Kaeppel’s Big vs Small Strategy
This is a follow up to Jay Kaeppel’s Big vs Small Strategy (h/t The Whole Street).
We covered this one back in 2009, and while I think it’s too early to call this long-term strategy superior to conventional trend-following, the strategy has performed reasonably well (out of sample) since.
[growth of $1, logarithmically-scaled, frictionless]
The red line in the graph above shows the results of going long the S&P 500 when the S&P 500 (large-caps) has outperformed the Russell 2000 (small-caps) over the previous 252-days (1-year), otherwise to cash.
Like Kaeppel, I’ve assumed a one day lag in executing trades (i.e. a trade triggered at Monday’s close will be executed at Tuesday’s close), and for comparison, I’ve included buy & hold in grey.
Geek note: I used the price-only S&P 500 index to determine when to enter/exit trades, but used dividend-adjusted data to show performance. Also, I’ve assumed a return on cash of half the nearest 13-week Treasury.
For the anal amongst us, note that I’ve replaced Kaeppel’s Russell 1000 (RUI) with the S&P 500 in this test because of their similarity and because more data is available on the S&P 500.
. . . . .
Like conventional trend-following approaches, the benefit of this strategy has not been generating outsized returns, but improving risk-adjusted returns and reducing drawdowns.
Importantly, as the table below shows, the strategy has worked using a broad range of lookbacks other than 252-days. Below I’ve shown results for 12, 9, 6, and 3 month variations of the strategy.
There has been a sweet spot right around the 9 month lookback, but the outperformance of that particular value (as opposed to 3, 6 and 12 months) is probably more curve fitting than anything else.
. . . . .
I’ve never been too enamored with this strategy only because of how little data we have to consider relative to how infrequently it trades, but I continue to keep it on my radar.
FWIW, performance of the strategy since I first described it in 2009:
[growth of $1, logarithmically-scaled, frictionless]
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Trading Strategies | 7 Comments
A topic recently on me noggin has been day-of-month seasonality (read more, more, and more). Using a simple walk-forward test to minimize hindsight bias, I showed that trading the days of the month that have been strong historically has consistently led to much stronger returns in the future. That’s as true today as it was in 1950.
Below is the DOM seasonality calendar for next month, broken out by quartiles (read why), with quartile 1 indicating the strongest days and quartile 4 the weakest.
Real-time results since I began sharing the calendar in October have been mostly inline with the historical test, with the S&P 500 averaging 0.12% (35% annualized) on the best half of days versus 0.01% (3% annualized) on the worst half.
Quartile 4 days (the worst of days) have been particularly bad, with an average return of -0.29%.
As I stressed when I introduced the concept, day-of-month seasonality never justifies a trade all by itself, but I do think it deserves to be one of many tools in the trader’s toolbox.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Time-based | 5 Comments
A topic recently on me noggin has been day-of-month seasonality (read more, more, and more). Using a simple walk-forward test to minimize hindsight bias, I showed that trading the days of the month that have been strong historically has consistently led to much stronger returns in the future. That’s as true today as it was in 1950.
Below is the DOM seasonality calendar for next month, broken out by quartiles (read why), with quartile 1 indicating the strongest days and quartile 4 the weakest.
I made just two exceptions: 03/20, the last day of the FRB meet, and 03/28, the day before Good Friday.
As I’ve shown previously, Fed days have been consistently bullish events for the last two decades, so even though 03/20 was a quartile 4 day, I’ve highlighted it dark green. And the day before exchange holidays tends to be slightly bullish, so even though 03/28 was originally a quartile 3 day, I’ve left it a question mark.
. . . . .
Real-time results since I began sharing the calendar in October are mostly inline with the historical test, with the S&P 500 averaging 0.09% (24% annualized) on the best half of days versus 0.01% (3% annualized) on the worst half.
Quartile 4 days (the worst of days) have been particularly bad, with an average return of -0.32%. Quartile 1 days (which should be the best of days) have so far been middling, with an average return of just 0.02%.
As I stressed when I introduced the concept, day-of-month seasonality never justifies a trade all by itself, but I do think it deserves to be one of many tools in the trader’s toolbox.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Time-based | 2 Comments
A reader asked an interesting question: what does the VIX futures term-structure (contango versus backwardation) say about future stock market returns?
My off-the-cuff answer was that the VIX term-structure is only predictive for volatility ETFs like VXX or XIV (because of the impact of the underlying VIX futures converging to the VIX spot), and that it probably isn’t predictive for the stock market itself.
But I had never given the question much thought and felt it deserved some ink.
[logarithmically-scaled, growth of $1, frictionless]
The graph above shows the result of buying SPY at today’s close when VIX futures will end the day backwardated (month 1 > month 2) in grey, or contangoed (month 1 < month 2) in red, since 03/2004. Trades are held until a switch in the term-structure.
Numbers for the number lovers…
Based on the limited data available, the state of the VIX term-structure hasn’t been consistently predictive of future stock market returns. The market has been very bullish even when backwardated (see 2007/08), and very bearish even when contangoed (also see 2007/08).
The only observation that has been consistent is that future SPY returns when VIX futures are backwardated have been significantly more volatile. This should make sense given that backwardation is usually the result of market volatility/crises.
In short, my off-the-cuff response was confirmed.
. . . . .
An interesting side note.
Recall my previous post showing the results of a common strategy trading VXX/XIV by blindly following the VIX futures term-structure (long VXX when backwardated, long XIV when contangoed). More importantly, recall the horrific 90% slide such a strategy would have suffered in 2007/08:
[logarithmically-scaled, growth of $1, frictionless, updated through 12/2012]
The first graph in this post illustrates why that slide occurred.
When VIX futures were contangoed (long VXX), the market steadily rose, which was bad for VXX. When futures were backwardated (long XIV), the market fell, which was bad for XIV.
That’s a nice reminder that the VIX term-structure can’t be the only consideration when buying XIV/VXX in size; beta matters. I think the fact that the blind term-structure trade has worked so well in recent years has fooled a lot of traders in to forgetting that.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: VIX & Volatility | 6 Comments
Readers know I’m not one to preach trend-following (at least relative to other more active, more effective strategies), but I have to stand up and defend the honor of this granddaddy of quantitative strategies from some recent negative press, such as the latest from Mark Hulbert.
Hulbert makes his case that the 200-day moving average (the most commonly quoted trend-following indicator) has become a poor predictor of future stock market returns because it’s done a poor job generating return in recent decades.
Regardless of whether that’s true, it completely misses the point of what trend-following indicators are and have always been useful for: reducing drawdowns and volatility (NOT generating above average returns).
I did a long numbers-centric post on the subject a couple of years back that I would suggest interested readers digest. It was about 50/200-day crossovers (aka “the Golden Cross“), but the conclusion was the same.
Long-term moving averages have never been useful for generating above average returns timing the stock market for any length of time. But they have consistently done a good job at capturing most of the stock market’s returns, while reducing volatility and sidestepping the worst of its drawdowns.
That’s been as true for the last decade as at any point since my test began in 1930.
. . . . .
Pulling an example from Hulbert’s post, Hulbert states that since 1990, a trader would have only returned 3.8% annualized by trading the DJIA when it was above its 200-day moving average, versus 7.3% for buy & hold.
Below I’ve run my own test (trend-following in red, buy & hold in grey). Note that I’ve used the S&P 500 in place of the DJIA, and I’ve adjusted returns for dividends, but the point is the same:
[logarithmically-scaled, growth of $1]
Yes, trend-following has underperformed in terms of return, especially over the last couple of years, but that shouldn’t come as too big a surprise. Again, read my previous numbers-centric post. It’s par for the course.
More importantly, look at what it’s done to reduce loss/volatility: max drawdown was halved at -27% versus -55%, Sharpe Ratio increased by 20%, UPI increased by 30%, etc. That’s also par for the course.
. . . . .
I don’t use trend-following indicators in my own trading today, so I don’t have any skin in this game, but I recognize that the type of short-term strategies we’re trading are just too active for a lot of investors.
And for those folks I think trend-following, applied to the stock market as part of a diversified portfolio, is a huge improvement over buy & hold.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Trading Strategies, Trend-Following | 14 Comments
