Equity consumer research
Stock signal win rates: duplicates, open alerts and missing records
Reconcile issued, unique, executable and resolved populations before interpreting an equity win-rate headline.

Give instructions and events different identifiers
One stock idea can generate an initial alert, amendments, target notifications and a closing note. Create a parent identifier and attach those events rather than counting every notification as a new independent trade. Preserve repeated candidates as well: deduplication can be appropriate for a particular evaluation, but its rule must be stated before the results are selected. Deleting repeated losing messages while retaining repeated winning ones is not a consistent population policy.
Keep a state inventory before calculating a rate
Separate executed closed winners and losers from open positions, cancelled instructions, unfilled orders and unavailable source records. There may be several useful denominators, provided each is named. Resolved-executed win rate answers a narrower question than performance across all issued instructions. Neither is an account return. Missing records are unknown observations, not implied winners, and cancellation without execution is not automatically a realised trading loss.
Reconcile timestamps and quantity evidence
For each classified closed row, request the original instruction and the event that actually resolved the intended quantity. A target-touch screenshot can describe a reference path without proving that the subscriber filled or exited. Define the win criterion after costs or before costs explicitly. Preserve breakeven treatment and partial exits; rounding can turn a small net loss into an apparently flat display if the underlying ledger is not available.
Publish what the population cannot answer
A rate can be exact within a declared small fictional sample and still say nothing about future reliability. When a real seller supplies a rounded percentage and total, do not reverse-engineer an exact win count unless the precision and population make it defensible. A reconciliation table helps a reader see which records remain unresolved. It is not a substitute for independently supplied records or a reason to publish a confidence interval for invented outcomes.
Completed fictional case
Fictional stream: 20 issued unique alerts comprise 8 closed wins, 6 closed losses, 3 open, 2 cancelled without execution and 1 missing outcome. The inventory balances: 8 + 6 + 3 + 2 + 1 = 20. Closed-resolved win rate is 8/14 = 57.1429%; closed winners as a share of issued alerts are 8/20 = 40%. The second number is not a claim that every other alert lost. Three target notifications attached to one of the wins do not increase the unique alert count or produce three additional winning positions.
| Record | Represented finding | Boundary |
|---|---|---|
| Inventory | 20 = 8 + 6 + 3 + 2 + 1 | Every declared state retained |
| Closed rate | 8 / 14 = 57.1429% | Resolved closed population |
| Issued share | 8 / 20 = 40% | Other states are not all losses |
Primary references and inspection scope
Public-document inspection: 12 October 2026. The case arithmetic is original fiction, not a sourced provider outcome. No paid software, selected leader, broker route or subscriber account was tested.