margin-ta

margin-ta

Read-only technical-analysis pipeline — generate multi-horizon stances with indicator consensus, tiered S/R and market/sector risk dashboard.

한국어

Read-only technical-analysis for US and Korean equities and crypto pairs. Combines 30+ vectorized indicators, multi-horizon daily/weekly/monthly stances, indicator-consensus layer, tiered support-resistance detection, and a market-dashboard with sector risk into one CLI — also the analysis engine behind an MCP server. The pipeline never places orders; every output is decision-support data.

• Python 3.10+, yfinance + pykrx public data by default
• 30+ vectorized indicators (SMA/EMA, MACD, RSI, Stochastic, ADX, Bollinger, ATR, Ichimoku and more) with TA-Lib patterns
• MCP server ready, JSON + chart PNG + TradingView deep links

Quick start — multi-horizon analysis with Python and technical-analysis

This quick start uses the Python pipeline for multi-horizon stances and technical-analysis indicators, including stock-analysis signals and quant scoring.

Install from GitHub Release tarball (no PyPI registry)

gh release download v3.0.0 --repo ianlyoo/margin-ta --pattern "margin-ta-*.tar.gz"
pip install ./margin-ta-3.0.0.tar.gz

When gh is unavailable — build the distribution tarball locally:

python -m build --sdist
pip install ./dist/margin-ta-3.0.0.tar.gz

Clone, install, and run

git clone https://github.com/ianlyoo/margin-ta.git
cd margin-ta
pip install -r requirements.txt
python scripts/margin_ta.py AAPL --json --quiet --no-tv --no-market
python scripts/market_risk.py --sectors

Ticker auto-detection: 005930, 005930.KS, AAPL all work. Add --chart for PNG and --save to persist JSON; --flow, --ownership, and --options-* layer on additional data.

Use cases for market-dashboard and trading with risk-management

Trading research where indicator consensus and tiered S/R guide position planning, and where market-dashboard plus risk-management context shapes exposure. Suitable for screening, watchlist ranking, and pre-trade review.

• Screen by Entry Score (0–100, capped per category) and indicator-consensus agreement
• Inspect near / intermediate / major support-resistance tiers
• Review market/sector risk 0–100 and regime calm/caution/stress/crisis before sizing

What this is not: an execution engine, a signal bot, or investment advice. All outputs are heuristic decision-support.

Architecture: indicator-consensus and support-resistance pipeline

flowchart LR
  A[Ticker] --> B[L1 Data: OHLCV + metadata]
  B --> C[L2 Indicators: 30+ vectorized]
  C --> D[L3 Signals: S/R + consensus + multi-horizon]
  D --> E[L3 Risk: market-dashboard]
  E --> F[L4 Pricing: triggers + ATR stops]
  F --> G[L5 Output: table + PNG + JSON]

Layer	Module(s)	Responsibility
L0 Orchestration	margin_ta.py	Ticker validation, provider selection, pipeline flow
L1 Data	layer1_data, layer1_market, layer1_kr_market	2y daily OHLCV + metadata; yfinance/pykrx public data; optional Toss/KIS hooks; market-regime fetch
L2 Indicators	layer2_indicators	30+ indicators via pandas ta and TA-Lib
L3 Signals	layer3_signals, layer3_consensus, layer3_horizons, layer3_liquidity	Tiered S/R, Entry Score, consensus, multi-horizon stances
L3 Risk	layer3_risk, layer1_market	Market/sector risk scoring
L4 Pricing	layer4_pricing	Trigger-based entry plans, ATR stops, tiered targets
L5 Output	layer5_output	Rich console tables, chart PNG, JSON, TradingView links

Multi-horizon stances (short/mid/long) collapse into aligned_bull, aligned_bear, mixed_pullback, mixed_rally, or mixed; horizons with insufficient history report insufficient_data.

Benchmark: Entry Score and indicator-consensus in measured runs

Qualified evidence only. No return or execution claim is made.

Setup (adjacent limitations): Synthetic watchlist of 30 tickers, one run per score snapshot, data as of 2026-08-25, --no-tv and --no-market for offline determinism, local cache only, no live brokerage, no forward-return validation. Scores are heuristic; thresholds may change with indicator tuning.

Watchlist	Median Entry Score	Consensus bullish share	Data completeness
US large-cap 15	42	33%	15/15
KR large-cap 15	38	27%	15/15

• Every figure is reproducible via python scripts/scan_nightly.py --top 30 --json on the same cached OHLCV snapshot; raw outputs are not financial advice.
• Verify locally:

python scripts/margin_ta.py AAPL --json --quiet --no-tv --no-market | jq .consensus
python scripts/market_risk.py --json | jq .regime

Limitations restated: offline snapshot, one run, no brokerage execution, no return prediction, public data may be delayed or incomplete, heuristic thresholds, no investment advice.

Validation methodology

• OHLCV from public yfinance/pykrx endpoints at run time (no stored API keys)
• Deterministic indicator calculation via pandas ta and TA-Lib (pinned versions in requirements.txt)
• Spot-check against TradingView via --no-tv toggle for manual review (not an automated oracle)

Responsible use

Market data is delayed and heuristics change; risk thresholds and Entry Score caps may evolve. Validate on your own data and do not extrapolate beyond the measured watchlist above.

Configuration

Env	Effect
MARGIN_TA_TOSS_IMPORT	Import path to Toss Securities client module
KIS_ENV_FILE	Env file with KIS credentials
MARGIN_TA_DATA_DIR, MARGIN_TA_CHARTS_DIR	Cache and chart output directories

See README body for full env list. Credentials are read from environment only — never hardcoded.

Project links

• Repository: https://github.com/ianlyoo/margin-ta
• Issues: https://github.com/ianlyoo/margin-ta/issues
• Pages: https://ianlyoo.github.io/margin-ta/
• License: MIT

License

MIT — see LICENSE.

Social preview

Social preview image (1280×640, solid background, high contrast): docs/assets/social-preview.png — Pages canonical https://ianlyoo.github.io/margin-ta/assets/social-preview.png — rebuild with node scripts/build-social-preview.mjs.