Backtested 8-K "special situation" filings against real outcomes — early-stage signals actually beat confirmed deals

I built a tool that scans SEC EDGAR daily for 8-Ks signaling special situations, and backtested it against \~1,900 historical filings to see what actually happens after each type of signal — using each company's own subsequent SEC filings (Item 2.01 for completed acquisitions, Item 1.03 for bankruptcy, Form 15 for deregistration) to confirm outcomes, rather than relying on stock price data alone, which is unreliable for delisted names.

A pattern that stood out, sorted by average gain when a deal actually completed:

* **Early-stage signals** ("exploring strategic alternatives," still-early filings): **+19% to +26%** average gain when it led to an acquisition
* **Formal process underway** (special committee formed, financial advisor engaged): **+10% to +17%**
* **Already-confirmed deals** (signed/definitive merger agreements): **+3% or less**, sometimes slightly negative

This actually makes sense once you think about it: by the time a merger agreement is signed, the market has already priced the stock close to the deal value, so there's little room left to move. The real re-rating happens earlier, when the outcome is still genuinely uncertain.

The one that surprised me most: **going-private proposals only completed 24% of the time** in this dataset, with a **-12.7% average outcome** — notably worse than every other category, both in completion rate and price.

Special committee filings had a real minority bad outcome too — **\~5.4% ended in bankruptcy** rather than an acquisition, worth knowing before assuming "special committee = getting bought."

**Important caveat:** this is still an early, actively-being-debugged project — I've already caught and fixed a couple of real methodology bugs (ticker reuse causing wild outlier numbers, thin-liquidity exits being unreliable) as I've dug into this, and I fully expect these exact numbers to shift somewhat as I keep refining it. Treat this as a directionally useful first pass, not a final answer.

Built this into a small tool for my own use that now emails me these filings same-day. Sharing in case it's useful to anyone else here too — [https://cbfiling.github.io/filingbell-site/](https://cbfiling.github.io/filingbell-site/). It's an early, one-person project — no instant welcome email yet, your first digest lands the next time it runs.

**Genuinely open to suggestions** on what to fix, refine, or add — whether that's other trigger phrases worth tracking, better ways to classify outcomes, or anything that looks off in the numbers above. This is very much a work in progress and outside eyes from people who actually trade this stuff would help a lot.