Testing the Duration Rotation Model Back to 1990 Using Proxy Data

The Quantifiable Edges Duration Rotation Model is currently available to new subscribers. Subscription information can be found here, and a detailed webinar recording is also available. The webinar was given earlier this month. This post is follow-up information with additional research.

When I designed the Duration Rotation Model in early 2020, I did so using data up through 2019. I settled on 3 different durations with which to implement the model. Those durations, and the associated ETFs, can be found below:

  1. 20 year treasuries (TLT)
  2. 7-10 year treasuries (IEF)
  3. 1-3 month t-bills (SHV)

One limitation of the design was that I could only run tests back to 2007. This is because SHV’s inception was January of 2007. Still, as I discuss in the webinar, I took care to make sure the system was robust, and even tested on alternative hypothetical data designed to mimic a bond bear market. I began trading the model near the end of May 2020. At the end of August I expanded the trading to many of my Eastsound Capital Advisors clients. Results thus far have been inline with what I had hoped for.

After attending the webinar, a prospective client sent me daily total return data for 30-year, 10-year, and t-bills since 1990, and asked me to see how the system might have worked using these instruments going back further in time.

The data showed close correlation to ETFs I use for the model, so it appeared to be a good way to test the effectiveness going back further. Before going back further in time, I decided I needed to compare returns using his data vs returns using the standard data and ETFs since 2008. The chart below shows growth of $1000 since 12/31/2007 for 1) 30-year bonds (blue), 2) Duration Rotation using the new proxy data (orange), 3) TLT (grey), and 4) standard Duration Rotation (yellow). Note for all tests I used the standard parameters for Duration Rotation. There were no adjustments made or optimizations run for any of these tests.

Several things to note about the chart:

  • 30-yr and TLT show strong correlation. 30-yr outperformed very slightly over the long run. It certainly appears to make a suitable proxy for looking back further in time.
  • DR-90, which is the Duration Rotation model using the proxy data back to 1990, outperformed TLT and 30-yr quite nicely. Keeping up with them while lowering the drawdown is an objective of the model, so this was encouraging.
  • Drawdowns were larger using the proxy data than using the standard ETFs, which hurt returns, especially recently. In comparing the trades, it appears this is because SHV tends to trigger more often than the t-bill proxy.
  • The outperformance during the live period over the last several months is primarily due to higher use of SHV. I find it encouraging that the actual securities manage to perform favorably vs the proxy in both backtested and live environments.

I decided to look at drawdowns in the Duration Rotation Model using proxy data vs ETF data a little closer. A comparison can be seen in the chart below.

What we see here is that it is not just the “max drawdown” that improved, but nearly every drawdown has been reduced with the standard ETFs vs the proxy data. I don’t have a great explanation for why SHV tends to work more effectively than the t-bill proxy data, but it certainly seems to.

Overall, the proxy data appeared suitable, and the proxy model appeared comparable, though perhaps not quite as good, as the standard model. Therefore, I went ahead and ran the model back to 12/31/1990, and compared the returns to the 30-yr treasury data. (This was a nervous moment, since you never know what will happen when you add 17 years of new data to an already established model.) Results are below.

This was everything I could have hoped for. The Duration Rotation Model, even using the proxy t-bill data rather than SHV, would have managed to keep pace with TLT and provide reduced drawdown over the long run. Overall, these additional tests have provided me even greater confidence in the Duration Rotation Model.

Monday & Tuesday Both Up – Did Thanksgiving Bullishness Just Get Front-Run?

A few days ago, I noted the seasonal bullishness around Thanksgiving, especially from Tuesday’s close to Friday’s close. One potential issue with seasonal studies is that they can sometimes get front-run, and then the edge weakens. It is now Tuesday, and the market is moving up strongly for the 2nd day in a row. So I decided to look back at other times where SPX closed higher on the Monday and Tuesday before Thanksgiving. The table below shows all instances since the inception of the modern S&P 500 in 1957.

Monday and Tuesday Both Up Prior To Thanksgiving. 2-day forward odds.

In this case, it does not appear that the Monday-Tuesday rally has eliminated the seasonal bullish tendency. In fact, it appears that the momentum has carried well through the holiday. Happy Thanksgiving to you and yours!

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

Updating Thanksgiving Week Historical Odds

The time around Thanksgiving has shown some strong tendencies over the years – both bullish and bearish. I have discussed them a number of times over the years. In the updated table below I show SPX performance results based on the day of the week around Thanksgiving. The bottom row is the Monday of Thanksgiving week. The top row is the Monday after Thanksgiving.

Monday and Tuesday of Thanksgiving week do not show a strong, consistent edge. But the data for both Wednesday and Friday looks quite strong. Both of those days have seen the S&P 500 rise at least 70% of the time between 1960 – 2019. The average instance managed to gain about 0.3% for each of the 2 days. (This is shown in the Avg Profit/Loss column where $300 would equal a 0.3% gain.) That is a hearty 1-day move. Meanwhile, the Monday after Thanksgiving has given back a good chunk the gains that the previous 2 days accumulated. It has declined 65% of the time and the average Monday after Thanksgiving saw a net loss of 0.34%.

In 2018 I gave a few more thoughts on the Wednesday edge and the importance of Tuesday’s action for Wednesday.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

One Look At Monday’s Massive Rotation

Monday saw a massive market rotation. It could be noted by the performance in the IWM vs the QQQ, or in looking at performance among S&P 500 sectors, where Energy beat Technology by 15% on Monday. But to really see how strong the rotation was, you’d need to take a look at individual stock performance within the SPX. Below is a list of the Top 10 S&P 500 stocks, ranked by YTD performance through Friday, 11/6.

Top S&P stocks and their Monday 11/9 performance

Every one of the Top 10 stocks closed down on Monday, with 6 of the 10 losing over 5%. Now let’s look at how the worst stocks of the year did on Monday.

Bottom YTD S&P stocks and their 11/9 performance

Every single one of these were up at least 14% on Monday, with most of them rising between 20% and 40%. That is a massive rotation. One way to play momentum is to get long strength and to get short weakness. That can be a profitable strategy. But on days like Monday, when the public thinks the “market” did well, you can really get served a world of hurt. Anyway, the rotation was enormous. I’m not sure what it means for the major averages going forward, but I certainly found it notable.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

SPX Performance After Big Weekly Reversals

After losing 5.6% this in the week ending 10/30/20, the S&P 500 completely reversed the losses this past week with a 7.3% gain. That is a fairly remarkable turnaround. Below is a look at all other times the S&P 500 lost 5% or more one week, and then made up for the losses and more the next week.

Instances are very few. Three in the 1930s, three in the 2007-2009 bear market, and then once in 2010 before this past week. Results are mostly mixed, with some struggles occurring over the next 13 weeks. Does not appear to be a great signal, but expecting the turnaround to continue to rocket higher from here would seem to be overly optimistic.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

The Last 4 Days Show Nice Momentum

I saw some chatter on Twitter on Thursday about how rare it is to see SPX close up > 1% for 4 days in a row. In fact, when filtering to see times it has occurred since 1970 with SPX closing above the 200ma, there was only one other instance – 10/11/82. But we did have a study in the Quantifinder that looked at strong 4-day win streaks that required at least 3 days close up 1% or more. That gave a better sample size, and can be seen below.

Numbers here are lopsided to the bullish side. The market is certainly short-term overbought, but this study suggests a good chance at becoming more overbought.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

Recent Afternoon Selling Could Be Setting SPY Up To Bounce

Despite attempts to put in rallies over the last few days, SPY has been persistently weak in the afternoon. It has not managed to hold gains and has closed in the bottom 25% of its daily range for the last 4 days in a row. That may not sound all that extreme, but it is quite rare. Below is a look at how the market has performed following past occurrences. It looks back to the inception of SPY in 1993.

SPY Closes in Btm 25% Range 4 Days In A Row

The numbers are very lopsided in favor of the bulls. Below is a list of all instances assuming a 5-day holding period. The bears have owned the afternoons lately. It appears in the next few days the bulls will have their chance, and this study suggests an afternoon that does not go to the bears could turn into a sizable move higher.

I do not see anything here that alarms me. The bears have owned the afternoons lately. It appears in the next few days the bulls may have their chance. This study suggests an afternoon that does not go to the bears could turn into a sizable move higher.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

The Next 5 Weeks All Are Among The Weakest – And Strongest – Of The Year

October is a month that is known for volatility. And that is a well-earned reputation. Crashes in 1929, 1987, and 2008 all occurred in October. But volatility cuts both ways. If you break the year down into 1-week periods, October also contains some of the strongest seasonal edges of the year, both bearish and bullish. Breaking the year down by week is something I have done numerous times over the years, and it has provided some interesting insights. The table below shows stats back to 1985. I chose 1985 as the start date because SPX options trading began in 1984, so 1985 is the 1st full year where there was an options expiration schedule. We have found over the years that options expiration can generate strong edges. Action on and around options expiration, which occurs on the 3rd Friday of each month, helps to generate some seasonal tendencies. So this study encompasses the full range of time that SPX options have been in existence.

Strongest and Weakest Weeks of the Year

You’ll note that the highlighted weeks are the October weeks. It’s amazing that all 5 potential weeks in October are included in either the Best 10 or the Worst 10 weeks of the year. Weeks are ranked based on Avg Trade (last column).

Weeks following the 1st and 3rd Friday in October have been among the worst on average. I’ll also note that the 2 largest “max losing trades” occurred following the 1st and 3rd Friday in October. They were a 12% and an 18% drop.

Weeks following the 2nd, 4th, and 5th Fridays in October made up 3 of the best 10 weeks of the year on average. And the 2nd and 4th week show two of the largest winners as well, with 7.3% and 10.5% max gains.

As I said, October can be volatile. Traders may want to keep the patterns of the last 35 years in mind. Seasonality will be swinging back and forth between bullish and bearish. Paying some attention might help your accounts survive through Halloween.

I’ll also note that Tom McClellan wrote an interesting piece on October seasonality during election years.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

Is the Weakest Week Still Weak When There is Weakness Prior to the Weakest Week?

As I have shown many times in the past, there isn’t a more reliable time of the year to have a selloff than this upcoming week. I have often referred to is as “The Weakest Week”. Since 1960 the week following the 3rd Friday in September has produced the most bearish results of any week. Below is a graphic to show how this upcoming week has played out over time.

2019-09-20-1

As you can see the bearish tendency has been pretty consistent over the last 60 years. There was a stretch in the late 80’s where there was a series of mild up years. Since 1990 it has been pretty much all downhill. Below is a table showing results of buying Sept. op-ex Friday and then selling X days later from 1990 – 2019.

2019-09-20-2

The consistency and net results appear quite strong. I note the only instances that didn’t post a lower close at some point during the following week was in 2001 and 2017. The 9/11 attacks certainly made for unusual circumstances in 2001, and 2017 did not see a decline, but it only rose 2 points, so it was not much of a victory for the bulls.

But this year seems a little different, because this is the 1st time since I have been running these tests that we are headed into the “weakest week” having already closed down 2 weeks in a row. (It has actually been three weeks.) That had me wondering whether the seasonal weakness may have come a little early this year. So I ran a test to look at other instances since 1960 that SPX had at least 2 losing weeks in a row leading up to the post-3rd Friday setup.

This certainly does NOT suggest the seasonal weakness came early. Nine of ten instances where there were 2 down weeks leading up to the “weakest week”, will still saw the “weakest week” close lower. And all 3 instances where there were 3+ down weeks still had at least one more down week to go. So I am viewing poor seasonality as a bearish short-term factor this upcoming week.



Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

Looking at 7-day Win Streaks

The recent rally has left the market short-term overbought by most measures. Short-term overbought often triggers some studies that suggest a downside edge, but when the overbought condition gets very strongly overbought, then those downside edges often disappear. And at some point, rather than strength leading to weakness, the strength will beget more strength. The strong move higher over the last several days has turned the market so overbought that we are seeing this scenario begin to unfold. The study below looks at the 7-day win streak in the Dow Jones Industrial average. It exemplifies this concept.

Performance after 7 up closes for DJI. Bullish.

There is not much of an edge over the 1st few days. But once you get out a little further, the stats appear solidly bullish. Below is a look at the profit curve assuming a 19-day holding period.

7 up days. 19-day later profit curve

The strong move from lower left to upper right appears to offer some confirmation of the bullish tendency. In last night’s subscriber letter, I also looked at the same setup with SPX, since SPX has also posted 7 higher closes. Results were very  similar. Short-term momentum has been so strong recently that it appears likely to carry through some on an intermediate-term basis.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

SPX Golden Cross History Since 1928

SPX will post a Golden Cross on Thursday afternoon. A Golden Cross occurs when the 50ma crosses over the 200ma. Having the 50ma above the 200ma is commonly considered a bullish market condition – and generally it is. In the 4/2/19 blog post I looked at SPX Golden Crosses dating all the way back to 12/31/1928. I have updated that research tonight with Amibroker Software and Norgate Data. Below is a list of all Golden Crosses since then. (Note that prior to 1957, S&P 90 data was used. The S&P 90 is considered the predecessor to the S&P 500.)

SPX Golden Cross results since 1928

The most recent Golden Cross formation got slammed by the March COVID crash. But since the 1961 trigger, the Golden Cross has generally served as a fairly good timing device to sidestep large portions of bear markets.  Prior to that it was not nearly as effective. This can be seen in the drawdown chart below.  (Note: Overnight Fed Funds rate was used to calculate nightly interest starting in July 1954, which is as far back as my data goes. Prior to that, no interest rate was assumed when out of the market.)

While drawdowns have been mostly fairly moderate since the mid-50s, prior to that there were some very large drawdown to endure. Of course, the 2020 drawdown is the biggest since 1940 – and we still have a good ways to go for the system to dig out of it.  Despite some fairly sizable drawdowns, the Golden Cross would have beaten “Buy and Hold” handily.  It is a bullish long-term trend indication.  But it is not a bulletproof long signal. 

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

First Day of Month Based on Month

Since the late 80s there has been a tendency for the market to rally on the first day of the month.  One theory on why this occurs is that there are often 401k inflows that are put to work on the 1st of the month.   I examined this tendency and broke it down by month on the blog in 2013 and 2009.  I thought it would be interesting to take another look at it today.  Below is an updated version of the study.

SPX 1st day of month broken down by month

July has been the most reliable month in term of closing positive. And when looking at the size of the average gain, it comes in 2nd.  You also note that August is the worst on both counts.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

What Friday’s Action Suggests About Short-Term Volatility

It was unusual on Friday to see such large opening gap (+2.54%) filled, and then SPY still managed to close positive. This can be seen in the study below, which I showed in the subscriber letter over the weekend.

SPY 2% Gaps Up that fill and then close positive.

Four instances closed higher, and 6 closed lower over the next 2 days. So there does not appear to be a strong directional bias. But the takeaway is that the volatility did not just disappear. At the bottom you’ll see I circled stats looking at avg “run-up + 1 standard deviation” and avg drawdown – 1 std deviation”. What those mean is that it would not be unusual to see the market rise as much as 6% or decline as much as 9% over the next couple of days. That is a huge spread for a 2-day time span. It speaks to the potential risk of being wrong.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.

SPX Historically Bullish On Thursday After Memorial Day

Thursday after Memorial Day has been a day that has exhibited a bullish bias for many years. I last showed this on the blog a couple of years ago. The chart below shows updated results.

SPX Perfromance on Thursday After Memorial Day

Single-day seasonality can certainly be overrun by other forces, but the Thursday after Memorial Day has been a good one for many years. That may be something that traders want to keep in mind for Thursday.

Want research like this delivered directly to your inbox on a timely basis? Sign up for the Quantifiable Edges Email List.