The thesis
Both extremes give you variables. We give you a verdict.
Pay institutional prices, or wade through retail noise. Either way, you're stitching twelve signals together in your head. We weight them once and ship a single number.
Institutional
$24,000/yr
OPRA + Dark Pool feeds. Powerful, but raw and unaffordable.
OptSec
The middle$20/mo
Institutional flow, distilled to a single conviction score.
Retail
~3%signal
Free, social-media-driven, late by the time it surfaces.
Dashboards: twelve variables, yours to combine
Premium, sweep flag, Vol/OI, IV, IV rank, days-to-expiry, strike distance — and ten more columns. You mentally combine them on the fly. That's exactly what trips traders mid-session.
C-Score: one number, weighted by us
We've already done the weighting — transparently, with weights you can see. Your job is to act or pass. Cognitive load drops from twelve variables to one.
The IP
One score. Six factors. No dashboards.
The Conviction Score (0–100) is the only number you need. Six weighted factors, recursively tuned. Here's exactly how it builds.
Live computation
NVDA 0DTE 122P · Sweep
Formula
Aggression
3.5
Size
3.0
Vol/OI
2.5
Scroll to walk through each factor. Final score lands at 87 — Conviction-grade.
Trade Aggression
Sweeps mean urgency — multi-exchange execution to fill fast. Blocks mean negotiated positioning. We weight Sweeps heavier because hesitation rarely shows up first.
Size & Premium
Trade premium normalized against the option's average daily premium. >$1M or >500% of daily average earns a max score. Smaller trades decay logarithmically — meaningful, but only relative.
Volume vs Open Interest
Vol/OI above 1.5 is fresh risk being placed — not existing positions being closed. The cleanest tell that something new is happening.
Time to Expiration
0–14 DTE means event-driven, catalyst-aware capital. 365+ DTE means deep conviction LEAPS. We weight both ends — the middle is usually noise.
Custom Watchlist Boost
Tickers on your watchlist get a small relevance boost (0–3 pts). The signal still has to earn its weight — we just put your context first.
CopyCat Deduction
If the same strike is already being chased internally (>2.5:1 call:put) or trending on WSB/X, we deduct — the trade is already crowded. Protects you from being last in line.
Worked example
NVDA 0DTE 122P · the math.
A real-shape signal traced through every factor. The Conviction Score lands at 87.
| Factor | Input | Weight | Δ C-Score |
|---|---|---|---|
| Trade Aggression | Sweep · above ask | × 3.5 | +32 |
| Size & Premium | $1.42M · 580% daily avg | × 3.0 | +23 |
| Volume vs OI | Vol 14,200 · OI 4,400 (3.2x) | × 2.5 | +20 |
| DTE | 0 days (event-driven) | × 1.0 | +10 |
| Watchlist boost | On user watchlist | +0–3 | +5 |
| CopyCat deduction | Low crowd saturation | −5 to −15 | −3 |
| Final C-Score | clamped to [0, 100] | 87 |
Factor weights are static. The model around them is not — see recursive tuning below.
The machinery
The system around the number.
Four systems exist only to feed, verify, and tune the Conviction Score. They are not features alongside it — they are how it earns its weight.
Post-score verification
Print Confirmation
After the C-Score crosses your threshold, we wait for the stock print to confirm direction. Then the alert fires.
F₆ — Crowd deduction
CopyCat Index
When WSB or X lights up on a strike, the C-Score drops 5 to 15 points. This is the F₆ term in the formula above.
Tuning source
Signal Archive
Every C-Score is archived with its outcome. The overnight job rebuilds the factor curves against the archive.
Your alert floor
Custom Thresholds
Define the minimum C-Score that earns an SMS. Below your floor, the score is computed but silent.
Recursive tuning
The model self-tunes.
The weights are fixed. The thresholds beneath them are not. Every signal is archived with its outcome — directional move, magnitude, time-to-fill — and an overnight job recalibrates the factor curves against the Archive.
This is Phase 2 of the OptSec algorithm: community-wide tuning that keeps the model honest against real-world outcomes, and per-user tuning that weights factors slightly differently for traders whose watchlists skew tech vs. energy vs. financials.
Tuning cycle
Signal fires
C-Score computed, archived
Outcome observed
Direction + magnitude over 1d / 3d / 7d windows
Cohort analysis
Per-factor accuracy vs. baseline
Threshold drift
Curve parameters adjust within bounded limits
Deployed at 04:00 ET
Pre-market, before the next session
The boundary
What we deliberately don't score.
The Conviction Score is opinionated about what it ignores. These exclusions are intentional — and noted in every alert.
Directionless trades
Straddles, strangles, condors. The model is designed for directional conviction. Multi-leg structures get a flag, not a score.
Illiquid options
Sub-$50K daily premium, OI < 100. The signal-to-noise floor is too high to compute a reliable C-Score.
Earnings-week IV crush
We exclude the 3 trading days before earnings. Implied vol distortion overwhelms the flow signal during that window.
Penny options
Anything under $0.10 premium per contract. The size factor breaks down at that scale.
When the Conviction Score is unavailable for any of the reasons above, OptSec surfaces the underlying flow as an unscored signal with the specific exclusion noted. You stay informed; we stay honest.
Pricing
One score. One price.
Private beta. Invite required. The full tier table lives on the pricing page.
Get started
Two ways in.
Already have an invite code? Activate below. Otherwise, request access and we'll be in touch within 48 hours.