8K backtest

Inspiration

“Smart money in the options market may know which way a stock is leaning.” We started with that intuition and asked a narrower, testable question: does unusual call-versus-put implied volatility become more informative around a company’s public disclosure? We use an 8-K as event context, not as a bullish or bearish signal, and test whether the option-market clue adds value beyond ordinary days.

What it does

The project measures 180-day at-the-money call and put implied volatility before SEC Form 8-K Item 7.01 disclosures. Item 7.01 is the Regulation FD category companies may use to make information broadly public; it covers many subjects and does not indicate a direction by itself. We compare the high-skew filing group with matched ordinary-day controls.

The notebook evaluates the five standard option strategies—long call, covered call, protective put, collar, and cash-secured put—and four bonus structures: risk reversal, bull call spread, short strangle, and short straddle. The saved in-sample comparison favors the protective put. The event-minus-control gap is positive, but small and statistically uncertain; the current evidence does not establish a reliably profitable edge.

How we built it

We use Massive market data and 8-K filings. For each company and session, we select the standard option expiry nearest 180 calendar days, pair a call and put at the strike closest to spot, and use recent valid bid/ask midpoints. A QuantLib American-option model converts those prices into implied volatility using Treasury rates and dividend carry. We subtract each company’s trailing 60-session call-minus-put IV average and standardize the residual across companies.

The event is tradeable only from the first session after the filing date. A same-company placebo group, explicit transaction-cost assumptions, fixed holding horizons, and reusable Python modules make the comparison inspectable. The submission notebook shows the full workflow and imports the event, signal, pricing, and portfolio code as modules.

Challenges we ran into

Historical open interest was unavailable, so we did not label option volume as open interest or treat volume as a substitute signal. Option quotes can be sparse, and daily option bars are only an approximation of executable fills. Item 7.01 filings vary widely in content, while the study’s fixed large-cap universe may introduce survivorship bias. We also had to separate pre-filing signal construction from post-filing trade timing and keep the reserved evaluation window out of development analysis.

Accomplishments that we're proud of

We turned the idea into a readable, date-bounded workflow rather than a single opaque backtest. It includes the five required strategies, four advanced comparisons, matched controls, cost and horizon analysis, yearly breakdowns, and a clear IV180 construction diagram. The strongest bonus, the risk reversal, outperforms several weaker standard structures in the saved comparison, while the protective put remains the strongest standard strategy.

What we learned

The 8-K plus high-skew group has higher average protective-put P&L than its matched placebo in the saved sample: a 0.35 percentage-point gap at 10 sessions and 0.40 points at 21 sessions. The corresponding t-statistics are 0.62 and 0.44, so those differences are not statistically convincing. The 33-cohort annualized Sharpe is 0.45, and the event-minus-placebo result changes sign by year. The practical lesson is that an interesting market story still needs strong controls, realistic costs, and repeatable evidence.

What's next for data-is-all-you-need

Complete the remaining in-sample sensitivity checks for expiry, entry timing, and filing category, then verify a clean-kernel run on the 2022–2024 development window. Keep the 2025–2026 evaluation configuration sealed; its IV feature warm-up uses only the 100 calendar days before the requested start, and evaluation must run only when authorized. Any future result should be judged net of costs, with uncertainty, drawdown, and year-by-year stability reported alongside Sharpe.

Built With

• massive
• quant