What happens if you think you have struck gold? - page 5

To add comments, please log in or register
Peter
120
Peter  

Have started again from Y2000

Up to 5m after 6 months

Recommendations for further testing?

William Roeder
20233
William Roeder  
  1. Your modeling quality is 25%. DL some real (M1) history and use a converter to create all TFs used.
  2. What spread are you using? Broker's current or setting a value like 20 (2pips)
Ian Venner
2397
Ian Venner  

My recommendation is, forget about further backtesting. Make a copy of it with the trade functions replaced with trade simulation functions. They are easy to write to keep a running total of the profits that would have been accumulated if the trades had been opened and closed for real, and some others if you need to simulate calls to OrderOpenPrice() etc, Then attach it to a LIVE price feed and leave it to run for at least several weeks, and see if it appears to behave the same way as it did on backtesting.

Then when you have those results, clear your chart history (save it first), let mt4 download new chart history from your broker, then run the EA on a backtest over the same period as your live test. If the results are the same, or close to the same it would suggest the rest of your backtests, and the brokers chart history you tested on are probably accurate. If that is the case you might make the capital you need by signing up to be a signals provider.

Peter
120
Peter  

2000 - 2002

Peter
120
Peter  
WHRoeder:
  1. Your modeling quality is 25%. DL some real (M1) history and use a converter to create all TFs used.
  2. What spread are you using? Broker's current or setting a value like 20 (2pips)


Thank you for your comment, please forgive my ignorance,

I spoke to Alpari and asked about historic data. They said that the data I downloaded (M1) is their real history. The EA runs on M1 - How and why would I convert? Help appreciated here

The spread is set at 2 pips

Peter
120
Peter  
SDC:

My recommendation is, forget about further backtesting. Make a copy of it with the trade functions replaced with trade simulation functions. They are easy to write to keep a running total of the profits that would have been accumulated if the trades had been opened and closed for real, and some others if you need to simulate calls to OrderOpenPrice() etc, Then attach it to a LIVE price feed and leave it to run for at least several weeks, and see if it appears to behave the same way as it did on backtesting.

Then when you have those results, clear your chart history (save it first), let mt4 download new chart history from your broker, then run the EA on a backtest over the same period as your live test. If the results are the same, or close to the same it would suggest the rest of your backtests, and the brokers chart history you tested on are probably accurate. If that is the case you might make the capital you need by signing up to be a signals provider.


Thanks SDC

As you can see below I have tried further back testing, I'm leaving it to run just for fun. Putting money to one side for a minute, being able to write some code that behaves in this way is an achievement for me, it is like trying to crack a puzzle and I have already gained a great deal of satisfaction from being able to look at the chart and see it going up all the way.

I will try and implement your simulation

Would you recommend writing buy, sell and close and stoploss to a text file? of just leaving arrows on the graph. Not sure of the best way to implement.

Because of the way it is built, it should work on other pairs, I'm hoping this to be the case and will try after current back test complete.

Peter
120
Peter  
WHRoeder:
  1. Your modeling quality is 25%. DL some real (M1) history and use a converter to create all TFs used.
  2. What spread are you using? Broker's current or setting a value like 20 (2pips)


Modelling quality won't get better than 25% on 1 minute bars.

Please see

https://www.mql5.com/en/articles/1513

Peter
120
Peter  
On the back testing the Biggest loss so far is £504.40 Biggest Profit is £2392.80
Simon Gniadkowski
17214
Simon Gniadkowski  
tootrue:
On the back testing the Biggest loss so far is £504.40 Biggest Profit is £2392.80
What is 100 * (average loss / (average loss + average win) ) vs Win Rate ?
Peter
120
Peter  

Win Rate 78.61% for Short positions and 88.57% for long positions this is taken from the Strategy Tester report Short Positions won(%) and Long Positions won(%)

Average 83.59

I make it 128.9219 vs 83.59

To add comments, please log in or register