About·QuantivWhat options
are saying.
Quantiv reads the options chain like a tape. For every print we measure what the market is paying for movement, what the stock has actually delivered across the last twelve quarters, and where today's premium sits inside its own 52-week history.
Three lenses on every print.
Implied move
Realized track record
ML forecast
The pricing engine, in seven lines.
Every chart on the ticker page traces back to one of these formulas. We show the math because the assumptions behind it (log-normal returns, constant volatility over the horizon) matter for how you read the output.
From an option quote to volatility
We match a call and put with the same symbol, expiry, and strike (nearest to delta ≈ 0.5), then average the two annualized IVs from the chain. Zero-sided and crossed markets are excluded before selection; the nightly scoring run also rejects stale, illiquid, and excessive-spread quotes.
What dealers are pricing
ATM call mid plus put mid, divided by spot. That is the market's dollar cost of a two-sided move through expiry, expressed as a percent of the stock. Collect the straddle if you expect a smaller move than that price implies; pay it if you expect more.
Scale IV to the horizon
ATM IV is annualized. To compare it with the straddle move, we scale it down to the print expiry. Front-month IV bakes in earnings risk; the next expiry is your ‘quieter’ reference.
Black–Scholes with dividends
Standard Black–Scholes–Merton sensitivities (continuous dividend yield). We surface the chain’s published ATM Greeks per expiry — delta, gamma, vega, theta — so you can see how delta-flat a position is, how much it moves on a 1-vol jump, and how fast theta accelerates into print.
Rich versus what actually printed
Compares today's implied move to the average |close-to-close| over the last four prints. ≥ +20% = options are pricing the print at least a fifth richer than recent history. Sample size is small; treat as a quick prior, not a signal.
LightGBM quantile ensemble
Five gradient-boosted models, one per quantile of |move|, trained walk-forward across every observed earnings event in the universe with no look-ahead. The 80% band P10–P90 is the model's confidence interval, not a guarantee.
Straddle exceedance
The ticker chart compares the ATM straddle move s with the validated P10, P25, P50, P75, and P90 absolute-move forecasts. It linearly interpolates only between those quantiles and reports bounds outside them. This is a market-relative model readout, not a sixth model or a return forecast; spreads, fees, and the post-event volatility change are not included.
From chains to decisions, hourly.
Chain
Hourly OPRA chain snapshots covering every listed expiry and every strike. Landed into a DuckDB-backed parquet warehouse.
Math
ATM IV, straddle and IV-scaled expected move, skew, term structure, and published Greeks from the chain. Density bands and the LightGBM 80% forecast on the print expiry.
History
Realized close-to-close moves bracketed by Finnhub-grade earnings timing (BMO/AMC). Twelve quarters per name; EPS / revenue overlay where available.
Score
Rich-vs-hist edge, IV rank vs trailing 52w, and the LightGBM ensemble’s edge over options. Names ranked so the interesting ones rise to the top of the screener.
Only validated snapshots reach the product.
Market and earnings features retain the observation time used for scoring.
Schema, duplicate-key, missing-record, and coverage checks run before scoring.
Versioned model artifacts must pass calibration, baseline, and IV checks.
A snapshot is published only when every required control passes.
Quantiv