A System.

Cheers LM. I notice quite a few people add more indicators, especially macd and mas. I won't be doing this, i don't think there's any point, either the HA alone has something or it doesn't.

As LV has already stated, more discretion may be needed.

One thing i can do is change the size of the trailing stop. Another is to discount some trades based on the range of the last candle.

All trades will be entered at the same fractions though. 0.00, 0.25, 0.50, 0.75.

The stop size at the mo is 0.5, it could move to 0.75 or 0.25, or even 1.25.
 
Cheers LM. I notice quite a few people add more indicators, especially macd and mas. I won't be doing this, i don't think there's any point, either the HA alone has something or it doesn't.

As LV has already stated, more discretion may be needed.
I''ll pop the test of HA up after :)
I'll include a chart plot of 2 X ema1 (essentially zero lag ema) each with input of HAopen and HAclose.
That will graphically illustrate that HA is little better than a moving average cross
with each ma having same value but one offset by 1-3 bars.
HA works great when markets trend, but there is nothing built in
to differentiate between trending and ranging markets.

Dinos uses HA, but I personally think he is using greater timing finesse based
on m1 with a keen eye on general news, sentiment and market conditions for that session.

That will always be the bugbear of automation if you go down the indicator route.
Pure automation is binary logic - either the rule is satisfied or not, there
are no shades of grey.
Thats why the vast majority of indicators are quite often worse than random
in this environment, they are already little better than 50-50, throw in the lag
and you're screwed.
 
I''ll pop the test of HA up after :)
I'll include a chart plot of 2 X ema1 (essentially zero lag ema) each with input of HAopen and HAclose.
That will graphically illustrate that HA is little better than a moving average cross
with each ma having same value but one offset by 1-3 bars.
HA works great when markets trend, but there is nothing built in
to differentiate between trending and ranging markets.

Dinos uses HA, but I personally think he is using greater timing finesse based
on m1 with a keen eye on general news, sentiment and market conditions for that session.

That will always be the bugbear of automation if you go down the indicator route.
Pure automation is binary logic - either the rule is satisfied or not, there
are no shades of grey.
Thats why the vast majority of indicators are quite often worse than random
in this environment, they are already little better than 50-50, throw in the lag
and you're screwed.



Mmmm. Would we be better off just putting in trades/orders at regular time intervals (or non-regular)?
 
If the statement, "a trend is more likely to continue", is true. This must surely be the base logic?

Any coding would revolve around set prices, times and parameters.
 
Mmmm. Would we be better off just putting in trades/orders at regular time intervals (or non-regular)?

I'm not going to elaborate too much on that, but yeah
you're on the right track.

Think of market constants - open, close, lunch, bond close,
typical fixed news release times, session times and so on.
They never change (within reason anyway).

I'm not saying trade AT those times, but there or there after, the liklihood
of something occurring is greatly increased.

I'm not going to detail exactly what I do, but that is much closer to the approach
I take than using indicators to pick entry.
 
If the statement, "a trend is more likely to continue", is true. This must surely be the base logic?

Any coding would revolve around set prices, times and parameters.

Yeah, but now you have to actually define an ambiguous term such as
"a trend is more likely to continue" in binary logic.
Nowhere near as easy as you think.

An algo will literally do whatever you tell it with no interpretation.
Tell it to cross the road, and supply no other instructions, it will
walk straight in front of a bus.

Thats the issue, binary logic has SFA to do with human logic.
It takes a while to adjust to that mindset and think in binary logic.
 
I'm not going to elaborate too much on that, but yeah
you're on the right track.

Think of market constants - open, close, lunch, bond close,
typical fixed news release times, session times and so on.
They never change (within reason anyway).

I'm not saying trade AT those times, but there or there after, the liklihood
of something occurring is greatly increased.

I'm not going to detail exactly what I do, but that is much closer to the approach
I take than using indicators to pick entry.


Come on LV, elaborate a little;)
 
Come on LV, elaborate a little;)

Session times for me.
Its entirely dependent on what you are doing and your goals.
Here's 2 screens of the HA ema plots I was talking about,
You can see what the HA open/close look like as ema plots
Its the chop that kills it:
 

Attachments

  • HA trend EOD.png
    HA trend EOD.png
    71.5 KB · Views: 228
  • HA ranging EOD.png
    HA ranging EOD.png
    96.8 KB · Views: 264
Kimo, you still interested in that HA test, seems to have gone off the boli now
so no point bothering?
 
daily.jpg

Have a butchers at this LV. Ordinary candles, between may 31 and aug 14, the trend didn't change until the daily % moves to the downside increased.

A simple rule could be implemented to avoid chop.
 
View attachment 149452

Have a butchers at this LV. Ordinary candles, between may 31 and aug 14, the trend didn't change until the daily % moves to the downside increased.

A simple rule could be implemented to avoid chop.

TBH you'd need to start adressing multi time frames, levels,
key action times, all sorts of stuff that add complexity and parameters.
Curve fitters paradise.

Seriously, you're better off looking at the stuff mentioned earlier.
its much more simple, and the bulk of the focus is on trade management and exits.
The reason that is simpler is because you have a hard metric to base
your code on - open PnL.

Anyway, I'm off soon, MOTO GP Valencia on TV :)
 
TBH you'd need to start adressing multi time frames, levels,
key action times, all sorts of stuff that add complexity and parameters.
Curve fitters paradise.

Seriously, you're better off looking at the stuff mentioned earlier.
its much more simple, and the bulk of the focus is on trade management and exits.
The reason that is simpler is because you have a hard metric to base
your code on - open PnL.

Anyway, I'm off soon, MOTO GP Valencia on TV :)


(y)
 
Cheers LM. I notice quite a few people add more indicators, especially macd and mas. I won't be doing this, i don't think there's any point, either the HA alone has something or it doesn't.

as lv has pointed out, i also don't believe there's much there to get excited about in ha alone, run the lonewolf ea from the link posted and you'll see why as you get crushed. maybe that's why, over all the years that this has been discussed, people add in stuff to try boost any perceived potential.
 

Attachments

  • lonewolf.JPG
    lonewolf.JPG
    158.8 KB · Views: 196
  • lonewolf2.JPG
    lonewolf2.JPG
    75.7 KB · Views: 171
well having just said the above, gone through the timeframes i have to say find some success with this on the daily, eurusd this time, not stellar but a slight nose in front for the year so far :clap:
 

Attachments

  • lonewolf daily.JPG
    lonewolf daily.JPG
    71.1 KB · Views: 161
Heken Ashi EOD backtests EURUSD.
Backtest doesn't show much promise on EOD.
Effect of reversing signals shown - pretty much inverse of original.

The only period in which the EOD variant shows any promise is june 07 - jun 08.
Not really :LOL:
I included that as its a pretty good demonstraion of how its possible to get
pretty decent backtest results over a certain period where the parameters
perfectly align with market conditions.
In other words a perfect curve fit :LOL:

Nothing positive to see so far.
Next up H1.
 

Attachments

  • HA test 50 point trail.xls
    20 KB · Views: 161
  • HA test 50 point trail.png
    HA test 50 point trail.png
    38.3 KB · Views: 204
  • HA test 50 point stop.xls
    20 KB · Views: 189
  • HA test 50 point stop.png
    HA test 50 point stop.png
    41 KB · Views: 216
  • HA test NO STOP rev signals.xls
    20 KB · Views: 265
  • HA test NO STOP rev signals.png
    HA test NO STOP rev signals.png
    41.4 KB · Views: 213
  • HA test NO STOP.xls
    20 KB · Views: 198
  • HA test NO STOP.png
    HA test NO STOP.png
    41.7 KB · Views: 212
  • HA test NO STOP jun07-jun08.png
    HA test NO STOP jun07-jun08.png
    33.1 KB · Views: 206
  • HA test NO STOP jun07-jun08.xls
    20 KB · Views: 189
OK H1 now.

Control entry times via market session windows.
UK - 07:00-10:00
USA 13:00-15:30
Potential for tweaking.

All in all nothing of interest again.
Barely cover transaction costs - if you're lucky.

However, it does show market session timing does reduce trade frequency
and improve consistency slightly.

At the end of the day, still worthless - due to barely overcoming trading costs.
Don't forget these tests do not include spread comms or slippage.
 

Attachments

  • HA test H1 NO STOP uk & usa sesh.png
    HA test H1 NO STOP uk & usa sesh.png
    46.8 KB · Views: 252
  • HA test H1 NO STOP rev signals.xls
    20 KB · Views: 171
  • HA test H1 NO STOP rev signals.png
    HA test H1 NO STOP rev signals.png
    46.1 KB · Views: 215
  • HA test H1 40pt STOP uk & usa sesh.xls
    20 KB · Views: 205
  • HA test H1 40pt STOP uk & usa sesh.png
    HA test H1 40pt STOP uk & usa sesh.png
    43.4 KB · Views: 172
  • HA test H1 NO STOP uk & usa sesh.xls
    20 KB · Views: 153
Last edited:
Top