Skip to content
stocks-llm

Percentile rank

Where one company stands in a field, expressed from 0 to 100 rather than in the metric’s own units — the device that lets a margin, a growth rate and a valuation multiple be averaged into one score without any of them dominating.

A percentile rank answers "how many of the others does this one beat", not "by how much". A company at the 90th percentile on operating margin has a wider margin than about nine tenths of the companies it was measured against; the number says nothing about whether that margin is 12% or 60%. This is the whole reason the Top-10 scores use percentiles instead of the raw figures: a margin is a fraction, a market capitalisation is in the trillions, a growth rate can be negative, and averaging those directly would let whichever metric happened to have the largest numbers decide the result.

Two details change the answers enough to be worth stating. Direction is declared per input, because for some of them lower is better — a low price-to-earnings ratio earns a high valuation percentile, not a low one. And the inputs are winsorised at the first and ninety-ninth percentiles before ranking, which means an extreme outlier is pulled back to the edge of the field rather than allowed to define it; without that, one company with a distorted figure could compress everyone else into a narrow band and effectively own the pillar.

The field matters as much as the number, and this is where percentiles are most often misread. On a market-wide list a company is ranked against every eligible company; on a sector list it is ranked against its own sector only, so the same company legitimately carries two different percentiles on the same day and neither is wrong. That is the point of running sector lists at all: a utility compared against software companies on growth tells you almost nothing, while a utility compared against utilities tells you where it stands among its actual peers.

Ask about this →

How the Top-10 scores are built →

Related terms

  • The Long-term Score — A 0–100 number that is the weighted average of a company’s percentile ranks on six pillars — profitability 20%, growth 20%, financial health 15%, valuation 15%, shareholder return 10%, long-term trend 20% — and the thing the Long-term 10 is sorted by. It is a published formula and its output, not a recommendation.
  • The Short-term Score — A 0–100 weighted average of five technical pillars — trend 25%, momentum 20%, setup 20%, volume and participation 15%, catalysts 20% — describing where a share stands on the tape today. It describes past price behaviour; it does not forecast the next move.
  • Rebalance — The scheduled moment when a list’s membership is allowed to change — rebalanced on the first trading day of each month for the Long-term 10, rebalanced every Monday for the Short-term 10. Scores move every night; membership only moves on the calendar.
  • Hysteresis (why a list does not churn) — The rule that a current member keeps its place while its rank stays within 15 even though only 10 are published — a deliberate gap that stops the same two companies swapping places every period on differences too small to mean anything.
  • Equal weight — Treating every holding in a list as the same size, so a company’s influence on the result comes from its own move and not from how large the company is — the basis on which the Top-10 lists are tracked against the market.

See more terms in the stocks-llm glossary.

Informational only — NOT financial advice. This is an educational definition, not a recommendation to buy or sell anything. Metrics on stocks-llm are delayed data and may be missing or stale. Always verify information independently and consult a qualified financial professional before making any investment decision.

Research stocks in your own words.

Informational only — NOT financial advice.

ScreenersBlogPrivacyiOS app