Backtesting reliability issues

Dear all,
Lately I've encountered some backtesting problems.
Testing my system with data downloaded from http://www.alpari-idc.com/en/dc/databank.php form the beginning of 2007 till now gave a 1,500 pips profit.
Reinstalling of Metatrader, new import and period conversion of the history data gave a totally different outcome - only 100 pips.
This is not the first time of inconsistency in backtesting I\m encountering with.
Has anyone experienced similar phenomena?
Can backtesting be accountable?

Mik

You should post this in the metatrader section

There are two obvious things to check.

a) Is the data that you used EXACTLY the same. Brokers do change things retrospectively, and candles sometimes go missing, or missing candles are added etc. I suggest writing something that checks each OHLC on a candle by candle basis, the smallest of differences can have quite large results

b) Did the backtest occur over EXACTLY the same time period, if the periods differ, then results will certainly differ

variability in back test results from different releases of metatrader with exactly the same data has been an issue in the past, personally I hand test everything, its just too unreliable.