SENTRA INTELLIGENCE

How Sentra Works

Institutional-grade insider signal detection, powered by academic research.

OUR THESIS

Investors try to make money in the market through all different types of comlicated means; whether it's analyzing financial statements, building large language models, or even using satellite imagery to count the number of cars in a retailer store parking lot. But the single most powerful signal has always been hidden in plain sight: insiders.

CEOs, CFOs, and major shareholders see earnings before they're announced, contracts before they're signed, and product pipelines before the market has any idea. They have an edge that no analyst, no algorithm, and no hedge fund can replicate. And fortunately for us, every time they act on that conviction, they have to tell you about it.

By law, every insider trade must be reported to the SEC within 2 business days. The information is public. The signal is hiding in plain sight. So why isn't everyone using it?

The problem is noise. Insiders file thousands of trades every month: outine stock grants, scheduled 10b5-1 plan sales, compensation exercises. None of trhat means anything. Finding the trades that actually matter requires knowing the difference between an insider following a calendar and an insider acting on conviction, isnt easy.

But that's the entire thesis of Sentra. We built a system that reads every Form 4 filing across the S&P 500, classifies each insider using a methodology developed by Harvard and MIT researchers, and surfaces only the trades where someone with real conviction is putting real money to work.

Academic Foundation

The Research Base

Seyhun (1986)
Journal of Financial Economics
Insiders' Profits, Costs of Trading, and Market Efficiency
Foundational study establishing that insider purchases predict positive abnormal returns and insider sales predict negative returns, proving insiders systematically exploit private information.
HOW SENTRA USES THIS
Sentra adapted the finding that purchase size relative to company size is the strongest predictor of abnormal returns. We use this to weight relative purchase size most heavily in our expected move calculation, giving more significance to buys in smaller companies than the same dollar amount in a mega cap.
Jeng, Metrick & Zeckhauser (2003)
Review of Economics and Statistics
Estimating the Returns to Insider Trading
Insider purchases earn roughly 6% abnormal return over 6 months on a value-weighted basis. Insider sales show no significant predictive power; buys are the signal.
HOW SENTRA USES THIS
Sentra adapted the finding that insider purchases earn roughly 6% abnormal return over 6 months, with buys carrying the signal and sales carrying none. This anchors our expected move estimates and is why Sentra only tracks open market purchases, not sales or option exercises.
Cohen, Malloy & Pomorski (2012)
Journal of Finance
Decoding Inside Information
Insiders who trade outside their historical calendar patterns generate significant market-adjusted alpha. Insiders on predictable schedules generate essentially nothing. This distinction is the core of Sentra's methodology.
HOW SENTRA USES THIS
Sentra adapted the core distinction between opportunistic and routine insiders. Insiders who buy outside their historical patterns generate the meaningful signal. This is why Sentra's classification system filters for opportunistic buyers and discounts predictable calendar-driven transactions.

Our Methodology

From Filing to Signal

01
Ingest SEC EDGAR Form 4 filings
Daily ingestion of Form 4 filings across 522 S&P 500 stocks. Every transaction is parsed, normalized, and stored with full insider metadata.
02
Classify insiders as Routine or Opportunistic
3-year rolling calendar analysis per insider. Trades that align with a predictable monthly or quarterly pattern are flagged Routine. All others are Opportunistic.
03
Score every signal 0–100
Our conviction engine weighs role seniority, trade size, sector alpha history, insider clustering, track record, and known high-alpha insiders.
04
Surface actionable alerts
Only signals scoring ≥70 are surfaced as trade alerts. Scores of 60–69 are flagged for monitoring. Routine trades and low-conviction signals are filtered out.

The Conviction Score

0–100 Signal Scoring Engine

Opportunistic classification (CMP 2012)Insider trades outside their historical calendar pattern
Base: 50 pts
Routine classification (CMP 2012)Insider trades on a predictable schedule
Hard cap: 1–49 pts
Open market purchase (Form 4, Code P)Required condition — signal is void if absent
Required: +0 pts
Transaction size $500K–$5M
+8 pts
Transaction size $5M–$25M
+15 pts
Transaction size >$25M
-10 pts
Purchase as % of market cap 0.01–0.05%
+5 pts
Purchase as % of market cap 0.05–0.1%
+10 pts
Purchase as % of market cap >0.1%
+18 pts
CEO or Executive Chairman role
+14 pts
10% Owner role
+20 pts
CFO or COO role
+10 pts
Director role
+6 pts
Insider cluster — 2 buyers, same 30d window
+5 pts
Insider cluster — 3 buyers, same 30d window
+9 pts
Insider cluster — 4+ buyers, same 30d window
+12 pts
Communication Services sector
+12 pts
Financial Services sector
+8 pts
Industrials sector
+7 pts
Consumer Cyclical sector
+6 pts
Stock in bottom 20% of 52-week range
+6 pts
Stock in bottom 40% of 52-week range
+3 pts
Stock in top 20% of 52-week range
-4 pts
Prior opportunistic buy, same insider, within 180d
+5 pts
Maximum score (opportunistic)
100 pts
Minimum score (routine, any criteria)
49 pts

Scores of 70 or above trigger a Sentra signal. Scores of 85 or above are classified as high conviction. Routine traders never exceed 49 regardless of transaction size or cluster activity.

Backtest Results

88,044 Trades · 2015–2026

88,044
Insider transactions analyzed
59.7%
Win rate, all opportunistic
90%
Win rate at conviction ≥70
100%
Win rate at conviction ≥85
+12.9%
Avg alpha at 180d (score ≥70)
11yr
Dataset spanning 2015–2026
BEST ROLE
10% Owner
100% win rate at 180 days
BEST SECTOR
Communication Services
81% win rate

The Trading System

Automated Signal Execution

Qualifying signals are executed automatically on Alpaca paper trading via a daily cron job. Every trade follows a strict rules-based framework:

ENTRY
Score ≥70
Opportunistic only
STOP LOSS
−15%
Hard exit within 30d
TAKE PROFIT
+25%
Limit order attached
HOLD PERIOD
90–180 days
Role and sector adjusted
TIER 1
$10,000 split
Score ≥70, equal weight
TIER 2
$500 flat
Score 60–69, monitor size

Sentra is a research and paper trading platform. All signals are for informational purposes only. Past performance does not guarantee future results.