InsiderNode

HOW IT WORKS

DECLASSIFIED METHODOLOGY & DATA SOURCES v1.0

The Alpha Score™

The Alpha Score (0-100) is a percentile ranking that summarizes an insider's historical ability to beat the market after their BUY filings. It combines 4 measurable components to help you prioritize which insider trades to pay attention to.

Component 1: Performance
Alpha vs SPY (180D)

Did the insider's stock beat SPY at 180 days? Higher average alpha = higher score.

Component 2: Consistency
Win Rate (smoothed)

How often did they beat SPY? We use (wins+10)/(n+20) to avoid overfitting on small samples.

Component 3: Confidence
Sample Size (n_180d)

More historical BUY trades = more reliable. Insiders with 15 trades score higher than those with 2-3.

Component 4: Skin in Game
Transaction Value (ln)

A $10M BUY signals more conviction than $50K. We use log-scaling so value matters but doesn't dominate.

IMPORTANT: The Alpha Score is a 0-100 percentile ranking computed exclusively for BUY transactions. SELL transactions are shown for informational purposes but do not receive an Alpha Score, as we focus on identifying conviction buys that may generate positive alpha.

Alpha vs SPY & Win Rate

We measure insider performance at the 180-day horizon (≈126 trading days) to capture medium-term conviction.

ALPHA CALCULATION:

alpha = stock_return - SPY_return

entry = close price on filing date
exit = close price 180 calendar days later

stock_return = (exit_price - entry_price) / entry_price
SPY_return = (SPY_exit - SPY_entry) / SPY_entry

• Alpha > 0 = insider's stock outperformed SPY
• Example: Stock +15%, SPY +8% → alpha = +7%

WIN RATE (SMOOTHED):

win = (alpha > 0)

smoothed_win_rate = (wins + 10) / (n + 20)

• Win Rate = % of trades that beat SPY at 180D
• Smoothing (Bayesian prior) prevents misleading stats from small samples
• Example: 8/10 wins → (8+10)/(10+20) = 60%

NOTE: All metrics computed BUY-only. We track conviction signals, not SELLs (which often reflect diversification, tax planning, liquidity needs).

Confidence Levels

Not all insiders have the same track record length. We classify confidence based on n_180d (number of historical BUY trades in our database):

High Confidence
n ≥ 15

Large sample. Alpha Score and Win Rate are statistically reliable.

Medium Confidence
5 ≤ n < 15

Moderate sample. Metrics are directional but use caution.

Low Confidence
n < 5

Small sample. Metrics may not be representative. Early-stage tracking.

TIP: On the Notable Signals page, you can enable "High-Confidence Only (n≥10)" to filter for insiders with proven track records in Big Players and Top Alpha categories.

Data Integrity

We source raw data directly from SEC EDGAR filings via institutional-grade APIs (Finnhub, FMP).

Unlike other platforms that show "Option Exercises" (free shares given as salary), InsiderNode filters strictly for Open Market Transactions (Code P and S). This ensures we track only real capital deployment.

Stop Trading Blind.

Access the Alpha Score and Real-time Alerts.