Post New

Flaws in timing with GTR1

The way to run a screen is:
1) The night before trading day
**** Compute the IN/OUT timing signal
**** Generate the screen picks
2) On the morning of trading day
**** If the timing signal is OUT, sell everything and go to cash
**** If the timing signal is IN, do the sells & buys according to the screen picks.
3) Wait until the next trading day. Do not sell or buy or look at the timing.


The timing in the below GTR1 screen does not work this way. It is fundamentally flawed for the above steps.
I have looked at the GTR1 steps but until today I did not see.
The GTR1 field is "Close all liquid positions and hold cash on any market dates where ..."

GTR1 will go to cash and then to the investments on any day, regardless of where it is in the holding period.
We do not trade a screen that way.

BTW, GTR1 evidently imputes interest on cash.


Screen:
https://gtr1.net/2013/?~nas100%20top5%20with%20BCC...

For full details, check these two boxes
Portfolio Values: Download daily portfolio values for all cycles of the focus variant.
Signal Values: Download daily values of any defined signal fields for the focus variant.

Here are a few examples.
The first shows that it goes to cash and back to stocks in the middle of a holding period.
The second shows it going to cash in the middle of one holding period and then back to stocks in the middle of the next holding period.
There are even times where it would flip back and forth every few days.

Date       Value    day# in holding period        BCC signal
10/04/94  $88.88    8                             IN
10/05/94  $91.76    9                             IN
10/06/94  $90.11    10                            OUT
10/07/94  $90.13    11                            OUT
10/10/94  $90.16    12                            OUT
10/11/94  $90.18    13                            OUT
10/12/94  $90.19    14                            OUT
10/13/94  $90.20    15                            IN
10/14/94  $90.40    16                            IN


11/19/07  $2,896.86 17                            IN
11/20/07  $2,913.03 18                            IN
11/21/07  $2,875.24 19                            OUT
11/23/07  $2,875.72 20                            OUT
11/26/07  $2,876.47 0                             OUT
11/27/07  $2,876.71 1                             OUT
11/28/07  $2,876.96 2                             OUT
11/29/07  $2,877.20 3                             OUT
11/30/07  $2,877.43 4                             IN
12/03/07  $2,850.50 5                             IN
12/04/07  $2,839.55 6                             IN

Here are all the OUT periods, the number of days in a row of OUT and the number of times that occurred.
It was OUT for just 1 day 19 times. It was OUT for 10 days twice.
Half of the OUT periods were less than 18 days.
The holding period is 21 market days.

Sequential days     Count
          out
          1         19
          2         7
          3         8
          4         2
          5         2
          6         4
          8         1
          9         1
          10        2
          15        1
          17        1
          24        1
          28        1
          35        1
          39        1
          40        1
          43        1
          49        1
          52        1
          55        1
          120       1
          196       1
          246       1
          413       1
          635       1

If anybody wants to see the spreadsheet that gathers all this information, let me know and I'll upload it somewhere. Or email it to you.

GTR1 will go to cash and then to the investments on any day, regardless of where it is in the holding period.
We do not trade a screen that way

Some people may trade this way. I had an email conversation with Robbie a long time ago about one of my screens. I had the BCC as a step, not cash if. He thought it better for to be a cash if command as it gave better results. I responded that the way I trade for example is for a 15 day hold screen is to have 15 consecutive trade days and let BCC control only that trade date. Psychologically, I couldn't go to 100% cash in one day. He thought otherwise.

Aussi

The screen referenced:

https://gtr1.net/2013/?~nas100%20top5%20with%20BCC...

may not be performing the timing that is intended. It is using a 2 day lag.

I downloaded GSPC and checked for the current price/sma(1,325) <1. One instance was 20250403 GPSC =5396.52, sma91,325) = 5494.899

Using this value on 20250403, the portfolio should be sold on 20250404 and show zero stocks. The referenced screen shows stocks on 20250404 and going to zero stocks on 20250405 two days after the crossover.

You can use GTR1 to check the timing by using the screen:

https://gtr1.net/2013/?~nas100%20top5%20with%20BCC...

This screen uses the same timing system, but uses it as a step rather than an import. When you do a stock count, it goes to zero on 20250404 as expected.

I sugest, if you want the timing to be effective the next trading day to change the following:

lf-1lp-1h1::BCC:gt0:BCC:linear(1,ifgt(ratio(gprc(0),sma(0,325)),1,1,0)){U:{!GSPC}}

change gprc(1) to gprc(0) and change sma(1,325) to sma(0,325)

change imports(0,BCC,2,-1) to imports(0,BCC,0,0)

However, if you want a 2 day lag on the timing, ignore evryuthing I wrote.

Aussi

You can use GTR1 to check the timing by using the screen:

It's the same screen. If you repost the right one I'll grab the data and analyse it.


...may not be performing the timing that is intended. It is using a 2 day lag.

I don't think the exact lag is important. You are just comparing " 'round about now" to "waaay back when". Today, yesterday, a couple of days ago, whatever. Just so long as you aren't using today's close to decide what to do today. We are making a coarse decision anyway -- the signal generated just before trade day determines if we are IN or OUT for the entire month.

Jim has suggested comparing the current 5 day average with the 5 day average NN days ago instead of using the instantaneous prices.

When the comparison is right at the cutoff, that's when any little wiggle can change the signal. Which is a good reason to add some hysteresis, a small difference between the sell & buy thresholds.

Post New

Mechanical Investing FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.