A frequent characteristic of residual z-scores is that they are not symmetric – the distributions above and below the axis cover different ranges.
Take this example of China Petroleum & Chemical Corporation, aka Sinopec, (SNP) vs PetroChina Company Limited (PTR), both ADRs on the NYSE.
Applying the same trigger of, say, ±2 standard deviations (with a couple of conditions covering direction) and then opening positions on the underlying share prices works fine. But there are not an awful lot of signals:
Go custom with, say +1.5 and -1.3 and things get busier (and alot more profitable):
Of course, this kind of playing with parameters is curve fitting (with enough effort even the losing open position could be repaired). One way around much of the danger in that is to do an in-sample vs out-of-sample comparison (and customizable dates in ArbMaker permit this).
We’ll be bringing out upgraded back testing features (and a few others too) in the next three weeks to do this type of analysis more easily at the portfolio level.
As for this pair itself, not bad. Both Sinopec and PetroChina are set to benefit from a loosening of China’s windfall tax on energy firms (although this is probably of greater benefit to PTR).
Hardly an irresistible catalyst but, in hand with the Sinopec out performance since July 2011, it makes considering another profitable reversion to the 5 year mean worthwhile. They run on the same kind of gas, right?
Finally, that open loss (under -1.5%) looks like this as of yesterday’s close:
Sell SNP @ $101.99 on 29 November (now $115.77)
Buy PTR @ $123.79 on 29 November (now $139.94)
We’ll have to revisit…