Winning Insider-Cluster Signals Take 25 Days to Peak. Losing Ones Fail in 4.

Published September 8, 2026 · ClusterSignal Research

Updated October 5, 2026: refreshed with the current public dataset (one more alert reached 30 days) and corrected entry prices on two March alerts that had been recorded at the insiders’ own purchase price instead of the market price at alert time. Updated in September 2026 after we retired 29 historical alerts from the public record.

Every insider-cluster BUY alert we send comes with the same implicit question: how long do you actually hold it? We went back through every real BUY alert publicly visible on our Results page since March 2026 — 26 with complete 30-day price data — to find out. The answer turned out to be a pattern, not a guess.

The pattern: winners take time, losers show up fast

Of the 26 alerts with complete 30-day data, 19 were profitable at the 30-day mark — a 73.1% win rate. But the more useful number isn't the win rate itself, it's when each group hit its best price:

• Winning alerts peaked at a median of 25 days after the alert — almost the full 30-day window.
• Losing alerts peaked at a median of just 4 days — an early bounce that faded fast, not a slow climb.

In other words: a genuine winner rarely peaks in the first week. If a position is already fading by day 4–5, that's not always noise — historically, it's the losing group's actual pattern.

What this doesn't mean

It doesn't mean "always hold 25 days no matter what," and it isn't investment advice — past performance on 26 alerts is a real sample, not a guarantee. A cluster of insiders buying is one input, not a certainty. What the data does support is a specific, falsifiable expectation: if you're evaluating whether a ClusterSignal BUY alert is working, day 4 is too early to judge it by historical pattern — and day 25 is closer to when a real winner tends to show its hand.

What happens after 30 days?

We track alerts well past the 30-day mark, and early signs suggest some 30-day "losers" aren't finished — a few of the alerts sitting negative at day 30 in this dataset had already turned positive by day 60 or later. We're not ready to put a number on that yet; there are only a handful of losing alerts total so far, which is good news for win rate but too small a group to draw a real conclusion from. We're tracking it and will publish a proper follow-up once there's enough data to trust.

Methodology

26 real BUY alerts currently public on the Results page have complete 30-day price data, out of every alert publicly tracked since March 2026. Deduplicated by cluster (a re-detected alert for the same ticker within 30 days counts once, following the same logic as the Results page), with one known outlier (AMST, a +790% analyst-upside penny stock that skewed early recalibration work) excluded to avoid one trade dominating the averages. A 2026-09-14 cleanup retired 29 historical BUY alerts that no longer pass today's scoring criteria from the public record — the main reason this sample is smaller than when this post first ran. "Peak day" is the date each alert's best 30-day closing price actually occurred, not the 30-day return itself. Every number here is pulled live from the same database that powers our public Results page — nothing here is cherry-picked or backtested after the fact.