end. Optional explanations available on the ticker page may summarise an already-computed score in plain language — they never change the score itself.
Three layers of audit run on every commit: 800+ unit tests covering the factor pipeline, 9 public structural invariants checked live at /coherence (Σ posterior = 1, weights sum to 1, correlations within bounds, etc.), and walk-forward cross-validation on every weight change with results visible at /backtest. A deeper 59-check engine self-test runs admin-only at /diag/engine — internal calibration constants are kept private, the structural laws stay public.
The full engine — 13-factor scoring on 1,000+ tickers, nightly refreshes, all sector + pattern pages, backtest harness, and the /diag invariants — is free permanently. Pro tier launches Q3 2026 for institutional features (alerts at scale, custom universes, API access). The free tier will keep the core scoring after that.
The architecture (which factors, which math, which papers) is what makes the engine inspectable and trustworthy — it has to be public for our claims to be verifiable. The calibrated values (per-factor weights, scale parameters, regime hazard rates) are what makes the engine an edge — those stay proprietary. Same model as a published-but-secret-recipe kitchen.
Yes, on two layers. (1) Macro catalyst calendar: FOMC, CPI, jobs, GDP, OPEC dates are tracked, and the conformal interval widens meaningfully as a high-importance print approaches — a published score is honest about being less certain the day before CPI. (2) Commodity sensitivity: per-ticker rolling beta vs WTI, gold, copper and the dollar index is computed weekly. Energy and miners load meaningfully on at least one; tech sits near zero. The macro calendar is live; commodity sensitivity is accumulating data and rolls into the per-ticker composite as each calibration window matures.