2002issue C051-5
A same-session pair book with a morning-fixed volatility envelope
This archive write-up treats a two-name day book as one closed pairs-trading procedure. A volatility envelope fixed at the session open around the normalized-spread marks when the first name is rich or cheap, when both legs come off at a return to zero, and when the book is aborted.
- Pairs-trading is written here as a market-neutral book: a long in one stock and a short in a related stock opened together, so the wager is on relative displacement rather than market direction.
- The live line is a normalized-spread: last price of the first name over its prior close, minus last price of the second name over its prior close.
- Upper and lower envelopes are computed at the session open from thirty-day historical-volatility and the pair correlation-coefficient; the default volatility-factor of 1.5 standard deviations turns the one-sigma band into the entry threshold.
- Both legs exit together when the spread recrosses zero, or when it extends beyond the volatility-factor times the envelope.
A two-name market-neutral book
Pairs-trading, in this write-up, is a market-neutral book that opens a long in one stock and a short in a related stock together, so the wager is on relative displacement rather than on market direction. The two legs are opened at the same time on the premise that a temporary decorrelation later mean-reverts.
Mean-reversion is the working assumption that two names that usually travel together will, after a temporary breakdown in co-movement, pull a normalized spread back toward its typical level.
The book is specified for highly correlated pairs with a correlation-coefficient of at least 0.5, where plus one means tandem movement and minus one means opposite movement.
Reading the live spread
The normalized-spread is the live relative-value line formed by scaling each stock's last price by its prior close and subtracting those two ratios. The live spread is last price of the first name over its prior close, minus last price of the second name over its prior close.
Envelopes fixed at the session open
Bollinger-bands are used here as a pair-level volatility envelope, computed at the open from each name's historical volatility and their correlation, and used as the quantitative baseline the live spread must breach and then recross. Upper and lower volatility envelopes are computed at the start of the session. A move to the upper envelope treats the first name as rich versus the second, and a move to the lower envelope treats the first name as cheap.
Historical-volatility is a lookback standard deviation of close-to-close percentage change, typically read over thirty sessions and treated as an input to the daily pair envelope. A 20-dollar name with a 20 percent reading is described as spending about 68 percent of that window between 16 and 24 dollars under a normal curve.
A one-day pair envelope is formed from the two 30-day volatilities, a 252-session year scale of 0.063, and one minus the pair correlation. Volatilities of 0.57 and 0.70 with correlation 0.60 produce a one-sigma band of 0.032.
The volatility-factor is a multiplier, defaulted to one and a half standard deviations, that turns the daily pair envelope into the entry threshold and a still-wider stop. The default signal uses 1.5 standard deviations, stated as giving prices at least an 80 percent chance of reverting toward the mean, which turns the 0.032 one-sigma band into 0.048.
Entry after a recross
A long-first-name entry waits for the spread to cross back above the lower envelope, then buys the first name one tick above the next bar high and shorts the second name on that fill. The reverse entry waits for a cross below the upper envelope and shorts the first name one tick below the next bar low.
A joint target and stop
Both legs exit together when the spread recrosses zero as the target, or when it extends beyond the volatility factor times the envelope as the stop.
Liquidity, pairing and news
Operational constraints in the write-up include screening for liquidity and volatility, preferring same-industry leader-laggard pairings, opening the short leg first when an uptick constraint applies, and standing aside after significant news until the spread begins to revert.
A leader-laggard-pair is a same-industry pairing of a faster growth name against a slower value or blue-chip name so the spread can travel even when both prices move in the same direction.
All readings on this track · 36 readings
- 1986A futures fade as one range, order, and secrecy procedure
- 1992Constructing the mass-index range-reversal procedure
- 1993Switch trend following and mean reversion with an equity-curve filter
- 1994Evaluating weekly trend-following and mean-reversion timing rules
- 1996Dual-horizon bands for a precious-metals cash switch
- 1997Constructing a moving regression oscillator
- 1997Regime-dependent long and short rules in mechanical systems
- 2002A same-session pair book with a morning-fixed volatility envelope
- 2004Combining noncorrelated trend and reversion systems
- 2004Failed-breakout overlays on trending markets
- 2004Rank rotation after a path split, then Robustness testing
- 2004Range-bound tape as a filter for trend and oscillator rules
- 2005A moving-average short pullback that is only in scope in a decline
- 2006Constructing an adaptive price zone from a double-smoothed range
- 2007Two-period relative strength index versus a one-week universe baseline
- 2008Building ETF mean-reversion entries with a two-bar washout
- 2008Rebuild a short-period stochastic as a premier stochastic oscillator
- 2008A three-market regime map for equity bounces and dollar cycles
- 2009Option trade adjustment as one testable procedure
- 2010Implied volatility as a May 2010 market-regime lab for the S&P 500
- 2011Treat a large one-day move as a classified event
- 2011Long-call exits, volatility regimes, and spread assignment
- 2011Pairing same-horizon oscillators with a walk filter
- 2012Two-bar band extreme entries with trailing stops
- 2012An eight-month average as a monthly gate for high-yield bonds
- 2014Complete the checklist before the trade
- 2014Coded rules should face one test, not a kinder sample
- 2015Build a mean-reversion basket from one correlation path
- 2015Index dip reversion is horizon and regime dependent
- 2016Treat the end of a trend as a handoff, not a broken system
- 2017A testable half-swing pullback for trend continuation
- 2017Evaluating four swing detection rules for mean reversion
- 2018Intraday breakout and mean reversion as one rule set
- 2018Evaluating rare consecutive-close mean-reversion entries
- 2020Moving-average baselines, price vetoes, and mean reversion
- 2020Two-dimensional FX scaling for trend and reversal systems