The best method of data is by direct import straight off the server rather than one big file dump that usuall has gaps etc. Gathering live is 'as it happened' tick by tick.

Where to get? I would go to FXDD (if they are not already your broker), open a demo, scroll back on all timeframes per currency until the chart data stops flowing (do this by either holding down the page up key or with wheel on mouse). Before doing this ensure you go tour MT4 tool bar and chose TOOLS, OPTIONS and then set max bars in histor and max bars in chart to 99999999999999. This will then default to the MT4 maximum nd ensure you grab ALL data that is available off the broker server.

Is the backtest as good as live/forward test? Answer - Some will say never and 98% I agree. It depends both on quality of data but mainly on what triggers your system. If you have a highly mechanical system that runs real time with non lag indicators the results can be very close to real. I know because I have done both. Built EA, run live, then a month later on a different MT4 station (to ensure no bias) run a backtest of same EA. The results mirrored the live almost to pip and minute with exception of few extra trades.

Getting back to Alpari. Data must be near 100% otherwise it is like trying to calibrate an engine with faulty measuring instruments. I optimized to alpari and then ran backtest on real data. The results were thousands $ difference (or not profitable at all). In the end I knew to believe broker data because results ended up same as forward test.

Hope this helps

WW