So, if you are a smaller trader you do not need to consider tax.
When choosing how to trade a particular instrument, you can forget about futures, which have excessively high minimum capital requirements. CFDs and spread bets have roughly similar cost levels and spread bets usually have smaller minimum sizes relative to CFDs. So, in a fight between CFDs and spread bets, spread bets nearly always come out better.
But what about larger traders? Should they trade futures, which are cheap, or spread bets, which are exem pt from CGT?
Let’s look at a worked example for a gold trader with several hundred thousand pounds in capital (well above the minimum capital for gold futures, which is £88,000). For the Starter System, which is expected to trade 5.4 times a year, gold futures have risk-adjusted trading costs of 0.006, versus 0.029 for gold spread betting. To get the annual cost as a percentage of our capital we multiply these figures by the risk target of 12% (I explain where this comes from in a few pages). That translates to a cost of about 0.07% for futures (0.006 ×12% = 0.07%) and 0.35%
for spread bets (0.029 ×1 2% = 0.35%).
The expected pre-cost return of the Starter System on both futures and spread bets is about 4.9% a year (again, this is calculated later in the chapter). So, a gold futures trader with
£339,327 in capital can expect to make 4.9% – 0.07% = 4.83% a year after costs, or £16,383. After allowing for the tax-free allowance of £11,700 they pay tax on 16,383 – 11,700 = £4,682.
CGT at a rate of 20% would cost 20% × 4682 = £937 leaving them with £15,446 in after tax profits.
A spread betting punter with the same capital would make 4.9% –
0.35% = 4.55%, or £15,446. They have no further tax to pay. You can see why I chose the rather odd figure of £339,327. This is the break-even point where the tax benefits of spread betting exactly match the hindrance of extra costs, at least in this example. With capital of less than £339,327 you are better off trading cheaper futures. With more than £339,327, the tax savings of spread betting give it the edge.
This break-even figure will vary depending on the level of taxes, the precise cost difference between products, and your expected returns. But clearly only extremely wealthy traders can benefit from the tax advantages of spread betting. Indeed, a trader with that sort of capital may struggle to find a broker willing to take his bets. This is doubly true if they are successful and trade relatively infrequently, making them both a large risk and an unprofitab le customer.
Although my own trading account is relatively large, I still prefer futures to spread bets. Firstly, because my orders are passed through to the exchange, which is preferable for reasons I explained back in chapter two. Secondly, because costs are relatively predictable, but returns are not. This makes the tax advantages of spread betting more uncertain than the h igher costs.
When should we open positions?
What makes a good opening rule?
To decide when we should open positions, we need a rule. Most people assume that finding the right opening rule is the only factor differentiating happy profitable traders from sad losers.
Actually, this is not true, as you will discover in the rest of the chapter. Still, it’s important to have a decent op ening rule.
What makes a good rule?
Let’s take a look at the some of the characteristics of good op ening rules.
The Starter System opening rule: A moving average crossover The rule we are going to use in the Starter System is a moving average crossover . Moving average crossovers are part of a larger group of trading rules known as momentum, or trend following, rules. I’ve chosen this rule because it is commonly used by many people, including myself and a substantial number of professional traders. Because it is so ubiquitous, your favourite trading software or charting website will probably calculate it for you. It is also easy to calculate with a spreadsheet.
These rules are intuitive : if something is going up we buy, expecting the upwards trend to continue. If the price is falling, we sell. They are explainable : there are several credible 20 20
reasons as to why these rules might work. 75 75 They are profitable
, with considerable evidence that momentum has worked in the 20 20
past, both from back-tests 75 76 and the successful track records of professional traders using this type of strategy. Finally, the rule is simple and objective . It can be calculated with a short formula and requires no subjecti ve judgment.
Now f or the rule.
First, I need to define a moving average. The moving average of an instrument price is the average price over the last n periods, where n is the moving average length . In this book I use daily prices, so n is in days. A moving average with length n ( MA n ) is equal to the average over the last n prices at time t as follows:
Formula 12: Moving average
MA n t = ∑(P t + P t-1 + P t-2 + …. + P t-n+1 )/n The following is the moving average crossover for a pair of speeds f (fast) a nd s (slow):
Formula 13: Moving average crossover (MAC) MAC f,s t = MA f t – MA s t
When prices change, a faster moving average (small n ) will react to the change quicker than a slower moving average (large n ).
Hence, in an uptrend the faster moving average will move higher than the slower, and in a downtrend their positions will be reversed. So, if the MAC is positive, we will be in an uptrend, and the rule will want to go long. When the MAC is negative, we are in a downtrend, and would be going short: MA f > MA s ; MAC f, s >0: Go long MA f < MA s ; MAC f,s <0: Go short Figure 3 shows how a version of this rule worked in the 2008
financial crisis, trading the S&P 500 equity index. A clear sell signal is generated in June when the 16-day MA falls below the 64-day MA. Then in April 2009 a buy signal appears when the 16-day MA goes above the 64-day.
Figure 3: Moving averages over the financial crisis What crossover lengths?
How do we choose the right pair of moving average lengths, the optimal values of f and s?
As with selecting an instrument, the choice of crossover lengths will depend on both pre-cost returns and costs. Moving average crossovers with lower values for f and s will pick up more short-lived trends, but will trade frequently, giving them higher costs. Higher values of f and s will only detect a few slower trends but will be cheaper because they trade less often.
Firstly, let us examine pre-cost returns across a selection of 20 20
crossovers. 75 77 Figure 4 shows the back-tested average Sharpe ratio for several crossovers, with the average taken across 37
instruments that I trade in my own portfolio. As I did with instruments in figure 2, I show the uncertainty of the Sharpe ratio estimates by displaying the values in a box and wh iskers plot.
Figure 4: Comparison of moving average crossover lengths The rules go from relatively fast (lower values of f and s (on the left-hand side)) to relatively slow (higher values of f and s (on the right)). The rule I chose for the Starter System lies in the middle (MA 16 64). All the rules have very similar performance, shown by their overlapping boxes and whiskers, except for the very fastest rule (MAV 2 8), which is significantly worse than both MAV 8 32 and MAV 64 256.
With the possible exception of the very fastest rule – which looks significantly worse – there is no reason to pick one set of crossovers based on pre-c ost returns.
Now let’s consider costs: transaction and ho lding costs.
First, I calculated the number of trades done each year by the trading rule. The level of transaction costs will depend on the instrument and product: it will be the cost per trade, multiplied by the number of trades done each year. I also need to add on the holding cost, which is the same regardless of the trading rule that is used. Table 8 shows us the figures for each crossover, 20 20
and for a selection of instruments an d products. 75 78
I have also included the ‘speed limit’ cost threshold for each trading rule: one-third of the average back-tested, pre-cost S
&nbs
p; harpe ratio.
Table 8a: Faster crossovers are more expensive, but it depends on the product
Estimated risk-adjusted trading costs for different length moving average crossovers.
• Product is too expensive to trade with rule (trading cost greater than speed limit: 1/3 of pre cost Sharpe ratio values shown in sec ond column).
Values shown in body of table are total cost calculated as H +
(TC×T) where T is the number of trades per year (first column), TC is the transaction cost per trade and H is the annual holding costs; all costs are risk-adjusted. T has been estimated as an average over back-tests for all the instruments in my own data set. Costs calculated using the formulas in appendix B.
Table 8b: Faster crossovers are more expensive, but it depends on the product
Estimated risk-adjusted trading costs for different length moving average crossovers.
• Product is too expensive to trade with rule (trading cost greater than speed limit: 1/3 of pre cost Sharpe ratio values shown in sec ond column).
Values shown in body of table are total cost calculated as H +
(TC×T) where T is the number of trades per year (first column), TC is the transaction cost per trade and H is the annual holding costs; all costs are risk-adjusted. T has been estimated as an average over back-tests for all the instruments in my own data set. Costs calculated using the formulas in appendix B.
Corn futures are really cheap instruments which can trade all the crossovers, without spending too much on costs. However, we cannot trade the fastest crossover (2,8) using a Euro Stoxx 50
dated CFD. From the table it would cost 0.139 Sharpe ratio (SR) units to trade including holding costs, but from the second column, the most we should be prepared to spend on costs for the fastest crossover is 0.023, one-third of the pre-cost back-tested SR (which is 0.069). In fact, the two fastest crossovers (2,8 and 4,16) are too expensive for all the non-futur es products.
You can now see why I’ve avoided choosing any of the faster crossovers. They would be prohibitively expensive, except for the very wealthiest traders who can trade futures. In contrast, MAC
16,64 is relatively affordable, with costs that are comfortably below the ‘speed limit’ of one-third of expected SR for all the instru ments shown.
But why haven’t I chosen either of the two slowest crossovers (32,128 and 64,256)? In figure 4 both have similar pre-cost performance to the crossover chosen for the Starter System (16,64), and they would be a little bit cheap er to trade.
I have two reasons for ignoring the very slowest crossovers.
Firstly, for many instruments their performance will end up being similar to a static long position in the underlying instrument.
One example: US bonds have mostly gone up in price over the last 40 years, so in the back-test for US 10-year bonds the MAV 64,256
crossover has been long around 72% of the time.
Secondly, with so few trades per year there is a risk that you will get bored. Some level of tedium is to be expected with system trading, but excessive boredom is problematic. Bored traders tend to end up ignoring their system and making their own trades. Hopefully the crossover MAV 16,64 is just active enough to keep you interested.
From figure 4, the pre-cost Sharpe ratio of the Starter System using the MAV 16,64 crossover is 0.24 and, from table 8, the expected number of trades per year is 5.4 . I will use these values throughout the rest of this chapter to calibrate the rest of the system (in fact, I have already used them to determine whether an instrument has a viable level of tra ding costs).
How large should positions be?
The next component of the system concerns position sizing . This is a vital part of any trading system. Many traders end up losing money because they take on too much risk. Even if a trading rule is 99% accurate, you will still eventually blow up your account if your position sizing i s incorrect.
Calculating position sizes for a given trade is a two-s tep process.
Step one : determine the required notional exposure in your home currency for your chosen instrument. For example, we may want to take £7,500 of long exposure to the Euro Stoxx 50 equity index.
Our desired notional exposure is the same regardless of the product we are using.
Step two : calculate what that exposure corresponds to in units of the relevant product : how many futures or CFD contracts; or how many £ or $ per point of spread bets.
How do we determine our notional exposure?
The formula for working out the required notional exposure is relati vely simple:
Formula 14: Notional exposure from risk and capital Notional exposure = (target risk % × capital) ÷ instr ument risk
%
The target risk is the annual standard deviation that you want on your account . I explain how this is calculated in a moment. Your capital is the total amount of money you are currently risking, denominated in your home currency (the currency used to value your account). If you have made profits since you began trading, 20 20
you may want to add these to your capital. 75 79 Any losses must be deducted. For most people capital will be equal to the current value of their trad ing account.
The notional exposure will also be in your home currency.
Finally, the instrument risk is the annualised standard deviation of returns. I’ll discuss how you can actually measure risk in chapter six.
As an example, suppose you have a risk target of 12% a year (I’ll explain where this figure comes from shortly – page 99 if you’re curious) and £10,000 in capital. For the S&P 500, which has instrument risk of 16% a year, your notional exposure targ et would be:
Notional exposure = (12% × £10,000) ÷ 16% = £7,500
20 20
This is a leverage factor 76 70 of £7,500 ÷ £10,000 = 0.75. Because this particular instrument has a higher risk than your target, you end up with a leverage factor of less than 1, with a notional exposure that is less than your capital. If the instrument risk was lower than the target, the leverage factor would be hi gher than 1.
How do we determine our target risk?
Formula 14 is pretty simple, but determining the right level of target risk is a little more difficult. Target risk should be the set at the lowest, most conservative, value from the fol lowing list:
maximum risk possible given leverage allowed by brokers or exchanges
maximum risk possible given prudent lev erage limits maximum risk given your own personal appet ite for risk optimal risk level given the expected profitability of your tr ading system
Maximum risk possible given leverage limits The maximum possible risk target given a particular levera ge limit is:
Formula 15: Risk target possible given maximum leverage Risk target = (Maximum leverage factor × inst rument risk) As an example, consider a margin trade in the SPY ETF, which tracks the S&P 500 index. The maximum leverage factor allowed for a margin account is usually 2, and the instrument risk is around 16% (annual standard deviation of returns). This implies that the maximum risk tar get will be:
Risk target = (2 × 16%) = 32%
Leverage is more generous in derivative markets. To trade the S&P
500 as a spread bet requires 5% margin with my current broker, equating to a maximum leverage of 100% ÷ 5% = 20, and a maximum ris k target of:
Risk target = (20 × 16%) = 320%
As you might expect, margins tend to be higher for more volatile products and lower for safer products. My broker requires a 50%
margin to trade spread bets on the cryptocurrency Bitcoin, which has instrument risk of around 100%. That equates to maximum leverage of 2, and a maximum risk target of 200%; not so different than the limit for S&P 500. Generally, for a given product the effect of margin and volatility will roughly cancel out, giving similar maximum risk targets across different instruments.
However, as I will explain, even a risk target of 32% is quite high for the Starter System. Unless you are an extremely aggressive trade
r, leverage limits will not cramp your style.
Risk possible given maximum prudent leverage From 2011 to early 2015 the Swiss franc to euro exchange rate had extremely low risk, as the exchange rate was strictly controlled by the Swiss Central Bank. Many traders took advantage of the 200:1 leverage allowed by certain FX brokers, and speculated that the banks would continue holding the rate down. Unfortunately, the Central Bank changed their minds. They allowed the FX rate to move freely in January 2015, when it increased from 0.83 to over
1.00 in a matter of minutes. This wiped out the accounts of thousands of traders who used too mu ch leverage.
Broker and exchange leverage limits are usually far too generous.
Sensible people should set their own leverage limits that reflect the worst possible scenario. You should estimate (i) the worst possible loss on an instrument that could occur before you had a chance to close your position, and (ii) the highest bearable loss to your account. The maximum prudent leverage w ill then be: Formula 16: Prudent leverage factor
Prudent leverage factor = Maximum bearable loss ÷ worst possible ins trument loss
Now you can calculate the maximum implied risk target: Formula 17: Maximum risk target given prudent leverage Prudent maximum risk target = Prudent leverage factor × ins trument risk
For example, it is possible that the US stock market could crash 25% in a day, as it did in October 1987 (worst possible instrument loss = 25%), and suppose that you could not bear losing more than a third of your capital in one day (maximum account loss = 33.3%). The instrument risk of the US stock market is around 1 6% per year.
Prudent leverage factor = 33.3% ÷ 25% = 1.333
Prudent maximum risk target = 1.333 × 16% = 21.3%
Personal appetite for risk
Can you cope with seeing an annual standard deviation of returns of 24% on your account? Would you prefer 12%? You have probably responded with a shrug and a blank look. It’s difficult for most people to relate to the rather abstract measure of risk which I use in this book.
Leveraged Trading: A professional approach to trading FX, stocks on margin, CFDs, spread bets and futures for all traders Page 10
