etorotrade

Yahoo Finance analyst data, aggregated into BUY / SELL / HOLD signals for equities and ETFs, with position sizing and CSV+HTML reports.

Created by plessas, an eToro Popular Investor. This tool provides analysis only, not investment advice. Every investment decision is the user's own responsibility.

etorotrade is a Python-based investment-signals engine. It pulls analyst consensus, price targets, and fundamentals from Yahoo Finance (yfinance + yahooquery), applies tier and region-specific thresholds from config.yaml, and writes color-coded reports to yahoofinance/output/.

The universe it scans is the ~12k tickers offered on eToro, plus your own portfolio. A full nightly pass runs on GitHub Actions in six parallel shards and commits the fresh CSVs back to the repo.

• eToro users who want a systematic, second-opinion read on their positions.
• Retail investors who prefer analyst consensus + fundamentals over discretionary picks.
• Developers who want a reproducible, testable signal pipeline they can extend.

• Not investment advice, not a robo-advisor, not an automated trading system.
• Not a source of proprietary data. Everything downstream of Yahoo Finance.
• Not a promise of returns. Historical signal accuracy is measured by the built-in backtester, and results are advisory.

flowchart TD
    A[Ticker universe<br/>yahoofinance/input/*.csv] --> B[AsyncHybridProvider<br/>yfinance + yahooquery]
    B --> C[Analysis engine<br/>tier + region gates from config.yaml]
    C --> D{BS classifier}
    D -->|B| E[buy.csv / .html]
    D -->|S| F[sell.csv / .html]
    D -->|H| G[hold.csv / .html]
    D -->|I| H[market.csv / .html]
    C --> I[etoro.csv<br/>full scored universe]
    E & F & G --> J[signal_log.jsonl]
    J --> K[Backtester<br/>T+7 / T+30 vs SPY]

Python 3.10, 3.11, or 3.12.

git clone https://github.com/weirdapps/etorotrade
cd etorotrade

# Create a venv and install the pinned, SHA256-hashed lockfile.
# Same install path CI uses. Poetry is NOT needed to install; only to change deps.
python3 -m venv venv && source venv/bin/activate
pip install --only-binary :all: --require-hashes -r requirements-dev-lock.txt

Or run scripts/dev/setup.sh, which does the same and also installs pre-commit hooks and copies .env.example to .env.

1. Export your positions from eToro (Portfolio, Export to CSV).
2. Save the file as yahoofinance/input/portfolio.csv. See yahoofinance/input/portfolio.csv.example for the schema.
3. Run:

python trade.py -o p

Results print to the console and land in yahoofinance/output/portfolio.csv and portfolio.html.

trade.py is the single entry point. It runs interactively when called with no arguments, or takes an operation via -o and a target via -t.

python trade.py                      # interactive menu
python trade.py -o p                 # portfolio analysis
python trade.py -o p -t n            # portfolio, fetch fresh eToro data first
python trade.py -o m                 # market screening
python trade.py -o m -t 10           # market, first 10 tickers
python trade.py -o e                 # full eToro universe scan (~12k tickers)
python trade.py -o t -t b            # BUY opportunities (from etoro.csv, excludes holdings)
python trade.py -o t -t s            # SELL opportunities (portfolio holdings with S signal)
python trade.py -o t -t h            # HOLD opportunities (from etoro.csv, excludes holdings)
python trade.py -o i -t AAPL,MSFT    # ad-hoc analysis for specific tickers
python trade.py -o b                 # backtest signals (T+7 / T+30 forward validation)
python trade.py -o sc                # signal scorecard
python trade.py --validate-config    # validate config + exit

python scripts/analyze_geography.py       # ETF geographic-exposure decomposition
python scripts/analyze_industry.py        # sector-allocation analysis
python scripts/refresh_etoro_universe.py  # refresh the eToro ticker universe
python scripts/market_snapshot.py         # quick market snapshot

Reports are written to yahoofinance/output/:

The scored files share a common schema:

TKR, NAME, CAP, PRC, TGT, UP%, #T, %B, #A, AM, A, EXR, B, 52W,
2H, PET, PEF, P/S, PEG, DV, SI, EG, PP, ROE, DE, FCF, ERN, SZ, BS,
SIGNAL_TRACK, SIGNAL_HORIZON

Key fields:

• TKR / NAME / CAP: ticker, company, market cap.
• PRC / TGT / UP%: current price, mean analyst target, implied upside.
• #T / #A: number of analyst targets and analysts.
• %B / AM: percent BUY consensus, analyst momentum.
• EXR: expected return (upside * buy% / 100).
• PET / PEF / PEG / P/S: valuation multiples.
• DV / SI / EG / PP: dividend, short interest, earnings growth, price performance.
• ROE / DE / FCF: return on equity, debt/equity, free-cash-flow yield.
• ERN: next earnings date.
• SZ: recommended position size (from PositionSizer).
• BS: the signal itself, one of B, S, H, I (INCONCLUSIVE).
• SIGNAL_TRACK / SIGNAL_HORIZON: metadata for the backtester.

1. Universe load: tickers are read from yahoofinance/input/etoro.csv, portfolio.csv, or market.csv depending on operation.

2. Data fetch: AsyncHybridProvider calls yfinance for the bulk of fields and falls back to yahooquery for the fields yfinance drops (PEG, some fundamentals). Rate limiting and a disk cache sit in front.

3. Tier + region gating: each ticker is classified into one of five buckets. Cutoffs live in config.yaml:

   Tier	Market cap
   MEGA	>= $500B
   LARGE	$100B to $500B
   MID	$10B to $100B
   SMALL	$2B to $10B
   (below $2B)	Marked INCONCLUSIVE, hard floor

   Universal analyst gates apply on top: $2B to $5B needs 6+ analysts, $5B and up needs 4+ (per-region tier blocks tighten further, e.g. US MEGA/LARGE/MID min_analysts: 8, HK MEGA: 15). Region blocks exist for US (baseline), EU (.L, .PA, .AS, .DE, .MI, etc.), and HK (.HK).

4. Signal classification: BS is set to B, S, H, or I based on the tier/region-specific buy and sell blocks in config.yaml (upside, buy%, PE, PEG, ROE, DE, beta, FCF yield, analyst momentum, and more).

5. Position sizing: PositionSizer in trade_modules/trade_engine.py produces SZ based on market-cap tier, expected return, and risk constraints.

6. Persist: results are written to CSV and HTML, and appended to signal_log.jsonl for backtesting.

All thresholds live in a single ~1,000-line YAML file: tier gates, per-region buy and sell blocks, position-sizing parameters, calibration metadata. Parameters are calibrated quarterly against the T+7 / T+30 backtester and edited only by human review.

yahoofinance/input/portfolio.csv (mirror of the eToro export):

symbol,totalInvestmentPct,totalNetProfitPct,instrumentDisplayName
AAPL,5.2,12.5,Apple Inc
MSFT,4.8,8.3,Microsoft Corporation

Optional. Copy .env.example to .env.

trade.py                    # CLI entry point
trade_modules/              # Trading logic
  trade_cli.py              # argparse + async orchestration
  trade_engine.py           # TradingEngine, PositionSizer
  analysis_engine.py        # signal generation
  config_manager.py         # ConfigManager, ticker substitutions
  backtest_engine.py        # T+7 / T+30 forward validation
  signal_scorecard.py       # per-signal accuracy scorecard
  committee_*.py            # optional multi-agent research committee
  signals_v2/, riskfirst/   # next-gen signal pipeline (WIP)
yahoofinance/               # Data layer
  api/providers/            # AsyncHybridProvider, AsyncYahooFinance,
                            # AsyncYahooQuery, AlphaVantage, Polygon
  analysis/                 # StockAnalyzer, market filters, tiers
  core/                     # DI container, logging, errors, config
  utils/                    # trade criteria, async helpers, market utils
  presentation/             # console + HTML renderers
  input/                    # portfolio.csv, etoro.csv, region files
  output/                   # committed CSV + HTML reports
scripts/                    # standalone analysis + ops scripts
  dev/                      # setup.sh, test.sh, lint.sh, format.sh, relock.sh
config.yaml                 # buy/sell thresholds per tier/region
docs/                       # USER_GUIDE, TECHNICAL, POSITION_SIZING, CI_CD, FAQ
tests/                      # unit, integration, e2e, benchmarks

Nine workflow files in .github/workflows/:

There is a forward-validation backtester in trade_modules/backtest_engine.py. Since Yahoo Finance does not expose historical analyst recommendations, the engine uses signal_log.jsonl (accumulated by every run since January 2026) and compares each signal against actual price movements at T+7 and T+30 trading days.

python trade.py -o b

Output:

• Per-signal accuracy (BUY, SELL, HOLD).
• Tier and region breakdowns.
• Comparison against SPY as benchmark.
• Reports in yahoofinance/output/backtest_*.csv and backtest_report.json.

The weekly-backtest.yml workflow runs the same pipeline on GitHub Actions and commits the reports.

pytest tests/                              # full suite
pytest tests/unit/                         # unit tests only (fast)
pytest tests/integration/                  # requires network
pytest -m "not slow"                       # skip slow tests
pytest --cov=yahoofinance --cov=trade_modules --cov-report=html
scripts/dev/test.sh                        # wrapper with coverage

CI runs the suite with --cov-fail-under=58.

scripts/dev/lint.sh          # black --check + isort --check + flake8 + mypy
scripts/dev/format.sh        # black + isort auto-format

Line length: 100. Enforced formatter and linter: ruff (ruff-format + ruff check, via .pre-commit-config.yaml). The legacy scripts/dev/lint.sh and format.sh still call black/isort/flake8, but they are not CI-gated. Type checker: mypy runs in lenient mode and does not gate CI.

Three requirements-*-lock.txt files (production, production+dev, production+smoketest) are exported from poetry.lock, committed, and enforced by the lockfile-sync CI job. To bump a package:

# 1. Edit pyproject.toml
# 2. Regenerate poetry.lock + the three exported requirements files
scripts/dev/relock.sh
# 3. Commit pyproject.toml + poetry.lock + requirements-*-lock.txt together

relock.sh takes two optional modes:

Plain poetry lock keeps every already-locked package pinned, so it will not pull transitive dependencies forward on its own. Reach for --regenerate when the goal is "everything at latest".

A fourth lockfile, requirements-universe-lock.txt, is not generated by relock.sh. It is a small hand-maintained hash list (5 packages, Python 3.11 / manylinux2014_x86_64) used only by weekly-universe-refresh.yml; regeneration instructions are in its header comment.

Poetry is only needed to change dependencies. Installing them uses pip directly with --require-hashes.

A Dockerfile is provided for containerised runs.

• User Guide: getting started, common workflows.
• Technical Architecture: system design, providers, DI.
• Position Sizing: the risk-adjusted sizing algorithm.
• CI/CD: pipeline stages, quality gates.
• Examples, FAQ, Troubleshooting.
• Evidence-based threshold changes and Large/Mega threshold changes: historical calibration decisions.

See SECURITY.md. Do not open a public issue for a vulnerability; report it privately through this repository's Security tab (Report a vulnerability).

MIT, see LICENSE.

This tool is designed for quantitative analysis and research. It does not constitute investment advice. Users should conduct their own due diligence and consider consulting a qualified financial advisor before making investment decisions. Past signal performance does not guarantee future results.