The Official Blog of Retail Backtest
About     Recent Entries

Whipsaws and Stop-Loss Trading

You hear occasionally, in trader talk, the word "whipsaw". While the word can refer to any sharp oscillatory price action in the market that is being traded, it also refers to traders panicking out of losing positions and then having to get back in at prices worse than the exit prices. "Getting whipsawed" is bad!

So we'll go over that today, but this blog entry isn't some idle chit-chat about the storied business of trading securities. The Retail Backtest project is instead dedicated to the task of making what is exciting for others really boring— by sharply reducing downside volatility in account equity with algorithms that never complain about anything. Today I'm going to explain something that is very important to understand, if ever you contemplate adopting some policy that amounts to instituting stop-loss trading to limit downside volatility.

The basic message is that the whipsaws are an unavoidable complication— their deleterious effects on your account equity are difficult to minimize and impossible to utterly eliminate (if you trade in and out of positions in a stop-loss manner). And I'm going to explain that a famous formula actually quantifies whipsaw losses and makes them into the price of a certain contract, via an algorithm of course.

What is Being Called Stop-Loss Trading Here?

If you have some sort of plan that involves getting out of a long positions in securities when the prices drops by too much, and later getting back in after a recovery has started, that's generally what I mean by stop-loss trading. Of course there is such a thing as a "stop-loss" order to a broker, which requires that the security be sold if the price falls to a certain level, and that is one of the blunt mechanisms that are in general use so as to attempt to stop losses.

What Exactly Is a Whipsaw?

Ahem... Well, this is somewhat painful but take a gander at the chart below which pertains to Retail Backtest's version of "momentum", the concept that market price trends are at least somewhat likely to persist.


This chart is for the "International" portfolio. Please read the notes in green on it. It is a line chart of monthly data, so the points at which the straight-line segments intersect are the actual price points (and the same is true for position sizes in the "% at Risk" part of the chart at the bottom).

The red line is basically the buy-and-hold line for the portfolio, and it dips by over 10% in May of 2012, only to very nearly fully recover in just about three months. You could call that whipsaw price action. The black line is what hypothetically would have happened to your money had you been investing back then with the Retail Backtest momentum program. So the program would have taken a beating back then: it sold out at the bottom and had to buy back in later at a significantly higher price. That's getting whipsawed by a whipsaw in price action.

Well, momentum is something that made it's way into the Retail Backtest project because of the substantial amount of academic support that it has garnered. In spite of its limitations— the Retail Backtest implementation suffers a bit from the fact that it only conducts trades once a month— it seems that it would have been remarkably successful overall at avoiding debacles such as the 2000 dot-com crash and the 2007-2008 Lehman Brothers/subprime crisis. You can view the long-term prospects for Retail Backtest's momentum program and you can get to the fully-interactive chart for it on the Retail Backtest site here. Of the three Retail Backtest momentum portfolios, two did not do well through this particular whipsaw.

No, I Mean... What Is a Whipsaw?


As this photo by Eric A. Hegg (1867-1948) shows, it's a two-man saw. Note that the saw will go up and down like the market but that the overall progress will be sideways, like the market sometimes; but if you get whipsawed with hapless stop-loss trading your progress will be downward, not up or sideways.


Momentum not having been a conception of my own, I readily went out and explored other possibilities. The chart below covers the same time period as on the chart above and pertains to the same list of international ETFs, but shows what would have happened with Retail Backtest's New Program. It involves weekly trading and it's not based on the momentum concept. (The chart here is a daily chart, not a monthly chart.)


Note especially that the RB Portfolio would not have faltered (and that goes for the other two portfolios of the RB New Program as well); the market, represented by the Rebalanced Benchmark, did. That is, with the RB program the dip in May of 2012 would not have been as bad as that of the market. The portfolio did not suffer due to the market's whipsaw but temporarily benefited from it (relative to the market) with much-reduced downside volatility.

A Rather Precise Theory

First let's indulge in a less-than-precise idea. Or rather, it sounds precise but the outcomes will unexpectedly vary. Let's say that to minimize losses you decide to enact a "stop price", a price somewhere just below the current market price, the starting price, below which you will no longer hold the security (we're talking now about long positions only). But if the market price moves higher then you'll want to eventually move that stop price higher... a "trailing stop" as it's generally called.

To further spell out your scheme, let's say that you will do at most one trade per day, will reset the stop price every month, and that if at any time during that month you are stopped out you will get back in as soon as the market price is back above the stop price. Your goal is therefore to lose not one dime more than the difference between the starting price and the stop price, yet to fully participate in the upside (if any).

Any guesses as to how this turns out? Well the problem is that you will fairly often get whipsawed— you'll find yourself buying back in at a higher price than the stop price (and also getting out at a price somewhat below the stop price). And over time these whipsaw losses will accumulate.

This very scheme is called by me a "poor man's call option" in this Retail Backtest article. At the top of page 2 of the article is an interactive chart that allows you to repetitively try out the poor man's call option, using real data. It does turn out that the whipsaw costs make the scheme non-viable. Just click the Resample button multiple times on that graph and you'll see on the table below it that the cumulative Sharpe ratio for the Poor Man's Call option, which is a sort of risk-adjusted return, tends to be rather miserable compared to that of Buy-and-Hold.

So much for the less-than-precise idea. The precise version is what is afforded by simply buying a listed call option on the target security instead of buying the security itself and effecting a stop price. The contract for the call option contains detailed provisions pertaining to exercise rights and specifies a "strike price" and an expiration date. And dividends paid on the underlying security pose a minor complication. But, for the purposes of the present discussion we can pretend that at expiration the options simply pay the holder the difference between the price of the security and the strike price if the former exceeds the latter, and nothing otherwise— where the strike price takes the place of the stop price of the poor man's call option. Some call options such as cash-settled "European style" options on stock market indexes pretty much work just that way.

Now the poor man with his stop price and one-month period until resetting the stop price expected to get the same outcome as would have been obtained with a listed call option having one month until expiration and a strike price equal to his stop price. But he got less than that, in the long run. And so, are listed call options free? No. Their price is called the "premium". The correct premium that should be paid for a listed call option was actually computed by Black and Scholes who assumed that "it should not be possible to make profits by creating portfolios of long and short positions in options and their underlying stocks."

Black and Scholes showed that they could simulate the performance of an option by trading in and out of the security as the market price for it fluctuated in relation to the strike price. However, unlike our poor man, they found a unique scheme that involved continually scaling out of the long position in the stock in small decrements as the market price for it approached the strike price from above, with about half the position still held when the stock price equaled the strike price and with complete scaling out occurring only when the stock price meandered much further below the strike price (and vice-versa when the stock price rose). With that unique scheme their outcomes were certain: the payout of a listed stock price option was exactly simulated but for a fixed deficit, which they took to be the proper price of the listed option.

To conclude, the work of Black and Scholes does pertain to a form of stop-loss trading and it conclusively illustrates that whipsaw losses are not spurious outcomes afflicting inept traders but are instead innately associated with the various forms of stop-loss trading due to the stochastic, random-walk nature of price histories. Testing shows that Retail Backtest's programs, which are not at base explicitly devised so as to have a purely stop-loss character, would show some whipsaw-like losses during bull-market phases but would tend to very substantially avoid large losses during bear markets (and would then not fail to get back in at prices lower than the exit prices).

© 2017 Michael C. O'Connor ∅ All Rights Reserved