99 signals scored. Here's what the data showed.
Between January and March 2026, the Ivolution signal engine ingested and scored 99 Form 4 open-market purchase transactions filed with the SEC. Each filing was assigned a conviction score (0–100) based on a rules-based model weighing insider tier (officer vs. director vs. 10% holder), transaction size relative to compensation, transaction type (open-market purchases weighted above option exercises and awards), and historical accuracy of the reporting insider. Price outcomes were then measured at the 7-day mark using closing prices from the filing date.
The central finding: scoring works. At the 70–79 conviction tier, 73.1% of signals showed positive price movement within 7 days (p=0.004), a result that clears the 99% confidence threshold. The moderate-conviction tier (50–69) performed at 58.1% — meaningfully above chance but below the threshold for statistical significance. The highest-conviction tier (90–100) posted 77.8% accuracy on a small sample of 9 signals; we flag this as directionally encouraging but statistically inconclusive pending a larger sample. The 80–89 tier data is currently undergoing a quality review; we will publish corrected figures in the next update.
Positive outcome = closing price at T+7 days above closing price at date of filing.
Win rate = proportion of signals where T+7 close exceeded filing-date close. p-value computed via one-sample binomial test against null hypothesis of 50% (chance).
TKO Group Holdings, Inc.
Verifiable on SEC EDGAR · Accession: 0001127602-25-005792 · This is a historical example, not a current recommendation.
All filings sourced from SEC EDGAR Form 4 electronic submissions. Data ingested via the EDGAR full-text search API. Only open-market transactions are included in accuracy calculations; derivative exercises, awards, and tax-withholding sales are excluded from win-rate computation. The signal engine processes approximately 10,400 filings across the covered period; the 99 scored signals represent transactions meeting the minimum position threshold for scoring.
Signal scores (0–100) are computed by a rules-based model. Primary inputs: (1) insider classification — officers score higher than directors, directors higher than 10% holders; (2) transaction type — open-market purchases score highest; (3) transaction size — scored relative to the insider's estimated annual compensation and prior filing history; (4) timing — cluster purchases (multiple insiders buying in a short window) receive a bonus. Derivative transactions and sales are scored but excluded from the accuracy analysis published here. The model does not use price momentum or any forward-looking variable.
Win rate = proportion of signals where the adjusted closing price at T+7 trading days exceeded the adjusted closing price on the filing date. Price data from public market sources. Results are not adjusted for market beta or sector performance. Signals that were halted, acquired, or delisted within the measurement window are excluded from calculations.
This report covers Q1 2026. The signal engine is live — new Form 4 filings are ingested and scored daily.
View live signal feed