Achieving accurate backtesting results on short time frames

[naxo79]

Tnx, it seems I got it working that way.

Is it possible to have both dataseries plotted in the backtest chart so I could verify the fills visually?

[NinjaTrader_Bertrand]

Glad to hear - that's unfortunately not possible, it would plot the primary series only in backtesting.

[naxo79]

Still something wrong with the executions, giving too optimistic results.

Exactly same strategy run on NT gives PF 1.7 and run on Multicharts gives PF 1.2.
Both use fx bid and ask data from IB.
Unfortunately MC's backtests have historically proven to be more reliastic.

[NinjaTrader_Bertrand]

naxo79, have compared executions across both backtests?

The IB connection would not offer historical bid / ask data, so likely the dataset you're testing on is inconsistent would be my guess here -

http://www.ninjatrader.com/support/helpGuides/nt7/historical_data.htm

[naxo79]

Your table is wrong regarding IB.

I can get both BID and ASK minute data for Multicharts and for NT up to a year back.

One cannot get that comprehensive data from Kinetick, even with the institutional package.
I tried today.

I compared executions and there are some small variations in fill prices, but over 300 trades so its not really feasible to compare them all.

[NinjaTrader_Bertrand]

Thanks for the feedback naxo79, we will check into it and amend the helpguide here. What kind of strategy are you using, are you sure the resulting trade signals would be comparable across platforms via the code and methods implemented - if you like me to test that would be of course possible as well - you could reach me directly under support at ninjatrader dot com Attn Bertrand

[naxo79]

The signals are calculated the same way.
Unfortunately I cannot share the strategy.
I can when I have time try adding the bid/ask test to a sample strategy and then share it and investigate the difference between Multicharts and Ninjatrader when trying to backtest on BID and ASK. But right now I dont have time.

If someone else stumbles upon this thread: I did not succeed in backtesting accurately on BID/ASK on shorter timeframes.