Rating Methodology & Limitations

A Mintr score is a percentile rank, not an absolute grade. A stock scoring 82 on a given horizon isn't "82% good" in some fixed sense — it means that stock currently ranks better than roughly 82% of the covered universe on that horizon's weighted blend of factors, recalculated fresh on every scoring run. The universe is the S&P 500, plus Nasdaq-100 names not already in it, plus a curated set of other large- and mid-cap U.S. names ranked by market cap — roughly 530 U.S. stocks in total. See the live count on the site footer, since it drifts as either index reconstitutes and as the curated set is revisited.

Because it's a percentile, two things follow directly. First, a score is only meaningful relative to the same universe on the same day — it isn't comparable across scoring runs the way a price or a fixed 1-10 grade would be, and it says nothing about stocks outside the covered universe (there is no ETF or fund coverage yet). Second, in a broad market decline every score is still measured against that day's universe, so the top band is the least-bad band that day, not necessarily a strong one in absolute terms — the percentile basis is the reason a "Top-Tier" score can still mean "the best of a genuinely weak field."

Every stock gets three separate scores — Swing, Mid, and Long — because the same stock can be a reasonable short-term trade and a poor long-term hold, or the reverse, and collapsing that into one number would hide the disagreement. The three horizons don't share one formula; each is its own independently weighted blend of the same ten pillars, described in §3.

Every score is built from the same ten pillars. Each pillar is itself an average of several sub-factors (shown below); a pillar's own value is a 0–100 percentile too, computed the same way the final score is — see §4. Five pillars are price-based ("trading" factors, available for essentially every covered stock every day); five are fundamentals- or estimates-based ("investing" factors, which depend on a company's own reported financials and analyst coverage, and can be unavailable for a thinly-covered or recently-added name).

This is the part of the model that makes the three-horizon framing real rather than cosmetic: each horizon weights the same ten pillars differently, based on which factors actually predicted forward returns at that horizon in testing (see §4 for how "tested" is defined here). The exact numeric weighting is proprietary — it's the output of ongoing backtesting and is revised as that testing continues — but the relative emphasis at each horizon is:

Reading the Swing row: momentum, relative strength, and proximity to highs dominate, and valuation and quality carry little to no weight — a two-week trade doesn't hinge on whether a company is statistically cheap or its balance sheet is pristine. Reading the Long row, that inverts almost completely: valuation and growth carry the most weight, quality and balance-sheet strength follow, and short-term technical/momentum factors are reduced to a small residual. Mid sits between the two. Every pillar contributes something at every horizon — none is weighted to exactly zero — but the ordering above is what actually separates one horizon's rating from another's for the same stock.

Step 1 — raw factor values

Each of the ~40 sub-factors listed in §2 is computed from real price history or the company's own reported fundamentals — nothing here is estimated or simulated.

Step 2 — cross-sectional percentile, per sub-factor

Each sub-factor's raw values across the entire eligible universe are winsorized (the most extreme 1% at each tail is capped, so one outlier company can't distort the scale for everyone else) and then converted to a 0–100 percentile rank within that day's universe — this is the same percentile logic described in §1, just applied one factor at a time before the factors are ever combined.

Step 3 — pillar scores

A pillar's score is the plain average of whichever of its sub-factor percentiles are actually available for that stock that day. A pillar with some missing sub-factors still produces a score from what it has — the model doesn't zero out a whole pillar for one missing input, though (see §5) a stock missing enough data has that reflected in its Confidence figure.

Step 4 — the weighted composite

The horizon's pillar weights from §3 are applied to that stock's pillar scores. If a stock is missing an entire pillar (e.g., a recent IPO with no analyst coverage yet), that pillar's weight is redistributed proportionally across the pillars that do have data, rather than either penalizing the stock with a zero or silently changing what the weights are supposed to represent.

Step 5 — final calibration

The weighted composite is itself converted to a percentile rank across the universe (the same winsorize-free ranking as Step 2) and rescaled to a 1–99 range, with the single top-ranked stock on that horizon reserved for 100 — so the top of the list never shows a cluster of tied 100s, and a score of 100 means "the single best-ranked stock today," not "at the very top of a tied group."

The technical gate (Swing only)

A Swing score can only reach "Top-Tier" if that stock's own Technical pillar is itself strong — a stock with weak technicals is capped well below 100 on Swing regardless of how well it scores on everything else, on the reasoning that a 2–30 day call is inherently a price/technically-driven one, so a weak technical picture shouldn't be overridden by strength elsewhere. A stock capped this way shows a "Weak Technicals" flag.

Liquidity and reversal flags

A stock trading below the site's liquidity/volume eligibility threshold has its score capped and carries a "Low Liquidity" flag rather than being silently excluded. A stock whose 5-day return sits in the top 2% of the entire universe is flagged as a possible short-term reversal candidate — informational only, it doesn't change the score itself.

Confidence

The Confidence figure shown alongside a score is the share of that horizon's pillar weight (not just pillar count) that actually had real data behind it — a stock missing a heavily-weighted pillar for that horizon (Valuation, at Long) shows a materially lower Confidence than one missing a lightly-weighted one, even if both are missing "one pillar." Confidence describes data completeness, not a forecast of how the stock will perform.

A percentile score is relative, not absolute, and not stable over time in the way a price is. The same underlying business could show a different score on different days purely because the rest of the universe moved, even if nothing about that company changed. Comparing a score today against the same stock's score from months ago is comparing two different rankings against two different universes, not a like-for-like trend.

Ratings are generated from historical and delayed data. They describe where a stock currently ranks on measurable, backward-looking factors — they are not a prediction of future price movement and are not a guarantee of any outcome.

Coverage is limited. The universe is the S&P 500, Nasdaq-100 names not already in it, and a curated set of other large- and mid-cap U.S. names — roughly 530 U.S. large/mid-cap common stocks in total. There is no ETF, fund, small-cap, or foreign-listing coverage yet; a holding outside this universe simply won't have a Mintr rating.

Sentiment and Analyst pillars depend on third-party coverage that isn't uniform. A thinly-covered stock can show a lower Confidence on these pillars specifically, through no fault of the company itself.

This is not personalized investment advice. Mintr does not know your financial situation, goals, tax situation, or risk tolerance, and a rating is the same for every viewer regardless of what they hold.

All investing carries risk of loss, including loss of principal. A high score is not a guarantee against loss, and a low score is not a guarantee of decline.