GridDancer / GridMaster

[fxstrategie]

I got a PM from chimoong from Malaysia (Welcome to the Forum, chimoong!) that went like this:

I thought that this was novel idea ... and it only required a few small changes to GridDancer ... So, in Post 1, you will find GridDancerv1.54. Let's see if we can keep the Balance-Equity-Gap under control with this new Dancer.

Thank you Chimoong for your first contribution!

Edit:

Below are the results after putting this on a demo account for four hours during the Asian session:

Will need much more testing to draw any conclusions ...

taken up your Idea and setting up a 24/5 demo for this.
1K Cent Account:

[renexxxx]

taken up your Idea and setting up a 24/5 demo for this.

Unfortunately, this idea didn't live up to its expectations ... the Equity-Balance-Gap kept growing, like it did with the other versions. Of course, you are welcome to verify this for yourself.

Since this version (v1.54) of GridDancer I have been experimenting with many new and different ideas (currently I'm up to v1.59), and I will be more than happy to share a new version if I believe that it has some chance to survive the test of time. Having said all that, my focus has now moved to GridMaster, and this is why: No matter in what fashion GridDancer opens the stack of BUYs and SELLs, due to the closing action of GridMaster, the stack always ends up in a losing state: many BUYs above the market price and many SELLs below the market price. This band of trades is getting wider and wider, and it is getting increasingly harder for GridMaster to find a set of BUYs and SELLs that can be closed in profit.

Thus, in the next iterations of GridMaster, I'm attempting to find ways to (1) Reduce the width of this band, by closing to the top BUY and bottom SELL in tandem (2) Whilst at the same time, finding a basket that has a non-negative total profit.
So, in a way, the priority of GridMaster has shifted in favour of leaving the 'state of the system' in a better position, instead of just following a greedy approach by blindly picking up a set of trades that are currently in a net profit.

As always, if I find something that is worth sharing, SHF will be the first to know about it.

[dokopy]

Just an idea. I assay similar EA as follows: Only GBPJPY and EURUSD pairs. After the Frankfurt Open run EA. I close all trades around 19:55 GMT. 8 weeks without disaster. (Apology for the translation).

[dydynamic]

taken up your Idea and setting up a 24/5 demo for this.

Unfortunately, this idea didn't live up to its expectations ... the Equity-Balance-Gap kept growing, like it did with the other versions. Of course, you are welcome to verify this for yourself.

Since this version (v1.54) of GridDancer I have been experimenting with many new and different ideas (currently I'm up to v1.59), and I will be more than happy to share a new version if I believe that it has some chance to survive the test of time. Having said all that, my focus has now moved to GridMaster, and this is why: No matter in what fashion GridDancer opens the stack of BUYs and SELLs, due to the closing action of GridMaster, the stack always ends up in a losing state: many BUYs above the market price and many SELLs below the market price. This band of trades is getting wider and wider, and it is getting increasingly harder for GridMaster to find a set of BUYs and SELLs that can be closed in profit.

Thus, in the next iterations of GridMaster, I'm attempting to find ways to (1) Reduce the width of this band, by closing to the top BUY and bottom SELL in tandem (2) Whilst at the same time, finding a basket that has a non-negative total profit.
So, in a way, the priority of GridMaster has shifted in favour of leaving the 'state of the system' in a better position, instead of just following a greedy approach by blindly picking up a set of trades that are currently in a net profit.

As always, if I find something that is worth sharing, SHF will be the first to know about it.

My dear Renexxxx,
Your observation is very correct, and that is the reason we have high drawdown which ultimately blows up an account.....
I prefer the approach of closing a group of unhedged sell trades that are in profit and above a group of buy and sell trades that are hedged as price drops or a group of buy trades that are in profit and are below a group of hedged sell and buy trades as prices rally upwards,we should avoid closing any trade that is hedged as this is a recipe for disaster but those group of sell trades from above can be closed together with a group of hedged trades as the market drops or a group of buy trades can be closed together with a perfectly hedged buy and sell trades as market rockets upwards. It is obvious that a group of unhedged sell trades from above will always be in profit as prices drop, likewise a group of unhedged buy trades from below will always be in profit as prices rally upwards....The major drawback of the present system of offsetting that we have now is that it offsets even in the middle of a group of hedged trades that is supposed to provide a kind of balance,if there is a way we can code offsetting to offset from the extremes leaving the hedged trades in the middle untouched,that will be awesome, kind of like swing trading....... :hi:

[woodlands]

Dear Renexxxx,
Are we trying to take too much?
I also set up test for this version but took advice from my son who trades fib retracements with stoch on the M1, successfully, for small pips, scalping really, he calls it 'nibbling at the market', I set GM to take these small bites. Only one day but we will see if he has any logic here.......

GD_nibble.png

Gridmaster_nibble.set

Thank you for the work you have put into this, I hope that this testing is of some small help, I will let it run for the week and post again.
Regards
Tony

[acostafulano]

Ive tried this approach many times, but trading costs (spreads commisions) always end up eating away all the gains. How do yu tackle this??

Dear Renexxxx,
Are we trying to take too much?
I also set up test for this version but took advice from my son who trades fib retracements with stoch on the M1, successfully, for small pips, scalping really, he calls it 'nibbling at the market', I set GM to take these small bites. Only one day but we will see if he has any logic here.......





Thank you for the work you have put into this, I hope that this testing is of some small help, I will let it run for the week and post again.
Regards
Tony

[woodlands]

Ive tried this approach many times, but trading costs (spreads commisions) always end up eating away all the gains. How do yu tackle this??

It may be an academic question anyway as it is moving in the opposite direction this morning, I'll let it run and see where we end......

[Boulder]

Hey guys just an idea that might be worth considering and experimenting with. The idea involves using Heiken Ashi candles on M5, M15 or M30 charts (the best TF is also yet to be found out).

Now the idea is for the GridDancers to trade in the following way:

• Previous candle is red and current candle is also red: only trade downwards
• Previous candle is red and current candle is blue: trade both ways
• Previous candle is blue and current candle is blue: trade only upwards
• Previous candle is blue and current candle is red: trade both ways

Another equally interesting variation would be for the GridDancers to trade in the direction of the colour of the previous bar regardless of the current bar's colour. This alternative thus wouldn't trade both ways.

The real job is put on the shoulders of GridMaster though, which shall attempt to "let the winners run". He should:

• If the previous candle is red and the current candle is red: don't use the sell trades for offsetting until we reach X sells. When we reach X sells, look for interpair offsets first. If none, then look for intrapair offsets. The buy trades of this pair are available for offsetting from the onset of course.
• If the previous candle is red and the current candle is blue: look for interpair offsets of the existing sells first. If none, then look for intrapair offsets.

Analogously for the reverse case.
Furthermore it should also keep its global basket TP in cash (or % of equity as an alternative) feature.

[renexxxx]

Hey guys just an idea that might be worth considering and experimenting with. The idea involves using Heiken Ashi candles on M5, M15 or M30 charts (the best TF is also yet to be found out).

Thank you Aram for this suggestion! There are 101 ways (and more) to configure the dancers and the master, and I have been demo-ing another idea since August 2nd, with the following results:

GMGD.PNG

The vertical line UP on August 11th, had to do with the RBNZ news that day. I used different strategies for both GridDancer and GridMaster:

- GridDancerv1.59: This Dancer is completely braindead: When it starts, it draws a line in the sand (called the anchor line), and creates a grid of equidistant price levels around that line. The distance is based on the ATR. It will only BUY above and only sell below this line (i.e. when price reaches a new grid level).
- GridMasterv1.7: This Master has a bit more "thinking" to do. It attempts to break down these hedge scenario's you always get after having run this type of strategy for a while, something like:

PerfectHedge.PNG

So, for each pair, it gathers the Biggest Losing BUY and the Biggest Losing SELL, to form an array of losing tuples. Its objective is to close the biggest losing tuple, such that the band from the top most losing BUY to bottom most losing SELL is reduced (at least for one pair at a time). GridMaster aims to keep that band as narrow as possible, as otherwise, the "system" gets into a state from which it is extremely hard to recover. As before GM finds some winning trades to sacrifice so that it can close that biggest losing tuple together with the winning trades in a nett profit.

I have been running GridDancerv1.59 on all the US-pairs (AUDUSD, GBPUSD, EURUSD, NZDUSD, USDCAD, USDCHF, USDJPY) with GridMaster1.7 closing trades, all OOTB. The results (as seen above) were encouraging, despite some scary drawdowns, so I thought that it was worth sharing this with the forum.

Overall, I think that the GridDancers/GridMaster concept has a lot to offer, if we just can get the formula right.

I have added these new versions to Post 1

[Boulder]

Beautiful work Rene :hi: :hi: You keep outdoing yourself. I'll be testing too