Im getting same results like u, and the numbers do look decent, but looking at the code this just seems super optimized. A supertrend 6,1? Aroon 20 and 76?

I think the sample size is way too small.
Have u done a walk forward? – I have, i checked only the Aroon numbers, and result was -27% WF. It just broke.
Have u checked other similar markets? – I have, NQ, WS, US 500, France, Spain.. Results are not good.

Another red flag is: No exit strategy. Meaning u hit stop loss or target every time. And these variables have also been optimized. Seems sketchy. If it works, im doubting its gonna work for a long time.. only one way to find out tho 😉

This are my 2 cents about this strategy.

If we look at the code it is a breakout strategy with 3 filtres (Ema, aroon, supertrend). One filtre could simple the code and avoir over optimization.

The High Windows rate is due to the narrowed tp and the sl 4 Times farther.

The problem is just that Jan-Aug 2019 is not alot of data. For example on 1h algos i like to optimize on 2006 -> 2013, and keep 2013 -> 2019 out of sample.

Now thats 6 years of  Out of sample data.

You have 8 months (be it with very choppy/trendy flat price) its still so little. I would love to see this algo from 2007 -> 2019 for example. I think its just way too overoptimized.

Another thing is the walk forward test. It sucks. And i only tested 2 of your variables. You have like 4 more you could throw in there. WF is pretty important, everyone agrees on that. (including pro’s)

Ive made 100’s of these types of systems that rely 100% on the variables thats put into it.  Its not hard to create, just 1 filter and 1 trigger and optimize that bad boy. But the problem is when it dosnt pass the WF test, and it looks pretty bad in other similar markets, it dosnt seem robust.

If you algo dosnt work on other markets, well this is more like a Red flag, rather than a “DONT EVER TRADE IT!!” warning.. In my eyes an algos EQ curve should look at least similar in other markets. if its just total shit in all other markets, theres a huge chance its curvefit for that 1 spesific market.

ive listened to tons of podcasts and read books where the experts disagrees. Some want the results to be so good in other markets, they can basicly run it in the other markets. Other experts dont give a shit how it looks in other markets because in their eyes, every market is unique and operates in its own way..

My opinion is that the better it looks in other similar markets: The more robust your algo is. The more robust = longer life b4 loosing edge, and less curvefit.

Edit: Also wanna add that you are on the correct path, its just that these types of systems, i think most of us on the forum with a few years of experience  have made very similar codes to this one.

Editedit: I thought ive seen this code before, in this forum.

Barney posted this here (scroll down a bit): https://www.prorealcode.com/topic/dax-5-min-longshort/

Its more or less exactly the same.. The system is more or less negative since the date he posted it, unless im mistaken here.. ?

Barneys code:

DEFPARAM CumulateOrders=False
defparam flatbefore = 090000
defparam flatafter = 173000

indicator1 = ExponentialAverage[46](close)
indicator2 = ExponentialAverage[10](close)
c1 = (indicator1 > indicator2[1])
indicator3 = AroonUp[20]
indicator4 = Aroondown[76]
c2 = (indicator3 CROSSES OVER indicator4)

indicator100 = close
indicator101 = Supertrend[7,1]
c100 = indicator100 > indicator101

IF c1 and c2 and c100 THEN
BUY 1 CONTRACTs AT MARKET
ENDIF

set stop ploss 20
set target pprofit 10

Also maybe T-trader’s code, not Barneys? not sure.

And he said:

“Hey guys,

Me and Barney are friends.

Let´s talk about the last code.

In may this year I optimized it on 70% of the avalible data on 100k and since may I´ve been running it live. Her´s a picture of how it looks and also the time configuration.

Despair, Thanks for your 200k backtest but it dosen´t look quite right if you compere to Barneys 100k test, or am I wrong? ”

And again, its negative today.