Mean Reverting Strategy on AUDCAD on h1 timeframe

Viewing 15 posts - 1 through 15 (of 37 total)

1 2 3 →

Author

Posts
03/05/2018 at 11:27 AM #64053 quote
Francesco78
Participant

Master
This is a mean reverting strategy that I currently run live on AUDCAD.

Enjoy, comment and improve 🙂
```
//-------------------------------------------------------------------------
// Main code : audcad_meanrevert_v2
//-------------------------------------------------------------------------
defparam cumulateorders = false

n = 1
m = 3
l = 90


cs= summation[n](close>open) = n
cs = cs  and close>bollingerup[20](close)  and close>dlow(2)
cl= summation[m](close<open) = m
cl = cl   and close<bollingerdown[20](close) and close<Dhigh(2)


size = 2
//entry conditions
if cs   then
sellshort size contract at market
endif

if cl  then
buy size contract at market
endif

//exit conditions
if shortonmarket and close<average[l](close) then
exitshort at market
endif


if longonmarket and close>average[l](close) then
sell at market
endif
```
audcadmrh-1519740014c4pl8.png audcad_mr_h1_stat.png
03/05/2018 at 11:37 AM #64416 quote

Nicolas
Keymaster

Master

Thank you Francesco, going back to mid 2010, the strategy is still quite good in tick by tick backtesting. Did you try some WalkForward analysis since you seem to have optimized the variables? I know there are only 3 of them, but should be interesting to know how it performs with WF. I launched one just now, I’ll get back with results.

03/05/2018 at 12:10 PM #64424 quote
juanj
Participant

Master
I threw my Heuristics Algorithm in there to optimize the average[l] variable;

Now results are looking good on other AUD pairs like AUDUSD as well.
```
//-------------------------------------------------------------------------
//-------------------------------------------------------------------------
// Main code : audcad_meanrevert_v2
//-------------------------------------------------------------------------
defparam cumulateorders = false

// Heuristics Algorithm Start

once OPT = 1

If OPT < 2 and onmarket then
OPT = 0
ElsIf countofposition = 0 Then
OPT = 1
EndIF

If OPT = 0 Then
optimize = optimize + 1
OPT = 2
EndIf

////////////////////////////////////////

once periods = 100
Increment = 14
MaxIncrement = 4
Reps = 20 //Number of bars to use for analysis
once PIncPos = 1 //Positive Increment Position
once NIncPos = 1 //Neative Increment Position
once Optimize = 0 ////Initialize Heuristicks Engine Counter (Must be Incremented at Position Start or Exit)
once Mode = 1 //Switches between negative and positive increments
once WinCountB = 0 //Initialize Best Win Count
once StratAvgB = 0 //Initialize Best Avg Strategy Profit

If Optimize = Reps Then
WinCountA = 0 //Initialize current Win Count
StratAvgA = 0 //Initialize current Avg Strategy Profit

For i = 1 to Reps Do
If positionperf(i) > 0 Then
WinCountA = WinCountA + 1 //Increment Current WinCount
EndIf
StratAvgA = StratAvgA + positionperf(i)
Next
StratAvgA = StratAvgA/Reps //Calculate Current Avg Strategy Profit

If StratAvgA >= StratAvgB Then
StratAvgB = StratAvgA //Update Best Strategy Profit
BestA = Periods
EndIf

If WinCountA >= WinCountB Then
WinCountB = WinCountA  //Update Best Win Count
BestB = Periods
EndIf

If WinCountA > WinCountB and StratAvgA > StratAvgB Then
Mode = 0
ElsIf WinCountA < WinCountB and StratAvgA < StratAvgB and Mode = 1 Then
Periods = Periods - (Increment*NIncPos)
NIncPos = NIncPos + 1
Mode = 2
ElsIf WinCountA >= WinCountB or StratAvgA >= StratAvgB and Mode = 1 Then
Periods = Periods + (Increment*PIncPos)
PIncPos = PIncPos + 1
Mode = 1
ElsIf WinCountA < WinCountB and StratAvgA < StratAvgB and Mode = 2 Then
Periods = Periods + (Increment*PIncPos)
PIncPos = PIncPos + 1
Mode = 1
ElsIf WinCountA > WinCountB or StratAvgA > StratAvgB and Mode = 2 Then
Periods = Periods - (Increment*NIncPos)
NIncPos = NIncPos + 1
Mode = 2
EndIf

If NIncPos > MaxIncrement or PIncPos > MaxIncrement Then
If BestA = BestB Then
Periods = BestA
Else
Periods = (BestA+BestB)/2
EndIf
NIncPos = 1
PIncPos = 1
EndIF

Optimize = 0
EndIf

GRAPH Periods AS "Periods"

// Heuristics Algorithm End

n = 1
m = 3
l = periods//90
 
cs= summation[n](close>open) = n
cs = cs  and close>bollingerup[20](close)  and close>dlow(2)
cl= summation[m](close<open) = m
cl = cl   and close<bollingerdown[20](close) and close<Dhigh(2)
 
 
size = 2
//entry conditions
if cs   then
sellshort size contract at market
endif
 
if cl  then
buy size contract at market
endif
 
//exit conditions
if shortonmarket and close<average[l](close) then
exitshort at market
endif

if longonmarket and close>average[l](close) then
sell at market
endif
```
Francesco78, Nicolas, manel and GraHal thanked this post
03/05/2018 at 12:11 PM #64425 quote

Francesco78
Participant

Master

Hi Nicolas, yes I did WF test of course.

Ill be curious to see the results since 2010. Many thanks

Francesco

WFtest.png

03/05/2018 at 1:50 PM #64435 quote

Nicolas
Keymaster

Master

Sorry, I forgot to attach the result in my first post.

You’ll find attached 3 different pictures, first with WF non anchored, second one with WF anchored and the last one with the variables you are using but with 200k bars backtest.

As you can see all the OOS periods make huge gain while the IS period is not performing so well, it is a sign of overfit. It doesn’t mean that the strategy is bad, it means that it should need refining with some extra filters or more OOS periods to optimize more often.

@juanj Thank you, will study this version.

Francesco78 thanked this post

WF-non-anchored.png WF-anchored.png normal-backtest.png

03/05/2018 at 9:31 PM #64503 quote

Francesco78
Participant

Master

Thank you Nicolas, can you claryfy for my own benefit why you think that adding filter in this case could be effecting on reducing overfitting?

Thanks!

03/06/2018 at 10:45 AM #64559 quote

Nicolas
Keymaster

Master

The question is, why does the strategy is not performing as well in first IS as in OOS? 2012 crisis? I meant to study entries depending of the market behavior in this case, of course, more variables mean more overfit… But as I said, maybe it could be solved by adding more IS+OOS repetitions.

03/15/2018 at 8:45 AM #65330 quote

Nicolas
Keymaster

Master

Well, I think it could be now added to the library, with the WFA attached, what do you think @Francesco78 ?

03/15/2018 at 11:13 AM #65358 quote

Francesco78
Participant

Master

@Nicolas

Thank you, I am fine with it, let me know if you want me to make some amendments on my original post.

03/17/2018 at 12:35 PM #65490 quote

Francesco78
Participant

Master

@Nicolas

What shoud I do to put it on the library?

Thank you

03/17/2018 at 6:03 PM #65528 quote

Nicolas
Keymaster

Master

Hey Francesco, I’ll do it for you on Monday 😉

Francesco78 thanked this post

04/23/2018 at 8:54 PM #69022 quote
Dymjohn
Participant

Senior
For what it is worth I tried tackling the large value of drawdowns compared to winning trades. I noticed the losing trades tend to have a larger number of bars so I optimized bars to < 80 bars which on 100000 bars reduced profit slightly 11384.4 down to 11015.20. It did reduce the maximum loss to 415.2o from 761.00.

The code for this change is attached
```
defparam cumulateorders = false

n = 1
m = 3
l = 90
once Trade = 1

cs= summation[n](close>open) = n
cs = cs  and close>bollingerup[20](close)  and close>dlow(2)
cl= summation[m](close<open) = m
cl = cl   and close<bollingerdown[20](close) and close<Dhigh(2)


size = 2
//entry conditions
if cs   then
sellshort size contract at market
endif

if cl  then
buy size contract at market
endif
if Barindex - Tradeindex < 80 then
Trade =1
else
Trade = 0
endif

//exit conditions
if shortonmarket and close<average[l](close) or Trade = 0 then
exitshort at market
endif


if longonmarket and close>average[l](close)or Trade = 0 then
sell at market
endif
```
Francesco78 thanked this post
05/04/2018 at 8:19 PM #69796 quote

Jan
Participant

Veteran

Thank you very much, Francesco, for the strategy, which seems solid and robust !

I tried this on AUD CAD 30 minutes and it works also quite well.

Extended variables : S for short average close

With 15.000 bars from 7 February till 4 May 2018 it gives a robust profit of 2.945 CAD in 3 months (I did no IS/OOS testing) with M=3, N=2, L = 50 and S = 40, in 398 trades, spread 1,7 pips calculated

I also tried a maximum of one trade per day, with adding the condition ” OneTradePerDay = IntradayBarIndex < (Barindex – TradeIndex(1)) ”

This reduces the amount of trades with more then half, 180 trades, giving a robust profit of 2.354 CAD, with M=3, N=5, L = 50 and S = 90 (also I did no IS/OOS testing so far)

========================================================

For other currencies and indices I am still testing

So far for the FTSE with 2 contracts, exposure more than 30.000 GBP (!), it gives interesting profit of 3.900 GBP on the 15 minutes bar chart with 15.000 bars, from an a maximum of one trade per day, from 6 september 2017 till 4 May 2018, with M=1, N=3, L = 50 and S = 40 (also I did no IS/OOS testing so far !!)

The risk of overfitting for this kind of short periods is very big !! When testing for 100.000 bars the profit especially in the beginning, completely deminish !!!!!!!, but the endresult is quiet the same !

When testing the FTSE for 100.000 15 minutes-bars with not restricted trades per day, the NON- robust profit is a bit lower as when restricted to one trade per day, but not robust and a bit lower.

Kind regards,

Jan

Francesco78 thanked this post

05/04/2018 at 8:22 PM #69798 quote

Francesco78
Participant

Master

Thank you Jan!

05/12/2018 at 9:00 AM #70240 quote

M. Pettersson
Participant

Average

Really nice strategy !

I’m running it in my demo account to try it out.

Thanks Francesco.

Francesco78 thanked this post
Author

Posts