How scores work

Filters come first. Stocks that fail your filters (sector, size, price, trading volume, or any metric with Filter turned on) are set aside before scoring. A filter without a value yet has no effect.

Every scored stock needs complete data. If a stock is missing a figure the model uses, it is left out rather than guessed, and the page says how many were left out and why.

Most scores are relative. All models except the F-Score compare stocks with each other, so the same stock can score differently in a different screen. At least 10 stocks are needed to score fairly, and fewer than 30 shows a warning.

Everything updates instantly. Scores are recalculated in your browser whenever you change the model, a filter, or a weight.

Above the results, four counts show how many stocks are left after each step, from left to right. Each step can only keep or remove stocks, so the numbers never go up.

1. 1

   Loaded

   Every stock in the current data: the largest US-listed companies by market cap (about 2,000), refreshed daily. Only common stocks are included, never ETFs, funds, bonds, preferred shares or warrants.

   Left out:

   • Before this count, a stock whose latest price is more than 10 days older than the rest is left out as likely delisted or halted.
2. 2

   In universe

   Loaded stocks that pass the Which stocks filters: sector, minimum market cap, minimum price and minimum traded per day. With none of these set, this equals the loaded count.

   Removed here by:

   • Being in a different sector, or below a minimum you set.
   • Having no figure for something you set a minimum on, such as a stock with no market cap when a minimum market cap is set.
3. 3

   Passed filters

   Stocks in the universe that also meet every metric with Filter turned on, such as P/E below 20. A filter switched on but without a value yet has no effect.

   Removed here by:

   • Failing any active metric filter.
   • Having no figure for a filtered metric: a company with no P/E (for example, one losing money) cannot pass “P/E below 20”.
4. 4

   Ranked

   Stocks the chosen scoring model can actually score. Each one gets a 0–100 score and a rank, and these are the stocks shown in the results.

   Removed here by:

   • Missing data for a scored metric, rather than guessing it. Banks have no gross margin, for example, and recent listings lack the price history for trend metrics.
   • Not being a company the model covers: the Magic Formula leaves out financial and utility companies, the F-Score leaves out financial companies and those without two years of statements, and the Income model only scores dividend payers.
   • Too few stocks: at least 10 must remain to rank fairly, or nothing is ranked.

The line under the counts says how many stocks passed the filters but were not ranked, and why, including how many are recent listings still building price history.

In the comparison view, ranks count only the stocks both models rank, so they can differ from each model's own list.

The default model. For every metric you turn on, the stocks are lined up from best to worst and given a score from 0 (worst) to 100 (best) based on their position, a percentile rank. For lower-is-better metrics such as P/E, the order is reversed. Equal values share the same score.

The overall score is the weighted average of the metric scores. Set each metric's weight from 1 to 10 with the − and + buttons: a metric with weight 3 counts three times as much as one with weight 1. Next to the buttons, the screen shows each metric's share of the score, its weight divided by the total of all the weights you have chosen, so you can see at a glance how much it matters.

Best for

Building your own recipe. The style presets (Balanced, Value, Quality, Growth, Momentum) all use this model, and Custom is your own mix: it lights up whenever your metrics and weights match none of the styles, and tapping it brings your mix back after trying a style.

Target metrics

RSI and payout ratio score 100 inside a healthy range (RSI 50–70; payout 25–65%) and lose points the further they stray, instead of being ranked.

Good to know

Position is all that counts: a stock that is far cheaper than the rest scores about the same as one that is only slightly cheaper.

Uses the same metrics and weights you choose, but measures how far each stock is from the average, in standard deviations (a z-score), instead of just its position. A stock twice as far ahead earns twice the credit.

Each metric's z-score is capped at ±3 so one extreme value cannot dominate, flipped for lower-is-better metrics, and combined with your weights. The combined result is then shown on the familiar 0–100 scale using the normal distribution: 50 is average, about 84 is one standard deviation better than average, and about 98 is two.

Best for

Rewarding genuine standouts, such as a company with far higher returns on capital than its peers.

Good to know

Scores bunch near 50 when stocks are similar, and a single strong metric can lift a stock more than under weighted ranking.

Scores each metric 0–100 like weighted ranking, then groups them into areas (valuation, profitability, growth, financial health, dividends, price trend) and averages within each area. The areas are combined with a weighted geometric mean rather than a plain average.

The effect: a weak area drags the score down much more. A stock scoring 90 on value and 30 on quality averages 60, but scores about 52 here. A stock scoring 60 on both still scores 60.

Best for

Avoiding lopsided picks, such as a very cheap stock with poor quality.

Good to know

It only differs from weighted ranking when your metrics span more than one area. The "+ pts" in a stock's breakdown show each metric's share of its score.

Joel Greenblatt's formula from The Little Book That Beats the Market. It looks for good businesses at bargain prices using just two measures:

• Earnings yield: operating income ÷ enterprise value. How much the business earns for its full price, including debt.
• Return on capital: how much profit the business makes on the money invested in it.

Stocks are ranked on each measure and the two ranks are added, which is the same as averaging the two percentile scores. The best combined rank scores highest.

Best for

A simple, rules-based value strategy. Greenblatt suggests holding 20–30 top-ranked stocks for about a year.

Left out

Financial companies and utilities, as in the original formula.

Good to know

Return on capital here is FMP's after-tax return on invested capital, a close cousin of Greenblatt's pre-tax version. Your metric Score switches are not used, but filters still apply.

Professor Joseph Piotroski's nine yes-or-no tests comparing a company's last two fiscal years. Each test passed scores a point, and the score out of 100 is the share of tests passed (7 of 9 is 78). Unlike the other models, it is absolute: a stock's score does not depend on the other stocks.

1. Profitable: net income was positive last year.
2. Cash-generative: operating cash flow was positive.
3. Return on assets improved: net income ÷ total assets rose from the year before.
4. Cash beats earnings: operating cash flow exceeded net income (high-quality earnings).
5. Less leverage: long-term debt ÷ total assets did not rise.
6. Better liquidity: current ratio (current assets ÷ current liabilities) rose.
7. No dilution: diluted share count did not grow.
8. Gross margin improved: gross profit ÷ revenue rose.
9. Asset turnover improved: revenue ÷ total assets rose.

Best for

Checking financial strength and improving fundamentals. Piotroski designed it to separate winners from losers among cheap stocks, so try it with a valuation filter such as P/E below 15.

Left out

Financial companies, whose balance sheets these tests do not fit, and companies without two years of statements.

Good to know

Many stocks share the same score; ties are listed alphabetically. The sector switch does not apply.

Ranks dividend-paying stocks on four measures, each scored 0–100 and then weighted:

• Dividend yield (weight 3): yearly dividends as a share of the price.
• Dividend growth (weight 2): yearly growth in dividends per share over three years.
• Payout ratio (weight 2): share of profit paid out. Scores 100 between 25% and 65%, and 1.5 points less for every percentage point outside that range, so both stingy and unsustainable payouts lose points.
• Dividend cover by free cash flow (weight 2): how many times spare cash covers the dividend.

Best for

Investors who want income that is generous today and likely to keep growing.

Left out

Companies that pay no dividend.

Good to know

Banks' cash flows swing widely, so their dividend cover can look poor even when the dividend is safe.

Normally every stock is ranked against all the others. Valuation and profitability differ a lot between industries, though: utilities nearly always look cheaper than software companies, and banks have no gross margin at all. Turning on Compare within each sector ranks each stock only against stocks in its own sector, so the top scores go to the best companies in each sector.

Works with

Weighted ranking, Z-score composite, Balanced, Magic Formula and Income. Not the F-Score, whose tests are pass/fail.

Small sectors

A sector with fewer than 5 scored stocks is ranked against the whole list instead, because ranking within a tiny group is unreliable.

Choose Compare with another model under the model picker. Both models score the same filtered stocks, and the results switch to a side-by-side view. Ranks in this view count only the stocks that both models score, so they can differ from each model's own list.

Agreement

How similarly the two models order the shared stocks (a rank correlation). 100% means identical order, 0% means unrelated, and negative means they tend to rank stocks the opposite way.

Top 20 overlap

How many stocks appear in both models' top 20.

Scatter chart

Each dot is a stock, with the first model's score across and the second's up. Dots near the dashed diagonal get similar scores from both; the top-right corner is where both models like a stock.

Groups

Both like it: top fifth under both. Model A prefers: top fifth under the first model but bottom half under the second (and the reverse). Biggest disagreements: the largest rank changes. Each row shows the rank change; ▲ means the second model ranks it higher.

Two classic accounting models flag possible trouble. A stock with a flag shows a warning triangle next to its ticker, and the stock's detail panel explains it. Both are also available as metrics you can score or filter on.

• Altman Z-Score estimates bankruptcy risk from working capital, retained earnings, operating income, sales and market value relative to assets and debts. Above 2.99 is the safe zone, 1.81–2.99 is a grey zone, and below 1.81 is the distress zone, which triggers a flag.
• Beneish M-Score looks for patterns common in manipulated accounts, such as receivables growing faster than sales or profits outpacing cash flow. A score above -1.78 triggers an earnings-quality flag. It is a reason to look closer, not proof of wrongdoing.

Left out

Financial companies, which these models were not built for. The Altman Z-Score also leaves out utilities and real-estate companies, whose heavy borrowing is normal for regulated, asset-backed businesses.

Financial Modeling Prep supplies trailing twelve-month ratios and key metrics, annual income, cash flow and balance-sheet statements, and end-of-day prices. Sectors use FMP's classification.

EBITDA and earnings yield use operating income over the last four quarters, so one-off investment gains or write-downs do not distort them. Companies that report semi-annually or in another currency use FMP's figures. Growth figures are capped at ±1000%, because growth from a near-zero base produces meaningless multiples; capped values show as “1000%+”. Ratios that rely on negative earnings, EBITDA or equity are left blank rather than ranked, and banks and insurers are left out of ratios built on EBITDA, debt or gross profit.

Only stocks are included: bonds and notes that trade on stock exchanges are screened out, and each company appears once. Stocks whose latest price is more than 10 days behind the rest are excluded as likely delisted or halted. Reported financial data can be amended or restated. Scores are for research, not investment advice.