Tapetide Score · formula v2.0.0

One number, and every step behind it.

The Tapetide Score rates every liquid Indian stock from 0 to 100 using six weighted pillars and a governance check, measured against a frozen, sector-relative reference. It is a fixed model over public data, so the same inputs always give the same score, and the whole method is open source.

Read the full method on GitHub See the ranking How scores changed each month

Also as Markdown and llms-full.txt

How a score is built

Five steps, each feeding the next. The order matters: normalization happens before weighting, and the governance check happens after, so a red flag can never be averaged away by good ratios.

  1. 1

    Who gets scored

    Eligible universe: actively traded NSE/BSE stocks with sufficient price history.

  2. 2

    Every metric on one scale

    Each raw metric is winsorized then normalized against a versioned, sector-relative reference distribution.

  3. 3

    Six pillars, weighted

    Pillar sub-scores combine metrics evenly; the composite is a weighted sum across pillars.

  4. 4

    Red flags pull down

    A governance overlay can reduce the score and hard-cap severe red flags.

  5. 5

    Held steady

    Composite is smoothed and held in a small dead band to avoid day-to-day whipsaw.

Six pillars, weighted by evidence

Each pillar is the plain average of its metrics. The composite is the weighted sum below, renormalized over whichever pillars a stock has. The widths are the shipped weights, not approximations, and the colours are the ones the ranking uses for the same pillars.

Weights sum to 100

Quality 22
Return on capital and equity, margins, cash conversion. Positive on every out-of-sample test date.
Valuation 24
Earnings yield, P/E, P/B, EV/EBITDA, free-cash-flow and dividend yield against sector peers. Cheap plus fragile is treated as a trap, not a bargain.
Growth 9
Trailing-twelve-month sales growth only. Three-year growth rates predicted lower forward returns in testing, so they are no longer scored.
Financial health 9
Leverage, interest coverage and a distress score. Kept as a risk control; it did not predict returns on its own.
Momentum 12
Six-month return and distance from the 200-day average, on corporate-action-adjusted prices.
Ownership 24
Direction of promoter, FII and DII holding changes. The strongest pillar out of sample, which is why it carries the weight it does.

Governance overlay

Multiplies the composite. Never adds.

Exchange surveillance flags, credit-rating downgrades, promoter pledges, heavy insider selling and audit qualifications each subtract from a multiplier that is floored at 0.5, so the overlay can at most halve a score. The most severe flags also hard-cap it: a stock under graded surveillance, or with more than three quarters of its promoter holding pledged, cannot score above 40 no matter how good its ratios are.

Reading the score

Scores concentrate in the middle of the 0-100 range: the observed all-time range is 12 to 76 and nothing has ever scored 80+. A 1-3 point gap between two stocks is inside the noise, so bands are the intended unit of comparison rather than exact rank.

37
45
55
66 0 100

Outlined: every score ever published falls between 12 and 76.Ticks: band cutoffs.

Strong 66 and up 3%
broad-based strengths and clean governance
Good 55–65 29%
above average, with real strengths across pillars
Average 45–54 42%
roughly typical for its sector and size
Weak 37–44 19%
meaningful caveats worth checking
Poor below 37 8%
multiple weak pillars and/or active red flags

What the score is

  • Data-driven, deterministic, reproducible — no LLM opinions.
  • Same inputs always produce the same score (byte-stable, auditable).
  • Standardized within a reference version: two stocks scored against the same frozen reference are directly comparable, and re-running the same inputs reproduces the same number exactly.
  • Sector-relative: a bank and an FMCG company are scored fairly against their own sector's history, not against each other directly.
  • Red flags can pull the score down, never average up into a positive number.

What it is not

  • Not investment advice. Decision-support only.
  • A high score is not a forecast — it summarizes past + current evidence on traded liquidity, fundamentals, ownership and governance.
  • Fundamental inputs are the latest reported figures, not the figures as they stood on a past date. A historical score therefore reflects later restatements, so the score history is best read as a trend rather than as a replayable point-in-time backtest.
  • Scores are comparable within a reference version. The reference is frozen and versioned, but it is calibrated from one day's cross-section, so comparisons across different reference versions are approximate.
  • The governance overlay is built from published regulatory and exchange disclosures. Those records begin partway through our history, so the penalties and caps are grounded in disclosed facts but have not yet been validated against a full market cycle.
  • Eligibility requires enough traded days and enough populated metrics. Thinly traded stocks are deliberately left unscored rather than scored on sparse data.

When it changes

Cadence
Once per trading day.
Timing
Recomputed each evening after that day's market data and fundamentals are consolidated, typically around 19:45 IST.
Coverage
Roughly 4,500 stocks are scored per run — every actively traded, sufficiently liquid NSE/BSE stock that clears the eligibility gate.
Score date
A score is stamped with the latest completed TRADING date, so weekends and market holidays reuse the previous session's score rather than inventing one.
Stability
Scores are smoothed and held within a small dead band, so a typical day moves the average stock by well under a point. Large single-day jumps indicate a real change in the inputs or a new red flag, not noise.
If a run fails
A run that sees materially less than the recent norm is refused rather than published, and rankings fall back to the last complete day. A truncated run can never be served as a full picture.

Questions people ask

What is the Tapetide Score?
A 0 to 100 rating for Indian listed stocks, computed by a fixed mathematical model from published financial data, prices, ownership filings and exchange disclosures. Six weighted pillars add up; a governance overlay can only subtract. There is no analyst opinion and no language model in the calculation.
Why is Ownership weighted as much as Valuation?
Because it earned it. In the out-of-sample backtest behind the current formula, the direction of promoter and institutional holding changes was the strongest of the six pillars and positive on every test date. The weights are chosen from that evidence, not from intuition, and the full table is in the open-source repository.
Why is Growth only 9 points?
Three-year sales and profit growth rates predicted lower forward returns in testing, a well-known mean-reversion effect, so they are no longer scored. Growth now rests on trailing-twelve-month sales growth alone, and a pillar carried by one metric is kept at a low weight on purpose.
Does a high score mean I should buy?
No. The score summarizes past and current evidence; it is not a price forecast and not investment advice. Tapetide is not a SEBI-registered research analyst or investment adviser.
Why do scores cluster between 40 and 65?
The composite is a weighted mean of pillars that are themselves means of normalized metrics, so a very high number needs a company to be exceptional on almost everything at once. Across every run to date the range is 12 to 76 and nothing has scored 80. The band cutoffs are set on that real distribution, which is why 66 is Strong.
How is a bank compared with a manufacturer?
Each is scored against a frozen reference built from its own sector and size group, using a metric template suited to how the business works. Banks and NBFCs are not scored on operating margin, EV/EBITDA or cash conversion, and their coverage is measured against the metrics that apply to them.
Why does a stock show “Insufficient data”?
A score is published only when at least four of the six pillars can be computed and at least 60% of the applicable metrics are present. Thinly traded stocks fail the liquidity gate on purpose: price ratios built on a stale last trade are worse than no number.
How often is it updated?
Once per trading day. Recomputed each evening after that day's market data and fundamentals are consolidated, typically around 19:45 IST. A score is stamped with the latest completed TRADING date, so weekends and market holidays reuse the previous session's score rather than inventing one. Roughly 4,500 stocks are scored per run — every actively traded, sufficiently liquid NSE/BSE stock that clears the eligibility gate.
Where is the full calculation?
Every weight, metric definition, normalization constant, eligibility rule and governance threshold is published at https://github.com/Tapetide-hq/tapetide-score, together with the validation method, the factors that were tested and rejected, and a changelog of every formula version. Corrections and proposals are welcome as issues or pull requests.

Tapetide is not a SEBI-registered research analyst or investment adviser. Consult a qualified advisor before investing.