Short bias - with an attached super profitable EA | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Help with profitable EA (attached for your own research) 5 replies

Hallett Norman - Mental Fitness For Traders (Attached) 2 replies

Pulos Bill - The Truth About Fibonacci Trading (Attached) 4 replies

The Pocket Book Of Economic Indicators (Attached) 2 replies

One More Zero - How To Trade FX Like A Pro (Attached) 0 replies

Attachments: Short bias - with an attached super profitable EA
Exit Attachments

Short bias - with an attached super profitable EA

  • #1
  • First Post: Edited Dec 31, 2008 12:43pm Dec 30, 2008 2:44pm | Edited Dec 31, 2008 12:43pm
I have been forward testing a new AI system since yesterday with good results but I am seeing a consistent bias towards opening short positions instead of Long positions.

The EA creates a market netural order table with all available currency pairs (in this case 24 pairs * 8 indexes * 3to6 positions for the hedge totaling between 576 and 1152 positions) then trades these positions based on what the AI algos determine is the best solution for each available position. The order table has about a 6 to 1 ratio of short to long positions but the actual trading is about a 15 to 1 ratio (see attachment). The only reason I can think of is that the AI is calculating everything based on the bars which only represent the Bid pricing and do not include the spread.

So the question is ... how do you handle the bias towards short positions in your systems ?

Also, how would you adjust your indicators to correct this bias ? this is particularly important in oscillator type indicators...
Attached Image (click to enlarge)
Click to Enlarge Name: testing snapshot.jpg Size: 93 KB
Keep it simple stoopid....
  • #2
  • Dec 30, 2008 2:54pm Dec 30, 2008 2:54pm
What's the testing period of this EA/system?

The market has just about gone through a retracement of the summer trend.
  • #3
  • Dec 30, 2008 3:16pm Dec 30, 2008 3:16pm
thats a live testing detailed statement (demo account) since yesterday.
Keep it simple stoopid....
  • #4
  • Dec 30, 2008 3:36pm Dec 30, 2008 3:36pm
it could just be that there was a downward trend. if there was an upward trend, you would probably have a long bias.
  • #5
  • Dec 30, 2008 3:36pm Dec 30, 2008 3:36pm
Use the middle price in your calculations? Maybe something like...

myPrice = (Bid + Ask) / 2.0;
nav++
  • #6
  • Dec 30, 2008 3:51pm Dec 30, 2008 3:51pm
Quoting blacksun1
it could just be that there was a downward trend. if there was an upward trend, you would probably have a long bias.

There would not be a downward trend in all 24 pairs though...
Keep it simple stoopid....
  • #7
  • Dec 30, 2008 3:55pm Dec 30, 2008 3:55pm
Quoting gvroom
Use the middle price in your calculations? Maybe something like...

myPrice = (Bid + Ask) / 2.0;
This method works when "triggers" are set on pricing indicators (indicators in the chart window) but the indicators themselves are skewed and do not account for the spread.

for example the stocastic indicator has a short bias because it is calculated using the pricing of the bars. Even if you calculate using HIGH LOW pricing you still have this problem. The indicator itself has a short bias.
Keep it simple stoopid....
  • #8
  • Dec 30, 2008 3:57pm Dec 30, 2008 3:57pm
Ah, now I understand your point. If you are right about the issue and it's cause then it sounds like you may end up having to write custom indicators...
nav++
  • #9
  • Dec 30, 2008 4:04pm Dec 30, 2008 4:04pm
Quoting Kurka Fund
There would not be a downward trend in all 24 pairs though...

lol. nevermind then. haha. that is pretty weird.
  • #10
  • Dec 31, 2008 12:16pm Dec 31, 2008 12:16pm
Anybody here want to discuss this fundamental bias in indicators ? It seems that there should be more interest in such a topic.

Oh wait I know how to get people on this board interested ... attach an old EA with a good backtest.
Attached Image (click to enlarge)
Click to Enlarge Name: TSF trader.JPG Size: 101 KB
Attached File(s)
File Type: mq4 Kurka Trader - TSF Secure.mq4   21 KB | 1,304 downloads
Keep it simple stoopid....
  • #11
  • Dec 31, 2008 12:22pm Dec 31, 2008 12:22pm
Indicators may appear biased. But consider this:

A market goes down faster than it goes up. So, with that in mind, perhaps this feature of bear markets is what appeals to the EA/indicator/system?
  • #12
  • Dec 31, 2008 12:24pm Dec 31, 2008 12:24pm
Quoting Ronald Raygun
Indicators may appear biased. But consider this:

A market goes down faster than it goes up. So, with that in mind, perhaps this feature of bear markets is what appeals to the EA/indicator/system?
Your basing this claim on ??? please explain
Keep it simple stoopid....
  • #13
  • Dec 31, 2008 2:23pm Dec 31, 2008 2:23pm
Quoting Kurka Fund
Anybody here want to discuss this fundamental bias in indicators ? It seems that there should be more interest in such a topic.
This is an interesting topic I've been noticing for a while, but uncertain how to address. I see the short pricing bias in all my pricing estimate indicators.

Things I've tried:
- midpoint pricing
- allocating an even proportion of spread to bid and ask
- allocation of spread only to ask, no spread on bid

In all these cases my indicators have resulted in a sell bias. The only time I've seen a buy bias at all, it was a while ago when I was pulling prices via DDE from a non-MT4 broker (thinkorswim).

Is it possible this is a structural aspect of the markets that will change as the volatility adjusts? Cynically, it occurs to me that retail traders and investors tend to have a long bias, so perhaps the indicators are reflecting that?

In attempting to dial the bias out of my work, I've thought about applying the spread disproportionately to Ask vs Bid, but other than the nice mental image of a balanced indicator I don't have any empirical basis for doing so.
  • #14
  • Dec 31, 2008 2:25pm Dec 31, 2008 2:25pm
Quoting Kurka Fund
Oh wait I know how to get people on this board interested ... attach an old EA with a good backtest.
Hah! Thanks, you do such nice work. :-)
  • #15
  • Dec 31, 2008 2:32pm Dec 31, 2008 2:32pm
Quoting k1_
This is an interesting topic I've been noticing for a while, but uncertain how to address. I see the short pricing bias in all my pricing estimate indicators.

Things I've tried:
- midpoint pricing
- allocating an even proportion of spread to bid and ask
- allocation of spread only to ask, no spread on bid

In all these cases my indicators have resulted in a sell bias. The only time I've seen a buy bias at all, it was a while ago when I was pulling prices via DDE from a non-MT4 broker (thinkorswim).

Is it possible this is a structural...

a bias toward long positions makes sense in equity markets not in the derivatives
Keep it simple stoopid....
  • #16
  • Dec 31, 2008 3:50pm Dec 31, 2008 3:50pm
Quoting Kurka Fund
a bias toward long positions makes sense in equity markets not in the derivatives
Understood, but does it not making sense mean it doesn't exist? Options are derivatives, and show a clear market bias toward the long side (most of the time...)

But I don't want to argue for a structural bias, as it doesn't help me do better work. I've noticed the short bias you mention, and have struggled with figuring out why it exists in my indicators. At first I thought it was simply accounting for the spread incorrectly, but the problem still exists when using midpoint.

So the question is why that would be? Is there a hidden truncation going on in the math that stacks up in the calculations? It seems unlikely, but I'm having trouble seeing a thread to pull on.
  • #17
  • Jan 5, 2009 2:00pm Jan 5, 2009 2:00pm
How come any thread that requires thought seems to die quickly on this board ?

Anyone here interested in solving the Short Bias problem with me ?
Keep it simple stoopid....
  • #18
  • Jan 5, 2009 2:55pm Jan 5, 2009 2:55pm
There is no general one-size-fits-all explanation for directional bias. That's the problem. We also don't know the system fully, so the only thing which can be done is speculate. And let's face it, speculating is boring.
  • #19
  • Jan 5, 2009 2:59pm Jan 5, 2009 2:59pm
Quoting Ronald Raygun
There is no general one-size-fits-all explanation for directional bias. That's the problem. We also don't know the system fully, so the only thing which can be done is speculate. And let's face it, speculating is boring.

Ok then how about more specifically addressig the short bias in a particular indicator ... say Stocastics. What is the cause for the short bias ? Ill write the code and share it just need some ideas that might explain this .
Keep it simple stoopid....
  • #20
  • Jan 5, 2009 5:04pm Jan 5, 2009 5:04pm
The main reason for any bias is usually a trend.

As for stochastics, I've traded an MTF stochastic system and it seems perfectly unbiased to me. But it was trading the cross of the 50 line.
0 traders viewing now