Open & Close Trade on Same Bar

New comment
 
I have seen really impressive backtest results of an EA that opens and closes trades on the same M15 bar. However, the EA does not test well live. So it seems that the back-tester overestimates the profit when an EA opens and closes trades on the same bar.

I don't understand why the backtester would make this mistake when it has M5 and M1 data available, and modelling quality is 90%. It should be able to see the time history of prices during the M15 bar, since it has the M1 data available.

I have learnt to not trust backtest results when the trade opens and closes on the same M15 bar, but I don't understand why. Does anyone understand how the backtester seems to get it wrong? Is there an article somewhere explaining it?

Thanks for any help.
 
Hi Barri.

I'm not sure but from what I've read in forums, MT4 backtester has only OHLC real data.

The '90% modelling quality' simply mean that your higher timeframe data was generated fully from M1 data. If your M1 history data doesn't reach your higher timeframe period it will take M5, M15, etc. data which generally go further in past and your modelling quality will decrease.

As it has no true intrabar (tick) data it generates intrabar motion using some fractal interpolation of the OHLC M1 data. Here comes the problem!

Apparently this fractal interpolation algorithm generates a much more volatile intrabar motion than what one expects in true ticks.

As a consequence, an EA that operates intrabar will present much more TP filling in backtests than in live operation.

That's what I learnt from more experienced traders' opinions.

Take care

Renato
 
i had some really bad results from testing on H4 bars. I found that when a single 1M bar is large it would trade over and over on the same bar.
I hardwired my EA to run H4 on any timeframe and ran it 1M. It worked much better, closer to reality.
Instead of using High,Low,Close, etc. use iHigh,iLow,etc and hardcode time frame you want to run it with.
 
link to the topic of the similiar problem with my answer...
'Made $ 258 663.50 from $3000 using Problematic Backtesting by just buying at open (Fractal ZigZag)'
 

Thanks to all who responded for your insights. My take-away conclusion is that if we see an entry and exit on the same bar in backtests, then the interpolator is probably helping us to get better results than we will in real life trading, so don't believe backtest results when we see entry and exit on the same bar. It seems that some of you are becoming very knowledgeable about back-testing and the current shortcomings, so maybe it's time MetaQuotes turned on the ability to use tick data in backtests.

 
They won't. They are only developing a software. It is not Metaquotes, who give you real time data either, it is your broker. Well, this is usual with financial softwares: actually, when trading futures, you need to pay even for quality realtime data, not to talk about historical. So we are just getting spoiled by free demos, and free real time data.
New comment
Reason: