Ichimoku Trading system – my 1st attempt

bmentink · Mon, 24 Jul 2017 07:51:13 +0000

Hi Guys, This is my first attempt at a trading system based on Ichimoku, including some trend indicators, slope indicators and trailing stops & a bit of magic 😉 etc .. The system is optimized for $ or Pound/Yen and most work well. (The profit over 3 months {10,000 ticks} is from 800%[worse] to 1600%[best] ), see attached pictures and equity graph. I don’t have a handle on if this is good or bad for my 1st attempt. I know the drawdown is higher than what I would like, but that could be managed with an overall initial stoploss at the expense of some profit. What do you all think? I have just started a couple of robots on my demo account to see what will happen, and will report back. A question: Can someone explain how the forward testing works? I don’t understand the % ratios between back/forward testing and their importance .. Thanks, Bernie

This topic has 39 replies, 7 voices, and was last updated 6 years ago by juanj.

Viewing 15 posts - 16 through 30 (of 40 total)

← 1 2 3 →

07/28/2017 at 10:13 AM #41886

Despair

When you ran WF tests, which variables did you optimize? Maybe a stupid question but I’m not familiar which all the ichimoku theory.

Despair

Blocked

Topics: 97

Replies: 1034

Been thanked: 217 times

07/28/2017 at 10:19 AM #41888

bmentink

Hi,

As mentioned in my post “They all need optimization of the stop level and the slope threshhold. (see comments in code)”

They are up the top of the code in the variable section, It is the following

SlopeThreshold = 2.5 // From back/forward testing optimization
TGL = 30 // also from back/forward testing, note TGL and TGS the same,

Cheers

PS: You will have different values, I was playing with them,

NOTE: These variables I am optimizing are nothing to do with Ichimoku, just the EMA Slope threshhold and the Stop level.

bmentink

Participant

Topics: 8

Replies: 27

Been thanked: 4 times

07/28/2017 at 10:31 AM #41889

Despair

So only slope and trailing distance, I see. And you run this on 10min chart, correct?

So far I find promising results for CL (oil) with the version you posted. The Yen pairs don’t give so good results but the version you uploaded is maybe optimized for CL and has to be adapted to the pairs?

Despair

Blocked

Topics: 97

Replies: 1034

Been thanked: 217 times

07/28/2017 at 10:48 AM #41890

bmentink

The trading system I posted works well on EUR/YEN mini, 10min 10000 units, make sure it has at least 10,000 yen capital to start as there is a bit of drawdown, I get the following with the code.

I am off to bed, talk tomorrow ..

bmentink

Participant

Topics: 8

Replies: 27

Been thanked: 4 times

07/28/2017 at 11:04 AM #41893

Despair

There we have the problem. You tested only on 10000 bars which is in 10min bars not much time. See the backtest with 200K bars for EUR/JPY with 1.5 pips spread.

Anyway for CL the test is fine even with 200K bars. 🙂 I will try to adapt other securities during the weekend. I have not so much time today.

Despair

Blocked

Topics: 97

Replies: 1034

Been thanked: 217 times

07/28/2017 at 12:58 PM #41897

juanj

@bmentink thank you for posting your very well structured and commented code.

I think like @despair pointed out it is important to at least make use of the bars you have available when optimizing (i.e. 100k).

I see you kept to the basics of the Ichimoku strategy. Have you ever looked into Ichimoku Number and wave theory?

Those are very important factors to consider when trading the Ichimoku strategy. You can find some nice resources on it here:

https://2ndskiesforex.com/trading-strategies/forex-strategies/ichimoku-price-theory-an-introduction/

https://2ndskiesforex.com/trading-strategies/forex-strategies/ichimoku-wave-theory-an-introduction/

https://2ndskiesforex.com/trading-strategies/forex-strategies/ichimoku-number-theory-an-introduction/

I will have a look at your code in more detail over the weekend.

I have found in my own time developing Ichimoku that ADX volitility checks and RSI divergence checks also really help sieving out the false signals.

juanj

Participant

Topics: 51

Replies: 505

Been thanked: 170 times

07/28/2017 at 7:25 PM #41917

Inertia

Hi, 3 pts spread for the Oil instrument, 1.5 for the DAX.

Inertia

Participant

Topics: 40

Replies: 223

Been thanked: 52 times

07/28/2017 at 10:02 PM #41925

bmentink

@despair I get your point. (I need to get a faster computer, WF takes forever on my machine testing back that far.)

@juanj Thanks for the kind words. Feel free to add any code that will weed out the false signals if you have time.

bmentink

Participant

Topics: 8

Replies: 27

Been thanked: 4 times

07/29/2017 at 3:02 AM #41928

bmentink

@despair I did an WF optimization on CL and got better results with 4.3 for slope and 36 for stoploss, check it out .. also WF efficiency 70%

Total gain over the period was 1298%

bmentink

Participant

Topics: 8

Replies: 27

Been thanked: 4 times

07/29/2017 at 9:36 AM #41934

Despair

Very nice! As I said, CL looks promising. I will try to improve the code today and see what i can do.

Despair

Blocked

Topics: 97

Replies: 1034

Been thanked: 217 times

07/29/2017 at 8:01 PM #41979

Despair

I spend some time with your code today but the result got worse if you are unwilling (like me) to hold positions over the weekend. All my other changes were minor improvements. I will post if I find an interesting result.

Despair

Blocked

Topics: 97

Replies: 1034

Been thanked: 217 times

07/30/2017 at 12:50 PM #42010

juanj

@bmentink you can add the following addition to your code to your CL result;

// ------------------------------- Additional Criteria -----------------------------------

R = RSI[15](close) //RSI Period to determine momentum

If R > R[22] Then
If close < close[22] Then
BDIV = 1 //Possible Bearish Divergence Present
SDIV = 0
EndIf
EndIf
 
If R < R[22] Then
If close > close[22] Then
SDIV = 1 //Possible Bullish Divergence Present
BDIV = 0
EndIf
EndIf

AX = ADX[12] > 11
// ---------------------------------- Market Stuff --------------------------------------

if BOK=-1 and BDIV = 0 and AX then
sellshort NbrContracts contract at market
endif

if BOK=1 and SDIV = 0 and AX then
buy NbrContracts  contract at market
endif

// ------------------------------- Additional Criteria -----------------------------------

R = RSI[15](close) //RSI Period to determine momentum

If R > R[22] Then

If close < close[22] Then

BDIV = 1 //Possible Bearish Divergence Present

SDIV = 0

EndIf

If R < R[22] Then

If close > close[22] Then

SDIV = 1 //Possible Bullish Divergence Present

BDIV = 0

EndIf

AX = ADX[12] > 11

// ---------------------------------- Market Stuff --------------------------------------

if BOK=-1 and BDIV = 0 and AX then

sellshort NbrContracts contract at market

endif

if BOK=1 and SDIV = 0 and AX then

buy NbrContracts contract at market

endif

Also be aware that according to my knowledge Senkou-Span A and B should be formulated as:

SSpanA = (tenkansen+kijunsen)/2
SSpanB = (highest[52](high)+lowest[52](low))/2

1 2	SSpanA = (tenkansen+kijunsen)/2 SSpanB = (highest[52](high)+lowest[52](low))/2

Reason being that they are projected 26 periods ahead. By calculating them as you did (26 periods back) you discount a big part of the system.

SSpanA and SSpanB have 3 components to them if calculated as mentioned above.

The leading edge (26 periods ahead) – SSpanA > SSpanB = Bullish and SSpanA < SSpanB = Bearish
The current cloud (26 periods back from leading edge of cloud) – used to determine Kumo break.
The trailing cloud (5 2 periods back from leading edge of cloud) – used to determine resistance (current close/chikou must be clear of the trailing cloud.

1 user thanked author for this post.

Despair

juanj

Participant

Topics: 51

Replies: 505

Been thanked: 170 times

07/30/2017 at 8:56 PM #42031

bmentink

@juanj Many thanks for that. A couple of questions:

Did you actually try my system with those modifications?
Regarding the RSI and ADX parameters, did you optimize those to get to those values? If not, where did the numbers come from?
Regarding SSpanA and B formula, I got those off the Forum, so maybe they are incorrect, I don’t know.

Anyway, I will try those mods out when I get home. Thanks again.

bmentink

Participant

Topics: 8

Replies: 27

Been thanked: 4 times

07/30/2017 at 9:14 PM #42033

juanj

@bmentink

Yes, it improved both the win rate and drawdown
It was optimized for CL although generally you want ADX to be above 10
The way you got it is not incorrect, it simply discounts the leading edge of the cloud

juanj

Participant

Topics: 51

Replies: 505

Been thanked: 170 times

07/31/2017 at 5:04 AM #42045

bmentink

@juanj

I tried your additions to the code, but although the drawdown did decrease, the reduction of %profit was well down on what it was originally. I have attached output of latest. (the original profit was in the order of 39% on CL ..)

It would have been better if you had posted your code version and output of your testing, so we are on the same page.

EDIT: And adding the changes to SSpanA/B that you suggested, dropped %profit down further to 2%

Cheers,

bmentink

Participant

Topics: 8

Replies: 27

Been thanked: 4 times

Author

Posts

Viewing 15 posts - 16 through 30 (of 40 total)

← 1 2 3 →

Find exclusive trading pro-tools on