Modelling Quality - Back-test Advice

New comment
 

Hi everyone.

I am testing my EA and, with my broker's history quotes, I get a quality of about 57% on XAUUSD H1.

Surfing the net I have discovered a program called Tickstory that seems good to achieve a 99% modelling quality.

I have two questions:

  1. My EA trades only on new bars, so I actually should not need every tick, with 99% of accuracy could it change so much?
  2. Has everyone tried Tickstory and it is worth the price? Could anyone suggest me another (maybe cheaper) software with the same features?
Thank you very much!
 
Anyone?
 

If your EA trades on bar open, I suspect you won't see much difference on your backtest entries. I guess the difference will come on how it manages your trades e.g. if the smallest timeframe candle available hits both your SL and TP, which came first?

I used Tickstory a lot back when it was free (I gather you now have to pay for it). I thought it was a good product considering it was free. No idea what it's like now. 

 
honest_knave:

If your EA trades on bar open, I suspect you won't see much difference on your backtest entries. I guess the difference will come on how it manages your trades e.g. if the smallest timeframe candle available hits both your SL and TP, which came first?

I used Tickstory a lot back when it was free (I gather you now have to pay for it). I thought it was a good product considering it was free. No idea what it's like now. 

Hi honest_knave, thanks for your reply.

It is exactly what I thought. I have a lot of knowledge and work experience with financial modelling but I am not so deep with MT4's goodness parameters calculation so I was a bit confused about. Anyway I figured it out that it should not have too much difference because of my EA's structure.

Around the web a lot of people have very positive opinions about this software, maybe it is worth the price.

Thanks again :) 

 
The modeling quality is just the quality of "modeled" ticks, whatever, this means nothing, because the ticks are modeled and not real and therefore it´s totally saying nothing when you do some backtesting with them and deal inside a bar. That´s why real results vary heavily from backtest results in such cases.
New comment
Reason: