Pair Trading

• You have identified two historically correlated instruments (e.g., stocks in the same industry, or BTC/ETH) and want to trade their relative divergences rather than outright direction.
• You want a market-neutral, long-short hedge where one leg is always long and the other is always short, reducing overall market exposure.
• You are backtesting equity pairs on Tushare-supported A-share markets or crypto pairs via OKX and need a ready-made Z-score signal pipeline.
• You prefer a simple, interpretable entry/exit rule driven by configurable Z-score thresholds and a rolling lookback window.

• You need to trade more than two instruments simultaneously; this skill requires exactly two instruments and does not support baskets or portfolios.
• You require a precise hedge-ratio calculation (e.g., via cointegration or OLS beta); this skill uses equal 50/50 capital allocation only.
• Your two instruments have unreliable or misaligned date indexes, as ratio calculation depends on synchronized price series.

Computes a rolling mean and standard deviation of the price ratio (leg A / leg B) over a configurable lookback window, then derives a Z-score to quantify how far the current ratio deviates from its historical norm. Entry and exit signals fire at user-defined Z-score thresholds (entry_z and exit_z).

When a signal is triggered, leg A and leg B always receive opposite directions — one goes long while the other goes short. Each leg is allocated exactly 50% of capital, ensuring a balanced, market-neutral hedge at all times.

Three core parameters — lookback (default 60), entry_z (default 2.0), and exit_z (default 0.5) — can be tuned to match the mean-reversion speed and volatility characteristics of any instrument pair.

Works with both Tushare (for A-share equities) and OKX (for cryptocurrency pairs such as BTC-USDT / ETH-USDT), making the strategy applicable across asset classes without code changes.

Before the lookback window is fully populated, Z-scores are NaN and signals are automatically set to 0, preventing spurious trades during the initialization period.

Trade two highly correlated A-share insurance stocks (e.g., 601318.SH and 601628.SH) by entering when their price ratio Z-score exceeds ±2.0 and exiting when it falls back within ±0.5. This captures temporary mispricings driven by sector-neutral noise rather than fundamental divergence.

Apply the same Z-score logic to BTC-USDT and ETH-USDT on OKX, exploiting the historically tight correlation between the two largest cryptocurrencies. When one rallies disproportionately relative to the other, the skill shorts the outperformer and longs the laggard.

Use the configurable lookback, entry_z, and exit_z parameters to backtest different mean-reversion assumptions over historical data, comparing how tighter or wider Z-score thresholds affect trade frequency, drawdown, and returns.

Embed this skill within a broader portfolio to add a market-neutral return stream, since the equal-weight long-short structure reduces net market exposure and provides returns that are relatively uncorrelated with broad market direction.