Rationale behind the 20 days default holding period

[Ray Zhung]

Just wondering what's the rationale behind the 20 days default holding period? I've experimented stretching the holding period to 100 days in my backtests and overall there's not significant difference in trading results (actually more pairs have slightly improved results). And I've noticed that 20 days is shorter than the life cycle of most pairs' life cycle of mean reversion, judging by their half-life, therefore could terminate trades prematurely. However, I also realize variance is a function of time and longer holding period could induce additional volatility in the portfolio, which is exactly what happened during my live trading this year. I lost count how many pairs did not mean revert in time (many still haven't), which I kept on holding and resulted in oversize loss that negated large chunks of positive gains. Given the mean reverting nature of this trading strategy, it's particularly hard to stick to rather short holding period when the spreads are being stretched and could mean revert any day theoretically. Personally I think determining the right holding period is one of the final hurdles of successfully implementing pair trading strategy in all market conditions, so please share some insights regarding the default holding period, i.e. have there any systemic tests been done to conclude that 20 days is the right balance between avoiding outsize loss and cutting short profit? Thank you very much!

[admin]

The holding period serves two purposes:

1) as a time-based stop loss - in the case you trade pair which in average mean reverts in 12 days, there is no point to stay in position too much longer (as the market regime of the pair probably changed)
2) to release margin of underperforming pairs to be used by better pairs (in case your portfolio is larger that number of margin slots, which it should be)

You should set your holding period the way you are comfortable with, determined by half-lives of your strategies and backtests. If you trade more pairs than you have margin slots, it may be preferred to cut the pair off earlier and release the margin to another strategy than waiting for an underperformer to mean revert.