← Back to the projects
RESEARCH NOTES / UPDATED 14 NOVEMBER 2026
Selling premium.
Studying the tails.
Exploring systematic options premium selling: collecting option premium and seeking to harvest theta as time passes. The research question is how to manage the directional, gamma and volatility exposure accepted in exchange.
What this page showsResearch methods and selected headline findings. Historical numbers below are model outputs, not personal-account returns. Exact entry rules, defence thresholds and implementation details remain private.
3,077historical price-path sessions
100pricing / timing cases in one study
22dates in a separate recent replay
01 / THE INVESTIGATION
Income is only half the question.
PREMIUM & THETAWhat is being harvested?
Studying the relationship between premium received, time decay and the risk carried. Theta describes sensitivity to time; a positive theta position can still lose when price or implied volatility changes.
INVENTORY & TAILSWhat survives a defence?
Comparing how reducing inventory changes exposure to continued moves and subsequent reversals. Keeping accumulated realised losses in the portfolio calculation.
EXECUTION & RELIABILITYWhat can actually happen?
Investigating quote availability, execution costs, broker acknowledgements and failure scenarios. Treating a modelled exit and a confirmed fill as different evidence.
02 / HISTORICAL MODEL STUDY
A trade-off, not a free improvement.
One completed study compares a reference management policy with two anonymised loss-reduction candidates on the same 3,077 SPX price paths from 15 November 2010 to 29 December 2023. It is a fixed-size SPXW model with a $100 multiplier, including model fees. This is a sampled historical universe, not complete calendar coverage.
CANDIDATE A / VERSUS REFERENCE52.3% smallerWorst model day loss
THE ASSOCIATED COST9.7% lowerModel average daily P/L
DAILY-CLOSE PATH45.1% smallerModel maximum drawdown
Why this is not a trading-return forecastSynthetic option prices, incomplete minute paths, recovery-timing proxies and expiry-reference assumptions limit the result. A deliberately discontinuous severe-repricing stress reversed the candidate ranking and made candidate averages negative. That stress is not a calibrated daily market process, but it exposes material sensitivity to pricing and costs. A favourable central simulation alone does not establish an edge.
Download the headline comparison CSV ↓
03 / RECENT OBSERVED-DATA REPLAY
Checking the idea against a tougher window.
A separate replay covers 22 matching ES dates, 15 October to 13 November 2026. Each policy makes its own simulated decisions; these are neither actual personal-account returns nor a direct SPXW historical comparison. This recent window remained negative under both execution assumptions.
Missing quote/path observations and unverified official settlement limit all four estimates. The adverse simulation follows a different path and is not a guaranteed lower bound. Daily-close drawdown does not measure the full intraday risk. No deployment recommendation follows from this table.
Download the recent comparison CSV ↓
04 / A REAL DATA-ENGINEERING WORK SAMPLE
One day of collection, made inspectable.
The downloadable one-day audit comes from an actual recorder session dated 13 November 2026 in America/New_York. It groups 189,352 recorded frames into 229 hourly product/outcome rows: 74,297 ES, 112,251 MES and 2,804 system frames. This is collector telemetry, including unsuccessful attempts and health outcomes, not 189,352 valid or executable quotes.
Use it in Excel or Python to count qualified versus unsuccessful observations, inspect hourly collection activity and practise building a coverage report. Empty CSV groups mean no recorded frames for that group, not proof of an exchange outage. The recorder day crosses UTC midnight; each bucket is labelled UTC.
Raw bid/ask prices, Greeks, account identifiers and contract identities are excluded. The file demonstrates genuine collection-quality analysis without redistributing the underlying quote feed. Coverage here is the ES/MES hedge recorder, not the complete collector universe.
Evidence boundariesPersonal-account real-money trading is part of the owner's stated activity, but private live-account performance is not published or verified on this page. Figures come from saved research outputs. Publication date: 14 November 2026. Exact strategy configurations remain private; headline files support arithmetic inspection, not independent reproduction of the proprietary strategy.
Try the interactive payoff example →