stevefromnaki
07 Jun 2015, 16:43

RE: RE:

Yes, now that you mention it, I agree completely! This is an issue I have struggled with too. My robots seem to run really well in backtesting, and then struggle in forward testing. Of course, this could just be a result of me curve fitting to maximise the returns in the sample market conditions, and then the market conditions change and the robots results are then different. I only use tick data to backtest and my data only goes back 15 months so this is a bit of an issue for me. I even ran the backtesting on tick data only, and in the real account (not a demo like i used to) to see if that made a difference. It didn't.

With respect to the actual data, this is fed from the broker (not from spotware) so I'm not 100% sure if it a data issue or a backtesting/software issue. I suspect it is a data issue. In either case, you are right, how are we to rely on the data.

I have been thinking about building a strategy that is not as reliant on specific tick moves, so it takes bigger moves on higher timeframes, then minute data should suffice.

Stermy said:

Perhaps, but as I said my big fear is that it is indicates a more deep seated Bug in cAlgo. What if it is often giving back invalid results, but they are more subtle and not noticed? 

I have only been using cAlgo for about a month, after a decade of MT4, and overall I really like cAlgo. But none of my Bots do anything like as well in 'real time going forward' as they do in Backtests. Is it because the backtest data is sometimes, or somehow, corrupted? I know that it often was on the MT platform, so in the end I built my own system, collected Ticks and analysed them externally. The results sometimes matched in MT but quite often there were big discrepancies, some of which really hurry my EA's and sucked my pocket almost dry.

I dream of not needing to rely too much on Backtests, but they are tempting. However it is awfully hard to avoid, 'Curve Fitting'  ;}

PS - the sample of 'future results' I posted are easy to recognise as 'noise', but where did it come from? Also if I scroll backwards a bit some of it looks VERY normal, but is still impossible.

stevefromnaki said:

yeah i also notice results predicted into the future. so i cut off the tick data backesting at the end of may, and i think it is ok??

 

stevefromnaki
07 Jun 2015, 11:20

yeah i also notice results predicted into the future. so i cut off the tick data backesting at the end of may, and i think it is ok??

stevefromnaki
05 Jun 2015, 18:01

tick data backtesting seems to be working again for me.

Is it working again for anybody else?

stevefromnaki
04 Jun 2015, 14:40

Thanks for the update nobulart,

Looks like it is an issue for a few of us, but not all of us cAlgo users, hope there is a solution on its way :)

stevefromnaki
15 Oct 2014, 05:44

Thanks for your quick response.

I only use onBar methods and that's why I was getting confused with the different results between 30m and 1m open prices for backtesting. So I exported to excel the 30min and 1 min on bar OHLC and converted the 1m to 30m data and noticed there are differences. That's why I asked my original questions above. I would love to backtest using tick data but it only goes back about 6 months due to the high volume of data included in tick data.

My best option (which i am doing anyway) is to forward test on a demo and see how the actual results compare to backtesting over the same time period under the 1m and 30m data sets.

Is there much of a difference between demo and live prices?

Thanks again,

steve.

stevefromnaki
14 Jan 2014, 00:49