Is there an optimal look-back period for finding pairs? I have read that it is a good idea to test pairs over multiple time frames but I don't know if that matters.
Another thing that I have read is that we should select pairs that have similar betas.
Do you mean for correlation? I just use a simple 100 day lookback for correlation, seems to work fine for me.
How often should one look for new pairs? For example, I have many pairs set up in PairTrade Finder now. In six months time should I backtest again?
An excellent thread - thanks Stocktrader. This is something I've been thinking of trying but not yet had a crack at. The PDF is really useful. The examples imply that IG Index allows you to place non-integer stakes - is that the case ? eg: Long RDSA at £1.00 per point / short BP at £3.29 per point. I use Capital Spreads, who allow only stakes that are multiples of £1.
Thanks Lado. The case for the program (rather than manual methods) is stacking up further. Does it work with indices and commodities or just equities ?
Thanks Lado and Stocktrader. I was envisaging spreadbetting indices such as DAX / FTSE100 (or, strictly, their cash futures) rather than trading ETFs that are based on them - unless you recommend otherwise. I'm also looking at pairs such as gold / silver, which are very much out of kilter at present - again, ideally spreadbetting rather than using ETFs.
I guess if the software doesn't look at commodities then one could use certain ETFs as a very good proxy for them.