As the idea has come in Pathfinder topic, I open a new thread to make a review of old codes published here that have well passed the “real market” test. So the idea is to consider, one year after, or a few months for code with short timeframes, this non tested period, to see if they are interesting. Unfortunatly we see that most of them fail in the real market, but some succeed, and it would be very useful to list them. As a first contribution, I present the one I have in live trading now, and which is in gain for some time: it seems to come from Francesco (thanks!) but sorry I can not find the initial post. The instrument is EUR/USD with a 1h timeframe.
//-------------------------------------------------------------------------
// Code principal : ATR EURUSD francesco 1h
//-------------------------------------------------------------------------
// EUR/USD(mini) - IG MARKETS// TIME FRAME 1H// SPREAD 2.0 PIPS
DEFPARAM CumulateOrders = False
DEFPARAM FLATBEFORE =080000
DEFPARAM FLATAFTER =210000
indicator1 = Average[200](close)
indicator2 = Average[50](close)
c1 = (indicator2-indicator1)/close
c2= abs(c1)<0.012
atr = AverageTrueRange[30]
if c2 then
m = 2
profits = 70//100//130
losses = 40//40//40
// LONG
IF (abs(open-close) > (atr*m) and close > open) THEN
buy 1 CONTRACTS AT MARKET
SET STOP pLOSS losses
SET TARGET pPROFIT profits
ENDIF
//SHORT
IF (abs(open-close) > (atr*m) and close < open) THEN
sellshort 1 CONTRACTS AT MARKET
SET STOP pLOSS losses
SET TARGET pPROFIT profits
ENDIF
endif
One Year After library: (list of all strategies reported in this thread)
https://docs.google.com/spreadsheets/d/1sVas9Tz5gA94nwu-dYWgeGKY1cud7HwvqKkeqjP1aIo/edit?usp=sharing
Great idea @Aloysius!
I checked Francesco78 EURUSD Systems, but couldn’t find the code above; maybe Francesco will be along at some stage and tell us?
Anyway I reworked the values over 100k bars using the WF tool and selecting values showing against more recent OOS periods. Readers can tell me if they think v1.1 below is now curve fitted?
//-------------------------------------------------------------------------
// Code principal : FrancATR EURUSD 1H v1.1
//-------------------------------------------------------------------------
// EUR/USD(mini) - IG MARKETS// TIME FRAME 1H// SPREAD 2.0 PIPS
DEFPARAM CumulateOrders = False
DEFPARAM FLATBEFORE =080000
DEFPARAM FLATAFTER =210000
indicator1 = Average[160](close)//200
indicator2 = Average[15](close)//50
c1 = (indicator2-indicator1)/close
c2= abs(c1)< 0.012 //0.012
atr = AverageTrueRange[15]//30
if c2 then
m = 2
profits = 60 //40??//70//100//130
losses = 30//40//40
// LONG
IF (abs(open-close) > (atr*m) and close > open) THEN
buy 1 CONTRACTS AT MARKET
SET STOP pLOSS losses
SET TARGET pPROFIT profits
ENDIF
//SHORT
IF (abs(open-close) > (atr*m) and close < open) THEN
sellshort 1 CONTRACTS AT MARKET
SET STOP pLOSS losses
SET TARGET pPROFIT profits
ENDIF
endif
Hoping this Thread gets loads of entries over time then we will need a way to locate Systems so I hope you won’t mind Aloysius , but I’ve raised http://One Year After as an easy way to link direct to posts.
The link is available in Attachments at the top of each Page.
Better ideas for layout welcomed, anybody edit yourself if you want?
GraHal
Thanks Grahal for the idea of the link for a document, where anybody will be able to report his good codes that succeed in real market. About the optimization you propose, it seems useful, and I admit that my live market trading for this code is rather short (one month), so backtests are a good information too. But I think we should be conscious, in this thread, that we have to restrain ourselves from optimization, ortherwise we risk to fall again into some illusions about codes, and the principle of this thread is to avoid that. Or at least, we can distinguish, in the document, codes that have been optimized in backtest and not ran in live, and those that have made their proofs in live trading. As you did, the “x.0” version could be the one that have run in live, and the “x.1″, x.2”, the ones that have been optimized. Anyway, to preserve the spirit of this thread, optimization should concern only small ultime settings. And we are all well informed, responsible and free to choose the version we prefer.
Yes good ideas again, Aloysius.
Maybe I should not have optimised as you say, it is against the spirit of this Thread?
Let’s have more of a think during the course of today if this Thread is for pure unadulterated – from original – code? I can always ask Mods to delete my post?
Alternatively I can write a note in the Comments column in the database, e.g. original against v1.0 and optimised against v1.* etc.
What do others think would be best?
It would be great if when posting we could have an image of the first backtest results when the strategy was first created and then a second image of the performance since then. That would be great IS and OOS results for comparison. Hopefully @Francesco78 will come along at some point and direct us to the original thread for this first strategy.
I agree that we should resist the temptation to improve by renewed optimization. I’d rather see a slightly sloppy strategy result with IS and OOS that make profit rather than a newly optimized result that now requires OOS testing to prove to be beneficial.
I’d rather see a slightly sloppy strategy result with IS and OOS that make profit rather than a newly optimized result that now requires OOS testing to prove to be beneficial.
or optimize only on 70% of data history to get an OOS testing. I know that’s backtest, but job is almost already made that way… 😉
or optimize only on 70% of data history to get an OOS testing. I know that’s backtest, but job is almost already made that way…
Yes I suppose that as long as no major changes have been made and it is just a little tweak of one or two variables up or down a few points then that is really just a slight fine tune of the strategy and the 70/30 test will confirm that we have not broken the strategy. The problem for us is with strategies on shorter time frames because one year later we may have lost part or all of the original history that the strategy was tested on so we cannot confirm that it still works during this period.
http://One Year After
Mods, sorry to bother you, but please could you fix the link above?
The link in Attachment No 3 works good so you could copy from that.
Thank You
GraHal
Hi Guys, thank you this is a great idea!
I dont really remember that particular code, however it looks like a variation on the theme of this post
EURUSD volatility breakout strategy
with an additional filter represented by the differential of 2 moving average.
I really think there is value in archiving the oos performance of the strategies posted in the forum. I hope this threads will prolificate!
Attached is the picture with the performance this last year. of the identical code
