Skip to main content
No. of Recommendations: 1
Does anyone know if we had a new Coppock signal back in March/April? I realized I hadn't heard about the signal in a while.

John
Print the post Back To Top
No. of Recommendations: 8
No signal.
It it based in large part on the notion that a downturn takes a certain minimum or typical amount of time to play out.
It won't signal till it sees an upturn after a downturn which is a certain mix of long and deep.
The current dip isn't yet long at all.

Arithmetically, it signals a "buy" when a certain slow smoothed index oscillator starts rising from a negative value.
That value isn't negative yet.
So, not only is there no buy signals, but it is not yet certain that there will be one in this downturn.
(I expect that situation to change, but it hasn't yet)

Its last buy signal was at the end of May 2016, with the S&P at 2097.
That dip was not particularly deep at all, but it lasted long enough to qualify for the test: the market had been net flat for about a year and a half.

Jim
Print the post Back To Top
No. of Recommendations: 2
And just how were the values used in the standard
timing of the Coppock Curve determined
i.e., 14,11,10 used in stockcharts.com?

From Wikipedia-

"Coppock, the founder of Trendex Research in San Antonio, Texas, was an economist. He had been asked by the Episcopal Church to identify buying opportunities for long-term investors. He thought market downturns were like bereavements and required a period of mourning. He asked the church bishops how long that normally took for people, their answer was 11 to 14 months and so he used those periods in his calculation".

Scientific???
I don't think you would be accused of datamining using some different values- IMHO.
Try using some different values; EXCEL's Solver add-in will give you pleasing
results for the major indexes.
Print the post Back To Top
No. of Recommendations: 15
And just how were the values used in the standard timing of the Coppock Curve determined...
Scientific???
I don't think you would be accused of datamining using some different values- IMHO.


Heck no, I don't think anybody would call that scientific.
And despite the "grief" narrative in the story, I suspect it was primarily fit to data up to that point.

But I would be a bit hesitant to try different values, if I were you.

Its main value as a signal is the extraordinary out-of-sample validation: it was published with all details on a known date 57.6 years ago, on 1962-10-15.
And it has worked remarkably well since then, for its intended purpose: above average entry points for long holds of the broad market.
These are the figures since it was published:
The average S&P real total return if there has been a buy signal in the last 19 months: CAGR inflation+14.3%/year. (about 37% of the time)
The average S&P real total return in other periods: CAGR inflation+2.2%/year (about 63% of the time)
Difference: 12.0%/year
That's pretty good out of sample validation.
Plus, it's not exactly frenetic trading: on average there is only one buy signal every 4.1 years.

The 19 month forward return I quote is pretty arbitrary...it seems to have predictive power on average for up to 2.5 years after a "buy", give or take.
You get different figures but the same general conclusion for other comparable look-ahead time frames.
For example, if there has been a signal in the last 30 months, real CAGR 10.9%; if not, real CAGR 1.2%; difference 9.7%/year.
If you just want to tag the times with the bad omen, the "lowest CAGR when not bullish" is if there has been no buy in the last 31 months, real CAGR 0.9%.

As published there was no sell signal, only buys. But it's not hard to imagine what it would have been.
The buy happens when the oscillator turns upwards from a negative value.
A continuous signal can be made which is bullish when it's rising or positive. Bearish if it's falling and negative. Again, checked on month ends only.
That wasn't in the original article, but seems to be the most obvious elaboration.
Since the article was published, that would have returned real CAGR +8.8% when bullish and real CAGR -6.8% when bearish.
That's certainly better discriminating power than most timing signals manage post publication.

Jim
Print the post Back To Top
No. of Recommendations: 2
Does anyone know if we had a new Coppock signal back in March/April?

Speaking of signals, if the Naz goes up just a tiny bit more we'll have a 99-day reset signal on QQQ (which, IIRC, is what Zeelotes liked to use).

DB2
Print the post Back To Top
No. of Recommendations: 0
As published there was no sell signal, only buys. But it's not hard to imagine what it would have been.
The buy happens when the oscillator turns upwards from a negative value.
A continuous signal can be made which is bullish when it's rising or positive. Bearish if it's falling and negative. Again, checked on month ends only.
That wasn't in the original article, but seems to be the most obvious elaboration.
Since the article was published, that would have returned real CAGR +8.8% when bullish and real CAGR -6.8% when bearish.


Jim,

Do you have the % of time to date since the original Coppock article that the signal would have been bullish vs. bearish, using the definitions you described above? (Back on 8/23/15, I found that you reported it would have been bullish 86% of the time up to that point.)

Also, I found that a contributor 'whafa' used to post monthly signals using a spreadsheet you had created.

I *think* I was able to re-create the spreadsheet, but wanted to check my Coppock signal values against what you have for this year if I might:

1/31/20   0.236756
2/29/20 0.241614
3/31/20 0.175164
4/30/20 0.162019
5/31/20 0.150660

Do these values look accurate to you?

Thank you as always,

Todd
Print the post Back To Top
No. of Recommendations: 0
I *think* I was able to re-create the spreadsheet, but wanted to check my Coppock signal values against what you have for this year if I might:


1/31/20 0.236756
2/29/20 0.241614
3/31/20 0.175164
4/30/20 0.162019
5/31/20 0.150660


Do these values look accurate to you?


I get the same numbers. But I show 1/1/20 for your 1/31/20, 2/1/20 for your 2/29/20, etc.

Elan
Print the post Back To Top
No. of Recommendations: 1
BarryDTO posted a “Coppock timing in GTR1” in 2016 were he compares it to the individual BCC signals.
The thread:
https://boards.fool.com/coppock-quottimingquot-in-gtr1-32145...
His Coppock Run to current date:
https://gtr1.net/2013/?s19740104lf-1lp-1h1i1f0.05::pref%28sp...
1974 till present with pref(sp500.a,sp90.a) = 1; 12.3% CAGR, 16.0 GSD MDD -33.5, UI 8.55

RAM
Print the post Back To Top
No. of Recommendations: 6
Do you have the % of time to date since the original Coppock article that the signal would have been bullish vs. bearish, using the definitions you described above?

Bearing in mind that the bull and bear version wasn't in the original article, sure.
It has been bullish 89% of the time since January 1978.
It was bullish 78% of the time 1930-1977.

Do these values look accurate to you?

Yup, perfect match for me.
Bullish if the number is positive, bullish if it's higher than it was the previous month, or both.

Long run returns of a long/cash strategy are actually higher if you let it be bearish even less of the time.
Try replacing the ">0" with ">-1.5%"
This only changes a few months, so it's far from being statistical support for the change.
But it does backtest a whisker better, so it might be in the realm of "harmless overtuning".

Jim
Print the post Back To Top