stockproof · 有据

An LLM reads stock news, earnings reports and company announcements for you, with every point linked to its source. A research toolkit then checks each signal against years of prices, and the app tells you what that check found. US stocks, China A-shares and Hong Kong. English and 中文.

· 中文说明

Try it offline in 30 seconds (no API key, no time limit). The demo opens the full dashboard on bundled real data; only buttons that need the internet or an AI model are turned off:

pip install stockproof
stockproof demo

No GitHub account or git needed. In mainland China, a mirror is faster: pip install stockproof -i https://pypi.tuna.tsinghua.edu.cn/simple

Liked it? The 10-minute tour takes you from the demo to your own stocks, a backtest and a portfolio check-up.

What it does

Daily stock watching

	Command	What you get
News mood	scan	Headlines for each stock, scored by an LLM for relevance and sentiment. Stories about other companies and repeats of the same event are dropped.
Earnings	earnings	The company's own release (SEC 8-K/6-K, Eastmoney for China) read by the LLM: tone, guidance, key points and risks, plus the earnings surprise (SUE) from point-in-time EPS.
Announcements	dashboard	A-share and HK filings are classified by importance (insider selling, regulator letters, buybacks, big contracts and more), and the important ones are summarized.
Why did it move?	dashboard	For an unusual move, the LLM explains it using only the news, filings and reports published around that day, cites each one, and says so when nothing explains the move.
What changed?	dashboard	The latest two reports compared: results, outlook, new risks, business and payouts.
Daily digest · calendar · schedule	daily, calendar, schedule	A daily summary of your watchlist, upcoming reports, and an automatic run every evening (macOS, Windows, Linux), with a desktop notification.

Screen and scorecards

	Command	What you get
Whole-market screen	screen	Every A-share ranked by earnings surprise, growth, ROE and trend, with at most 3 per industry and plain-language reasons. Any stock's rank can be looked up.
Evidence scorecard	scorecard	0–100 per dimension, with cited evidence, positives and risks. Each dimension shows whether history tests found it useful, and only tested dimensions count toward the total.

Research and the quant lab

	Command	What you get
Research	research	Event studies and cost-aware backtests: does any of this predict returns?
Policy signals	policy	新闻联播 since 2016, read for industry policy: stance and strength on the wording ladder (研究 → 稳步推进 → 加快 → 大力支持).
Quant lab	lab	Backtest any idea described as a small spec (earnings, 业绩预告, policy signals, factors) on point-in-time history, with an illustrated report. The last 2 years are locked until you confirm once, and every variant you try raises the bar for significance.

Your portfolio, and AI agents

	Command	What you get
Portfolio check-up	portfolio	Your holdings on real price history: concentration, holdings that secretly move together, where the risk comes from, and replays of the 2015 / 2018 / 2020 / 2021–22 / 2024 crashes on today's weights.
AI agents	mcp	23 MCP tools (DeepSeek Harness plugin in dsh-plugin/; Claude and Cursor work too). Every feature above can be asked in plain words ("what can you do?" lists them), and the agent can do research it can't do alone: run backtests on real data, screen 5,000 stocks, check your portfolio.

Everything runs in one local web page (stockproof dashboard). The Home tab shows what you can do, with clickable examples; the Lab tab backtests ideas from a form, no JSON needed; the Portfolio tab checks up your holdings and remembers them for next time. The first time you use stockproof in DeepSeek Harness, the agent lists everything it can do and suggests trying the dashboard; after that, ask "what can you do?", type /stockproof for ready-made requests, or say "open the dashboard". Say "open the dashboard's lab tab" to jump to a page. Running stockproof with no arguments prints the same grouped menu, and in DeepSeek Harness typing / lists ready-made prompts (check up a portfolio, backtest an idea, why did it move…).

What an LLM can't do alone — and can with stockproof

Ask DeepSeek Harness "A 股业绩大超预期的公司，财报公布后第二天买入持有 20 天，扣费后能跑赢吗？". The model turns the question into a spec, the lab runs it on ten years of point-in-time A-share data, and a report opens in your browser. The answer was no: −3.6% a year (t = −2.0), positive in 3 of 4 years before 2020 and negative every year from 2020 to 2024. Ask for a variant and the agent reports that the significance bar went up with the number of tries.

In DeepSeek Harness, type /stockproof for ten ready-made requests:

What the research found

Run on the author's data (2016–2026). Each result is a click away in the Research tab, with t-statistics and caveats.

Question	Answer
Do prices react to earnings surprises?	Yes, on the day. Good minus bad SUE: US large caps +1.7% (t = 5.4), CSI 300 +2.0% (t = 17.4). Dated by the 10-Q or legal deadline instead, the reaction is gone, so the real date matters.
Does the LLM's reading add anything?	Yes, for the reaction. On 472 US releases, tone explains the day's move where SUE does not (tone t = 3.1, SUE t = −0.5; R² 0.004 → 0.035). A-shares: negative tone −1.3% (t = −4.2).
…and does it predict what happens next?	No. Tone does not predict days 2–61 (t = 0.6).
Post-earnings drift?	Gone in US large caps since 2021. Still present in A-shares (+2.2%, t = 2.5), but long-only it earns +0.3%/yr after costs (t = 0.2): the drift sits on the short side, which A-share investors can hardly trade.
Does a fundamentals screen beat the market?	No. Top 30 by the composite score: +1.3%/yr vs the same universe (t = 0.5). Momentum alone: +6.4%/yr (t = 1.5, not significant).
Does 新闻联播 policy wording predict industry returns?	Not tradably. Over 8,490 new central measures, the strength of the wording is priced in before the open (+0.03%, t = 5.3; strongest level +0.12%). Afterwards, supported industries do slightly worse. Backtest: +0.6%/yr (t = 0.4).

The pattern holds across all of these: public information is priced fast. So stockproof helps you read and check evidence faster. It does not hand you a trading signal, and the app says this wherever it shows a score.

Principles

• Cited, not invented. The LLM may only use the numbered items it is given. Uncited points are dropped, and "unexplained" is an allowed answer.
• Scores carry their track record. ✓ useful in tests · tested, not significant · ? not tested yet.
• Point-in-time. EPS as first reported, real announcement dates, entry after the news. No peeking ahead.
• Any LLM. DeepSeek (default), OpenAI, Qwen, a local Ollama model, or any OpenAI-compatible API. Each item is read once and cached in SQLite, so daily use costs cents.
• No paid data. SEC EDGAR, Yahoo/Google News RSS, Nasdaq, and Eastmoney / Sina / CCTV / Shenwan via AKShare.
• Local and private. The dashboard listens on 127.0.0.1 only and refuses requests from other websites. API keys are read from the environment and never written to disk.

Install

Python 3.10+ (3.11+ for China markets).

Works on macOS, Windows and Linux (tested on all three in CI).

pip install "stockproof[cn,mcp]"

Then give it an API key (DeepSeek by default; stockproof providers lists the others). US earnings also need a contact for the SEC.

macOS / Linux — add to ~/.zshrc (or ~/.bashrc) so it is kept, then open a new terminal:

export DEEPSEEK_API_KEY="sk-..."
export STOCKPROOF_SEC_CONTACT="Your Name you@example.com"

Windows — in PowerShell (saved for your user; open a new window afterwards):

setx DEEPSEEK_API_KEY "sk-..."
setx STOCKPROOF_SEC_CONTACT "Your Name you@example.com"

Then start the dashboard:

stockproof dashboard

Add stocks in the page (AAPL, 600519, 300750, 00700 …) and click Refresh news / Refresh earnings.

10-minute tour

After installing and setting your API key:

1. Look around the demo (2 min). stockproof demo, then open each tab. The Home tab shows what you can do, with examples to click.
2. Your own stocks (3 min). stockproof dashboard → Stocks tab → type a few codes (AAPL, 600519, 00700) → Add → Refresh news, then Refresh earnings. The first refresh takes about a minute per stock. Click a row for its news, earnings read and announcements; click Why? next to an unusual move.
3. Test an idea (2 min). Lab tab → click a template, e.g. "Buy after big earnings beats, hold 20 days" → Backtest. Read the verdict, then change a condition and run it again: the significance bar goes up with each variant you try.
4. Check your portfolio (1 min). Portfolio check-up tab → type your holdings (600519 100, AAPL 50) → Check my portfolio. The report shows concentration, holdings that move together and how today's weights did in past crashes.
5. Ask in plain words (1 min, optional). stockproof mcp --install-dsh, restart DeepSeek Harness, type /stockproof or ask "what can you do?".
6. Keep it up to date (1 min). stockproof schedule --time 21:00 refreshes your watchlist every evening and writes the daily digest (macOS, Windows, Linux).

Commands

stockproof dashboard                         # everything in one local page
stockproof scan AAPL 600519 00700 --lang zh  # news mood report
stockproof earnings AAPL MSFT --open         # latest releases read by the LLM, with SUE
stockproof screen --top 30                   # whole-market A-share screen
stockproof screen --backtest --skip-screen   # test the screening rule on history
stockproof scorecard                         # evidence cards for the screen's top 10
stockproof scorecard 600519 NVDA             # …or for any stocks
stockproof research --market cn --open       # earnings event study + backtest (us / cn)
stockproof policy                            # 新闻联播 policy study (first run: a few hours of downloads)
stockproof daily                             # one refresh of the watchlist + daily digest
stockproof schedule --time 21:00             # run `daily` every evening (macOS, Windows, Linux)
stockproof calendar                          # who reports in the next two weeks
stockproof mcp --install-dsh                 # add to DeepSeek Harness on this computer (desktop app too)
stockproof lab caps                          # quant lab: what can be tested here
stockproof lab run '{"type":"earnings","filters":[{"field":"sue","op":">=","value":2}],"hold_days":20}' --open
stockproof lab confirm <id>                  # one shot on the locked holdout
stockproof portfolio "600519 100, 00700 300, AAPL 50"   # portfolio check-up
stockproof demo                              # offline demo, no key

Chinese codes can be typed in any common form: 600519, 600519.SS, 300750, 920001, 700, 0700.HK.

How the pieces work

News relevance

A feed for MRK also returns stories about the German Merck KGaA, a feed for GE returns GE HealthCare news, and NVDA appears in every "stock market today" article (in one scan only 2 of 14 headlines were really about Nvidia). The model gives each headline a relevance score (0–1) and a sentiment score (−1 to +1), with a reason. Off-topic stories stay visible but are not counted. When several headlines cover the same event, the model marks them as duplicates, so the event counts once.

Earnings and SUE

SUE = (this quarter's EPS − the same quarter last year) / standard deviation of the previous 8 such changes, using EPS as first reported. US quarters come from SEC XBRL, and Q4 = full year − Q1–Q3. A-share reports are year-to-date totals, so single quarters are derived by subtraction. Announcement dates come from the 8-K Item 2.02 filing (after 16:00 New York time counts as the next day) or from Eastmoney's 预约披露 actual disclosure date.

Screen and scorecard

• Screen: fundamentals for all ~5,200 A-shares come from Eastmoney's 业绩报表 tables (one download per report period).
  • Filters: no ST stocks, market value ≥ 50亿, at least 2 years listed.
  • Score: the average percentile of five factors.
• Scorecard: six dimensions plus industry policy, with one LLM call per stock per day.
  • Earnings quality asks whether growth comes from the core business and is backed by cash, or from one-offs.
  • Announcements count each kind of event once, since one buyback can produce ten filings.

Backtests

• Entry and holding: entry at the close after the reaction (earnings) or at the next open (policy), with capital split into slots and empty slots holding the market.
• A-share costs and trading rules:
  • Commission 0.025% per side, slippage 0.1% per side.
  • Stamp duty on sales: 0.1%, or 0.05% from 2023-08-28.
  • Price limits are respected, at 10% / 20% / 30% depending on the board.
• Benchmark: the equal-weighted average of the stocks studied, because today's CSI 300 members beat the index by 13% a year (survivorship).

Data sources and limits

• Sources: SEC EDGAR, Nasdaq, Yahoo/Google News RSS, and Eastmoney, Sina, CCTV and Shenwan via AKShare.
  • These are free, public, unofficial endpoints. They can change or rate-limit you.
  • stockproof slows down, retries, and stops with a clear message when a source refuses.
  • tools/*_probe.py checks which sources work from your network.
• LLM errors: models can misread. Every point links to the original, so check it.
• Not investment advice. The research above is part of the reason this tool doesn't give buy or sell calls.

Development

pip install -e ".[dev,mcp]"
pytest              # ~180 offline tests: fake feeds, fake LLM, fake AKShare; no key or network needed
python tools/make_demo.py   # rebuild the demo bundle from your own stockproof_data

MIT License.