Have you checked out Michael Harris' work? He sets out a framework (or mechanical method, if you prefer) for defining, designing, testing, mining, and scanning for signals based on price patterns. The book he wrote oriented toward stocks is this one:
traderslibrary.net
I haven't yet tried to use his techniques, and I didn't find them very practical for my own work, though I did find his theories and insights interesting. You might, for instance, use his approach to test various definitions of a reversal (or non-reversal) bar within a more or less extensively defined larger pattern, but to develop an approach that qualified as a "scientific" test within the larger Dunnigan framework you're exploring might be rather difficult, I think. Still, Harris at least sets out the criteria for such tests, and also describes how you might go about assembling a large enough number of examples to make a test meaningful. It's just conjecture on my part, but I think the conceptualizing, programming, and data-mining involved would amount to quite a project...
If you're serious about trying his approach out, you might want to consider his previous book (more oriented toward futures) as well, but I'm not really sure. |
