IBD RS Rating

Percentile-ranked relative strength ratings (1–99) for ~4,600 US stocks, recalculated every trading day.

Quick start · Documentation · API reference · How it works

What this is

Relative Strength (RS) Rating answers one question: over the past year, did this stock outperform more of the market than that one? A stock rated 90 outpaced 90% of the roughly 4,600 stocks it was measured against. A stock rated 50 was average. It is the metric William O'Neil built a strategy around, and it is the first filter most momentum investors apply.

Investor's Business Daily publishes RS Ratings behind a paid subscription. Open-source alternatives usually stop halfway — they compute a weighted return and call it a rating, skipping the step that gives the number meaning. A weighted return tells you a stock rose 18%. A rating tells you that 18% put it in the top 3% of the market. Only the second one is comparable across stocks, across sectors, and across time.

This project does the full computation: it collects daily closes for the whole US common-stock universe, computes each stock's weighted momentum score, and then ranks every score against every other score on that same trading day to produce a 1–99 rating.

from rs_rating import RS

rs = RS()
rs.get("NVDA")
# {'ticker': 'NVDA', 'date': '2026-03-19', 'close': 121.4, 'rs_raw': 0.1666, 'rs_rating': 70}

No account, no API key, no rate limit. The reading client is pure Python standard library — installing it pulls in nothing else.

Highlights

• True percentile ranking. Every rating is a stock's position within the full universe on that date, not a rescaled return.
• Zero-dependency client. rs_rating uses only urllib and json, so it drops into any environment without dependency conflicts.
• Sector and industry analysis. Rank sectors by average RS, or find the strongest names inside one industry — O'Neil's research attributes roughly half of a stock's move to its industry group.
• Honest gaps. A stock with under 252 trading days of history gets no rating rather than a rating built on thin data, and a trading day whose coverage falls below 90% of the universe is left unrated rather than published with a distorted denominator.
• Self-hostable. The full calculation engine ships in the same repository. Point it at SQLite for a laptop or any Postgres for production.

Installation

pip install ibd-rs-rating

Requires Python 3.11 or newer.

Quick start

from rs_rating import RS

rs = RS()

# One stock, latest rating
rs.get("AAPL")

# The strongest names in the market right now
rs.top(10)

# Everything in the top decile
rs.filter(min_rating=90)

# Head-to-head
rs.compare(["NVDA", "AMD", "AVGO", "INTC"])

# Momentum that is accelerating: biggest rating gains over 5 trading days
rs.movers(days=5, n=10)

# Which sectors are leading?
rs.sector_ranking()

Every call returns plain dicts and lists — no custom types to learn, and the output drops straight into pandas.DataFrame(...) if you want it there.

The first call fetches a short-lived anonymous access token and caches it for the session, so there is nothing to configure. See Getting Started for a full walkthrough and API Reference for all 15 methods.

How a rating is built

RS Raw = 0.4 × ROC(63) + 0.2 × ROC(126) + 0.2 × ROC(189) + 0.2 × ROC(252)

ROC(n) is the return over that stock's last n valid trading days. The most recent quarter carries five times the weight of the oldest, which is what makes the score respond to accelerating momentum rather than to a rally that ended nine months ago.

That raw score is then percentile-ranked against every other stock rated on the same date and scaled to 1–99. The ranking step is what turns a private number into a comparable one. Concepts explains each term; Architecture explains how the pipeline produces them.

Data universe

Roughly 4,600 US-listed common stocks (NYSE, NASDAQ, AMEX) with a market cap above $50M, excluding ETFs and shell companies, including ADRs. SPY and QQQ are tracked and ranked alongside individual stocks so you can see where the index itself falls in the distribution.

A snapshot of the latest ratings is committed to data/tickers.csv.

Self-hosting

The engine that produces the data is in the same repository, and it runs on your own database:

git clone https://github.com/tjdwls101010/IBD-RS-Rating.git
cd IBD-RS-Rating
pip install -e ".[engine,pg]"

python -m ibd_rs init      # download 2y of history and compute RS (20-30 min)
python -m ibd_rs update    # daily incremental update (~3 min)
python -m ibd_rs top 20    # inspect results

Without DATABASE_URL set, everything runs against a local SQLite file. Set it to any Postgres connection string to use that instead. Operations covers running it as a scheduled job.

Documentation

Full documentation lives in docs/wiki/:

Project status

Beta, maintained by one person. The public data pipeline runs automatically on weekdays and the client API is stable — the last breaking change was the 0.4.0 backend migration, recorded in CHANGELOG.md. Treat the hosted endpoint as best-effort: if you depend on this data operationally, self-host.

Contributing

Issues and pull requests are welcome — see CONTRIBUTING.md for setup and the test commands. To report a security issue, follow SECURITY.md instead of opening a public issue.

Disclaimer

Not affiliated with Investor's Business Daily or William O'Neil + Co. RS Ratings here are a reverse-engineered approximation of IBD's published methodology; the official formula and universe are proprietary. For official ratings, use IBD MarketSmith.

This is a research and educational tool. It is not financial advice, and nothing it outputs is a recommendation to buy or sell any security.

License

MIT — see LICENSE.