yigityildiz0 Technical Quant Analysis
Use indicators as measurements, not prophecies. Read references/data-contract.md, references/interpretation.md, and for strategy testing references/backtest-standard.md.
Workflow
- Verify instrument, exchange, currency, timezone, interval, data source, fetch time, delay, and corporate-action adjustment.
- Exclude incomplete candles. Reject duplicated, unordered, non-positive, or internally inconsistent OHLCV rows.
- Run
scripts/technical_indicators.pyon sufficient data. Do not report unavailable long-window indicators as if calculated. - Classify regime: trend, range, breakout, breakdown, volatility expansion/contraction, and liquidity quality.
- Inspect price structure, moving averages, RSI, MACD, ATR, volume, gaps, and multi-timeframe agreement. Avoid indicator-count voting.
- Express entry zone, invalidation, targets, and risk/reward as scenarios tied to confirmed levels and execution costs.
- Combine with fundamentals, valuation, macro, catalysts, and product mechanics when the decision is an investment rather than a chart exercise.
- For historical performance claims, require a point-in-time backtest packet and run
scripts/backtest_audit.pybefore interpreting results. - Route cross-asset rates/inflation/FX/credit/breadth regime work to
$market-regime-analysis; keep this skill focused on instrument data, signals and test quality.
Hard rules
- Never calculate from a screenshot when raw dated prices can be obtained.
- Never mix adjusted and unadjusted prices across splits, dividends, rights issues, or contract rolls.
- Never use an unfinished candle to confirm a close-based signal.
- Never call overbought an automatic sell or oversold an automatic buy.
- Never optimize on the full sample and report the same sample as validation.
- Include fees, spread, slippage, taxes, latency, delistings, survivorship bias, look-ahead, and data revisions where relevant.
- Distinguish exploratory signals from independently validated strategies.
Output
Report data timestamp and quality first, then regime, decisive levels, indicator readings, bull/base/bear paths, invalidation, modeled risk, and confidence. Say what evidence would confirm or reject the setup. Do not output a naked BUY/SELL score.