How our composite score works — and what it deliberately doesn't claim

Every screener result on getcoinsight carries a signed score from −100 to +100. It folds seven analytical pillars — technical consensus, momentum, candlestick and chart patterns, market structure, derivatives positioning, on-chain flows, and market sentiment — into one number with a per-factor breakdown you can inspect on hover. This post explains exactly how that number is built, because a score you can't audit is a score you shouldn't trust.

The mechanics

Each pillar produces one or more signed factors in [−1, 1]. A pillar's score is the mean of its factors, and the final composite is a weighted average of pillar scores. The weights depend on the profile you pick: a scalp profile leans on derivatives, an invest profile leans on on-chain valuation, and the stock profile swaps crypto pillars for P/E, P/B and dividend factors.

One consequence worth being honest about: because a pillar averages its factors, a second agreeing signal inside the same pillar can dilute it rather than amplify it. That damping is deliberate — it keeps the score bounded and comparable — but it means "more green checkmarks" does not mechanically mean "higher score".

The coverage badge

Not every coin has every data source: a token without a futures market has no derivatives pillar, and most long-tail assets have no on-chain metrics. When a pillar is missing, its weight is redistributed across the pillars that do have data. That keeps the scale consistent, but it also means two coins can show the same +62 while measuring very different things. This is why every score in the screener now shows a coverage badge like 4/7 — four of the seven weighted pillars had data. Treat low-coverage scores as a narrower claim, not a weaker signal.

What it is not

We ran a walk-forward calibration over 40 coins and four years of daily data (~31k observations) to test whether the composite predicts forward returns. The honest answer: at 3–30 day horizons, it is a state descriptor, not an alpha signal — daily crypto mean-reverts, and leaning on momentum actually hurt rank correlation out of sample. That's why the UI labels scores as "bullish confluence", never "buy". Use the score to find setups worth your attention and to see why a coin screens the way it does — then do your own work on the shortlist.