in my personal experience, cointegration on equities does not lead to any interesting results. Once I discovered that cointegration was useless, I tried to think about pairs that made sense from an economic point of view and instead of single stocks, I considered ETFs and futures. For instance, one can think of a lead-lag relationship between natural gas and ETF of gas supplier firms.
I definitely agree with you that just approaching *solely* from a pure quantitative perspective of co-integration isn't likely to yield results. An economic rationale of the pair is necessary, but co-integration can be used as a safety-check to confirm that the relationship is statistically significant.
That's a pretty interesting example and could potentially offset the drawback of low returns due to the leverage of the futures contract. I imagine when applying this to futures we'll need to also incorporate ratio sizes when considering the tick values per future contract.
So, we chose the 200-day average since it's generally a good baseline for medium-term trends. People commonly refer to the 50-200 day moving average crossover when forecasting short/medium term directions. Going with the 50-day average may also work, but since it is much shorter, it may trigger trades too early/often.
by the way very interesting this stat arb path, looking forward to your insights
in my personal experience, cointegration on equities does not lead to any interesting results. Once I discovered that cointegration was useless, I tried to think about pairs that made sense from an economic point of view and instead of single stocks, I considered ETFs and futures. For instance, one can think of a lead-lag relationship between natural gas and ETF of gas supplier firms.
Doing so, the results were much more interesting.
Hey Lorenzo,
I definitely agree with you that just approaching *solely* from a pure quantitative perspective of co-integration isn't likely to yield results. An economic rationale of the pair is necessary, but co-integration can be used as a safety-check to confirm that the relationship is statistically significant.
That's a pretty interesting example and could potentially offset the drawback of low returns due to the leverage of the futures contract. I imagine when applying this to futures we'll need to also incorporate ratio sizes when considering the tick values per future contract.
Thanks!
Thanks for the fantastic article! jw ~ any particular reason why you chose 200 as the window length for MA calculations?
Hi Jay, super glad you enjoyed it!
So, we chose the 200-day average since it's generally a good baseline for medium-term trends. People commonly refer to the 50-200 day moving average crossover when forecasting short/medium term directions. Going with the 50-day average may also work, but since it is much shorter, it may trigger trades too early/often.