Index | Recent Threads | Who's Online | Search

Posts: 25    Pages: 3    1 2 3 Next
Last Post
New Thread
This topic has been viewed 3965 times and has 24 replies
z8735
trying to implement the Hi-Lo method - what's wrong

An article by George Vrba suggests the diff between new high to new lows as a timing indicator, see:
http://imarketsignals.com/2015/avoiding-stock...hi-lo-index-of-the-sp500/

Trying to repeat his result, I created the Hi_Lo_3m custom series (see screenshots below) which seems to generate a chart behaving as expected. However when using this custom series in a screen (or sim) it does not generate any buying signal until the very end (the last year or so). What am I doing wrong?

Thanks,
Zvi

Attachment custom-series.png (39564 bytes) (Download count: 410)


Attachment custom-series-plot.png (66837 bytes) (Download count: 377)


Attachment Hi-Lo_3m_screen.png (117023 bytes) (Download count: 375)


The Value Devils: diversified and liquid R2G with top performance
https://www.portfolio123.com/app/r2g/summary/1174555

Aug 29, 2015 2:59:07 AM       
Market_Neutral
Re: trying to implement the Hi-Lo method - what's wrong

z - start your custom series at 1999 not 2011.
Steve

Aug 29, 2015 3:06:31 AM       
z8735
Re: trying to implement the Hi-Lo method - what's wrong

Hi Steve, thanks, but that doesn't explain why the screener only chooses data from 2014, clearly - as the custom series chart shows - the screen result should have been positive much earlier even if data started in 2011.

BTW when trying to change the chart to 1999 doesn't work - it immediately bounces back to the 2011 date.
Still, clearly there is a way to do it, the article even shows P123 sim screenshots. What are we missing?

Cheers,
Z.

The Value Devils: diversified and liquid R2G with top performance
https://www.portfolio123.com/app/r2g/summary/1174555

Aug 29, 2015 3:25:14 AM       
z8735
Re: trying to implement the Hi-Lo method - what's wrong

oops - got it, I neglected to save.
Once saved you get the full range of the chart which in this case is 2011 (+40 days). Here is the modified screenshot. Now the remaining question: Why wouldn't it allow me to set dates before 2011?

Thanks,
Z.

Attachment Hi-Lo_3m_screen-after-save.png (102017 bytes) (Download count: 371)


The Value Devils: diversified and liquid R2G with top performance
https://www.portfolio123.com/app/r2g/summary/1174555

Aug 29, 2015 3:31:43 AM       
Market_Neutral
Re: trying to implement the Hi-Lo method - what's wrong

Try 2000 instead of 1999. It might be something to do with lookback prior to the start of the stock prices.
Steve

Aug 29, 2015 5:56:00 AM       
geov
Re: trying to implement the Hi-Lo method - what's wrong

You need a sell rule as well. My series formula is a bit different, but your formula also works with a trigger value of 4.7%. With daily rebalancing and no trading costs the performance is shown in the picture and is the same as for my model.

Attachment H-_Lo_3m.png (156140 bytes) (Download count: 368)


Aug 29, 2015 9:03:29 AM       
geov
Re: trying to implement the Hi-Lo method - what's wrong

And here is a nice performance of the healthcare ETFs. Uses FXH and prior to FXH inception uses XLV. When not in the healthcare ETFs it switches to IEF. Return is 15% with max drawdown of -12%. Anybody should be happy with this performance seeing there were only 45 trades over the last 15.7 years.

Attachment FXH or XLV-IEF.png (156416 bytes) (Download count: 359)


Aug 29, 2015 9:57:15 AM       
Edit 1 times, last edit by geov at Aug 29, 2015 9:58:36 AM
z8735
Re: trying to implement the Hi-Lo method - what's wrong

Thanks George, very nice and simple system.
But I still cannot get P123 to go back beyond 2011 with the custom series, I wonder why. With my series P123 reverts the start date back to 2011 whenever I try entering an earlier date. Do you have your series set to 1999 and P123 accepts that, else how do you get the backtest to go all the way to 1999?

Sell rule: Since your article suggests simply using a fixed threshold, then a P123 Screen should do it, Screens have an implicit sell and just rescreen every week. Of course a sim has more options and require explicit rules but I am not there yet, I just want to see the general response curve.

Best... Z.

The Value Devils: diversified and liquid R2G with top performance
https://www.portfolio123.com/app/r2g/summary/1174555

Aug 29, 2015 12:05:00 PM       
geov
Re: trying to implement the Hi-Lo method - what's wrong

Your buy rule is wrong. I am using your custom series with trigger value= 5%.
Here is my screen with transaction costs set to zero and weekly rebalance. Return is the same as for the sim.

Attachment H-_Lo_3m Screen.png (167121 bytes) (Download count: 347)


Aug 29, 2015 2:23:15 PM       
z8735
Re: trying to implement the Hi-Lo method - what's wrong

Hi George,
Nop, my result from 2011 and onwards are pretty much identical, pixel for pixel on the image, to the image you posted so the buy rule is probably correct. You can try it by copying and pasting the images (with some scaling) into a good pixel editor. The problem has to be related to P123 not providing access to data prior to 2011 for whatever reason, maybe something wrong with my use of custom series, or maybe an account setting or another error. Perhaps Marc Gerstein or Marco can offer an answer here.

BTW to everyone who is following this interesting take on timing: Changing the custom series to SP1500 and using EMA (with some adjustments) seem to be even better, but they are all good. Indeed this indicator, which can be called "general stock market mood" is a powerful one. It makes sense that a breakdown will be proceeded by an extended deterioration of the general mood. That doesn't cover all possible instances of a breakdown but it is useful nevertheless. Nice work.

Best... Z.

The Value Devils: diversified and liquid R2G with top performance
https://www.portfolio123.com/app/r2g/summary/1174555

Aug 29, 2015 4:20:53 PM       
Posts: 25    Pages: 3    1 2 3 Next
 Last Post