Quark Research publishes daily and intraday signals — regime probabilities, factor-structure reads, alpha attribution, and per-asset weights — delivered via versioned REST, WebSocket, and CSV. Suited to systematic desks, multi-asset overlays, and risk teams that need a second-opinion architecture on top of conventional tools.
Every dataset is point-in-time-reproducible, versioned under /api/v1/,
and carries explicit deprecation headers when surfaces change. Webhooks deliver push events
whenever a regime, signal, or weight crosses a threshold.
Posterior probabilities for risk-on / neutral / risk-off / crisis, plus the dominant classifier and a confidence score. Daily refresh, intraday on stress events.
GET /api/v1/regime-probabilityRegime label, tail-risk read, recommended allocation summary, and risk parameters for use in overlays. Compact JSON suited to webhook delivery.
GET /api/v1/signals/currentFull per-signal contributions across the 18-factor stack: regime score, factor stability, mispricing intensity, correlation persistence, and ten more orthogonal reads.
GET /api/v1/signals/rawReplay any signal back through the available history window. JSON for systems, CSV for spreadsheet ingestion, both with consistent schema.
GET /api/v1/signals/history.csvRebuild what the model would have produced at a specific historical timestamp. Designed for backtest reproducibility and walk-forward validation.
GET /api/v1/signals/at?as_of=...Per-signal alpha contribution, normalized weights, and rolling attribution windows. Lets risk teams see exactly which signals drove the period's return.
GET /api/v1/alpha/attributionIntegration is a plain HTTPS API — three lines of standard-library
Python and no SDK install. Public endpoints (regime, current signal, benchmark) require
no key. The institutional surfaces and the unfiltered signal stack require a
qk_ token from your tier. A packaged Python client is in preparation and
will be announced here when it ships.
# no install needed — requests only import requests r = requests.get("https://api.quarkresearch.cc/api/v1/signals/current", headers={"Authorization": "Bearer qk_..."}) # public endpoints need no key print(r.json()) print(client.regime())
# Public — no auth required curl https://api.quarkresearch.cc/api/v1/regime-probability # Authenticated — institutional tier curl -H "Authorization: Bearer qk_..." \ https://api.quarkresearch.cc/api/v1/signals/raw
A sample report, a sample CSV from the public history endpoint, and the methodology whitepaper — enough to satisfy a portfolio-engineering review without a sales call.
One-page weekly publication covering regime classification, factor reads, and the recommended overlay tilt. Mirrors what Professional+ subscribers receive.
30-day historical signal feed, CSV format. Includes regime, tail-risk read, and
allocation summary per session. Free public tier — no key required. Full 730-day
history available via authenticated /api/v1/signals/history.csv with
a 7-day trial key at quarkresearch.cc/sandbox.
Plain-English notes on how the signal stack composes the regime classifier, factor stability monitor, and tail-risk module. Linked references for desk diligence.
The signal methodology, the API reference, and the published research remain available. There is no paid plan to sign up for at this time.
The signal stack composes a Markov-switching regime classifier with a factor-stability monitor, an order-flow imbalance read, and a tail-risk module that watches the tails of the return distribution rather than its centre. Each component runs independently and combines into a single composite read, with full attribution preserved on every cycle.
Models target one specific failure mode of vanilla quantitative tools: their tendency to flag regime changes after the fact. By tracking the structure of factor returns (not just their level) the architecture sees rotation events earlier in the process, while still anchoring to public benchmarks.
Every published signal is reproducible point-in-time. Methodology paper, full attribution backtests, and a sample data feed are linked below.
Read insights →Designed for desks that already have a primary quantitative process and want a structurally orthogonal second read.
The crash-barrier and topological-stress reads update before the conventional VIX / credit-spread proxies catch up. Risk teams use them as a leading indicator on the drawdown shield.
Live regime probabilities feed into dynamic asset-class tilts. The recommended allocation surface is a starting point for an in-house overlay or a direct lift.
Point-in-time signals make it trivial to replay any internal strategy against the live signal stack as a counter-factual — an honest answer to "what would Quark have said two years ago?"
Public endpoints are open for evaluation. The validation methodology, sample reports, and the methodology paper are linked below.