"AUNZ Matrix" - Arbitraging on AUDNZD

Avatar
  • updated

Hi fellow traders,


I'd like to introduce you to a trading strategy I've developed called the "AUNZ Matrix", that leverages the historical correlation between the Australian dollar (AUD) and the New Zealand dollar (NZD).

The Basic Idea: The core concept of the AUNZ Matrix is to create a grid of buy and sell orders on the AUDNZD pair. This currency pair has predominantly traded in a sideways range between 1.14 and 1.03 over the past decade. Each order within the grid comes with its own trailing exit. The primary objective is to generate consistent, albeit modest, profits by closing multiple orders within the grid. No Stop Losses are employed in this strategy.


the Dynamic Virtual Wall: While the basic strategy does work, it can result in a significant deposit load and a high drawdown percentage. To address this issue, I've incorporated a "dynamic virtual wall" into the strategy. This virtual wall divides the grid into two distinct zones: an upper zone and a lower zone. The wall is constructed using a weekly Bollinger Band with a 100-period SMA and a 0.5 deviation.

The virtual wall serves as a filter for the grid orders. In the upper zone, only sell orders are allowed, and in the lower zone, only buy orders are permitted. This adjustment dramatically reduces deposit load and drawdown percentage, albeit at a slight reduction in profit.  Additional filters can be implemented to optimize performance further.


Risk and Limitations: It's crucial to acknowledge the main risk associated with the "AUNZ Matrix" strategy. The primary risk factor is the potential decorrelation of the AUD and NZD. This can occur if the respective central banks change their monetary policy. Traders following this strategy must closely monitor central bank actions and be prepared to pause the bot if any significant policy changes are anticipated.

This precaution may result in some losses if the bot has open orders at the time of a potential decorrelation event. However, it's a necessary measure to protect your capital and align the strategy with changing market conditions.

Set File: I've attached a set file that you can use to test this strategy on your own. It serves as a starting point for further experimentation. You have the flexibility to tweak the parameters of the "martingale" grid or incorporate additional indicators and techniques to enhance its performance.


Exploring Future Possibilities: In my ongoing exploration of this strategy, I'm interested in developing a volatility-based grid, where the step size of the grid adjusts dynamically according to market conditions. Unfortunately, I haven't found a way to implement this with this EA yet. It appears that the "change steps" parameter does not function correctly when individual exits are used. It also seems that individual take profit, with volatility coef does not work properly. 
If anyone has insights into achieving this, please share your knowledge. Incremental step adjustments could be intriguing for incorporating very light lot multiplier techniques. After all, based on the core concept of this strategy, raising the aggressivity when the price action reach the boundaryes of the ranging channel could make sense.


CP Section used:

"Martingale Properties", for creating the grid.

"Individual Order Properties", for creating the individual exits.

"MA Filter #1 Prop." & "Volatility for Ma": for creating the Bollinger Bands

That's all.


I hope you find the "AUNZ Matrix" strategy intriguing as i do. Let's try to improve its performance and create a nice "ATM" by arbitraging the Kiwi Aussie correlation together. So feel free to share your thoughts and results.


Happy trading! ^^

Image 5107

Image 5106



AUNZ Matrix v1.0.set

MT version:
Set for MT5 only
EA version:
2.55.5
Symbol:
AUDNZD
TimeFrame:
Weekly
Avatar
Giamma
Quote from Ulises Cune

It's interesting... They don't close losing operations but they keep non-positive operations open for a long time.

Chart Daily:

Image 5117

Thank you for having tested the strategy. Can you please explain it better? You mean there are no stop losses or something else? 
Then, what initial capital and timeframe you used for the backtesting?

Avatar
Giamma
Quote from jay hu

Mine is like this before it blows. Grid system is not feasible with strong trending I think.

Image 5134

Hello! thanks for having tested it. 

Try to retest the strategy, using 10.000 as starting capital (not 300) and weekly or monthly timeframe, on the settings panel of the tester. Tell me how it goes :)

Avatar
jay hu
Quote from jay hu

Thank you for sharing the strategy and set, Giamma. I have encountered the same issue as Ulises had in that some individual positions did not close in time so bringing equity DD was very high during the 4H TF backtest from 2022 to 2023. The backtest failed but the forward test survived with DD around 70% in the Blueberry market MT5. Think about putting breakeven and partial close to lock the profit. Does anyone have an idea?

Image 5131

Image 5132

Image 5133

Mine is like this before it blows. Grid system is not feasible with strong trending I think.

Image 5134

Avatar
jay hu

Thank you for sharing the strategy and set, Giamma. I have encountered the same issue as Ulises had in that some individual positions did not close in time so bringing equity DD was very high during the 4H TF backtest from 2022 to 2023. The backtest failed but the forward test survived with DD around 70% in the Blueberry market MT5. Think about putting breakeven and partial close to lock the profit. Does anyone have an idea?

Image 5131

Image 5132

Image 5133

Avatar
Ulises Cune

It's interesting... They don't close losing operations but they keep non-positive operations open for a long time.

Chart Daily:

Image 5117

Avatar
Giamma
Quote from jay hu

I aware that Strategy tester use 99% or 100% history quality real tick data may give us a better test result? The test done on ICmarket only show 8% history quality data for this model test.

Yes, of course, with a low historically quality, the result could be distorted, but the test on ICMarket is not mine :D

Avatar
jay hu
Quote from Giamma

Thank you for your test ^^ Yes, the strategy does work to some extent with all timeframes and brokers, as it is based on fundamental concepts and a few indicators.


When working on the H4 timeframe significantly increase your earnings, you must also consider the difference in the Equity max Drawdown between your test and mine. Imagine if the New Zealand or Australian central bank were to make an extraordinary announcement regarding their monetary policy while you are in a 40% DD. This could potentially result in such a substantial loss. It's important to generate consistent profits while keeping both deposit load and DD% as low as possible. This way, you can ensure steady returns and minimize risk, even if you need to stop the bot abruptly one day.
So, in my idea, when optimizing this streategy it's really important to find out the best risk-adjusted return, instead of the absolute one. 

I aware that Strategy tester use 99% or 100% history quality real tick data may give us a better test result? The test done on ICmarket only show 8% history quality data for this model test.

Avatar
Giamma
Quote from Mat_cl

Hey nice set

This is ICmarkets demo account

Timeframe H4 using your set

Image 5108

Image 5109

Image 5110

Thank you for your test ^^ Yes, the strategy does work to some extent with all timeframes and brokers, as it is based on fundamental concepts and a few indicators.


When working on the H4 timeframe significantly increase your earnings, you must also consider the difference in the Equity max Drawdown between your test and mine. Imagine if the New Zealand or Australian central bank were to make an extraordinary announcement regarding their monetary policy while you are in a 40% DD. This could potentially result in such a substantial loss. It's important to generate consistent profits while keeping both deposit load and DD% as low as possible. This way, you can ensure steady returns and minimize risk, even if you need to stop the bot abruptly one day.
So, in my idea, when optimizing this streategy it's really important to find out the best risk-adjusted return, instead of the absolute one. 

Avatar
Mat_cl

Hey nice set

This is ICmarkets demo account

Timeframe H4 using your set

Image 5108

Image 5109

Image 5110