Successful Trend-Following Systems

As a guide, every successful trend-following system somehow seems to incorporate elements of the following method:

1. Setup and stock selection: Buy into a strong upward trend, immediately it resumes itself following a short-term pullback. Maxim: trade in the direction of the prevailing trend (to maximise both the probability of a win, and also the potential size of the subsequent price movement).

2. Timing of entry: earlier = higher risk of being stopped out, but greater profit (more favorable entry price) if successful. Excessively late risks missing the trend altogether.

3. Distance of stop loss exit: tighter stop = higher number of trades stopped out (lowers win rate), but keeps average loss smaller. Maxim to aim for: cut losses short.

4. Distance of profit exit: looser trailing stop = lower risk of profit being cut earlier in the trend, but gives back more profit when ultimately triggered. Maxim to aim for: let profits run.

5. Position size: the ultimate determinant of bottom line. Priority: manage risk. Apply capital so that risk is approximately equal (maximum 2%) across each trade undertaken. Keep sizes small enough to avoid emotional involvement.

6. Psychology: trade a proven, pre-defined plan consistently, with patience and discipline. Trade what you see on the chart, not what you think or feel.

There is sufficient randomness in short term movements to make minor variations in execution largely irrelevant, e.g. which indicators are used, variations in timing. However, it is important to prove your own idiosyncratic variations to your satisfaction, so that you can confidently trade your plan.

To point #1, I would add the following considerations, to further improve the probabilities:

a) Buy the strongest performing stocks in industry sectors that are outperforming the market index (use "relative strength comparative" or similar indicator).

b) Trade in the direction of the market index's trend (to reduce "market risk").

c) Buy stocks whose technical analysis exhibits the weakest and/or most distant areas of potential resistance (i.e. best reward to risk ratio, greatest freedom of price movement).

d) Buy stocks that are liquid enough to avoid slippage.

e) If you follow fundamentals, buy stocks exhibiting the best fundamentals.

f) As an alternative (or foil) to #1, also buy stocks breaking out of a sideways pattern (or narrowing wedge), with high volume.

g) For short positions, the reverse applies, but in general it's necessary to act more swiftly, because bear movements tend to be sharper.

To point #5, I would add the following consideration:

h) Consider diversifying capital across multiple uncorrelated stocks that meet the above criteria, to reduce loss, should one trade "earthquake" against you. Benefits of diversification and/or high throughput include: mitigation of risk, greater opportunity for "edge" to prevail, smoother consistency of income, accelerated learning curve, more frequent compounding of gains (and losses!), lower emotional involvement, accelerated statistical confidence.

Bottom line is that if you catch enough trends that move far enough to cover both dealing costs, and also profit consumed by their ultimate reversal, you will finish up ahead. Otherwise not. Simple enough in principle.... :-)

It's important to understand statistical significance. Winning 7 trades out of 10 proves little; winning 70 out of 100 gives a much greater measure of confidence. The more extensive your track record, the more confident you can be that your system will work, on balance, across all market conditions. Or perhaps more pertinently, under which market conditions (bull, bear, trending, sideways, volatile, inert, etc) your system works best. Back testing can be a fair proxy for lack of a real-life track record.

Now I'd like to use Intel (INTC) to illustrate some of the points that I made above.

In the above chart (as at May 2005) we see that the long term trend (thick blue lines) is up, because INTC formed a higher low at point 2 than point 1. This is arguably not a perfect situation, because the high at point 4 is slightly lower than the high at point 3. However, note that at the right hand edge, there is more room for the stock to move upward, in the area between the blue lines.

The red lines denote an intermediate term trend. Both red lines are sloping upward, the result of the stock making both higher highs and higher lows. We will come back to this area of the chart shortly...

...but first I want to look at the area between points 1-5-6-4-7. Point 6 forms a higher low than 1, suggesting at this time that a rally might be underway. If you were trading the longer term trend, then the wave between 5 and 6 is your "pullback", so you would be looking to enter as soon (after point 6) as you are confident that the upward trend is resuming, following the pullback. (See point #2 in my earlier post re timing of the entry - how early or late you enter depends on your own preference).

A longer term trader would then likely set a "loose" enough trailing stops to trade this part of the wave all of the way up to point 4, but then would get stopped out somewhere around point 7. Thus profit is being forfeited in determining that the upward trend is now reversing. As discussed in point #4 of my prior post, setting the trailing stop more tightly runs the risk of an exit earlier in the trend, but of course one is forfeiting less profit in the retracement wave. A delicate balance between risk and return.

OK, now let's zoom in on the area between points 6 and 4 on this chart, to see what a shorter term trader might be hoping to achieve:

Without labeling the chart, you should be able to see the classic zigzag pattern, how each upward movement is followed by a pullback, but that both successive local highs, and successive local lows, are higher than their predecessors, causing the yellow support and resistance lines move upward in step-wise fashion. The same principle applies for the shorter term trader, wait for the pullback, ensure that a new low has been formed that's higher than its predecessor (that's the "setup"), then wait until you're confident that the upward trend is resuming (again, how far is the "timing the entry" issue). There are 5 or 6 profitable upward moves in this section of the chart. Note how the length of the upward moves exceed the length of the downward (pullback) moves. That's how trading "with the trend" increases opportunity to obtain profit, the longer moves increase margin for error, because you're "swimming with the prevailing tide".

Remember, you are forfeiting profit at the start of the move, to convince yourself that the trend is upward, and also at the end of the move as proof that a reversal has occurred. There are also transaction costs that also erode profit. So it's important that the move is sizeable enough to cover all of this - hence the maxim to trade with the trend.

At point 7, a lower low is formed, which marks the start of the longer term correction wave (downtrend). To be fair, obtaining 5 or 6 successive profitable moves is really a "best case scenario", as patterns are not normally as clear-cut as this, but it illustrates the method behind a trend-following trading system extremely well. And of course, we are viewing all of this in hindsight!!

Moving a little forward in time, the above chart shows how points 4, 8 and 9 form a "triple top", which is generally a bearish reversal signal. Note how the low at point 7 is lower than its predecessor (unlabeled, sorry, but between points 4 and 8), further increasing the probability of reversal. We can see the downward wave with the yellow lines now moving in descending steps. Unless we are short selling, we would be steering clear of INTC here: there are plenty of other stocks to choose from. However tempting it might be to trade the small upward waves here, this would be trading against the downward tide, so we must exercise patience here.

Now I want to zoom back to the area at the right of the chart, which is the final up-down-up zigzag, at the hard right edge of the first chart. As discussed, both the long term (blue) and intermediate term (red) trends are up, so we have probabilities in our favor. If INTC's sector is outperforming the market, and the market indices (S&P, DJI, etc) are also rising, then we have further INDEPENDENT corroboration (see points (a) and (b) above).

Now I would like to discuss entry timing specifics. [Note: I have relabeled the points on the above chart].

Many traders use indicators to assist them, I am using nothing more than OHLCV. Notes refer to labeled points on the chart -

(1) Early entries at these points, following a full-bodied unfilled (close > today's open) blue (close > yesterday's close) candle, cause losses when the stop is triggered (presumably somewhere before point 2). This type of early (aggressive) entry carries extremely high risk, as there is little to confirm that the upward trend is in place. More often than not (in this case, three times) we will be stopped out. However, a tight stop means only a small loss.

(2) An unfilled blue candle. In hindsight, this is entry very close to the local low, giving the best possible entry point. However, in real time, the fact is that this qualitatively not much different from (1), i.e. extremely risky. An eagle eye might spot the following provisos: a) there has been a sharper downward movement immediately prior, b) a two candle uptrend is now in place, c) the first of which is a doji (reflecting a possible weakening in sentiment), and d) volume is high, denoting what Richard Arms calls a "washout", i.e. possible selling exhaustion.

(3) The upward trend is starting to mature. This is what I would call a "medium" risk entry.

(4) The safest entry, with the upward trend maturing, and surpassing all previous highs (see the horizontal black line) in the trough created by the prior downtrend.

The problem with all of this, of course, is that I've used a "perfect" specimen example to illustrate my points, and the patterns are not normally this clear-cut, and obvious. But we live in an imperfect world, and must content ourselves with whatever (hopefully) gains the market allows us. Otherwise, why bother attempting to trade?

Now some general remarks:

When analyzing the profitability of your system, there are two essential factors:
a) the number of winning trades to losing trades (e.g. a value of 60% means 6 wins out of every 10 trades, on average);
b) the average win size to average loss size (e.g. a value of 1.7 means that the average win is 1.7 times the size of the average loss)
Note that wins and losses are calculated after deducting costs.
Note also that combining (a) and (b) gives the profit factor (PF), "expectancy" or "edge" (call it what you will).

A good (and realistically attainable) target is to aim for > 50% value of (a), and > 2 for (b). The latter means you are choosing trades where the profit potential is > 2 x the distance from entry to your stop loss.

I've brought this up because I want to discuss some of the biggest issues surrounding entry and exit, irrespective of trading time-frame, market traded, instrument used, indicators used, etc. Irrespective because the bottom line is that you must necessarily enter on a price bar and exit on a price bar. Provided that your method involves some kind of trend following, all of the following points apply:

1. How soon after the trough (long position) or peak (short position) you enter.

The earlier you enter, the higher the risk, in the sense that a "genuine" reversal might not have occurred, i.e. it is just a minor deceleration ("noise") along the trend. As compensation, the earlier entry will generally attain a more favorable entry price.

Hence an earlier entry will, on average, increase (b) but decrease (a). In other words, your profit will be greater on the occasions that you do win, but you will win less often. This gets back to your original question about divergences. Every trend must necessarily decelerate before reversing, and divergences will tend to pick up these decelerations. But not every deceleration will actually result in a profitable reversal. Hence the higher risk.

Conversely, a later entry will, on average, increase (a) but decrease (b). The later entry increases the probability that a profitable trend is emerging; however, one is forfeiting eventual profit at the start of the trend, by entering at a less favorable price.

As I've said elsewhere, profit is determined not by WHETHER the price moves in the anticipated direction, but HOW FAR (and how quickly). Profit is attained only if the price moves far enough to overcome costs (see point 3 below). Because it is difficult to predict how far price will move (some traders attempt to use previous support and resistance, others Fibonacci retracements or space/time ratios, as a guide), if the trend turns out to be short lived, then delaying entry too long actually results in increased risk.

2. How tight your stop losses, and trailing stops, are.

It is widely accepted that the exit is more important than the entry. Here are some major reasons why.

If you set your stop loss too tight, you will get prematurely stopped out of trades that would ultimately result in a profit, more frequently. This is a double whammy: suffering a loss, AND forfeiting a potential profit. However, the tighter stop means that your losses will always be small. This results in an extremely poor (a) value, but a high (b) value.

The same applies to trailing (profit-taking) stops. A tight trailing stop will guarantee locking in some profit, but has the potential to prematurely curtail extended future profit. In other words, it can "cut winners short".

Of course, the converse applies if stoplosses, and trailing stops, are set more loosely. The "looser" the stop, the less frequently premature stop-outs will occur, but the greater the average loss (or forfeiture of profit) becomes, when the stop is triggered. Hence the value of (a) increases, but at the expense of (b).

Another major reason why stop-setting is crucial, is its effect on fixed fractional position sizing (the method that most traders seem to use). As an example, let's say your trading account balance is $10,000, and you are willing to risk 2% (the recommended maximum) on a given trade. 2% of $10,000 is $200. Let's suppose that stock XYZ gives a "buy signal" at $2.00, and you decide to set your stoploss at $1.90. Then if the stop gets triggered, you will lose 10c per share purchased. Given that you are willing to put $200 at risk, then you would buy 2,000 shares (i.e. 2,000 shares @ 10c loss per share = $200 loss).

However, if you were to set your stoploss at $1.95, and it is triggered, then you will lose 5c per share purchased. Given that you are willing to put $200 at risk, then you would buy 4,000 shares (i.e. 4,000 shares @ 5c loss per share = $200 loss).

In other words, the tighter the stoploss ($1.95 vs $1.90), the more shares purchased (4,000 vs 2,000), and the more capital "tied up" in the transaction (admittedly an irrelevance if one is trading derivatives on low margins). Assuming that one works to the above formula, the loss is the same ($200 vs $200), but the potential for return is doubled (4,000 shares x favorable price movement vs 2,000 shares x favorable price movement). Offsetting this is the greater likelihood that the stoploss will be triggered (as already discussed).

An alternative to using a trailing stop is to extract profit by closing positions "progressively", by selling off parts of the position immediately arbitrary target points along the way are reached. The advantage to this approach is that the profits taken are totally secure, whereas stops are more vulnerable to gaps and slippage, especially should a position suddenly "earthquake". The drawback is that potential profit is being sacrificed prematurely, should the trend proceed to "mature" a great deal further.

One final point: some traders don't use "mechanical" stoplosses. But to do so requires both the judgment to recognize when a trade has "gone awry", and the discipline to exit immediately at that point.

3. The three costs involved. These are:

(i) Transaction costs: brokerage, spread, slippage, financing costs, etc.
(ii) Profit forfeited between the trough/peak and the entry point (determined by how early/late the entry is - see point 1 above) .
(iii) Profit consumed between the trough/peak at the end of the trend, and the exit (determined by how tight the stops are - see point 2) .

If you can (on average) overcome all of these costs, you will profit; otherwise you will lose. That is the bottom line.

All of the above applies to any trend following system. Whatever your method, you should BACK-TEST it across a sample (large enough to be statistically significant) of historical data. Here's why:

(i) The overall results will give you values for (a) and (b), telling you how THEORETICALLY profitable (or otherwise) your system is. This is especially important if you plan to be a "mechanical" trader (i.e. entries/exits follow a set of iron-clad mathematical rules).

(ii) A study of the individual trades will help you gauge whether you are (on average) entering too early or too late, and whether your stops are too tight, or too loose. If you plan to be a "discretionary" or "intuitive" trader (i.e. entries/exits follow a set of general principles, applied with judgement), developing this kind of "feel" for what is likely, or unlikely, to happen in different situations, is absolutely vital.

(iii) Study the losing trades, and find out the reasons why. Vary your parameters, and see their effect on (a) and (b). What effect does eliminating/alleviating the losing trades have on the winning ones? Find the best compromise.

(iv) The most important reason to back-test (as if the above were not enough) is CONFIDENCE. You must believe that your system will continue to work, even during the rough times, if you are going to continue applying your system with DISCIPLINE. This is even more important for a discretionary trader, whose more "flexible" rules provide greater temptation to deviate.

(v) It's important to test different samples across differing market conditions (e.g. bull market, bear market, trending, ranging, volatile, inert). This will help tell you over which conditions your system performs best, and what adjustments need to be made to cope with the differing underlying "climates".

One caveat to back-testing. Unlike casino-based games of chance, the "markets" (in reality, crowd behavior driven by - amongst other things - greed, fear and other emotions) do not run according to precise probabilities. What has happened in the past is only APPROXIMATELY more likely, than not, and in some APPROXIMATE form, repeat itself in the future. So back-testing results are guidelines, no more. But if there is any non-random behavior amongst the chaos, then it can be mathematically approximated, and (God willing!) profitably exploited.

4. Effects of (a) and (b) on trading cashflow, and psychology.

Putting the math aside for some real-life practicality: Some traders are willing to live with a system that delivers a value of (a) 50%. As we have seen, we can time entries and set stoplosses to balance (a) and (b) to suit our personalities, goals and lifestyle. To deliver long term profit, the product of (a) and (b) must deliver positive expectancy. In other words, if one is willing to let (a) slip as low as 40%, then (b) must exceed 1.5, and so on. However, if trading is one's primary source of income, then there are other important considerations.

The first is cashflow. A low value of (a) could mean a long wait for the next big winning trade, necessitating careful household budgeting.

The second is psychology. Again, a low value of (a) could mean a long wait for the next big winning trade, leading to any of impatience, despondency, lack of confidence, and ultimately temptation to deviate from the system's rules and principles.

Finally, consider analyzing your losses. Are they of a statistically significant number that some patterns can be detected? For example, are you entering or exiting positions too early, or too late? Missing significant trends? Are you setting your stops too close, or too distant? Is your strategy working for some stocks, but not others? In a trending, or ranging market? Or have you just been unlucky? (i.e. need a higher transaction frequency, to allow the positive expectancy to shine through). Are you becoming impatient and trading low probability setups? If you assemble enough data, you should be able to discern patterns from the outcomes. That is a good starting point.