I just wasted 2 weeks effort on invalid historical data. As a robustness test I created a few MT5 accounts at various brokers to backtest the same EA on slightly different data sets. Well, "slightly" different, except for Darwinex which is wildly different! After getting very odd results in my backtests I took a closer look:
Note the spikes at every whole hour. They seem to lookahead and spike to a price level that the actual chart will reach later that hour. Needless to say, my EA perfectly discovered this "edge" to predict the future and showed unrealistic profits.
Compare for example with Dukascopy:
Same chart, but no spikes.
I know many people use Dukascopy data. I would like some additional alternative data sources. I took a look at ICMarkets and Pepperstone and they both look okish. Some gaps in their tick data, but M5 or higher should be fine for most backtests.
Can you recommend additional (free or affortable) data sources of reasonable quality?
Note the spikes at every whole hour. They seem to lookahead and spike to a price level that the actual chart will reach later that hour. Needless to say, my EA perfectly discovered this "edge" to predict the future and showed unrealistic profits.
Compare for example with Dukascopy:
Same chart, but no spikes.
I know many people use Dukascopy data. I would like some additional alternative data sources. I took a look at ICMarkets and Pepperstone and they both look okish. Some gaps in their tick data, but M5 or higher should be fine for most backtests.
Can you recommend additional (free or affortable) data sources of reasonable quality?