Strategy Results, Good? Bad? Okay?
-
the problem with testing out large sums of money is that it lies to you with your drawdown actual numbers. Just as you indicated, you would nearly need to start the account with $5000 just to maintain...that is some serious drawdown my friend.
As to the test only lasting 4 seconds...I can say for sure something is not right if its testing that fast. Are you on MT5 testing every real pip?
I'm not trying to discourage you, I'm just getting you to realize some things now that will save you time in the future when you build EA's. I've built some monsters in history....but when I moved them to MT5 from MT4 and have achieved 100% history (which I still think is wrong), I realized my EA's were not as good as I thought...even though MT4 said the history was 90% efficient. There are several other factors and nuances I have learned on the way, but that advice would be a good start.
-
@seb-0 I was just referring to some finance theory... Sp500 should be the ideal investment (you should put 100% of your money there), but its even better if you have 2 investments that dont correlate with each other (sp500 and your EA probably dont).
-
@jsauter86
5 seconds is about the simple test, not the parameters optimization, to be sure we are talking about the same thing.
And yes I am running it on MT5 real pip. The 3% bad quality is on a really short period out of 21 years so..
Also I am on FxPro as a broker, they are supposed to be specialized in this field, must be serious.
Finally the draw down is relative, which means I can lower the %DD by decreasing the lots, or by increasing my balance.
@miki @roar
Yes I was thinking that if possible, it would be better to run the EA on 2 uncorrelated instruments, rather than correlated ones. I am interested in the Sp500 ideal investment, however I tested the EA on the Sp500 and it behaves far as well as on the EURUSD.
When you are saying "the Sp500 and my EA are not correlated", do you mean 1.the behave of my EA on the Sp500 is uncorrelated to 2.the behave of my EA on any other instrument ?
Thank you tho -
@seb-0 said
When you are saying "the Sp500 and my EA are not correlated", do you mean 1.the behave of my EA on the Sp500 is uncorrelated to 2.the behave of my EA on any other instrument ?
Thank you thoI meant just going long on sp500 and then using your EA to trade forex, for example. It is unlikely that a negative day on sp500 always happens same time as negative day with your EA trading EURUSD, for example. The return series are uncorrelated. In theory, having many profitable but uncorrelated investments will smooth out negative shokcs in the whole portfolio, giving you better sharpe ratio.
-
@roar
Behavior*
I see thank you for the advice! -
Something just doesn't make sense for me for even a single scenario to be run. Almost as if its skipping a lot of data. I create very simple rules for EA's myself and they take significantly longer to take than a few seconds. For me to have a scenario last a few seconds I would have to test no longer than a month.
-
@jsauter86
Are you running the EA on "open prices only" ?
-
@seb-0 every tick based on real ticks
-
@seb-0 how does the results look with every real tick modeling?
-
@jsauter86
You are right about the fact that I should test the EA on real ticks instead, it gives more accurate results.
Ill tell you. -
@seb-0 i would expect vastly different results
-
@jsauter86
Nevermind, the EA is built to act only once per bar, running it on every tick doesn't change the final result at all. -
@seb-0 It should be different
I too use the once per bar function a lot. But when you use open prices only, that means it doesn't calculate closing prices. At least that is my understanding.
-
alright, updated EA based on price and time.
took @roar advice, incorporated sells and buys as well as a longer tested time frame...
-
i dont know if theres much I can do for history data. being that im going as far back as 2010...it is what it is. with 100% accuracy on a shorter time frame, it is still in the high 60 percentile win rate with a PF of higher than 1.25.
Anyone have a good free, or cheap resource to run this EA on a server that doesn't go down? Or anyone know the ability to load and run EA's from a mobile platform?
-
@jsauter86
The brut force optimizer can turn a random EA into a profitable one(especially if there are a lot of parameters) on a backtest.
But since the optimizer knows the data in advance it is cheating. The way you can verify either the robot WILL be profitable(with a certain probability)or not, is by running it on a period it hasn't been optimized on.
I just know some brokers offer good quality data -
@seb-0 not sure where you heard that before, but its total bs. If someone thinks they can put up a few blocks, allow the computer to change a bunch a variables, pay some $$ for the cloud processing power on MT5...I still think you would not be able to produce a good EA. A better understanding of price VS time would be more worth while.
I am just looking for somewhere to run my EA on a demo account because I don't have a dedicated computer that will be online.
-
@jsauter86 its not bs.
Imagine as a simplest example: an EA that checks if today is a certain date -> then buy.
When you run the optimizer using dates as the input parameter, your computer will surely find the best dates. The EA will look awesome in backtest, totally useless in real trading.Using many indicators (as we all do) to find the best dates is more complicated approach, but it has the same problem - statistical curve fitting by the optimization algorithm.
-
I think thats just another way to say statistics.
If statistically Mondays and Fridays are very volatile days, I want to avoid them due to unpredictability. That's an optimization parameter. But either way, that makes no matter to me or this EA example, I think I'm confused when he said that to begin with.
The rules for this EA are based solely on a single 4H time price VS price of day. 0 indicators. If I could get it to close the trades on gold now, it would be profitable there too. Just asking for a place to forward test this EA because I don't have a place with a constant computer to run it
-
@jsauter86
what i experienced so far is that some markets fit to different ea's the main thing is to find the right market for your ea or make your ea determine if it actually fits the market right now...
