Hi MichaelSorry but I should have put that in context. What I...

Hi Michael
Sorry but I should have put that in context. What I mean is that true raw tick data (at millisecond levels) is generally full of crap and given the problems with even dealing with one pair at a time I can see it being more difficult with several.

With raw tick data the spreads can jump from 0.2 pips to 5.0 pips to 1.5 pips in the space of say 200 milliseconds (or less) .. or maybe the entire sell side disappears for 0.5 seconds. When you want to look at several pairs then you have to also match up the timing periods between ticks for all the pairs. Its not impossible but hard.

One final point of context is that my only focus is short term (at the moment) and therefore I can not use pre-filtered timed intervals of data (ie the normal 1 minute bars or 5 minute bars) - hence why I thought of pricing and timing issues. But depending on your model it still may work with this type of data ... medium to low frequency are generally not so reliant on spreads.

hope that helps.
cheers