By the way, has anyone else noticed that the FSBv3.0 seems to find many more 'ambiguous bars' in the same data than previous versions did?

Are you saying it finds more ambig. bars with the same data AND the same strategy?

Perhaps, but what if the FSB gremlins don't restrict their activities to generating wonderful strategies and Sharpe ratios via the no-parameter Ross-Hook indicator?

One thing that you may do is read up on Ross Hook, it is a bar setup, it is not an adjustable setup. Joe Ross developed a very simple approach and that is what the indicator attempts to deliver.

I start with 3.0 and expect to get much much higher like 20. I guess it is individual preference, seems to me that Sharpe is important re future success.

Thanks!

Blaiserboy wrote:

Did you see the RC version below the one you downloaded, get the most recent and keep looking at that page as it is being update often.

Forex Strategy Builder v2.98.3 RC

Thank you so much. Yes, the RC gives me the "criteria" menu, and I can see many more options.  Looks like I can filter these options to give me a much more stable equity curve.

Last question Dave - when you filter strategies for sharpe ratio, what is the level you recommend? The default of 0.2 or something much higher as mentioned in some of your older posts?

I start with 3.0 and expect to get much much higher like 20. I guess it is individual preference, seems to me that Sharpe is important re future success.

For 8.5 years of data you have 53 trades. 53 trades for 3127 trading days. 1 trade in 3 months in average. Doesn't look too good. It seems to be exemplary case of curve-fitting to me (when I look at the numerical parameters) or the second option would be that the holy grail in trading exists, and the guru master himself has unravelled it, praise and glory to him, even though it is damn slow for a true grail.

Absolutely, Footon!

All the beauty of this resulting equity curve comes from the Detrended Oscillator being curve-fitted to death...

Changing "manually" the Detrend MA lengths immediately shows bad results within just + or - a few periods.

The combination with RossHook does help a lot though: replace RossHook Opening Point with a simple Bar Opening and the magic disappears, meaning that the Detrended Oscillator alone is useless.

You can even replace the Detrended Oscillator with an MACD(10,45,9) with the same entry condition (cross zero line up), and it gives a similar nice upwards balance curve (with higher DD though)...
And again, replacing RossHook with Bar Opening kills the mirage...

My conclusion would be that RossHook truely gives an edge, when combined with some other sauce...

I prefer postimg.org. The last right button on command toolbar above the post form..

Thanks for the link, i'll use that!

I reviewed the code of RossHook.cs, and i haven't found any bug in it so far.
The calculation of the up and down levels looks correct to me...

The entry levels shown on the indicator chart seem to second that: the rosshook levels seem to be correctly expanded to the right, and trades are entered at those levels as they should.

The RossHook implementation does not seem to be the issue after all...

Sorry jgp, you may have to accept that all those wonderful strategies are just... wonderful
You'll have to thank Mr Popov and all the dev team and contributors for their hard work and their patience...

Sq

Perhaps, but what if the FSB gremlins don't restrict their activities to generating wonderful strategies and Sharpe ratios via the no-parameter Ross-Hook indicator?

Stupid computer, I don't have those gremlins apparently...

J, can you make your gremlins at your loft generate me a wonderful short-term strategy, on M1 or max M5 timeframe, with a maximum Sharpe ratio possible. Very few losses wouldn't hurt either

How hard can it be with a faulty backtester.

Hi,

I tested jgp's last generated strategy on "my" FSB data, GU H4 from AlpariNZ.
Simply creating the strat' "by hand" from scratch from jgp's posted strategy description.
I adjusted the number of bars to match jgp (in Data Horizon : 18766, to match the 18710 "tested bars", although my GU H4 data can go back to 21071 tested bars)

Attached is a screenshot of what i get...

(too bad that we cannot upload images directly into the forum, i mean hosted by the forum, this would be a very nice addition to the forum!)

This DOES look too good to be true indeed...

Would something be wrong with how rosshook is being handled by the baktester ?

Sq

For 8.5 years of data you have 53 trades. 53 trades for 3127 trading days. 1 trade in 3 months in average. Doesn't look too good. It seems to be exemplary case of curve-fitting to me (when I look at the numerical parameters) or the second option would be that the holy grail in trading exists, and the guru master himself has unravelled it, praise and glory to him, even though it is damn slow for a true grail.

(too bad that we cannot upload images directly into the forum, i mean hosted by the forum, this would be a very nice addition to the forum!)
This DOES look too good to be true indeed...

I prefer postimg.org. The last right button on command toolbar above the post form..

Market: GBPUSD 4 Hours
Spread in points: 20.00
Swap Long in Points: 2.00
Swap Short in Points: -2.00
Commission per lot at opening and closing in Points: 0.00
Slippage in points: 0

Maximum open lots: 1.00
Entry lots: 0.10
Adding lots: 0.10

Intrabar scanning: Accomplished
Interpolation method: Pessimistic scenario
Ambiguous bars: 0
Tested bars: 3679
Balance: 3392 points (53392.20 USD)
Minimum account: -22 points (49978.20 USD)
Maximum drawdown: 274 points (273.90 USD)
Time in position: 33 %

[Strategy Properties]
     A same direction signal - Adds to a winning position
     An opposite direction signal - Reverses the position
     Permanent Stop Loss - None
     Permanent Take Profit - None
     Break Even - None

[Opening Point of the Position]
     Ross Hook
     Enter long at an Up Ross hook

[Opening Logic Condition]
     Average True Range
     ATR falls
     Smoothing method  -  Smoothed
     Base price  -  Bar range
     Smoothing period  -  21
     Level [points]  -  184
     Use previous bar value  -  Yes

[Opening Logic Condition]
     Accumulation Distribution
     AD falls
     Use previous bar value  -  Yes

[Closing Point of the Position]
     Trailing Stop Limit
     Exit at Trailing Stop Loss or at a constant Take Profit level
     Trailing mode  -  Trails once a bar
     Initial Stop Loss  -  4091
     Take Profit  -  398

Hi,

I tested jgp's last generated strategy on "my" FSB data, GU H4 from AlpariNZ.
Simply creating the strat' "by hand" from scratch from jgp's posted strategy description.
I adjusted the number of bars to match jgp (in Data Horizon : 18766, to match the 18710 "tested bars", although my GU H4 data can go back to 21071 tested bars)

Attached is a screenshot of what i get...

(too bad that we cannot upload images directly into the forum, i mean hosted by the forum, this would be a very nice addition to the forum!)

This DOES look too good to be true indeed...

Would something be wrong with how rosshook is being handled by the baktester ?

Sq

To upload an image, you may find it convenient to use an image site and upload the url for the image. That is what others have done in the past

All the more reason to find and correct if possible what may be a serious bug in FSB before marketing FST-FSB Pro to the masses.

I invite you to review the code so that you , as the expert, can point out the specific errors.

The code is available for you to examine as it is open source.

Let's see what you can discover as a flaw and you can enlighten all of us.