April 10, 2017
Why Trade FX?
April 10, 2017
How to Read Daily Reports
April 10, 2017

Methodology – Approach to Forecasting

April 10 2017      Barbara Rockefeller

Forecasting foreign exchange rates is notoriously difficult. Economists don’t have a standard model of what determines exchange rates, and even the most brilliant and closely-reasoned forecast can run off the tracks as events unfold. But multinational corporations and global fund managers must have forecasts in order to select foreign activity priorities and to protect against catastrophic loss.

Exchange rates tend to be more highly trended than other financial price series (such as individual equities). This is because the institutional and economic variables underlying exchange rates tend to move slowly, such as relative inflation rates.

While we study these fundamentals and write a 2500-word summary every day before breakfast of what is moving the FX market, far more reliable than human judgment is techncial analysis.

Technical analysis consistently identifies market trends.

Fundamental economists may get the direction of exchange rate moves right, but they hardly ever get the timing right. The premise of technical analysis is that the price alone contains all the information we need to know about market sentiment, and a trend once in place will continue until something comes along to reverse it. That “something” can be a government policy change, new information about the state of an economy, or just a fresh way of looking at a currency. Whatever factor is dominant at the moment doesn’t matter: the price reliably reflects the sum of market perceptions about the currency.

Whether you are managing a financiaql transaction (like a dividend) at a multinational firm or trading FX as a security class, foreign exchange management is a continuous process of trying to hit a moving target. The Rockefeller approach is to use all available tools, fundamental and technical, to achieve excellence in the management process. We define excellence not as “being right” more than 50% of the time, a common misconception, but rather always knowing the range of possible outcomes, i.e., reducing uncertainty over the chosen timeframe.

The Rockefeller Methodology

We employ a highly disciplined quantitative methodology. “Technical analysis” can be of several different types, including cycle theories, “neural networks,” and other schools of thought. RTS has designed its proprietary technical analysis system using a combination of well-tested statistical modeling techniques.

Each currency has a customized set of models, which are weighted and combined in a “confirmation approach.” Two of the models use the moving average crossover concept, whereby the current spot price rising above various moving averages generates a “buy” or “uptrend” signal. Moving average models are the workhorses of statistical analysis and are robust in identifying new trends– but they are backward looking, or lagging indicators.

The second valuable concept in statistical analysis is momentum, which captures the rate of change in a price series. This is a current indicator; a price moving up with greater acceleration is statistically more significant than a price moving more slowly.

We use two momentum models for each currency. Finally, we wish to measure volatility. A period of low volatility, in which a price moves significantly less than its statistically-defined “average true range”, usually precedes a breakout either up or down.

Volatility models are forward-looking; we use two volatility measurement models. We weight each of the models and generate buy/sell signals based on the sum of the weights. These models were designed and developed in 1990 and have been little changed since; they are simple and powerful, and tend to remain successful as the global environment changes over time.

Our modeling system is scientific in the sense that anyone using the same price series and the same statistical models could replicate it. Technical analysis out performs human judgment over long periods of time, even when new signals may seem to be counter-intuitive. We never override the system output with judgment. The decision to ignore, or override a technical signal belongs to the client, and we offer commentary and consulting to refine that decision.

Rockefeller Trading Philosophy and Methodology

Our approach is mostly technical, tempered by a thorough understanding of the fundamentals. See the FAQ for a discussion of why you really should know the fundamentals and keep up with them. In a nutshell, it’s to prevent unhappy surprises.

Aspects of Technical Analysis

We like to say that the purpose of technical analysis is to quantify trader sentiment. If you see a series of higher highs and higher closes, you have an uptrend and you want to be a buyer to take advantage of the movement of the crowd (or “bandwagon”). We have dozens of indicators to measure the rising and falling sentiment behind every move. Note that in FX, we lack volume data to confirm our guesstimates of trader sentiment.

But technical indicators are ALL based on the past behavior of the majority of traders. No technical indicator has predictive properties. Sometimes technical analysts in the grip of religious fervor or wishful thinking assert that Elliott Waves or the stochastic or some other indicator has a crystal-ball capability. This is never true, or rather it’s true less than 50% of the time and even then we have to wonder about coincidence.

As those of us who have burned the midnight oil performing back-tests will tell you, no indicator is reliably correct more than 50% of the time over any serious length of time. You may get 5-7 months when the MACD is right 100% of the time, followed by a 5-7 month period when it’s wrong 75% of the time, or right only after a horrible lag. Each indicator works best on its own timeframe. If you have a 20-period indicator but typically a 5-period holding period, your indicator will not work at all.

Granted, there are regularities in the trading of every security. At the end of every day, and this is especially marked on Fridays and month-ends, a trend will retrace as traders take profit. Many prefer to go home with no position or a pared-back position. After a big move on a Thursday and/or Friday, the following Monday sees a wobble that culminates in a sizeable pullback on Tuesday. We call this “pullback Tuesday” and we need to be alert to it (by widening our stop, for example).

Most of all, old-fashioned bar-reading is very valuable. These days it’s done in the form of candlestick reading, but candlesticks are really just a form of bar-reading and it has only very short-term applicability (1-3 days, usually). Let’s say you have a series of untrended bars going sideways in a narrow range and many of the bars have the open and close at the same level ,or near it (doji bars). Volatility shrinks and by definition, so does the Bollinger band, which is formed by drawing a two-standard deviation band on either side of a 20-day simple moving average.

This is “congestion” and always leads to a breakout. We can usually draw a support-resistance triangle around congestions, by the way. We want to be correctly positioned to take advantage of the breakout when it occurs, but there isn’t a single technical indicator that can guarantee which way the breakout will occur. We like to think that the pre-congestion trend will re-assert itself, but sometimes congestions are the precursor of a trend reversal. How do we know which way to bet? To be honest, we use judgment arising in part from fundamental analysis to assign a probability to whether the next move is up or down.

Remember that the purpose of technical analysis is to get a realistic and fact-based estimate of the next price move. Because human beings are prone to emotions like hope and fear, technical analysis is useful because it removes emotion. On the other hand, sometimes experience has informed emotion so that a bias, which is a form of emotion, has a basis in fact even if the user is not fully aware of it. For example, periodically the market takes out a hit on sterling. Everybody gangs up to sell sterling even if there is no specific crisis at hand. They manufacture the crisis. If you are an experienced sterling trader, you will respect this move beyond what the indicators are telling you. Perhaps some indicators are not reflecting the giant sell-off (all indicators lag—it’s only a matter of degree how much they lag). But an experienced trader will jump on the move even without indicator “proof.” After all, in technical analysis we use the preponderance of indicators. We hardly ever have “proof,” anyway.

We follow a handful of core rules:

1. FX is trended most of the time and trends are powerful. We prefer to trade with the trend and will trade against the trend (“fading” the trend) only if we feel very confident.

For example, let’s say the euro is clearly and obviously in a downtrend but has just bounced up by a very large amount. We know, or think we know, that a bounce normally lasts longer than one day, and in fact may end up retracing 25% or 50% (or some other percentage) of the previous downmove. We would want to buy the euro to take advantage of the expected upmove, or at least not become a seller into a rising market—right?

Sometimes, but usually we do not join the counter-trend crowd. The basic assumption of technical analysis is Charles Dow’s assertion that a trend once in place will remain in place until Something Big comes along to alter its course. This is an adaptation of Newtonian physics but it’s an excellent rule to live by in trading, since you never know when a new piece of news will come along. The probability is high that the fresh news will further the existing trend. If we are long the euro and the fresh news is euro-negative—and it was a series of euro-negative news that put it into a downtrend in the first place—the euro will stop rising and our long position will generate a loss. We wish to avoid losses and we especially wish to avoid foreseeable losses.

This is not to say we will not participate in countertrend bounces or try to estimate where the bounce will end in order to get short at a good level (see Footnote Rule 1), but it is to say we take trend-following very seriously. Variation from trend-following has to be careful and rule-based.

How do we determine trend direction? The core indicators are 20-day moving average and the slope of the linear regression channel. The 20-day moving average in particular seems to have magic properties. Traders observe carefully when a close moves over the 20-day the first time and then holds it for several more days. You will see reference to other moving averages, like the 25-week or 200-day. This is just made-up stuff from the imagination of guys with a lot of computing juice. The 20-day is the one to watch. Here’s another bit of FX technical analysis folklore: if the price breaks the 10-day, it will almost always go on to test the 20-day. Some traders do little else but trade on this rule.

2. We want always to have a positive expectancy of winning. Trading is not gambling, in part because we expect outcomes to be non-random, but we still want to obey the first rule of gambling—don’t bet unless you have a positive expectancy of winning. Using forecasted high-low ranges gives us a framework for positive expectancy.

We use three methods of estimating the expected future high-low range. We need to forecast the expected high-low range in order to know where to set our profit target and where to set our stop.

A. The first is the Bollinger band, mentioned above. Prices very seldom vary away from a moving average by more than two standard deviations. A variation that is bigger than that is statistically abnormal and constitutes a breakout. John Bollinger says you should always respect a breakout and trade in the direction of the breakout.

We find that is not so in FX. In FX, a breakout of a Bollinger band almost always results in a pullback in the other direction. The Rockefeller model treats an upside breakout of a B band top as a contrary signal—a sell signal. In fact, once a B band is broken to the upside, we expect a drop back to the midline (the 20-day moving average) and many times, all the way to the other band. The contrarian rule has to be judged in the context of other indicators, including the fundamental reason for the breakout. There is always a reason behind a breakout and it is our job to judge whether it’s a valid reason or just market hysteria.

An exception is when we get a double, triple or other multiple breakout. Let’s say the price broke the Bollinger band to the upside, a hand-drawn resistance line, the last intermediate high from 2-5 weeks ago, the ATR band and the linreg channel top (see below for the last two). That would be a quintuple breakout or a breakout by five technical definitions, and we would NOT expect the usual pullback that we would expect from a B band breakout alone.

B. A second way of determining the maximum daily price excursion is the average true range, or the recent high-low range over the past few days (incorporating any gaps). If the euro has traded in a range of 120 points over the past 5 days, we assume it will trade in a 120-point range today, too. That means it would be absurd to place a stop within today’s 120-point range since it will most likely get hit, and it would be equally silly to place a profit target outside the range because it is unlikely to be reached. We will write more about trading the range below because it is central to our ideas about risk management.

To be really conservative, to the actual average daily range over (say) 5 days, we add 30% of the range to the top of the band and add 30% to the bottom. If the actual 5-day high-low average of the euro is 120 points, we would place the top of the band at 156 points over the midpoint of the average daily range and bottom also at 156 points away from the midpoint. Now we have a really wide band of expected maximum outcomes. A breakout of this band must be taken very seriously.

An even better way to employ the ATR band is to adapt the percentage amount over/under the ATR number to the trend. If the currency is trending up, it should have to pass a harder test on the downside to trigger a stop. So, the stop could be set at ATR plus 40% of ATR or 168 points in this case. The problem with this approach is that currencies are not always trending and the process is cumbersome unless you are a programming whiz.

C. The third technique is the linear regression channel. The linear regression itself is a line that minimizes the distance between itself and every other data point in the set of data selected by the user. A linear regression channel is constructed by adding two standard errors to the linear regression on the top and also on the bottom. The slope of the linear regression is a reliable indicator of trend direction and the channel is a reliable forecaster of the maximum highs and lows that can be expected going forward into the future. When a linreg channel is drawn correctly—and drawing it correctly is to some extent a function of operator skill—it is stable, meaning the slope changes very little as more data is added. Also, the width of the channel is stable over time, meaning average true range is stable.

The linear regression channel acts as a support and resistance band. We can draw support lines connecting a series of rising lows or a resistance line connecting a series of lower highs, but these get broken all the time and have to be re-drawn. All our hand-drawn support and resistance lines are done in red to indicate they are potentially on-fire.

A breakout of the linreg channel is an Event in its own right. When we first started using this channel, it was not widely used by others. Nowadays, judging from glimpses of websites we sometimes see, it’s used by everybody, at least for display purposes. But we use it for more than display. We say a linreg channel breakout is one of the few indicators we require to name a trend change. That means skill at drawing the channels is essential. We usually start and end at a significant high or low, “significant” being a judgment call. We also extend the channel by hand into the future using dotted lines so you can see where we ended the channel. The dotted lines suggest the future min-max price rage if the slope and width of the trend does not change. This gives us a forecast, if too wide a range to be terribly useful.

Sometimes a linreg channel will fall on top of or near a hand-drawn support or resistance line, and we call it a trifecta when we also get a key moving average (like the 20-day) also tracking the linreg channel line. We consider it more than just luck or coincidence when all three are confirming direction.

3. Timeframe is everything. If you are trading in a very short timeframe like 30 or 60 minutes, you wouldn’t use the daily average range—you’d use the range appropriate to the expected holding period of the risk.

In our afternoon FX trading advice, we expect to enter and exit the position within 24 hours. In practice, it’s usually about 6-12 hours. Therefore, we start with the daily timeframe (and that is what is shown on the charts accompanying the report) but we also look at the 6-hour and 1-hour timeframes, too.

We assume that no one of the George Soros camp is following our trading signals, by which we mean that nobody can afford to have a long-term holding period (months). Surprises and fast reversals do happen in FX. Mr. Soros can live through a huge retracement when he is confident of a long-term trend, but you and I cannot—we have to pare positions and exercise close-by stop-losses. And the professionals know it. We can set a stop at a reasonable distance from the last close using a combination of our range work, but the pros can figure out where our stop is and go gunning for it. Pros set off cascading stop losses all the time. We can try to be a little tricky and place stops and targets away from obvious numbers, but on the whole, they think of that, too. We can always tell when the pros are targeting stops—we miss or get hit by 1-3 points.

What is your timeframe? To be honest, your timeframe is a function of how much money you have to trade with, your “initial capital stake.” (If you can’t afford to take any losses at all, obviously you shouldn’t be trading. You should be saving.) How much you can afford to lose in any single trade is a complicated subject, but on the whole, most beginners with small capital decide to keep their trades very short-term (under an hour or two) in order to avoid an unacceptable loss.

A 1-2 hour expected holding period doesn’t mean you can’t use our trading signals. First, note the primary direction—up or down, buy or sell. It’s perfectly okay to do the opposite if you have a good reason, but remember, our goal as a newsletter publisher is to identify the trend using tried-and-true techniques, and we have been doing it for 30 years. That doesn’t mean we don’t get it wrong sometimes, but mostly getting it wrong is a function of estimating the direction when a currency prices comes out of a sideways move, which is actually a trend in its own right. We quickly adapt when this happens. Just be sure to understand that if you are trading counter-trend, you are taking more risk.

Again, if you are trading in a 1-2 hour timeframe, our stops and targets will be too far away for you, since they are based on a longer timeframe. Adapt your stops and targets accordingly.

3. Risk management

Everybody wants a win loss ratio of 3-5, meaning they would make $3 to $5 for every dollar they lose. This is unrealistic. We don’t know anyone who can do that consistently over any longish period of time. In our case, the win-loss ratio ranges from 1.52 to 2.50 over the past 20 years. It is simple arithmetic that if you have a win/loss ratio of about 1.50, meaning you make $1.50 for every $1.00 you lose, you must have targets that are bigger in point and dollar terms than the stops. That’s if your winning trades are 50% of total trades and your losing trades are 50% of total trades.

If you have a higher proportion of winning trades than losing trades, like 60-40, you can afford to have some stops be bigger than some targets. Which stops and by how much?

Some math whizzes can no doubt figure it out on a rolling, cumulative basis, but we are not a math whizz. Instead we use seat-of-the-pants calculations. Here’s the reasoning: Let’s say we can see on the chart that the price is near the top of several channels and bands as described above. We have an uptrend. The probable lowest low is 150 points away at a channel bottom or support line or some other noticeable spot, and the probable highest high at the channel top or resistance gives us only a 40-point gain. In other words, if we lose, we lose 150 points and if we gain, we gain only 40 points.

From a risk perspective, this is a trade we should let go by without us. You don’t have to take every trade. But you can’t win if you don’t play, and our goal is to make money, not to get a zero-risk situation. A zero risk situation is a savings account. Besides, 40 points is a good amount. In euro futures, it’s $500. If we had to place initial margin of $4500 and we end up making $500, that’s 10% in one day. If we could do it every day (be careful here), over 240 trading days we would make $120,000 on an initial capital stake of $4500. This is a dumb way to think but for the moment, the point is that there is nothing wrong with 40 points.

Our goal, therefore, is to reduce the stop to a number that is less dire than the range-set 150 points. We would prefer the stop to be a multiple of the range because then we would be sure it won’t get hit, but this is almost never possible. So let’s go look for something else with which to modify the stop. Here’s a short list of Bar Events that allows us to do that. We don’t always succeed in getting the number of points in the stop closer to and under the number of points to the target, but the chart does offer realistic options:

Support line

Previous significant low
Midpoint of a big bar, especially a mini-breakout bar
Lowest low of a doji bar
20-period moving average
Linear regression line
Fibonacci retracement
B band limit
Kaufman adaptive moving average
Moving linear regression indicator

Traders today like to use pivots. We think pivots misrepresent the action defined by the bar by excluding the open. When the US market opens after the Asian and European action, we watch that first hour very closely—it tends to set the tone. When Europe has rallied a currency, New York will likely follow suit. Why would anyone not want to watch the open?

Besides, in FX, the “open” is a known number only in futures. In spot FX, the open (according to eSignal) is the New Zealand open at 6 pm New York time. We use it, but maybe the Singapore open would be more representative. The London open is more meaningful than the Frankfurt/Zurich open. We can say the same thing about the “closes.” The only close that makes sense is the 6 pm New York “close” and yet if you are a London trader, you are going to bed when it is being recorded.

Frankly, the popularity of pivot points is partly a function of some software featuring it. Well, traders must have demanded pivot points for the software designers to have included it, and it’s also true that when a technique becomes a fad, it also becomes a self-fulfilling prophecy. But we still think that today there is too much reliance on pivot points and the support/resistance lines drawn off them. They give the appearance of manifest destiny when fate has nothing to do with price movements. We prefer to draw our own “pivot” lines horizontally off major bar events. Our track record speaks to this working quite well.

Barbara Rockefeller

President + Author