Is this a Grail??

 

Hello everybody,

I have been working in this area for past few months and relatively new to this FOREX area. I see very wonderful topics in this forum and very exciting ideas.

I have been working on my EA on which I posted some back testing results sometime back. I had problems with it regarding optimization and was not satisfied about how much optimization is needed to make it profitable on the sample data out of optimized period. I came across some articles posted on NN and went through some of them. I was skeptical about using NN but when I went through and I stated experimenting I found really fascinating results with my EA. I tried to adapt simple NN and use it with my EA to get the better results. What I got is excellent results on out of sample data too and it really works. I am posting back testing results on my EA. I dont know about how you classify any EA as a Grail but does this sound like it? Please comment on the performance of this EA in terms of draw downs, consecutive wins and losses, profit trade percentage and so on.










Symbol EURUSD (Euro vs US Dollar) Period 1 Hour (H1) 2009.01.02 06:00 - 2009.09.25 05:00 (2009.01.01 - 2009.09.26) Model

Every tick (the most precise method based on all available least timeframes)




Bars in test 5421 Ticks modelled 9161331 Modelling quality 90.00% Mismatched charts errors 0 Initial deposit 10000.00 Total net profit 31208198.28 Gross profit 45656323.99 Gross loss -14448125.71 Profit factor 3.16 Expected payoff 134518.10 Absolute drawdown 841.00 Maximal drawdown 5974000.00 (39.57%) Relative drawdown 47.29% (13655.55) Total trades 232 Short positions (won %) 63 (74.60%) Long positions (won %) 169 (82.25%) Profit trades (% of total) 186 (80.17%) Loss trades (% of total) 46 (19.83%) Largest profit trade 3908500.00 loss trade -1000000.00 Average profit trade 245464.11 loss trade -314089.69 Maximum consecutive wins (profit in money) 19 (4592536.75) consecutive losses (loss in money) 2 (-1999500.00) Maximal consecutive profit (count of wins) 7563000.00 (6) consecutive loss (count of losses) -1999500.00 (2) Average consecutive wins 4 consecutive losses 1

 

Sorry not a Grail - maybe you have something but its very high risk :(

The drawdown is high in all respects - the account nearly died around trade 126...

Check results of starting the backtest 1-n weeks earlier and 1-n weeks later

I think you have ended up over-optimising for this period...

Good Luck and keep at it :)

-BB-

 

There's no way for you to be able to tell if the EA almost died around trade 126. Back when trade 126 was new, the line would have been nowhere near horizontal due to the increasing lots. So it's impossible to tell whether it was close to 0 or still way in profit at that point.

.

The EA looks good compared to many that I have seen. Although I wouldn't trade it with much money because of the alarming drawdown, that's the first place I'd look to improve. On another note, try running the EA for a different period with no increasing lots. Even though it's exciting to see those huge numbers off your creation, it only hinders your testing accuracy. Without increasing lots, the performance will be linear. This will let you be able to see much more clearly where it went down, why it went down, what you can do to stop it from going down in those cases, etc..

.

Bigger numbers do not necessarily mean a better expert.

Jon

 
Archael:

There's no way for you to be able to tell if the EA almost died around trade 126. Back when trade 126 was new, the line would have been nowhere near horizontal due to the increasing lots. So it's impossible to tell whether it was close to 0 or still way in profit at that point.

.

The EA looks good compared to many that I have seen. Although I wouldn't trade it with much money because of the alarming drawdown, that's the first place I'd look to improve. On another note, try running the EA for a different period with no increasing lots. Even though it's exciting to see those huge numbers off your creation, it only hinders your testing accuracy. Without increasing lots, the performance will be linear. This will let you be able to see much more clearly where it went down, why it went down, what you can do to stop it from going down in those cases, etc..

.

Bigger numbers do not necessarily mean a better expert.

Jon

Yes, it was nowhere near dieing. Here is the order data around that period:

All the orders were profitable and the balance was 538160.21 at close of 126 order. I tried to run with maximum size to see if it stop-out but it went through very well.

The important thing I tried out here was that it was still profitable on data sample out of training data and that is good thing about it.

Wish I had draw-downs low. May be need some more work to maintain low draw downs but still draw-downs are below 50% which may not be that bad.


246 2009.03.31 21:00 buy 124 80.40 1.3285 1.3185 1.3333


247 2009.04.01 02:54 s/l 124 80.40 1.3185 1.3185 1.3333 -80359.80 470336.21


248 2009.04.01 03:00 buy 125 68.20 1.3181 1.3081 1.3229


249 2009.04.01 03:27 t/p 125 68.20 1.3229 1.3081 1.3229 32736.00 503072.21

250 2009.04.01 04:00 buy 126 73.10 1.3223 1.3123 1.3271


251 2009.04.01 13:17 t/p 126 73.10 1.3271 1.3123 1.3271 35088.00 538160.21

252 2009.04.01 14:00 buy 127 77.90 1.3253 1.3153 1.3301


253 2009.04.02 08:21 t/p 127 77.90 1.3301 1.3153 1.3301 37508.85 575669.06


 
What is NN?...
 

Can you direct me to the NN site you used as a starting point?

sn

 

neural networks maybe?

https://www.mql5.com/en/articles/1565

 

As far as im concerened less than 50% drawdown is good. Most of the EA's ive seen have drawdowns far bigger than that and in any case you could reduce that drawdown just by lowering the trading lot size.

 
19730719:

neural networks maybe?

https://www.mql5.com/en/articles/1565


I couldn't get solid results from this - but I readily admit it was new territory for me... That is why I asked if there was something else out there my grey matter could absorb.

sn