CanOz's intermittent and sporadic trades

[CanOz]

walk forward Trade of the year....CL during the inventory announcement....I'm sure I'd need colocation to get these fills on a live account.

 

[CanOz]

View attachment 88503 Not allot of time for updates lately, but plenty of testing had happened.

Briefly, I have been using my trade activity filter with a long average. Last week I changed this average after I realised it was not able to pick high periods of activity due to the period I set it too. I changed the average and conducted tests with the dax and saw a dramatic improvement in win% and the size of the winning trade. I started testing it on the hang seng and got similar results initially, but then the results dropped off just as dramatically. Then today I had a moment and it hit me, I realised the reason for the poorer results was contract roll. Because I use volume charts and the hang seng rolls every month, the volumes would be throwing false signals. So whilst I didn’t get time to confirm this, I know in reality I won’t be able to trade the last 5 days of the month into contract roll.

I may get to post more substantial test results tomorrow, as the kids are in daycare. We’re building a house now and just finished the colour selection process with the builder. Hopefully now that’s over I’ll have more time for research.

I’ll leave this with today’s dax trades.....a fine example of how to trade a rotational day.

 

[CanOz]

Nearly 4000 lines of code but we've got the ability to add up to three levels to trade now....

 

[CanOz]

Cool Ted talk....

 

[CanOz]

Two trades on the Dax last night, a couple on the ES which I've just setup.

 

[CanOz]

Massive trend day....

 

[CanOz]

I’ve set the dax up to trade right from the open today, long and short, with a wide atr trail. The hang seng sold off late last night so I set that up short only from the open in anticipation of an open drive. The dax though should price most of it into the gap....

 

[lindsayf]

I’ve set the dax up to trade right from the open today, long and short, with a wide atr trail. The hang seng sold off late last night so I set that up short only from the open in anticipation of an open drive. The dax though should price most of it into the gap....

Doing some interesting work there Canoz.

 

[Roller_1]

have you done historical back testing can? Or are you just doing all walk forward testing?

 

[CanOz]

have you done historical back testing can? Or are you just doing all walk forward testing?

Yes i have done months of back testing but not years.

It was marginal when i first did it because of the usual suspects, noise, chop etc. So i'm walking forward and trying to write rules around the various scenarios such as opening in balance, out of balance, relative volume, initial balance range, average daily range, etc., with a view to getting a context filter coded. Ideally i would like this to test out consistently profitable in out of sample data and walk forward on its own but yet be albe to see some benefit from human involvement, mostly around context or benefit from machine learning.

Intra-day systems are notoriously difficult because of the range of conditions experienced, as you would know. None the less, if i as a human can adapt, then why can i not program the algorithm to adapt?.

That makes sense?

 

[CanOz]

I’ve set the dax up to trade right from the open today, long and short, with a wide atr trail. The hang seng sold off late last night so I set that up short only from the open in anticipation of an open drive. The dax though should price most of it into the gap....

The algorithm traded the dax today as I have so many times when you can’t get the opening drive....kept buying the lows until one stuck....

 

[CanOz]

For the HSI today, I've left the algorithm trading long and short but with a 4 multiple on the atr instead of a 6, in anticipation of a choppy day in search of balance. I was tempted to only take shorts below the overnight high so we'll see how that would have worked out...

 

[CanOz]

Here's an example of what I need to work on. A day after a big trend day is usually very choppy as the market trades around in balance. It's easy enough to justify coding it out, just not trade at all. Wide ranging day, closing within 30% of the low/high, don't trade. Or I could set my trade activity filter to a higher threshold....I'm not going to avoid all chop but this one is the one I want to avoid the most right now.

 

[CanOz]

I think I may have found a solution to one of my problems with my system and in the process inadvertently solved another problem...I’ve created a trend filter based on the cumulative delta of bids vs asks, the narrow ribbon in the first panel. I’ll use it to enter new trades but also add to open trades...

 

[CanOz]

After considerable testing I've developed some rules to try and deal with the dynamics of intraday markets. These rules should allow the algorithm to perform at its optimum without trying to curve fit it. If these work well then I'll get them coded into a context filter. This would be the final phase of development.

Trading Rules - semi autonomous imbalance algorithm

Must set up just after the open:

Opening in the prior days value area requires:
1.) waiting until the initial balance period before trading, set trading start time
2.) tight trailing stop, low targets or only trade extremes of value, set stops or trade zones
3.) give back rule of 25 % then stop trading. Use a day stop of 3 consecutive full stop losses.

Opening outside of the prior days value area
1.) trade from the open, adjust trading start time
2.) wide trailing stop to capture larger swings, wider target, adjust atr multiple
3.) give back rule, day stop 4 consecutive full stop losses


If we open inside value and then breakout then adjust atr stop multiple to wider setting

Monitor trading throughout the session with remote access:

Use naked pocs as targets, tighten up trailing stop as price approaches, by adjusting atr multiple.

 

[CanOz]

Regarding the pace of trade filter, it is different on every market but generally it works well in all scenarios in the fdax. For the HSI it helps performance on balanced days whilst switching it off on trend days is beneficial.

 

[CanOz]

Got my trend filter coded up and have tested it on the last month of data. Significantly better performance and ready to add the pyramid option, which will add contracts up to maximum size as long as the trend filter permits.

 

[peter2]

@CanOz I'm sensing an opportunity in copper again. Are we there yet?
ie Are we near the bottom of a volume profile?

 

[CanOz]

I haven't been watching It to be honest. I've only had a few markets up whilst testing...I'm curious though and I'll try and check it out.

 

[CanOz]

The latest version of my imbalance algorithm, with an option to add contracts up to maximum size....