No dataset yet

The page renders from web/data/data.json, which the pipeline builds from SEC EDGAR filings and Yahoo Finance prices. Nothing has been generated yet.

Start a first build with the Update data button in the sidebar, or from a terminal:

cd "C:\Local Disk\investment\stock-screener"
set SEC_USER_AGENT=Your Name your@email.com
python -m pipeline.cli update --universe sec --limit 300

The first run is the slow one. After that every refresh is incremental — unchanged filings answer with 304 Not Modified and prices only fetch the bars since the last stored date.

My account

Sign in to build and track a portfolio. Balances are hypothetical — no real money and no brokerage is involved.

Build a portfolio

Answer the risk questionnaire, review the proposal and its reasoning, then fund it with a hypothetical balance and track it.

Live trading

Real orders on Binance Spot, signed with credentials held in this machine's environment. Administrator only.

Everyone's Favourite

undervalued on a majority of the methods valid for that business

Top value candidates

Top growth candidates

Valuation labels across the universe

which strategist's method each company satisfies

Classification

Graham score distribution

criteria passed, 0–7

Dataset coverage

what the system currently holds

Industry coverage

by SIC description

Strategy indexes vs S&P 500

top N per approach, optimiser-weighted, benchmarked against SPY

Stocks per index Weighting Min market cap ($m)

Binance-tradable only

Historical backtest — pick a start year

selects each approach using only fundamentals filed by that date, then measures forward with no look-ahead

Start year As-of date End date Stocks per portfolio Weighting Min market cap ($m)

Binance-tradable only assesses the full universe — first run for a new date takes about two minutes, then it is cached

Valuation screener

every method is only applied where it is valid for that business

Valuation methods — pick the strategists you want

ANY finds names any strategist likes; ALL demands agreement.

Graham's defensive criteria — tick only what you want enforced

Filters

Minimum rating (0–100) How many results

exclude banks & insurers exclude unstable earnings Binance-tradable only

Earnings opportunities

reporting soon or just reported · scored with the reasoning shown

Days ahead Days back (drift window) How many Models to run

Binance-tradable only

Automatic shortlist

ranks the universe on criteria you set, and shows why each name appears

Profile How many Min market cap ($m) Review my holdings instead (comma separated tickers)

Binance-tradable only

Reporting calendar

Most important upcoming reports

ranked by market capitalisation and analyst coverage

Model estimates vs consensus

walk-forward models run on the largest upcoming reports

Each name refits its models, so this takes a few seconds.

Earnings forecast

walk-forward validated · every model scored against a seasonal-naive benchmark

Symbol

Fits on every quarter strictly before the one being predicted. No random splits.

Strategy backtest

signals execute at the next bar's open · costs charged both sides

Symbol Strategy Interval Capital Risk per trade Stop (ATRs, 0 = none) Commission bps Spread bps Slippage bps

walk-forward (out-of-sample) apply Binance order rules

Portfolio construction

models you parameterise · outputs are calculations, not orders

Symbols (comma separated) Method Lookback (bars) Max weight Risk-free Shrinkage

Current holdings (optional)

SYMBOL:QUANTITY:PRICE, comma separated — produces rebalance trades

Holdings Cash No-trade band

Screen filters

These control what the screener shows. The canonical Graham score and the Value / Growth labels are computed by the pipeline — to change those, set the matching environment variables from pipeline/config.py and re-run python -m pipeline.cli export.

How each number is produced

Data sources

Universe — Binance Stocks (--universe stocks) reads Binance's own US-equity symbol list from GET /sapi/v1/equity/market/exchangeInfo, giving the authoritative set of stocks and ETFs your account can trade, along with each symbol's tradability, fractional-share support, overnight eligibility, minimum order size and bStock code. That endpoint is MARKET_DATA: it needs your API key in the X-MBX-APIKEY header but is not signed, so no API secret is involved.
Other universes — --universe sec screens all ~8,000 SEC filers with no credentials at all; --universe file takes your own CSV.
A caution about the crypto endpoint — the general /api/v3/exchangeInfo is a crypto spot endpoint and must not be used to find equities. Its base assets collide with US tickers constantly: ENS is Ethereum Name Service not EnerSys, CVX is Convex Finance not Chevron, BCH is Bitcoin Cash not Bank of Chile. That legacy mode survives only as --universe crypto-spot and should not be used for screening.
ADRs and ETFs — some Binance-tradable symbols have no SEC 10-K filer behind them (foreign ADRs, most ETFs). They are tradable but cannot be screened on fundamentals, so they are reported and skipped rather than shown with empty ratios.
Fundamentals — SEC EDGAR XBRL companyfacts, annual (10-K) figures only, latest amendment wins per fiscal year.
Prices — Yahoo Finance daily adjusted closes, plus split events.
Split adjustment — adjusted closes are split-adjusted but a 10-K's EPS is not, so filed per-share figures are divided by every split that post-dates the fiscal year end. Without it Booking Holdings' 25:1 split showed a P/E of 1.0.

Graham's seven defensive criteria

Adequate size — sales ≥ $100m (industrials) or assets ≥ $50m (utilities).
Strong financial condition — current ratio ≥ 2 and long-term debt ≤ working capital; for utilities, debt ≤ 2× book equity.
Earnings stability — positive EPS in each of the last ten years.
Dividend record — an uninterrupted dividend for twenty years.
Earnings growth — at least one-third growth in per-share earnings over ten years, measured on three-year averages at both ends.
Moderate P/E — price ÷ average earnings of the past three years ≤ 15.
Moderate price to assets — P/B ≤ 1.5, or P/E × P/B ≤ 22.5.

Criteria 3, 4 and 5 need a decade of filings. EDGAR's XBRL archive begins around 2009, so a company can fail them for want of history rather than for want of quality — the detail view shows how many years were actually available.

Other valuations

PEG = current P/E ÷ EPS growth in percentage points.
DCF — two-stage free cash flow (operating cash flow − capex), five high-growth years at the historical FCF CAGR capped at 15%, then a 2.5% terminal growth, discounted at 10%, net of debt, per diluted share.
Graham formula value — V = EPS × (8.5 + 2g) × 4.4 ÷ Y, with EPS the three-year average, g capped at 15 and Y a AAA bond yield assumption.
Industry average — trimmed mean across companies sharing a two-digit SIC group, so one outlier cannot swamp the column.

Known limits

Ratios come straight from XBRL tags. Tagging varies between filers; financials and REITs in particular will show gaps where a tag does not exist.
Dividend history is derived from CommonStockDividendsPerShare* tags, which start well after 20 years ago for most filers — criterion 4 is therefore strict.
Prices are adjusted closes; intraday and corporate-action edge cases are not modelled.

This is a screening and analysis tool, not investment advice. Every figure is a mechanical calculation from public filings and should be verified against the original 10-K before it informs a decision.