Leaderboard numbers look simple; definitions are not. This guide explains how we score stated predictions, what dashboard backtests summarize, and which mistakes make any hit-rate statistic misleading.
Before scoring, translate the claim: asset, direction, start date, end date, and benchmark (absolute, sector ETF, or index). If any piece is missing, the score is undefined—not secretly bullish or bearish.
A horizon is the number of trading days from anchor date T to evaluation date. Changing the horizon after you see the outcome is cherry-picking. InsightMeter stores the horizon used for each scored call.
A directional hit might mean “stock return > 0.” A relative hit might mean “stock return minus sector ETF > 0.” The same forecaster can look brilliant on one definition and mediocre on another. Always read which definition powered the number you see.
If delisted tickers drop out of track records, past accuracy rises artificially. Our methodology aims to retain unresolved or delisted outcomes according to published rules rather than silently deleting failures.
Backtests on the dashboard let you filter date ranges and signal sources to see aggregate metrics under those filters. They are descriptive summaries—useful for orientation, not proof that a strategy would have been profitable after costs, taxes, or slippage.
Free tiers may show top-line scores or limited history; Pro unlocks deeper tables and exports. The scoring rules in this guide apply regardless of tier—only visibility changes.
Trust a number only after you can repeat its definition in one sentence. Link to our forecast track-record guide for reader workflows and to methodology for data handling.