CausalTrader uses PCMCI+ and VARLiNGAM to surface the news, sentiment, and macro signals that actually drive a stock — packaged into a comprehensive WSJ-style memo with a per-headline basis-point attribution table.
$29 thereafter, or $29/mo for 10 reports. Cancel anytime.
Built on peer-reviewed causal-inference research — PCMCI+ (Runge et al. 2019) and VARLiNGAM (Hyvärinen et al.) — with a frontier LLM writing the analyst memo on top of the evidence.
Two assets can correlate at 0.8 for a year and have nothing to do with each other. Trading on that confusion is how analysts get blindsided by regime shifts. CausalTrader is built around a single discipline: we only report a signal if it survives a causal test.
Every report ships with the full causal graph, edge statistics, and bootstrap confidence intervals — so you can audit the conclusion, not just read it.
Four agents collect their own slice of evidence in parallel: OHLCV bars and technicals, headlines from SEC EDGAR + Alpaca/Finnhub/GDELT, social-media sentiment via xAI/Grok, and macro proxies from VIXY, TLT, GLD, and UUP.
Multi-agent collectionPCMCI+ tests every signal against next-day returns and against every other signal, controlling for confounders and lags. Spurious correlates are dropped. VARLiNGAM then attributes a basis-point effect to each individual headline, with bootstrap CIs and FDR correction.
PCMCI+ · VARLiNGAMA WSJ-style analyst memo, the directed causal graph, a 4-panel technical chart with regime shading, a per-headline bps attribution table, and an indicator glossary. In your inbox within 24 hours of purchase.
Analyst memo + PDFThis is a real, unedited report we generated for INTC over a 90-day window: a directed causal graph from PCMCI+, a per-headline bps attribution table from VARLiNGAM, four-panel technicals with regime shading, and a comprehensive WSJ-style analyst memo.
View full PDF Every report is the same pipeline: four data agents collect evidence, a causal layer filters spurious signals, and an LLM writes the memo — each layer independently testable, statistically validated, and repeatable for the same ticker on the same window.
SMA, EMA, RSI, ATR, MACD, Bollinger Bands, OBV. Fed by Alpaca minute and daily bars.
Social-media sentiment scored by xAI / Grok and the Loughran-McDonald financial lexicon.
Headlines from SEC EDGAR (8-K), Alpaca, Finnhub, and GDELT — auto-falls back if a key is missing.
VIXY, TLT, GLD, and UUP ETF proxies for fear, rates, gold, and the dollar.
PCMCI+ tests every signal against returns and against every other signal. VARLiNGAM attributes a basis-point effect to each headline with bootstrap CIs and FDR correction.
A comprehensive WSJ-style memo plus per-figure captions, grounded in the causal evidence — not the prompt. Every claim is anchored to a directed edge in the report.
We launched cheap on purpose. Until our hit-rate page is live, we'd rather have your feedback than your wallet — but we still need to keep the lights on.
Try the product for less than a coffee. Same comprehensive deliverable as our standing report.
À la carte after your intro. Pay only when you need a deep dive on a specific name.
For active investors who want a steady stream of causal reads on their watchlist.
For research desks, RIAs, and quant teams who want the engine inside their workflow.
Not investment advice. Reports are research products. All purchases subject to a 7-day money-back guarantee.
S&P 500 names at launch — those have the deepest news and analyst coverage, which gives the causal layer the most signal. Mid-caps and ADRs are coming as we expand the news source chain.
Each purchase is a point-in-time analysis of the window you choose (default 90 trading days). Pro subscribers can refresh any prior report at no extra cost within the same month — we'll re-run the pipeline on the latest bars and headlines.
No. CausalTrader is a research product. Every report includes a clear disclaimer; nothing here is a recommendation, solicitation, or fiduciary advice. If you're managing client money, treat our output the way you'd treat any third-party research: as an input, not a decision.
Most factor models report correlations to a fixed factor library (value, momentum, quality, etc.). We don't assume the factors. PCMCI+ discovers the directed dependency structure for *this ticker, this window* — including signals that would never appear in a pre-baked factor table, like a specific earnings-related news cluster.
OHLCV bars and news from Alpaca; SEC EDGAR 8-K filings; Finnhub and GDELT for additional headline coverage; sentiment via xAI/Grok; macro via VIXY/TLT/GLD/UUP ETF proxies. Sources are listed on every report so you can verify the inputs.
Every report ships with its full causal graph, the directed edges that drove the conclusion, bootstrap confidence intervals on each effect estimate, and Benjamini-Hochberg FDR correction for multiple testing. The methods (PCMCI+, VARLiNGAM, market-model abnormal returns) are all peer-reviewed and cited inline. If the evidence doesn't support a claim, we don't make it.
You do. Reports are licensed to you for personal or internal-team use. Republishing or reselling is not permitted without a Team plan license that explicitly grants distribution rights.
Because we'd rather have your feedback right now than your money. The first $5 covers our API costs and gives us a Stripe customer record so the next purchase is one click. We'll raise prices when we have a public hit-rate page; until then, we earn the right to charge more.
Yes — the Team plan includes API access to the causal graph, raw attribution data, and white-label PDF rendering, plus a discovery call to scope custom universes (international tickers, private securities, factor portfolios). Email us to talk.
Drop your email for a sample chapter and a $5-off code on your second report — or skip the line and grab the intro report now.