🧭सही Insider Signals कैसे पहचानें (और जाल से कैसे बचें)

From "Insider Bought" to "Insider Bought CONVINCINGLY"

At a basic level, it's enough to know that only open market purchases (P-code) are tracked and that purchases by the CEO/CFO carry more weight. At an intermediate level, there's more to it: not every actual purchase is a valuable signal. The goal is to distinguish conviction from noise — and especially from traps that look like a purchase but aren't a signal.

Five Signs of True Conviction

1. Cluster Buying

When 3 or more different insiders buy within a narrow time window, it's much stronger than a lone buyer. One person can be wrong or buy for personal reasons. But when the CFO, COO, and two board members buy within a few days, they share the same inside view of the company. A cluster is one of the strongest signals there is.

2. Size Relative to Market Capitalization

Absolute dollars can be misleading. A purchase of $5 million in a $200 million company = 2.5% of the entire company = huge conviction. The same $5 million in a $500 billion company = a rounding error. Always measure dollars against the size of the company.

3. Purchase at Yearly Low

A purchase near the 52-week low is a contrarian, value-oriented signal. An insider buys when sentiment is at its lowest and the price is low — not when everyone is cheering. Conversely, a purchase after a big rise means you'd be entering far above where insiders bought.

4. Pure Accumulation

The strongest picture is when insiders only buy and no one simultaneously sells. If part of the management is buying and another part is simultaneously clearing positions, the signal is mixed. Pure accumulation without concurrent selling = uncontaminated signal.

5. Correct Hierarchy

A purchase by the CEO or CFO carries the most information. They are the most deeply informed and have the most "skin in the game."

Traps to Avoid

• 10% owners and VC funds clearing positions. A large shareholder or venture fund distributing shares is not conviction — often the opposite.
• Option exercises disguised as purchases. M-code (option exercise) looks like a "purchase" in raw data, but the insider didn't reach into their own pocket. A classic trap.
• Routine planned sales (e.g., 10b5-1 plans) — pre-planned, without informational value.
• One insider with a negligible amount. A symbolic purchase of a few thousand dollars by a single person is not a signal.
• Purchase AFTER a big rise. If the stock has already risen significantly, you're entering well above the insiders.
• Penny stocks and micro-capitalization. In very small companies, one purchase skews the whole picture and liquidity is fragmented.

What the ICS Score Means

ICS (Insider Conviction Score, 0–100) is a composite score by QMA that combines these principles into a single number:

• Hierarchy — CEO/CFO purchases weigh more.
• Cluster Timing — rewards clusters of multiple insiders in a narrow window.
• Value Quality — real dollars invested, only P-code.
• Sell Contamination — 100 = pure accumulation, 0 = strong concurrent selling.
• Price Proximity — purchase near or below the current price.

The higher the ICS, the stronger the conviction the signal describes. The score is a descriptive tool — it doesn't say "buy," it says "this is how strong the conviction was."

Confluence: Confirmation from Multiple Sources

Beyond insider buying, you can "layer" additional smart money clues:

• Congress (STOCK Act) — trades by U.S. lawmakers.
• Dark pool (FINRA) — off-exchange accumulation.
• Unusual options activity.
• 13F — institutional fund positions.

These layers act as confirmatory tilt, not as a standalone edge. In backtesting, congressional overlap was on average mildly additive; the specific figure is not published (see the correction note below). The alpha still stems from the conviction and quality filter — confluence only confirms it, doesn't replace it.

What Remains of the Methodology and What Was Withdrawn

🚨 CORRECTION NOTE (31 July 2026). The specific historical figures for this methodology

(alpha, win rate, sample size) are WITHDRAWN. Two reasons: they were computed through a

historical scores table whose older part was derived from later prices, and since 31 July 2026 a

stricter significance threshold applies — derived from every trial ever run, not from a single

pass. The figures are not replaced by lower ones: the input data was never valid as of the day

in question, so no honest replacement exists. The description of the rules remains an

educational account of the methodology, not a documented result.

In backtesting (bullish window 2024–2026, overlapping snapshots, approximated market capitalisation — take these caveats seriously) the HIGH conviction cohort did better than a base of companies where insiders were net sellers. The value lies in that gap between cohorts, not in an absolute return — in a bull window the absolute figure is inflated for both.

Summary

• True conviction = cluster + large size relative to the company + purchase at a low + pure accumulation + CEO/CFO.
• Traps: clearing 10% owners, options disguised as purchases, routine sales, one small purchase, purchase after a rise, micro-capitalization.
• ICS combines hierarchy, cluster, money quality, purity, and price proximity into a score of 0–100.
• Confluence (congress, dark pool, options, 13F) is confirmation, not a standalone edge.

---

*Past performance is not indicative of future results.