This is the first in a series on what timing is, what it isn’t, and some right ways to do it.
In recent days I have been experimenting with one of our proprietary indicators to expand its use. I named this indicator a “Risk Index” with the idea being when the indicator is high the risk are low and when the indicator is low the risks are high. As you can see in the chart below a higher percentage reading indicates a more favorable market and a lower reading indicates a less favorable market.
Our risk index is simply the percentage of out timing models that are bullish or bearish each week. For US equities we run 10 different models that look at trend, valuation, interest rates, inflation, sentiment, breadth, and intermarket relationships. I estimate that 53% of the individual components in the 10 models are equity trend based. There are a few reasons for this but the most important gets at the heart of timing. We use timing tools to help us first lower risk and in a distant second to increase returns. It turns out that trend following indicators while not a “Holy Grail” do a great job at keeping you in the big moves and minimizing your downside.
Our models come from many places. If you are familiar with Nelson Freeburg, Marty Zweig, and Ned Davis you would recognize a few of the models and would be able to see the inspiration in the other models.* Five of the 10 are straight from their material and the other five while homegrown take inspiration from their work. All the models have been backtested and while most of them slightly improve returns they all drastically improve drawdowns which is our primary goal.
So if each of these models is solid in its own right why would we take a consensus approach? There are several reasons but the two that stand out are that you never know when the market is going to change and invalidate a model. Now we can stand a prolonged period of under-performance but we cant handle a catastrophe. If a model underperforms for a long enough period of time we would take it out if we could see that something had changed. As an example I once created a breadth based system that I was able to backtest and it generated low 20% returns with the worst drawdown being just over -7%. Well I got to use it for about a year before decimalization came and within weeks the results when to hell. I suspected something was off but it took a few more months to confirm it. I still update it and monitor it as it displays a certain segment of market behavior but its risk/reward is no longer favorable.
Of course most of the models in the risk index are based on weekly data and are longer term in nature. Still the risk is very real that something changes and some of them cease to be useful. By taking a consensus approach any downturn based on a degrading model can be minimized.
We are not going to get into the specifics of each model but instead how almost any model, in this case a consensus model, can be used. Don’t worry because in a future post we will go over how to build a simple but effective long term timing model.
So we have a US Equity model that is based on the buy/sell signals of 10 separate timing models. How can we use it? We could backtest it and see what readings give the best risk/reward and trade it that way but what inspired this post was the idea that we would just invest X% of a portfolio depending on the reading. If the model said that 50% of the models were on buy signals we would invest 50% of the portfolio and change it each time the buy signals percentage changed. If that went well, it did, and sufficiently cut risk, it did, we could then experiment with different levels of leverage.
We did this with the data we had on hand and got the following results. Trading SPY-SP500 ETF, and using the total return series so that includes dividends, we got the following results. Buy and hold did fine on the upside but had a -50.77% drawdown. Timing trailed a bit on the upside but only suffered a -13.67% max drawdown. Finally by using a full 2X leverage we were able to cut buy and hold risk in half and increase returns by 1.89 times. In case you are wondering by using only 1.2X leverage you beat buy and hold by a few bucks but your max drawdown is still under -15%.
Looking at a chart of the equity curves for each of the strategies you can see how timing plus leverage killed buy and hold. Of course while max drawdown was far less the intermittent drawdowns were sometimes larger. Take 2011 for example when the market corrected just enough to turn the model down to 10% bullish only to rocket higher. That is the main risk to any system as you can get whipsawed in and out during a longer term trend. Of course anytime you are using leverage you can expect to have higher volatility at times as you are seeking higher returns.
Looking at the individual drawdown charts shows just how risky buy and hold is as the SPy-SP500 ETF was down over -50%. This of course requires a 100% return just to get back to breakeven.
Looking at the drawdowns for the timing without leverage equity curve you can see that while it has a lot of little drawdowns it has only had three double digit drawdowns since early 2008 with the worst one being -13.67%. It may have lagged in total return but not by much and as such would have been a lot easier to handle. Of course as we discussed one would only need 1.2X leverage to achieve equal returns with buy and hold with less than 1/3 the risk.
Finally we have the drawdown chart of the timing strategy but using 2X leverage. As you can see the worst drawdown was half of that of buy and hold. Of course the next two worst drawdowns also hit -20% in contrast to buy and hold which only had one more -20% drawdown. Still the overall risk has been cut in half and the returns almost doubled.
Why do we only have the risk index back to 4/11/08? We are working on extending it back a few decades but as we were building these we had some data limitations on two of our homegrown models. When we finish building them out we will share the results with our subscribers as well as the blog. For now however we think that capturing most of the carnage of 2008 along with the correction of 2011 does a decent job of what can be accomplished with a good timing model and a few different ways to use it.
One aspect of this model that we like is that is gives a specific allocation percentage instead of just a buy/sell signal. This will be the purpose of a future post but if you go back and read all the Marty Zweig stuff, and Zweig was a timer if there ever was one, he never said to go all in or all out.
“How should you, the reader of this book, react to the constantly changing circumstances? Basically, I think you should shun the idea of buy-and-hold. I consider it a fallacious strategy. In the coming decade we are likely to see more bear markets and deeper ones. To lower risk, there will be periods when you should peel back your investments, in the stock and bond markets. It’s a matter of degree. You don’t have to go 100% to cash but you should cut back as risk rises and invest as risk recedes. I believe my market-timing methods in this book will help you do just that.”–Marty Zweig from “Winning On Wall Street”
If you go read Howard Marks book “The Most Important Thing” you will find variations of the same concept. If you are a traditional value guy/gal your heart just skipped a beat as I said Howard Marks in the same post as “market timing”. The reality is that all active management has the same goal-minimize risk and maximize reward. Marks in his excellent book talks about assessing the range of future outcomes and discusses risk throughout both his book and other writings. Despite different approaches both Marks and Zweig have the same goal. be aggressive when their indicators-be they book values or how much the ZUPI moved-say to be aggressive and back off when things look risky.
*I can’t write this and not give credit where credit is due. Nelson Freeburg the late publisher of Formula Research was a fantastic guy and his publication as well as correspondence has had a great influence on me. In fact while the idea of combining timing models together was not new, the way in which he did it elevated my thinking to a new level in his January 15, 1998 issue “The Power of a Composite Stock Market Model”. The components of my risk index are very different but if you read that report you can not help but see similarities. Aside from that report however he put out more interesting and functional models than anyone I know of. If you can get a hold of any, or all, of them you will be better for it.