As I said earlier, the Collaboration is Good trade is a mean-reversion trade. This is actually a misnomer, as there is no "mean" the market is returning to; a more accurate term would be "trend-reversion", but the industry uses the terms it uses. In any case, the general principle is that when a market is in a long-term trend, and has a short period (the length of this period varies by market) of returns counter to that trend, a correction back toward the long-term trend is more likely than a continuation of the short term counter trend. I realize for most readers, that makes very little sense, so let me illustrate.
Example: A Rubber Band
Go get a rubber band - say a medium-sized one about 4 inches long, and hold it in your left hand at a stationary height. With your right hand, grasp its other end and stretch it a few inches downward and release it. It snaps back, but not too violently, right? Now stretch it a lot farther than you did before - to something like 12-14 inches. Notice how you feel the band begin to fight you, equaling out the force of your hand with its own elastic properties? OK, release it, and notice how much faster it returns to slack. In fact, it probably bounced up past the fingers of your left hand - in finance, we call this over-correction. It may have snapped your fingers, too, which provides a good object lesson on why you shouldn't take my blog commentary as advice.
The problem with the rubber band example is that the market is not being held stationary in someone's left hand. Because of the vast chaos and complexity that exists in the capital markets, there is plenty of room for contrary viewpoints. One of these is the Random Walk Hypothesis. In a nutshell, it states that the market cannot be predicted, and any apparent successes in doing so are simply luck.
Counter Example: The Coin-Flip Fallacy
Imagine we have a fair coin for flipping, one that gives heads 50% of the time, and tails 50% of the time. Now, if we flip this coin 3 times and then report its results in percentage terms without giving the flip count, it will look terribly weighted toward one or the other (67%/33%, or even 100%/0%). As we flip it more and more, though, we naturally expect it to approach 50/50. Does that mean that a long run of all heads makes tails somehow more likely? Of course not - any reasonably aware person can see that the results of any given flip of the coin are independent of all previous flips.
But when roulette players mutter about a particular number being "due", isn't that the exact same thing? As the number of variables increase, the complexity goes up; this, in turn, makes the game look more and more beatable. Now how much more complex is the capital market system than a game of roulette? If it is indeed just a random environment, no human has the ability to intuitively detect that - we're just not wired that way.
I was once forced to "prove" to a previous employer that die rolls in a craps game were independent of each other. I put it in quotes because the only thing that constituted proof, to him, was a complete model of the entire game that he could run various betting systems on until he managed to convince himself it was just a negative expected-value game of chance. This man was a trader - none of us are immune.
I personally prefer to believe that the market is not random - instead, it is the numerical result of millions of complex and confusing variables and herd psychology. Whenever I evaluate a strategy, I try to understand the underlying causes for the apparent pattern. If I can't identify any, I find it very difficult to trust the predictive nature of the strategy.
Real-World Data Sample
Starting in March, 2009, the S&P (proxied here by SPY) enjoyed a massive 14-month, 85% rally. One leg of that rally started in July and ended in October. Backtesting reveals that CiG would have signaled a buy on September 28, after 3 days of counter-trend price movement. That time period is visible here, with a couple of trendlines to help remove some of the noise (click the picture for a larger view). For the curious, yes there were other signals - profitable ones - near Aug 17 and Sep 2. But they had more bowl-shaped corrective rallies, and thus were not as clear an example as Sep 28, took its profits after a single day.
Here is a zoomed-in view of SPY at the time of the trade signal. I've overlayed a Relative Strength Indicator to help illustrate the stretching rubber band effect that CiG is designed to capture.
As you can see, SPY built up three days of upward corrective pressure, shown by the red counter-trend line. Finally, nearly all that pressure was relieved on Sep 28 with a 19-point (in the futures) rally. CiG would have signaled a buy on the close of Sep 25 (Friday), and a flattening sell on the close on the next trading day, Sep 28, booking a $950/contract profit.
Double-Down or Go Home
Essentially, the reasoning behind doubling one's investment in a trade that has lost money goes like this:
- I still think the trade will work out and be profitable; in fact, I feel that it is more likely to have a winning result now that the security I'm trading is even more incorrectly priced.
- Given that the probability of success is higher now than when I originally entered the trade, I should allocate more capital to it.
Even if it makes sense for the trade at hand, it ultimately is a Martingale. If a trader decides to take this approach, even after some soul-searching and healthy self-doubt, he ought to first clearly define his risk-management rules. The truly dangerous nature of a Martingale, to the plan-deprived, is that at any given moment during this trade, booking the loss looks worse than taking on a little more risk. Notice that the reasoning above doesn't ask two important questions: can I bet a greater slice of my financial life that my model is flawless and applies to this situation? and do I have more capital to allocate?
CiG doesn't attempt to answer these questions - good strategies shouldn't presume to be that comprehensive. Instead, it does not allow increasing position size within a security, and limits the list of securities traded; it also has predefined stop-loss limits. When the limit is crossed, the position is closed and the loss is booked. Period.
John Maynard Keynes famously said: markets can remain irrational a lot longer than you and I can remain solvent.
And that, my friends, is the Risk of Ruin.