I'm not sure how well any "pattern-recognition" software could work. I mean, the parameters would have to be pretty broad, wouldn't they? My limited experience has been that you get lots and lots of candidates that you have to sift thru to find anything truly worthwhile. I would think a better method would be to scan for certain criteria that's more definitive, first narrowing your universe of stocks by avg daily volume for instance, then maybe something like relative strength vs the parent index or vs that universe of stocks, or some kind of indicator like MACD or Stochastic extremes or crossings, and so on. Either way, you're still going to have to do a lot of eyeballing of the resultant charts to find suitable candidates. Granted, I haven't tried any such software in several yrs now and maybe the technology has improved, but patterns will always have a subjective aspect to them, won't they? H
Pattern-recognition software works well if it is software that can do that. AI and neural network can coupe with that. Suppose you selected a set of indicators that you think sufficiently reflect market up and down trends. You examine historical data and mark places where trends change their directions. Then you provide this data together with set of indicators to software as data for inductive training. Software assesses each area where trend changes its direction and forms its description based on selected set of indicators. Software can use accumulated descriptions as a base for prognosis of future situation at real market. Human being has quite limited potentials to do same work because he has difficulty to process more then 7 criteria at the same time. That is why he should narrow initial set of criteria that he would like to use for analysis.