June 12, 2019

Market Timing with a Canary, Gold, Copper, LQD, IEF and much more

One commonality in my strategies is the inclusion of a market timing component. This could be a signal to go into cash or reduce position size or enter a ‘safe’ ETF. This applies to my swing trading strategies, my monthly rotation strategies and my Tactical Assert Allocation strategies. As a researcher, I am always on a looking to improve this part of my strategies.

There have been a handful of market timing methods I have been wanting to test and compare with my current 200-day moving average version. I collected enough of them to test all at once and to compare the results.

The Test

The backtest is from 1/1/2004 to 12/31/2018 on the SPY, dividends included. I start in 2004 because that is as far back as I have data for two symbols that I need. For today’s post, I will focus on only looking at the signal at the end of the month. The reason for this is that both my rotation and TAA strategies trade like this and this might help the most there. In a future post I will look at the effects of trading on a different day of the month.

Buy Rule

  • Last trading day of the month
  • Market timing gives a buy signal
  • Enter on next Open

Sell Rule

  • Last trading day of the month
  • Market timing gives a sell signal
  • Exit on next Open

Nothing complicated. While not in the SPY we will be in cash earning no interest. Improving the cash portion will be a topic of a future post

The Strategies

200-day Moving Average

This is the most common one we have all read

  • Buy signal: Close is above the 200-day moving average
  • Sell signal: Close is below the 200-day moving average

 

200-day Moving Average 5 days

This is my current favorite. Will it change after this post?

  • Buy signal: Close is above the 200-day moving average for 5+ days
  • Sell signal: Close is below the 200-day moving average for 5+ days

 

1% Band

I developed this one years ago while working for Larry Connors.

  • Buy signal: Close is 1% or more above the 200-day moving average
  • Sell signal: Close is 1% or more below the 200-day moving average

 

Golden Cross

I had to include this popular one.

  • Buy signal: 50-day moving average is above the 200-day moving average
  • Sell signal: 50-day moving average is below the 200-day moving average

 

Jay Coppock

With a post titled, The Only Indicator You Will Ever Need(?), I had to test it. For more details read the post.

Coppock Guide = The sum of a 14-month rate of change and 11-month rate of change, smoothed by a 10-period weighted moving average

  • Buy signal: Coppock Guide closes this month above its level of 3 months ago
  • Sell signal: Coppock Guide closes this month below its level of 3 months ago

 

13-Week High/Low

From my Hi-Lo Index as a Market Timing Indicator post. That was on stocks but I must try this on the SPY.

hiloValue = 40-day moving average of 100*(# stocks making 13-week highs-# stocks making 13 week low)/500

  • Buy signal: hiloValue greater than 5
  • Sell signal: hiloValue less than 5

 

LQD/IEF Ratio

From Adam Robinson on the The Knowledge project podcas, a non-stock market podcast, he makes an offhand remark about the ratio of LQD (Investment Grade Corporate Bond ETF) divided by IEF (7-10 Year Treasury Bond ETF) being am good market timing indicator. He gave little details.

I had to play around a little with this given the lack of details

Ratio = LQD close divided by IEF close

  • Buy signal: ratio closes above the 200-day exponential average for 5+ days
  • Sell signal: ratio closes below the 200-day exponential average for 5+ days

 

Canary

This idea gave from Trend Following on Steroids post. For both BND (Total Bond Market ETF) and VWO (FTSE Emerging Markets ETF), calculate their momentum. From the post:

So far so good. But how do we identify the right canary assets? Well, we refer to our prior work for details, but in short, we searched for the best canary universe over the in-sample period from Dec 1926 – Dec 1970. We found a two-asset universe consisting of the US Aggregate Bond and Emerging Markets as the optimal canary universe to identify the trend of SPY in the next month. In the following, we will use two Vanguard ETFs — BND and VWO — to represent both assets.

What is the intuition for using BND and VWO as a canary universe? One hypothesis is that SPY is sensitive to yields (higher yields leading to lower SP), and a similar mechanism may hold for US currency (higher USD leading to lower SPY). And higher yields/USD are reflected in lower BND/VWO prices, resp. In addition, VWO may also signal emerging market unrest. Regardless, what matters is the out of sample results of the strategy.

Momentum = 12 times 1-month return plus 4 times 3-month return plus 2 times 6-month return plus 12-month return. Then divide by 4

  • Buy signal: momentum for BND & VWO are both above zero
  • Sell signal: momentum for one of BND or VWO is below zero

 

Copper/Gold Ratio

This is from the same podcast of the LQD/IEF Ratio. He commented about using the Copper and Gold ratio. I figured this had no chances on working. The reason he gave is that copper is used in lots of industries when they are doing well, this price goes up. While when the markets are bad, gold goes up.

Ratio = Copper (continuous future contract) close divided by Gold (continuous future contract) close

  • Buy signal: ratio closes above the 200-day exponential average for 5+ days
  • Sell signal: ratio closes below the 200-day exponential average for 5+ days

 

How are we going to compare?

I will use these statistics to compare the results

  • CAR – Compounded Annual Rate
  • MDD – Maximum System Drawdown
  • RAR – Risk Adjusted Return. This is CAR divided by exposure.
  • SR – Sharpe Ratio

Compounded Annual Rate

I included Buy and Hold on the SPY as a base comparison. The results fall into two general buckets. Those with CAR between 8.3 and 8.8 and they beat Buy and Hold. Those with CAR below 7.4. The results with CAR above 8.3 are very similar. The Canary method stands out because the high number of trades compared to the other methods. The Copper/Gold did poorly as expected.

Maximum System Drawdown

Ranking by drawdowns, I see three buckets. Now Copper/Gold has the best but this is probably partly due to the low exposure. Ranking this way, I like the LQD/IEF method. But the results in the middle buckets are all similar.

Risk Adjusted Return

Ranking by RAR, I see three buckets. Again, Canary method stands out. Interestingly Copper/Gold is near the top. Jays Coppock which had been in the bottom using the other two ranking methods moved near the top but the drawdown is by far the worst.

Sharpe Ratio

Our final ranking by Sharpe Ratio produces three buckets. The Canary method again at the top.

Potential Issues

The main problem with most market timing methods is the number of trades. The methods had between 6 and 26 trades which is not very much to decide how valid they are. Now I could test further back with most of the methods, but to me the farther you go back the more the markets differ from today. Even if I double the lookback and the trades doubled, it still would not be a lot.

Spreadsheet

File the form below to get the spreadsheet with lots of additional information. This includes top drawdowns, trade statistics and more.

Final Rankings & Thoughts

Taking the raking for each method for each statistic, I then add them up to see which ranked best overall.

The Canary, Copper Gold and the LQD/IEF methods are near the top of the rankings. These are not direct measures of the market which is interesting. I must spend more time with them to see if I want to incorporate them into my trading.

The MA200, MA200-5Days and 1% Band fell in the middle of the results. This is good to see for me since I use the MA200 and MA200-5Days.

The Golden Cross did worse than I expected.

Jays Coppock and HiLo consistently did not do well.

Buy and Hold fell at the bottom for most rankings. And this is why I use market timing methods.

What are your thoughts on these? Any favorites? Add to the comments below.

Follow on posts: Day of Month and Market Timing and Market Timing and Bond ETFs

Is your trading idea profitable?

Want to backtest your trading idea? Want your strategy evaluated? Contact me for a fixed bid on your project. 100% confidential.

 

Backtesting platform used: AmiBroker. Data provider: Norgate Data (referral link)

Good quant trading,

Fill in for free spreadsheet:

spreadsheeticon

 
Privacy Policy

Click Here to Leave a Comment Below

Fran - June 13, 2019 Reply

The guy is Adam Robinson on the The Knowledge project podcast. I found similar results for copper/gold ratio…not much of a buy signal, but a very good sell signal with potentially heavy drawdowns for risk assets. I also tested on a shorter time frame. I enjoyed this article. Thanks for writing it.

    Cesar Alvarez - June 13, 2019 Reply

    Thank you so much! I will update the post.

Chris - June 13, 2019 Reply

Might be interesting to see how an ensemble approach could perform.

Like say, if you take all the individual indicator strategies (or just a diversified mix of the most promising ones – the Moving Average ones for example are probably all very tightly related and will not give wildly different allocations) and average their signals into a meta strategy.

This way you’re not always just either binary all in or all out, but at varying levels of balanced exposure (for example if you have 10 individual indicator strategies, 6 of them say Buy, 4 say Close, you’d end up with a 60% SPY / 40% Cash allocation for that period). The number of concurring indicator signals on each side can thus be seen as sort of an expression of “confidence” (ie. the degree on which more indicators agree on a certain positioning).

Intuitively from experience with other ensemble approaches, this method will probably not improve returns (which will still be average and not as good as the best single indicator strategy) so much as improve the risk measure side of the equation.

Or if you still want to keep the binary “either fully invested or fully out” Allocation choice outcome, you could use the ensemble of indicator signals as a trade filter, for example “only buy/sell when more than half (or more than two thirds etc if you want a higher “confidence”) of the indicators agree on it”.

Just some ideas for future tests 🙂

    Cesar Alvarez - June 13, 2019 Reply

    Have you been spying in my research notebook? I am working on something like that. Maybe a future post.

Scott - June 13, 2019 Reply

Excellent work and recap; love to see more like this!

    Cesar Alvarez - June 13, 2019 Reply

    The next post will be on market timing with different check dates. And I have a potential idea for the post after that which Chris mentioned in the comments.

MARCO - June 13, 2019 Reply

Hi Cesar, I’ve never heard about the Canary strategy… It’s simply great! I believe intermarket strategies are the best ones!
Marco Simioni
http://www.nightlypatterns.blog

Ola - June 16, 2019 Reply

Hi Cesar, thanks for another great post!
I’m sure there are lots of potential variations to test. Using one method for entry and another for exit?
I’m looking forward to hear about ideas for the cash component.

Ryan Watson - August 7, 2019 Reply

fyi its not Jay Coppock but Jay Kaeppel

Thanks for article.

Mattia - August 20, 2019 Reply

Hi, thanks a lot for this post! I’ve actually found more than one post very interesting 🙂 One question about LQD/IEF Ratio:

“ratio closes below the 200-day exponential average for 5+ days”

The 200-day exponential moving average (EMA) refers to:
1. the moving average of the ratio (that is, EMA(LQD/IEF) ), or
2. the ratio of the moving averages (that is, EMA(LQD)/MA(IEF) ) ?

Thank you!

    Cesar Alvarez - August 20, 2019 Reply

    Good question. It is #1. I like your second one. I might have to test that.

greg - September 9, 2019 Reply

I got higher returns using jnk/iei ratio –

    Cesar Alvarez - September 9, 2019 Reply

    Thanks for sharing that research result

Chris Bern - October 24, 2019 Reply

Very interesting post and great work–thanks for sharing! Just a quick comment, I also heard the podcast with Adam Robinson, and I believe he uses Copper/Gold more as a predictor of rates (and therefore more likely as a trading tool for bonds, not equities). Basically he said Copper/Gold ratio is highly correlated with 10-yr yields (TNX), and in the last 18 years when those two indicators have diverged, that the direction of Copper/Gold was always correct and “early”, meaning yields would eventually follow. So for example at the time of the podcast (Aug-2018) Copper/Gold was near a 1-yr low and TNX was near a 1-yr high, and so his prediction was TNX would decline. And now we can say over a year later his prediction was correct, the 10-yr yield went from 2.99% at the time of the podcast all the way down to 1.76% today.

    Cesar Alvarez - October 24, 2019 Reply

    I need to relisten to the podcast to see if I come up with the same conclusion. THank you for pointing this out.

James Coltman - April 9, 2021 Reply

Super interesting post. When I look at the inception dates for BND (2007/04/03) and for VWO (2005/03/04) they are well after the start of your backtest (1/1/2004). What did you use to bridge the gap?

    Cesar Alvarez - April 9, 2021 Reply

    I typically find a mutual fund that is similar and use that data. But I cannot find my notes from this test so I cannot confirm that is the case. Really annoying. Sorry.

Leave a Reply: