Probablility in Investing
How Market Timing Works
Long-Only Versus Long & Short
---
All investing involves uncertainty and probability. Most of the time we don’t actually estimate numerical probability, but it’s always there. When we buy a stock or a mutual fund or an ETF, we never know whether it will be profitable, but we are implicitly betting that the probability of gain is greater than the probability of loss.
To build confidence in those probabilities, it pays to see many occasions of how similar bets worked out in the past. No two market bets are ever identical, but if the decision criteria are fairly definite, we can identify prior occasions that were pretty much the same as whatever opportunities we’re facing today. And if we follow those criteria rigorously, we “know” how things will turn out on average. And knowing that, we don’t have to worry about individual instances.
This is what an actuary does; an actuary applies criteria to estimate probabilities, and then relies on the law of large numbers to take care of the outcome. If your house has alarms, and is not too far from fire services (and other conditions), the actuary “knows” how likely it is to burn down. Then if the company insures many houses in similar condition, the actuary “knows” how many will burn. If the company insures your house today, and it burns down next week, it’s unfortunate, but not especially troubling to the insurer. It’s just one outcome of many.
Investors should think like actuaries. Build enough prior experience to create reliable decision criteria and use probability in investing, so you’ll know what to expect on average. Then follow those criteria closely and continuously, and let the law of large numbers go to work for you. One example of this principle can be seen in the timing component of our Cabot ETF Investing System.
Our Market Timing Model
The Cabot ETF Investing System uses a market-timing model to determine when to buy and sell. The ETF System uses a separate selection model to determine which sector ETFs are most likely to be stronger than the market. Performance is reported every Tuesday. In those reports, it’s easy to see how the selection model is doing; the average of our “Favored” sectors is tabulated along with the all-sectors average and the aggregate S&P index tracker.
But it’s not so easy to see how the timing model is doing. Every week is either an up-week or a down-week. It wouldn’t make sense to say the timing model is “right” whenever the indicator is positive and the market goes up for the week (or when the indicator is negative and the market goes down); nor would it make sense to say the timing model is “wrong” whenever the indicator is positive and the market is down (or when the indicator is negative and the market is up).
A timing model can only be assessed in its performance over multiple market cycles. And our longer-term ETF performance combines both timing and selection effects. As readers sometimes ask about the timing model, I have isolated the timing-only performance and summarized it below.
Market Momentum
First let’s see how the timing model works. The model is called the Cabot Tides indicator, and it’s based on market momentum. That means (essentially) that the model calls for continuation of whatever trend is currently under way: rising prices call for further rising prices, and falling prices call for further falling prices.
The model works because trend continuation is more prevalent than reversal. Of course, every trend eventually ends with a reversal (when the trend takes a new direction). Therefore momentum will always turn out to overstay a trend at some point. So the trick to a successful momentum model is to have the continuation period last long enough so that the give-back at the end doesn’t wipe out the gains. And the trick to doing that is to have trend-change criteria that are triggered fairly promptly…but not too promptly.
Momentum Criteria
Criteria for the Tides model are based on five market indexes and their moving averages. (The five indexes are the S&P 500 (large cap), S&P 400 Mid Cap, S&P 600 Small Cap, the Nasdaq Composite and NYSE Composite.) When three of the five components are above the lower of their 25-day and 50-daymoving averages (and that moving average is rising) the trend is defined as “up.” When three of the five components are below the lower moving average, the trend is defined as “down.”
Here is an example of the Mid-Cap index interacting with its two moving averages. The red line is the 25-day moving average and the blue line is the 50-day. The blue line has been the lower of the two (lower than the red line) since mid-December, and the Mid-Cap index stayed clear above it continuously…until a few days in mid-April. The index closed below the 50-day average for five days on April 15, 17, 18, 19 and 22.
Other indexes were doing much the same thing, and four of the five components crossed negative on April 18 giving a Sell. Then everything changed, the market advanced, indexes re-crossed, and we had a Buy three market days later on April 23.
This kind of sequence is called a “whipsaw,” and it’s a loss. The index closed at 1105 on the Sell date, then closed at 1137 at the Buy. So that was 2.9% lost (missed out on) in just three days. When this happens—a Sell followed within days by a Buy, and at a loss—it can be a real challenge to actually act on that second signal. After all, your brain thinks, the timing was “wrong” just days ago…maybe it’s going to be wrong again!
Indeed, maybe it will be wrong again. But this is when your inner actuary has to be firm, telling you sternly “We never know which signals will be right or wrong. And because we don’t know, we have to keep following the “underwriting” criteria, which is to say, keep following the Buy and Sell signals. (If it comes time to re-evaluate the criteria, that’s another thing. But so long as the criteria are in place, it’s important to take the positions they call for, and let the probabilities work.)
As it turned out in the April whipsaw, the quick re-purchase more than covered the three-day loss. The Mid-Cap index is up 6.55% from that April 23 re-Buy, and the S&P 500 is up 5.62%. Things don’t always work out well, but they do more often than not.
Timing-Only Trading
To see how the Tides timing model works over time, I’ve generated cumulative returns paths without any sector selection. One path follows the “Long Only” strategy, holding the S&P tracking ETF (SPY) under Buy signals, and holding cash under Sell signals.
The Long-Only trading (with no sector selection) was up 66% from 2001 through April 2013. That’s an annual rate of 4.23% a year, with substantially lower volatility than the SPY.
The second returns path is similar but sells short the SPY when there’s a Sell signal. The Long-Short returns path (below) was up 30% over the same span, averaging 2.19% a year, and again with substantially lower-than-market volatility. In particular, there are no deep draw-downs as experienced in a buy-and-hold. Of course, the Long-Short strategy does far better in bear markets.
But the real point—the “takeaway”—in these return paths is that even a successful timing model has its ups and downs, ebbs and flows. The net result is better than the market, but the sawtooth progress is inevitable. We never know which signals will be the best or worst. (If we knew that, there’d be nothing but up-up-up!) What we do know is how the system works over time. And that knowledge allows us to focus on the process, not on the individual trades—to think like an actuary.