vested. Some, believing the losses were more likely to continue than to reverse, bailed out. Had they been willing to endure a little longer, they would have experienced double-digit gains during the remainder of 2003 that would have restored and exceeded all of their losses. But of course there was no way to know that in advance.

Most timers won’t tell you this, but all market timing systems are *optimized* to fit the past. That means they are based on data that is carefully selected to *work* at getting in and out of the market at the right times. Think of it through this analogy. Imagine we were trying to put together an enhanced version of the Standard & Poor’s 500 Index, based on the past 30 years. Based on hindsight, we could probably significantly improve the performance of the index with only a few simple changes.

For instance, we could conveniently *remove* the worst-performing industry of stocks from the index along with any companies that went bankrupt in the past 30 years. That would remove a good chunk of the *garbage* that dragged down performance in the past. And to add a d

Market timing systems are based on patterns of activity in the past. Every system that you are likely to hear about works well when it is applied to historical data. If it didn’t work historically, you would never hear about it. But patterns change, and the future is always the great unknown.

A system developed for the market patterns of the 1970s, which included a major bear market that lasted two years, would have saved investors from a big decline. But that wasn’t what you needed in the 1980s, which were characterized by a long bull market. And a system developed to be ideal in the 1980s would not have done well if it was back-tested in the 1970s. So far in the 1990s, any defensive strategy at all has been more likely to hurt investors than help them.

If your emotional security depends on understanding what’s happening with your investments at any given time, market timing will be tough. The performance and direction of market timing will often defy your best efforts to understand them. And they’ll defy common sense. Without timing, the movements of the market may seem possible to understand. Every day, innumerable explanations of every blip are published and broadcast on television, radio, in magazines and newspapers and on the Internet. Economic and market trends often persist, and thus they seem at least slightly rational. But all that changes when you begin timing your investments.

Unless you developed your timing models yourself and you understand them intimately, or unless you are the one crunching the numbers every day, you won’t know how those systems actually work. You’ll be asking yourself to buy and sell on faith. And the cause of your short-term results may remain a mystery, because timing performance depends on how your models interact with the patterns of the market. Your results from year to year, quarter to quarter and month to month may seem random.

Most of us are in the habit of thinking that whatever has just happened will continue happening. But with market timing, that just isn’t so. Performance in the immediate future will not be influenced a bit by that of the immediate past. That means you will never know what to expect next. To put yourself through a *timing simulator* on this point, imagine you know all the monthly returns of a particular strategy over a 20-year period in which the strategy was successful.

Many of those monthly returns, of course, will be positive, and a significant number will represent losses. Now imagine that you write each return on a card, put all the cards in a hat and start drawing the cards at random. And imagine that you start with a pile of poker chips. Whenever you draw a positive return, you receive more chips. But when your return is negative, you have to give up some of your chips to *the bank* in this game. If the first half-dozen cards you draw are all positive, you’ll feel pretty confident. And you’ll expect the good times to continue. But if you suddenly draw a card representing a loss, your euphoria could vanish quickly.

And if the very first card you draw is a significant loss and you have to give up some of your chips, you’ll probably start wondering how much you really want to play this game. And even though your brain knows that the drawing is all random, if you draw two negative cards in a row and see your pile of chips disappearing, you may start to feel as if you’re on *a negative roll* and you may start to believe that the next quarter will be like the last one. Yet the next card you draw won’t be predictable at all. It’s easy to see all this when you’re just playing a game with poker chips. But it’s harder in real life.

For example, in the fourth quarter of 2002, our Nasdaq portfolio strategy, with an objective to outperform the Nasdaq 100 Index, produced a return of 5.9 percent, very satisfactory for a portfolio invested in technology funds only. But that was followed by a loss of 7.8 percent in the first quarter of 2003. Most investors in this strategy, at least those we know of, stuck with it. But they experienced significant anxiety at the loss and the shock of a sharp reversal in what they had thought was a positive trend. The same phenomenon happened, with more dramatic numbers, in our more aggressive strategies.

Some investors entered those portfolios in the winter of 2002, and then were shocked to experience big first-quarter losses so quickly after they had invested. Some, believing the losses were more likely to continue than to reverse, bailed out. Had they been willing to endure a little longer, they would have experienced double-digit gains during the remainder of 2003 that would have restored and exceeded all of their losses. But of course there was no way to know that in advance.

Most timers won’t tell you this, but all market timing systems are *optimized* to fit the past. That means they are based on data that is carefully selected to *work* at getting in and out of the market at the right times. Think of it through this analogy. Imagine we were trying to put together an enhanced version of the Standard & Poor’s 500 Index, based on the past 30 years. Based on hindsight, we could probably significantly improve the performance of the index with only a few simple changes.

For instance, we could conveniently *remove* the worst-performing industry of stocks from the index along with any companies that went bankrupt in the past 30 years. That would remove a good chunk of the *garbage* that dragged down performance in the past. And to add a do

rstand. Every day, innumerable explanations of every blip are published and broadcast on television, radio, in magazines and newspapers and on the Internet. Economic and market trends often persist, and thus they seem at least slightly rational. But all that changes when you begin timing your investments.

Unless you developed your timing models yourself and you understand them intimately, or unless you are the one crunching the numbers every day, you won’t know how those systems actually work. You’ll be asking yourself to buy and sell on faith. And the cause of your short-term results may remain a mystery, because timing performance depends on how your models interact with the patterns of the market. Your results from year to year, quarter to quarter and month to month may seem random.

Most of us are in the habit of thinking that whatever has just happened will continue happening. But with market timing, that just isn’t so. Performance in the immediate future will not be influenced a bit by that of the immediate past. That means you will never know what to expect next. To put yourself through a *timing simulator* on this point, imagine you know all the monthly returns of a particular strategy over a 20-year period in which the strategy was successful.

Many of those monthly returns, of course, will be positive, and a significant number will represent losses. Now imagine that you write each return on a card, put all the cards in a hat and start drawing the cards at random. And imagine that you start with a pile of poker chips. Whenever you draw a positive return, you receive more chips. But when your return is negative, you have to give up some of your chips to *the bank* in this game. If the first half-dozen cards you draw are all positive, you’ll feel pretty confident. And you’ll expect the good times to continue. But if you suddenly draw a card representing a loss, your euphoria could vanish quickly.

And if the very first card you draw is a significant loss and you have to give up some of your chips, you’ll probably start wondering how much you really want to play this game. And even though your brain knows that the drawing is all random, if you draw two negative cards in a row and see your pile of chips disappearing, you may start to feel as if you’re on *a negative roll* and you may start to believe that the next quarter will be like the last one. Yet the next card you draw won’t be predictable at all. It’s easy to see all this when you’re just playing a game with poker chips. But it’s harder in real life.

For example, in the fourth quarter of 2002, our Nasdaq portfolio strategy, with an objective to outperform the Nasdaq 100 Index, produced a return of 5.9 percent, very satisfactory for a portfolio invested in technology funds only. But that was followed by a loss of 7.8 percent in the first quarter of 2003. Most investors in this strategy, at least those we know of, stuck with it. But they experienced significant anxiety at the loss and the shock of a sharp reversal in what they had thought was a positive trend. The same phenomenon happened, with more dramatic numbers, in our more aggressive strategies.

Some investors entered those portfolios in the winter of 2002, and then were shocked to experience big first-quarter losses so quickly after they had invested. Some, believing the losses were more likely to continue than to reverse, bailed out. Had they been willing to endure a little longer, they would have experienced double-digit gains during the remainder of 2003 that would have restored and exceeded all of their losses. But of course there was no way to know that in advance.

Most timers won’t tell you this, but all market timing systems are *optimized* to fit the past. That means they are based on data that is carefully selected to *work* at getting in and out of the market at the right times. Think of it through this analogy. Imagine we were trying to put together an enhanced version of the Standard & Poor’s 500 Index, based on the past 30 years. Based on hindsight, we could probably significantly improve the performance of the index with only a few simple changes.

For instance, we could conveniently *remove* the worst-performing industry of stocks from the index along with any companies that went bankrupt in the past 30 years. That would remove a good chunk of the *garbage* that dragged down performance in the past. And to add a d

ugh a *timing simulator* on this point, imagine you know all the monthly returns of a particular strategy over a 20-year period in which the strategy was successful.

Many of those monthly returns, of course, will be positive, and a significant number will represent losses. Now imagine that you write each return on a card, put all the cards in a hat and start drawing the cards at random. And imagine that you start with a pile of poker chips. Whenever you draw a positive return, you receive more chips. But when your return is negative, you have to give up some of your chips to *the bank* in this game. If the first half-dozen cards you draw are all positive, you’ll feel pretty confident. And you’ll expect the good times to continue. But if you suddenly draw a card representing a loss, your euphoria could vanish quickly.

And if the very first card you draw is a significant loss and you have to give up some of your chips, you’ll probably start wondering how much you really want to play this game. And even though your brain knows that the drawing is all random, if you draw two negative cards in a row and see your pile of chips disappearing, you may start to feel as if you’re on *a negative roll* and you may start to believe that the next quarter will be like the last one. Yet the next card you draw won’t be predictable at all. It’s easy to see all this when you’re just playing a game with poker chips. But it’s harder in real life.

For example, in the fourth quarter of 2002, our Nasdaq portfolio strategy, with an objective to outperform the Nasdaq 100 Index, produced a return of 5.9 percent, very satisfactory for a portfolio invested in technology funds only. But that was followed by a loss of 7.8 percent in the first quarter of 2003. Most investors in this strategy, at least those we know of, stuck with it. But they experienced significant anxiety at the loss and the shock of a sharp reversal in what they had thought was a positive trend. The same phenomenon happened, with more dramatic numbers, in our more aggressive strategies.

Some investors entered those portfolios in the winter of 2002, and then were shocked to experience big first-quarter losses so quickly after they had invested. Some, believing the losses were more likely to continue than to reverse, bailed out. Had they been willing to endure a little longer, they would have experienced double-digit gains during the remainder of 2003 that would have restored and exceeded all of their losses. But of course there was no way to know that in advance.

Most timers won’t tell you this, but all market timing systems are *optimized* to fit the past. That means they are based on data that is carefully selected to *work* at getting in and out of the market at the right times. Think of it through this analogy. Imagine we were trying to put together an enhanced version of the Standard & Poor’s 500 Index, based on the past 30 years. Based on hindsight, we could probably significantly improve the performance of the index with only a few simple changes.

For instance, we could conveniently *remove* the worst-performing industry of stocks from the index along with any companies that went bankrupt in the past 30 years. That would remove a good chunk of the *garbage* that dragged down performance in the past. And to add a d

a row and see your pile of chips disappearing, you may start to feel as if you’re on *a negative roll* and you may start to believe that the next quarter will be like the last one. Yet the next card you draw won’t be predictable at all. It’s easy to see all this when you’re just playing a game with poker chips. But it’s harder in real life.

For example, in the fourth quarter of 2002, our Nasdaq portfolio strategy, with an objective to outperform the Nasdaq 100 Index, produced a return of 5.9 percent, very satisfactory for a portfolio invested in technology funds only. But that was followed by a loss of 7.8 percent in the first quarter of 2003. Most investors in this strategy, at least those we know of, stuck with it. But they experienced significant anxiety at the loss and the shock of a sharp reversal in what they had thought was a positive trend. The same phenomenon happened, with more dramatic numbers, in our more aggressive strategies.

Some investors entered those portfolios in the winter of 2002, and then were shocked to experience big first-quarter losses so quickly after they had invested. Some, believing the losses were more likely to continue than to reverse, bailed out. Had they been willing to endure a little longer, they would have experienced double-digit gains during the remainder of 2003 that would have restored and exceeded all of their losses. But of course there was no way to know that in advance.

Most timers won’t tell you this, but all market timing systems are *optimized* to fit the past. That means they are based on data that is carefully selected to *work* at getting in and out of the market at the right times. Think of it through this analogy. Imagine we were trying to put together an enhanced version of the Standard & Poor’s 500 Index, based on the past 30 years. Based on hindsight, we could probably significantly improve the performance of the index with only a few simple changes.

For instance, we could conveniently *remove* the worst-performing industry of stocks from the index along with any companies that went bankrupt in the past 30 years. That would remove a good chunk of the *garbage* that dragged down performance in the past. And to add a d

vested. Some, believing the losses were more likely to continue than to reverse, bailed out. Had they been willing to endure a little longer, they would have experienced double-digit gains during the remainder of 2003 that would have restored and exceeded all of their losses. But of course there was no way to know that in advance.

Most timers won’t tell you this, but all market timing systems are *optimized* to fit the past. That means they are based on data that is carefully selected to *work* at getting in and out of the market at the right times. Think of it through this analogy. Imagine we were trying to put together an enhanced version of the Standard & Poor’s 500 Index, based on the past 30 years. Based on hindsight, we could probably significantly improve the performance of the index with only a few simple changes.

For instance, we could conveniently *remove* the worst-performing industry of stocks from the index along with any companies that went bankrupt in the past 30 years. That would remove a good chunk of the *garbage* that dragged down performance in the past. And to add a dose of positive return, we could triple the weightings in the new index of a few selected stocks; say Microsoft, Intel and Dell. We’d get a new *index* that in the past would have produced significantly better returns than the real S&P 500. We might believe we have discovered something valuable. But it doesn’t take a rocket scientist to figure out that this strategy has little chance of producing superior performance over the next 30 years.

This simple example makes it easy to see how you can tinker with past data to produce a *system* that looks good on paper. This practice, called *data-mining,* involves using the benefit of hindsight to study historical data and extract bits and pieces of information that conveniently fit into some philosophy or some notion of reality. Academic researchers would be quick to tell you that any conclusions you draw from data-mining are invalid and unreliable guides to the future. But every market timing system is based on some form of data-mining, or to use another term, some level of *optimization.* The only way you can devise a timing model is to figure out what would have worked in some past period, then apply your findings to other periods.

Necessarily, every market timing model is based on optimization. The problem is that some systems, like the enhanced S&P 500 example, are over-optimized to the point that they toss out the *garbage of the past* in a way that is unlikely to be reliable in the future. For instance, we recently looked at a system that had a few *rules* for when to issue a buy signal, and then added a filter saying such a buy could be issued only during four specific months each year. That system looks wonderful on paper because it throws out the unproductive buys in the past from the other eight calendar months. There’s no ironclad rule for determining which systems are robust, or appropriately optimized, and which are over-optimized. But in general terms, look for simpler systems instead of more complex ones.

A simpler system is less likely than a very complex one to produce extraordinary hypothetical returns. But the simpler system is more likely to behave as you would expect.

To be a successful investor, you need a long-term perspective and the ability to ignore short-term movements as essentially *noise.* This may be relatively easy for buy-and-hold investors. But market timing will draw you into the process and require you to focus on the short term. You’ll not only have to track short-term movements, you’ll have to act on them. And then you’ll have to immediately ignore them. Sometimes that’s not easy, believe me. In real life, smart people often take a final *gut check* of their feelings before they make any major move. But when you’re following a mechanical strategy, you have to eliminate this common-sense step and simply take action. This can be tough to do.

You will have long periods when you will underperform the market or outperform it. You’ll need to widen your concept of normal, expected activity to include being in the market when it’s going down and out of the market when it’s going up. Sometimes you’ll earn less than money-market-fund rates. And if you use timing to take short positions, sometimes you will lose money when other people are making it. Can you accept that as part of the normal course of events in your investing life? If not, don’t invest in such a strategy.

Even a great timing system may give you bad results. This should be obvious, but market timing adds a layer of complication to investing, another opportunity to be right or wrong. Your timing model may make all the proper calls about the market, but if you apply that timing to a fund that does something other than the market, your results will be better or worse than what you might expect. This is a reason to use funds that correlate well you’re your system.

The bottom line for me is that timing is very challenging. I believe that for most investors, the best route to success is to have somebody else make the actual timing moves for you. You can have it done by a professional. Or you can have a colleague, friend or family member actually make the trades for you. That way your emotions won’t stop you from following the discipline. You’ll be able to go on vacation knowing your system will be followed. Most important, you’ll be one step removed from the emotional hurdles of getting in and out of the market.