Backtesting EA and closing trade at ma

Hi...chasing some wisdom

I'm building an EA for the M1 EUR/USD with Mt4 and having massive difficulty with INCONSISTENT back testing.

I can work on it for ages with progressively improving results as I tweak it, and get eg $100,000 nett from a $10k account testing back to 1.1.10.

Five minutes later it tests a shocking graph and only earns $20k or less.

I've re-loaded my mt4 platform. I've re-loaded the data in the history center. etc etc. ...now run out of ideas!

Really desperate and frustrated, because its very hard to improve or check an EA if the results are not reliable!

Any suggestions would be very appreciated.

(Also, on a different note. If anyone could help me with some code to close a "sell" if it goes into negative and the price crosses upwards above the ma on the M5 chart. This "sell" trade would have been initiated below the MA. Other trades that are initiated above the MA would be closed either in profit or with a normal stop loss)

Thanks heaps in anyone can help@

cb

maybe the second backtest performed with a worse spread? post your ea so we can debug it together

yeah that'd be one of the likely reasons.

i've backtested an ea that scored over 3mil from a $500 startup, then on another occasion only got to 70k, and so on. I'm also of the understanding that each backtest simulates each bars volume of ticks randomly, so depending on what maths you are doing and how the random ticks play out you can have times when the random nature of each bars ticks plays in your favour whereas other times not.

but it's likely more the spread issue, as the spread for the backtest is taken from the currently live chart and used throughout the backtest, so if you backtest on the weekend when the broker is closed and the spread has been left widened then you'll get a dog of a result compared to during the week at a time when the live chart has a tight spread.

Yeah, the only time I notice any inconsistency from one back test to the next is when the spread changes. So, that's usually the first thing I look for anymore when the results change and I know they shouldn't.

If you are scalping, you need to use tick data and consistent spreads. Backtesting isn't very reliable at best, although your experiences is more extreme that mine and others. I have seen variations but usually in the order of less than 10 percent. I accept the shortcomings of Strategy Tester don't let it bother me.

It sounds to me you need a better backtester or to skip any more backtesting and get on with live testing. Live testing is much more realistic anyway.

Edit: forgot to provide an example of how to use an MA as a stop loss. I can't remember the details of the coding, but I know it works from the feedback I continue to get. Use the code in post #10

http://www.forexfactory.com/showthread.php?t=204438

Hi Guys,

Thanks for your comments. Spread sounds like the big issue. Probably especially as I'm in Australia and that is the quiet time for the eur etc, so spread is probably more varied.

Also makes sense now why the weekends are so difficult to do any EA building.

Are there ways to fix the spread for back testing purposes?

As far as posting the EA here...good idea, but I figure any really good EA would become so popular that it would be self defeating as it would begin to change the market behavior that it is designed to take advantage off! Not that mine will probably end up much good anyway.

I figure its best for this kind of forum to help each other out in building stuff so that at the end of the day there are variations in the EA outcomes and market characteristics are reinforced, rather than altered.

I'm still chasing some way of coding to close a trade if it crosses the MA, so if anyone is kind enough to offer input....greatly appreciated.

Cheers

cb

Unless you are wanting to demostrate your skills or acquire some bragging rights, don't post your EA anywhere on the internet. You will just attract a lot of frivolous mod requests and requests for help in how to do MT4 basics like how to run an EA. Then you get the people who want to steal your code for their own purposes including commercializing it. Although I think the people who commercialize code usually select terrible systems so maybe yours wouldn't be of interest.

Just some advice from someone who has made the mistake of posting EAs.

Thanks for your wisdom on that one CodeMeister.... Sounds like a lesson learned the hard way...

cb