Quote:
Originally Posted by tdion
bah
if the EA is coded right, a back test is as good as a forward test... assuming there is no scalping... and timeframes are small enough to be realistic....
but most don't know how to backtest, see a billion dollar graph and get their lamborghinis in their head until their dreams are shattered, then blame the backtester.
|
That implies that the pricedata as well as the Backtester itself is OK - it isn't! Yes, I agree that the backtester is robust, but the data is another matter entirely. And it also depends on how you define 'scalping'. The official definition would appear to be 1-5 pips, though brokers have often shoved this into the 10-30 pip range, but most traders would disagree with this definition. So, if you went with the former definition, chose to backtest on EURCHF, and your EA didn't 'scalp' (10 pips TP) your end results run the risk of including an unacceptable number of bogus trades. Increasing this to 20 pips still has an unacceptable number of potential trades that would not occur in real life - see below. And other pairs such as GBPCHF are even worse! btw, this only checks adjacent bars opened at the 'correct times' - it does not include Close/Open prices that had missing bars in between, so what you see is a 'best-case' scenario. Also, notice the number of 'missing' bars - quality data? methinks not...
FWIW, EURUSD is surprisingly free from these problems, so should be the pair of choice for backtesting. Also, these errors drop off almost completely after Nov. '06, so are far as I'm concerned, this is the reason for the unrealistic results the majority of EA's show before this date - so imo, the only worthwhile data Metaquotes offers is from Nov '06 to present...