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 😉
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.
Maybe, but in that case I do something wrong when i op.
After I filled in my variables for supertrend (1-10 with step 1) then i probacktest my system.
In this case, I choose PRT to find the best variables until 1 January 2019.
When I get the best values for that period, I let a new probacktest up to today’s date. If it turns out that the algo goes well even after my optimization up to today’s date, I think I succeeded with the optimization.
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.
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:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
DEFPARAMCumulateOrders=False
defparamflatbefore=090000
defparamflatafter=173000
indicator1=ExponentialAverage[46](close)
indicator2=ExponentialAverage[10](close)
c1=(indicator1>indicator2[1])
indicator3=AroonUp[20]
indicator4=Aroondown[76]
c2=(indicator3CROSSES OVERindicator4)
indicator100=close
indicator101=Supertrend[7,1]
c100=indicator100>indicator101
IFc1andc2andc100THEN
BUY1CONTRACTsATMARKET
ENDIF
set stopploss20
set targetpprofit10
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 you should also use many iterations (more than just one = more OOS/IS periods), to get an idea on how and when the strategy benefit from a new optimization process.
You are very right in what you write. I found some of the code on this forum, may be true that it is Barney’s code. What made it more interesting than Barney’s code is that I found the following:
BuyPrice = HIGH + 4 * PipSize
SellPrice = LOW-4 * PipSize
On page 122 I think it was here on the forum that I found that interesting part.. Then the stops are different too.
Also true that it is very difficult with 1-5 min algo because you have so little data.
Everyone here at the forum probably agrees that it is very difficult to get an algo that makes profit year after year.
There are “rules” for how to optimize and what timeframe you should use to succeed.
But if you look at all the 100’s of algo found here on the library and on the forum, I won’t find anyone making a profit (if you ignore the ones that are newly made that are too early to evaluate yet)
Does this mean that it is basically impossible to be a regular private person to get an ago who makes a profit year after year?
There are many threads about it already where the question reads “is anyone making a profit on their algos?”
I myself am too bad at coding.
Therefore, usually find different parts of codes in the library and on the forum and modify them a bit and then I optimize them.
I simply get to take a course to learn how to code and see if there is any improvement to my “bad” algos that I post here on forums 🙂
I think grahal is keeping a trackrecord of forum-algos that actually have produced positive numbers. And if u check out some of the “Does anyone make any money here??” posts, there are plenty of replies that “yes indeed there are people who make profitable strategies”.
But its not easy. And thats why when you finally make one, you dont really want to give up your “secret sauce”. Even though it might be befinicial, and people can contribute to it, its kind of like giving up a patent for a really cool product youve spent hours/days/weeks/months/years making. So thats why you wont find so many “plug and play” profitable systems here. If you pick up a book or two, there are some containing many profitable systems. And if you just keep on working on making systems, u will probably eventually make your own profitable systems.
Yes it does take 1000 ideas to come up with one that might work. Yes it does take many long nights, and if you think thats boring, well then this is not for you.
However if you can find a passion in this subject, it can be very fun and also very rewarding.
We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will assume that you are happy with it.
.
