Variability in results when using different modelling techniques

 

Hi everyone,


I have been testing an EA using 

1) Open Prices

2) 1 min OHLC

3) Real Ticks


There is similarity between Open Prices and 1 min OHLC however there is a large difference when using real ticks. Well it's about 0.75% which I guess isn't too bad. It's just more than I was expecting. See the results below.

When inspecting the data it looks like it's just due to very small differences in entry and exit prices. 

The thing that worries me is that the real tick curve appears to be diverging as if it's a systematic error and not a random one. Maybe it's because the real spread is used in the real ticks case and minimum from the bar in the open prices and 1min OHLC cases?

Is this in line with what should be expected? I suppose looking in % terms it's not that bad. 


Comparison of equity curves with 1min OHLC and Real Tick modelling


For info this a comparison of all three models used.

Comparison of equity curves for 3 input models



Many thanks,


Ed

 

The differences depends of the strategy used.

Your results show your strategy needs to be tested on real ticks.

 
Alain Verleyen #:

The differences depends of the strategy used.

Your results show your strategy needs to be tested on real ticks.

The strategy is a very basic mean reversion strategy using Bollinger bands. Buy or sell when +/- 2 Std away from mean then opposite when crosses the mean. 0.1 lot size, no stop loss, no take profit. I will re run using provided MA EA and report back. I have a feeling that the difference is due to spread differences.

My understanding is that 1 min OHLC and open prices use the minimum spread seen within the bar.

 
EdFuk #:

The strategy is a very basic mean reversion strategy using Bollinger bands. Buy or sell when +/- 2 Std away from mean then opposite when crosses the mean. 0.1 lot size, no stop loss, no take profit. I will re run using provided MA EA and report back. I have a feeling that the difference is due to spread differences.


My understanding is that 1 min OHLC and open prices use the minimum spread seen within the bar.

I forgot to mention that the EA uses bar close price to make the decision to buy or sell which is why I was surprised at the difference. The trade is triggered when a new bar forms. 

 
EdFuk #:

I forgot to mention that the EA uses bar close price to make the decision to buy or sell which is why I was surprised at the difference. The trade is triggered when a new bar forms. 

If you are thinking your EA should give similar results with real ticks like 1 minute OHLC, then you could have to fix your EA. There is no point to post anything about MA EA. 

 
Alain Verleyen #:

If you are thinking your EA should give similar results with real ticks like 1 minute OHLC, then you could have to fix your EA. There is no point to post anything about MA EA. 

I want to see whether there is the same variability with a very basic EA to rule out the possibility it is to do with my EA. Having a quick look where the biggest differences in trades lie between the two methods it appears it's when liquidity is low and the spread is very high (close to out of hours periods) . I will do some more investigating this afternoon.