I previously introduced three different seasonal edges that are constant enough to be exploited, and that form the basic setups for this particular trading strategy (I suggest reading those articles to fully understand the following conclusion). The first is a bullish bias on days of an FOMC meeting, the second is a bullish bias on the first trading day of a month, and the third and final one is a bearish tendency between the open and the close on options expiration Fridays.
These setups are just easy examples of a multitude of different patterns that can provide an edge big enough to build a whole trading strategy around it.
Even though all of our setups proved to be constantly profitable, there is a lot more work to be done to turn it all into a solid and especially tradeable system.
One more safety net
First of all, we want to make sure that our system will endure a wide range of market conditions to deliver the stability we all seek.
A natural reaction would be to go through the basic rules of our setups and start to add other indicators or conditions that might even out some of the experienced drawdowns, but this would only slowly push us into over-optimized territory, speak we tweak and twist our own rules until the result looks promising, but our trading rules are too specific and won´t necessarily work out well in the future.
Instead we choose another more practical approach: analyzing the system's equity curve to determine whether a strategy should be traded or not.
There is one bittersweet truth about trading systems that most developers overlook when they start out searching for a "holy grail": There is rarely any system that will work perfectly through all kinds of markets without suffering longer drawdowns or flat spells. The secret rather is to know when to trade a system and when to abandon it.
This also highlights the advantage of having a set of say five to ten different systems on hand, of which some will perform brilliantly while others are temporarily laid off until more favorable conditions come around.
So instead of further analyzing the original underlying, we can take a look at the equity curve of the system itself in order to determine when to pull the plug on it and just accept the fact that now is not a good period.
To differentiate between those periods, one could plot a moving average on the equity curve and only trade it if the system closed above it. One could also calculate the trailing expectancy for the past trades or even just the hit rate to determine the validity of the traded edge.
In this case I simply took the rate of change of the equity curve. More important with this system is the chosen period, as all of the three subsystems generate only between eight (Fed-days) and twelve (First day of the month) trades per year, so plotting something on a one-year period will skew the results, as we have maximum a dozen different data-entries and the results will be too random.
To get a better glimpse on if the subsystems are in a favorable period, I decided to consider the rate of change over the past 20 trades; or in other words, a subsystem is only traded if its equity was higher than twenty trades ago. In case we skipped a trade, the theoretical result of course still had to be included in the next calculation to keep up with reality.
So here are the three subsystems before and after adding this safety feature:
The end result is effected in a slightly negative way, as there were barely any really bad periods for those subsystems in the past 25 years; but to point out the real advantage of this method, I want you to take a look at the blue system-equity in this graph:
(click to enlarge)
This curve is basically all over the place, with good and bad periods. It is the equity curve of the same "First day of the month"-edge we used as the second subsystem, only over the 25 years prior to our backtest-period (1963-1988).
After applying our good-and-bad-times-filter, the overall performance improves substantially, turning this edge into something tradeable after all (red curve).
A routine like that puts an extra safety net under a mechanical trading approach and prevents deep losses without compromising the simplicity of the original setup, which makes this technique worth a consideration for a vast range of trading styles.
One more thing to consider with this particular system is the lack of trades, as we only generate a maximum of 32 possible trades over a whole year (around 250 trading days), given that the 20-trade rate of change is positive at any given time.
Finding more edges or subsystems to integrate is one possible solution, parking the funds in something that actually earns interest is another. Especially the fact that we already know well in advance when we are going to take our next trade, makes this variation quite interesting.
It seems less attractive in our actual low rate environment, but consider that there were times one could actually add 3% p.a. or more which would be enough to make up for a major part of the transaction fees etc.
How to trade it
This is one of the most crucial parts in developing a functioning trading system, especially for a strategy that seeks to capture moves of less than 50 basis points in average.
There are three different instruments on my mind that could be possible candidates:
ETFs have quickly become an extremely popular tool for investors to participate in a vast range of indices, sectors and specialty strategies; and for a good reason. A majority of them are highly liquid combined with a very low expense ratio. What makes them particularly attractive for retail investors is the fact that they offer a easy way to get exposure to leverage and inverse price movements while they can be traded in the simplest of brokerage accounts and in low increments. Transaction fees for ETFs depend on the broker you choose. InteractiveBrokers for example charges half a cent per share (1 penny/roundturn) within its flat rate program, which is next to nothing on a $150 share like SPY, even considering the average expectation on our system's trades (0.3% on SPY = $0.45). Some ETF/brokerage combinations are even completely commission free; for example, TD Ameritrade and Fidelity do not charge fees on trades in IVV at all.
This leaves us with the last tricky issue to account for: slippage.
As all of our trades in this system take place either at the open or the close of a trading session, we are exposed to the most volatile times of the day; but the good news is that we are not following a momentum or breakout strategy, which means we randomly either trade with the order flow or against it, which concludes that we will experience negative and positive slippage.
There is not too much data available on this subject, but I think accounting for a 2 cent slippage at either end of the trade is fair, which will put us on total "costs" of around 4-5 cents a share and trade, which is still only about a tenth of our average trade expectation.
Futures are a good way for more sophisticated investors to trade this strategy, but less favorable for smaller accounts. They too are highly liquid and very cheap when it comes to transaction fees. The biggest hurdle for retail investors is to be eligible for and fund a futures account.
At let's say 1500 points, the value of a full S&P 500 contract sits at a stunning $375,000 and even the smaller e-mini future comes on $75,000; so those would be the needed account minimums to trade this strategy without leverage. Another disadvantage is the lack of scaling possibilities, as there is basically no way to trade fractional contracts and get a 1:1 exposure.
The costs are ridiculously low with IB charging $0.85 per contract and transaction and not applying any carrying fees for holding contracts overnight. This puts us at $1.7 per trade on a 75k transaction "value". Slippage again is a whole different chapter, and the same applies of what was said about ETFs; negative and positive slippage can occur. Some papers suggest an average slippage of one tick or 0.25 points in S&P Futures (0.5/roundturn), which equals about 0.03% of the face value and again about a tenth of our expected trade return.
CFDs (contracts for difference) are a relatively unknown vehicle, even though they offer a fair way to participate with very small accounts. CFDs are like small siblings of futures, where one unit usually represents an exposure of $1 for each point of change in the underlying which makes them perfect to scale in size. The required margin for these products lies between 1 to 4% on equity indices and if and how much transaction fees apply vary from provider to provider. What is important to know is that the pricing of CFDs is completely subject to the provider, so even though they are supposed to move with the underlying cash indices, there is no guarantee and therefore slippage is extremely hard to quantify.
Companies like GFT UK and Oanda finance themselves through the spread which lies at half a point on the S&P 500 equivalent, others like IB only charge a small transaction fee of 0.005% (0.01%/roundturn).
All three instruments offer different advantages, but concerning all different factors like pricing, fees, scalability and access through retail accounts I would call ETFs the best option for the average individual investor. Within this group, I consider SPY and SH (no leverage), SSO and SDS (for 2x leverage) and UPRO and SPXU (for 3x leverage) the best alternatives in the vast ETF universe to trade this strategy.
To show the full impact of transaction costs, we plot our combined edges system with a set of different assumed grades of slippage and trading fees.
The first set accounts for 0.5 cent/share fees and 2 cents per share slippage, on either end of the trade; which adds up to 5 cent per share and trade, or 0.033% on a SPY price of $150.
The second version adds another cent of slippage on either end, so 7 cents per share or 0.047%, while the last version calculates with 4 cent of slippage on either end which leads to 9 cents or 0.06% transaction costs.
(click to enlarge)
Obviously, the transaction costs make a dent in the overall performance of this system, but not to an extent that this strategy would be untradeable. There are for example a few things one could change to boost the results. First, there still is the fact that this system is inactive about 90% of the time and the unoccupied capital can still earn interest.
Second, one could apply some leverage to the system as this strategy runs quite stable and does not experience a lot or deep drawdowns. Of course, caution is necessary and every individual's risk appetite should be considered.
The system within a portfolio
One of the main objectives of this article series was to cover the advantages of adding mechanical trading systems into a regular equity portfolio, in order to smooth out the ups and downs of the broader market. A single strategy is not representative of a whole basket of systems and of course any portfolio looks better if one of the components is as stable as the one I just introduced. So the following chart has to be taken with a grain of salt, but it still is a good indication of why every investor should at least consider applying rule based strategies.
In this case, we took the System that accounts for a 4 cent slippage on both ends of the trade (the weakest version), and the S&P 500 representing a Buy and Hold equity portfolio. Both parts started out with a 50% weighting and got rebalanced to a 50% weighting every three months.
(click to enlarge)
While not compromising too much the overall performance, this approach smooths out the equity curve quite nicely and turns the drawdowns from disasters into bumps along the road.
As a final note, I just want to mention the Quantifiable Edges Blog that delivered a lot of the inspiration for this work and I want to make clear that I am not receiving any compensation in any form from the brokerage firms mentioned above. I have/had personal accounts with each of them and only referred to them as examples on how to turn this strategy from theory to live trading.