More On Parameterless Trading Model

I convey written before that my ideal trading model is i that has no parameters, too what ways in that place are to achieve this. Actually, I forgot to cite that a trading strategy proposed past times doc Andrew Lo discussed previously is inward fact parameterless, too the technique is so full general that it tin plough over notice hold upward applied to whatever mean-reverting strategy.

The technique is but this: keep a long (or short) portfolio amongst uppercase proportional to the distance betwixt a supposedly mean-reverting mensurate too its long-term hateful value.

For e.g. if you lot are pair-trading PEP vs KO, too you lot believe that the spread betwixt PEP too KO is mean-reverting, too so this spread is the mean-reverting mensurate you lot should employ.

As the spread moves away from its mean, perish along buying (or shorting) the spread inward equal dollar amount. And every bit the spread reverts, perish along selling (or buying) the spread inward the same dollar amount. What this dollar total should hold upward depends on: a) the full buying ability you lot possess, b) the expected maximum departure of the spread from its mean, too c) how oftentimes you lot intend to buy/short. Note that betoken c is non a parameter: it is arbitrary too express exclusively past times transaction costs, technology, too other operational issues. As for the expected maximum deviation, it tin plough over notice hold upward obtained past times observing the history of the spread since inception.

This scheme hence obviates the involve for entry or leave of absence thresholds, too amongst them, the possibility of data-snooping bias. (You may withal desire to impose an entry threshold based on transaction terms consideration - but that would non count every bit a complimentary parameter.)