Backtesting question

TambaTrader · Post by **TambaTrader** » Mon Jun 15, 2015 11:30 am

Hello,

I am trying to use the Empty4 backtester to work out some settings for an EA. It initially seems like it has worked with five years data and I have results but the modelling quality says n/a and there are mismatched chart errors.

What does that mean and how do I get rid of them? Do I need to get rid of them? I am only trying to see how often I get a big loss with certain settings.

Shona

P.S. would also be good if someone could tell me how I just copy and paste a screen print into a post on this forum - I tried to upload the whole report because I don't know how to just put a picture here, but it won't let me as it says it's a possible attack vector.

Nufty · Post by **Nufty** » Mon Jun 15, 2015 1:14 pm

The history in Empty4 is ordinary to say the least. It doesnt give a true reflection of the days trading

Here is how they work it out (if you actually care - I know I dont lol)

ModellingQuality =
((0.25*(StartGen-StartBar) +
0.5 *(StartGenM1-StartGen) +
0.9 *(HistoryTotal-StartGenM1)) / (HistoryTotal-StartBar))*100%;

Have a look at this

https://www.youtube.com/watch?v=R-rMOen4UHg

It will give you some hints on improving your backtest modeling quality, by introducing some more ticks to your data.

Personally I am not interested in the results of a backtest - I only use them to see of EAs get into and out of trades as expected - I then run them in forward test to check what trouble they get into. This data forgets important trading stuff like spread widening, slippage and commissions etc

One of the major failings I have of the back test - is it wont allow me to run ALL my pairs and show me what they would have resulted in. Where the forward test will. YES, thats a long time to wait for results!!

as for pics .. I cant help you - I cant get an image to work either

N

Post by **Radar** » Thu Jun 18, 2015 9:36 am

Hi Shona,

TambaTrader » Mon Jun 15, 2015 10:00 pm wrote:Hello,

I am trying to use the Empty4 backtester to work out some settings for an EA. It initially seems like it has worked with five years data and I have results but the modelling quality says n/a and there are mismatched chart errors.

What does that mean and how do I get rid of them? Do I need to get rid of them? I am only trying to see how often I get a big loss with certain settings.

The only thing Empty4's slacktester is good for is testing code, and even then, some errors slip by

The "Every Tick" setting is a lie... If you're testing on an H1 chart, it uses the M1 open & close data as "ticks" (I'm not sure if it uses the high and low as well, as I haven't coded and tested in quite a while)... If the M1 & H1 close prices are different, you get the chart mis-matched error.

I use Tickstory Lite to d/l tick data from DukasCopy, then export it (as all timeframes) to a copy of Empty4 that I use for coding/testing. I usually get a quality rating of 99% with that.

P.S. would also be good if someone could tell me how I just copy and paste a screen print into a post on this forum - I tried to upload the whole report because I don't know how to just put a picture here, but it won't let me as it says it's a possible attack vector.

Just paste your screenshot into Paint (or GIMP if you're using Linux), crop out what you don't want, and save... Upload the saved file, and you're done

Hope that helps

Take care,

&

Have fun!

Radar =8^)

TambaTrader · Post by **TambaTrader** » Sun Jun 21, 2015 7:45 pm

Ok I'm going to try and upload my backtesting picture. It may or may not work.

If it works you will see what I mean, the modelling quality is n/a and it has 1737 mismatched errors. So it sounds like you are saying that if my data that I uploaded was minute data (it was) but when I ran the backtesting I did it on the hour period (think I did) then it is possible some of my uploaded data overlapped with the history data that was already in Empty4 and that was inconsistent causing errors.

In which case I can probably fix it by running it with minute as a period, or maybe by making sure I delete all of the history data out, take it offline and then reuplaod my minute data. It should be fine to do that for my strategy.

It seems universally agreed that Empty4 backtester is rubbish. I am using it to try and test different settings for ALR (see ALR thread) over a long time period. ALR looks successful over the short term as it rarely loses but it's the long term that is important as when it does lose losses are big. I will treat the results with some skepticism. It is just so quick to do it in Empty4 compared to anything else I've tried.

Thanks for your help both.

Post by **SteveHopwood** » Sat Jul 18, 2015 8:25 pm

Shona, I am going to post here only once and will not interfere with you again so long as you do not introduce backtesting elsewere in this forum.

You can accept from experienced members of SHF that backtesting in Crapergy Tester is a waste of time (and only an idiot with a forum death-wish ignores them) or you can continue to bang your head against the wall of your ignorance. Up to you. Explanation of why backtesting in Crapergy Tester is a waste of time at
http://www.stevehopwoodforex.com/phpBB3 ... =28&t=4020

TambaTrader · Post by **TambaTrader** » Sat Jul 18, 2015 9:28 pm

You can use it to see what won't work though can't you? Do you agree with that? I am forward testing my strategies but I've written off lots of settings because you can see it gets big losing streaks in the backtester.

TambaTrader · Post by **TambaTrader** » Sat Jul 18, 2015 10:03 pm

P.s. just noticed in your linked post you said discussion of backtesting is informally banned. Therefore I won't post any more about it. Apologies.

Post by **SteveHopwood** » Sat Jul 18, 2015 10:14 pm

TambaTrader » Sat Jul 18, 2015 10:03 pm wrote:P.s. just noticed in your linked post you said discussion of backtesting is informally banned. Therefore I won't post any more about it. Apologies.

Sorry, I know I said I would not post here again but this means I have to.

Feel free to post about backtesting in your thread here as much as you want. I will take no action against you. Just do not export your thoughts further, ok? Continue to discuss them as much as you want here.

I am removing my subscription from this thread so my attention will no longer be dawn to it and I will no longer be seen as a threat here. It is now up to you whether you accept that backtesting in CrapergyTester is a waste of time or not.

Carry on having fun at shf. The rest of us are.