SMA Archives - Algorithmic and Mechanical Forex Strategies

Easy Ways To Enhance Simple Moving Average Crossover Systems

November 12, 2013 by Andrew Selby Leave a Comment

One of the first types of systems that many people experiment with are simple moving average crossover systems. While there is a wide range of different parameters that quantitative traders can use to build these types of systems, the basic idea behind them is fairly straightforward. This is what makes simple moving average crossover systems so appealing.

One of the biggest problems with these types of systems is the amount of whipsaws that can be generated when a market is trading within a certain range. Jeff from System Trader Success recently posted a few different ways that simple moving average crossover systems could be enhanced in order to eliminate some of those whipsaws.

Jeff took a simple moving average crossover strategy on the Euro and looked at different ways to reduce the number of whipsaws.

Jeff began his post by describing the baseline crossover system he would be using:

Our baseline system will consist of two simple moving averages (SMA) executed on a daily chart of the Euro futures.
I’m picking the Euro because it has demonstrated solid trending characteristics as opposed to the stock index markets which tend to be mean reverting.
If you will recall, signals are generated when a faster moving average (trigger SMA or trigger line) crosses a slower moving average (slow SMA or slow line).

Jeff sets his slow moving average to 50 and his fast moving average to 3 and backtests the simple system from May 2001 through the end of September 2013. The results indicate a profitable system that struggles with whipsaws at times.

The first method that he suggests to improve this is a delayed entry. He suggests waiting ten bars after the signal and then entering on the eleventh as long as the fast moving average does not cross back below the slow moving average. This forces the system to confirm its signal before taking a position:

The idea behind this method is if a new bull market is about to start, price should not fall back below the slow SMA.
In short, it’s another way to measure the amount of conviction for the next market phase. However, we will keep the exit the same

The results of this method reduce the overall profitability of the system, but also eliminate almost all of the whipsaws. The resulting equity curve is much smoother.

The next improvement Jeff suggests is another way to require confirmation of an entry signal:

For example, picture another band above the slow SMA that is 1 ATR above the slow SMA.
In order to open a new long position we require the trigger line to penetrate that ATR band above the slow line.
Now picture another band that is one ATR below the SMA.
This band represents our short trigger when we open a short position. We hope to eliminate some whipsaws by delaying our entry and forcing the market to show us some strength.

This method results in an equity curve that is somewhere in between the original and the first enhancement. This method allows the system to keep most of its returns, while reducing the volatility slightly.

Another method Jeff suggests is implementing a delay with a time decay feature:

Often you will notice a string of whipsaws on a moving average crossover system right after a great winning trade was closed. The market apparently is now morphing to a range bound market and will likely do this for sometime.
However, as the days or weeks wear on the likelihood of a breakout probably increases. Thus maybe we can reduce the delay amount as time goes by. After the close of a successful trade we begin looking for the next cross with our default X bar delay.

One final suggestion Jeff makes to enhance a simple moving average crossover system is probably the most obvious idea:

Maybe only taking long trades during a bull market or taking short trades during a bear market would improve results. This would be an interesting and simple test to perform.

None of these ideas on their own warrant the trading of real capital. However, they do a great job of encouraging traders to think about different ways to enhance their systems. The most successful systems are often very simple strategies that contain just a few key enhancements.

SPY and Dojis

October 10, 2013 by Shaun Overton Leave a Comment

Everyone here knows that I’m a fan of the dumbest possible strategy. The less complex that rules are, the easier it is to understand when something inevitably breaks.

The SPY and Doji strategy from paststat most certainly fits the bill. The SPY is the ETF that tracks the S&P 500 index.

They tested a simple idea: does a doji on the SPY contain any predictive value? The short answer is a qualified yes.

The dojis offered no predictive value on their own. However, teaming them up with a 200 period moving average showed much better results. Use the location above or below the moving average showed a clear relationship between profit and loss.

So am I not touting this to all my readers? The biggest problem with a strategy like this is the sample size: there are only 40 trades in the largest sample. Some of the posted results only analyze the outcomes of 12 trades. That’s not a strategy. It’s hardly an observation, either.

What I do like about the idea is that it hightlights something fundamental about how the stock market works. I always think of it as a pair trade. You buy 1 unit of stock for every X dollars of currency.

You’re really pair trading two asset classes: one is a stock, the other is dollars. The flow of the forex market dominates stocks. A surge in dollar strength almost certainly anticipates a down day in stocks.

The fits and starts of the forex market lead stocks to crawl their way upward. Bull markets last for years and years. Unless you’re counting Fed days, there’s no particular move that dominates in stocks.

SPY Strategy Idea

Contrast that with bear moves and the difference is night and day. Everyone knows about the 1987 crash when the market crashed 22% in a single day. Our more recent crisis in 2008 inflicted similar damage over a period of weeks.

Price crossing the moving average is one of my favorite strategies. It’s dummy proof and I’ve seen it work in several markets, especially forex commodity crosses.

The SMA filter that paststat applied in the doji strategy might just make for a better daily trading strategy altogether. I’d even be tempted to ignore the long trades and take short only signals. A simple filter like the put-call ratio might help filter out the worst of the noise trades.

What I like even more is that it’s a strategy that could apply in today’s market:

The stock market is trading on record margins of debt
Selling the idea of a crisis isn’t hard with the potential US default looming

We’ll have to take a look at this in more detail tomorrow.

Evaluating Your System’s Risk of Ruin

May 9, 2013 by Andrew Selby 2 Comments

What are the odds that your system blows up your entire account? Probably a lot higher than you think.

Risk of Ruin

Risk is one of the most important factors in successful trading. The problem with risk is that it comes in multiple levels. Many traders are able to evolve to the point where they are able to determine how much they are willing to risk on any single trade. The problem is that most traders stop there.

Too many traders never even consider the odds of blowing up their account. This was certainly an issue I dealt with early in my career. When you’re just getting started, you are naive enough to believe that success will be easy to find. I didn’t think I’d ever come close to losing my account. I didn’t even think I’d ever have a losing streak.

Is it only a matter of time before your account explodes?

As we evolve, we learn that all systems have drawdowns. All traders struggle at some point. The most important lesson in trading is to survive to trade another day. In order to do this, we have to calculate the odds of losing everything and adjust our trading to make sure we are able to survive. This is why we study risk of ruin.

The Three Components of Risk of Ruin

Win Ratio – The percentage of expected winning trades based on historical backtesting.
Payoff Ratio – The average profit of a winning trade divided by the average loss of a losing trade. Shaun also calls this the R multiple.
Percent of Capital Risked – The amount of capital your system puts at risk on any given trade.

How These Components Interact

Once you have these three components, it is easy to determine your risk of ruin using a table. The ideal situation is to have a risk of ruin as close as possible to 0. In the event that your Risk of Ruin is greater than zero, then you can adjust any of the three components in order to improve it.

The component that you have the most control over is Percent of Capital Risked. Simply adjusting down your position size will dramatically reduce your risk. The drawback to this is that you will also be adjusting down your profits.

The other alternative is to tweak your system in a way that improves your edge. This can be done by increasing the percentage of winning trades your system can expect or by increasing the Payoff Ratio. The Payoff Ratio can be increased either by having bigger profits on winning trades or less losses on losing trades. These components are more difficult to adjust because there are more factors outside of your control.

Using My 10/100 SMA System Example

I recently had a simple moving average system backtested. Using the historical backtest data for that system, we can calculate the risk of ruin. The system was set up to have rules that were extremely simple. It establishes a long position when the 10 day SMA crosses above the 100 day SMA and then goes to cash when the 10 day SMA crosses below the 100 day SMA.

Based on the results from historically backtesting this system on the SPY, I discovered that it would be profitable on about 41% of its trades. Its winning trades are a little better than 4 times as large as its losing trades, so I can put it in the 4-1 category.

According to the Risk of Ruin tables, if I was to use this system and risk 1% of my capital on each trade, my Risk of Ruin would be 0. Nothing is ever definite, but there is very little chance that I would go broke trading this system. The problem, however, is that this system also fails to beat out standard buy and hold strategies. There would also be an issue if the system’s Win Ratio fell below 30%. If that were to happen, Risk of Ruin would jump to 100%.

Scalper EA

March 5, 2013 by Shaun Overton 88 Comments

It looks like I accidentally stumbled into a scalping strategy. I posted the rules near the top of the page for those that don’t care about how I developed the expert advisor.

Warning: this EA cannot possibly profit on wide spreads. Do not follow this method if your broker’s spread and execution slippage average more than 2 pips.

Scalper EA Trading Rules

Chart: EURUSD 5 minute
SMA Period: 200
Moving Average Envelope: 1.0% of the SMA
Style: Counter-trend scalper

Entry rules

If the price crosses and closes below the lower envelope, then buy at market.

If the price crosses and closes above the upper envelope, then sell short at market.

Exit rules

If the price crosses and closes above the lower envelope, then exit long at market.

If the price crosses and closes below the upper envelope, then exit short at market.

Notice that the scalper strategy uses the same envelope for entry as it does for exit. The distribution of distance around the moving average is sticky when the price extends far away from the SMA.

The screenshot from NinjaTrader shows trades entering and exiting around the lower envelope

Get the code for MetaTrader 4 or NinjaTrader

Why the strategy works

The original intent for this research sought to uncover a range trading strategy based on the price crossing the moving average. Most traders think of distance in terms of pips.

Pips are valid for general context. The problem with modeling a strategy based on pips is that the the meaning of one pip’s movement changes over time.

Taking the idea to extreme lengths, the value of a pip in 1999 when the euro launched around 0.80 hardly compares with today’s price of 1.30. Keeping the discussion to percentages helps to fix the meaning of an idea like 100 pips over long periods of time.

I wanted to visualize how the price generally behaves relative to the moving average. A custom NinjaTrader indicator that my programming team wrote collects and analyzes the data in an Excel spreadsheet. Excel allows me to draw a graph of how frequently the price stretches away from the 200 period simple moving average.

Distribution of prices from SMA 200 on EURUSD M5

The frequency of percentage distances from the SMA 200 on the EURUSD.

The slope of the curve bends as the price extends further from the moving average. As you move left to right along the horizontal axis, the slope is steep until 0.75%. A kink in the curve forms at that point. The slope of the line flattens substantially from that point onward.

A big slope implies that the price will be anywhere but here on the next bar. A flat line means that the price isn’t likely to go anywhere. The distance is sticky at that level.

Scalping in a moving market doesn’t make sense. It’s only when we identify the sticky price condition of 1% away from the moving average that running a scalper EA makes sense.

Scalper EA Backtest results

I developed the strategy in NinjaTrader using data from 2011. My blind period was 2012, which was data that the strategy never saw in development. It was a pure walk forward test.

Results without spread costs

Profit: $5,740 on 108 trades trading 1 standard lot per signal
Profit factor: 2.18
Percent accuracy: 83.33%

NinjaTrader backtest, M5 EURUSD for 2011 without trading costs

Results with 2 pip spread costs

Profit: $1,420 on 108 trades trading 1 standard lot per signal
Profit factor: 1.24
Percent accuracy: 65.74%

A 2 pip spread substantially weighs on the performance

The scalper EA is incredibly sensitive to two assumptions: spread and slippage. I assume that anyone following this methodology trades at a reputable broker with good execution. The backtests assume that the combined cost of both spread and average slippage is 2 pips.

If your broker does not provide execution and spreads within that 2 pip window, then do not trade this strategy. I would not expect for you to walk away a winner.

Walk forward results without spread costs

Profit: $1,960 on 34 trades trading 1 standard lot per signal
Profit factor: 4.38
Percent accuracy: 88.24%

The NinjaTrader backtest shows walk forward results from 2012 on the EURUSD M5.

What is scalping?

Scalping refers to a short term trading style. Profits are very small and occur a large percentage of the time. When losses happen, they tend to be several times larger than a typical winner.

The high number of wins attracts traders of all stripes. The idea of consistently earning profits makes trading more fun and appealing. Traders with experience, which inevitably means traders that have suffered losses, also find the high percentage win rate appealing. It makes the emotional suffering far less difficult.

The emotional component that attracts traders to scalping strategies leads to illogical business decisions. Traders place the need to win frequently above the long term need to expect a profit.

Too many scalping expert advisors tap into the high winning percentage. Most fail to present a clear and obvious reason why it makes sense to scalp.

The EURUSD typically costs $2 to trade a mini lot. Many scalpers set narrow profit targets between 1-5 pips, which are worth $1-5.

The trader spends $2 to make $1-5. If this were a normal business, that would be the end of the game. You win. The game is only limited by the number of trades placed.

Trading, unlike other businesses, frequently results in losses. It’s very possible to build an expert advisor that could win trading for free but loses when costs enter the picture.

The best example is the difference in the scalper EA backtests for 2011. The first test showed a percent accuracy of 83% without including the spread. Adding the spread to the second backtest decreased the accuracy to 65%.

Trading costs make all the difference in scalping. Many scalping strategies live and die based on their trading costs.

The accuracy dropped because all trades had to subtract the spread cost from their simulated winnings. The number of trades that flipped from profit to loss because of a 2 pip spread shows how many trades profited by the narrowest of margins. The narrower the margin of profit, the more sensitive a strategy becomes to spread costs.

Do you think that I should have considered other ideas in the strategy? Suggest some ways to improve in the comments section below.

After-thoughts

This series eventually led to a profitable trading strategy. If you’d like to read through the journey, then I suggest reading the articles sequentially

The initial strategy idea
Selecting an appropriate time frame
A research plan
An annoying surprise in the initial backtests
An attempt at range trading
Range trading results
The moving average envelope scalper

Ranging SMA 200 Walk Forward Results

February 25, 2013 by Shaun Overton 4 Comments

We have come a long way. The original group trading strategy anticipated the need to resort to a complicated money management strategy. A simple range trading strategy emerged that made complicating steps unnecessary.

Testing from 2011 showed the best chance of earning a profit came from the following settings on the EURUSD:

30 minute chart
Price exceeds 1.5% or more of the 200 SMA
Buy/sell at market on an expectation of reverting to the SMA

The tests showed a hypothetical net profit of $1,310 trading 1 standard lot per signal. Assuming that you’re comfortable using 10:1 leverage, that makes for an annual return of 13.1%. Higher leverage increases the return at increased risk. Lower leverage decreases the return at decreased risk.

Walk forward results

The walk forward results are profitable! That is a huge relief, especially given the amount of effort put into the research.

The thing that disappointed me the most, however, is the huge drop in the number of trades. The original strategy placed 60 trades in a 12 month period. The walk forward test only traded 22 times.

The equity curve shows a substantial time lag in between trades. The blind test shows a profit.

The equity curve makes it obvious that nothing happens for large periods of time. Over six months pass between the first trade on January 19, 2012, and the next trade on June 29.

Thereafter, the pace really picks up. September 2012 showed the most activity with 12 total trades – more than half of the year’s activity happened in that individual month.

The gross outcome is a profit of $580. The cost of trading, assuming a 2 pip spread, is $440. The final return is $140, or a 1.4% return on 10:1 leverage.

Metrics

The results for the walk forward tests, as shown in NinjaTrader

Trading efficiencies of the 2012 walk forward results

The entry efficiency

The exit efficiency

Analysis

The substantial drop in the number of trades relates to something that we already knew. The trading strategy needs volatility in order to find trading opportunities.

Whenever volatility drops, which forex brokers bemoaned at the Forex Magnates conference in London, the number of trading opportunities drops, too.

The walk forward test indicates that this is a great strategy to keep in my pocket for when volatility picks up later this year. It held up on blind data. Although the higher profit facotr likely results from the small sample size, it feels reassuring when the metrics improve on the blind data.

I allowed myself to cheat slightly where I tested the 1% setting in 2012 instead of the 1.5%. The performance showed a similar drop in the number of trades and in the profit. Importantly, however, the numbers in that sample set improved.

I like this strategy because it is stable. Changing the setting from 1.3% to 1.5% does not cause a drastic U-turn in the performance. In other words, it’s steady and predictable.

I would feel far less confident in the outcomes if minor changes in the settings yielded enormously different profits or losses. Violent shifts in outcomes from small changes relate to chaos theory. That’s not a desirable trait in an algorithmic system.

Future improvements to the strategy might relate volatility to the entry setting. When volatility is low, the 1.5% threshold might lower to 1%. As volatility rises, a band of 2% might yield better outcomes in extreme scenarios. An obvious next step would be to relate the movement of the band to the volatility of the EURUSD itself.

After-thoughts

This series eventually led to a profitable trading strategy. If you’d like to read through the journey, then I suggest reading the articles sequentially

SMA 200 Range Strategy

February 5, 2013 by Shaun Overton Leave a Comment

After evaluating the SMA 200 reversion idea on the M1 charts, I realized that the strategy required a higher time frame to earn a profit. Trading costs were otherwise far too high.

The bottom axis shows the percentage distance of the price away from the SMA 200. The vertical axis contains the frequency that the price achieves that distance from the moving average.

The most interesting observation about these charts is how the shapes of the lines change as the time frames increase. Part of the effect is obscured because the horizontal scale is mushed on the M30 chart.

Blowing them up to full size reveals that the amount of curvature decreases as the time frame increases. The plot looks more and more like a straight line on higher time frames.

A distance frequency plot on the M5 chart.

Price percentage distance on the M15 chart

M30 price percentage distance from SMA 200

Price percentage distance on the M30 EURUSD chart

A viable strategy

The best result jumped out when I finally made it to the M30 charts. The pattern is obvious. Profitability potential jumps as the time frame increases.

The best candidates that I identified on the M5, M15 and M30 charts all coincide with the inflection points of the curves.

M5 Results

Results of crossing the 200 SMA on the M5 chart	Results of reaching 0.33% away from the 200 SMA on the M5 chart
Results of reaching 0.5% away from the 200 SMA on the M5 chart	Results of reaching 1% away from the 200 SMA on the M5 chart

M15 Results

Results of zero crossings over the 200 SMA on the M15 chart	Results of reaching 0.5% away from the 200 SMA on the M15 chart
Results of reaching 1% away from the 200 SMA on the M15 chart	Results of reaching 1.5% away from the 200 SMA on the M15 chart

M30 Results

Results of zero crossings on the 200 SMA on the M30 chart	Results of reaching 0.5% away from the 200 SMA on the M30 chart
Results of reaching 1% away from the 200 SMA on the M30 chart	Results of reaching 1.5% away from the 200 SMA on the M30 chart

Conclusion

I highlighted the final item in red. It appears to be the first candidate that can overcome trading costs. None of the backtests include the spread in the final results.

I assume that the best case final scenario after slippage and everything that goes wrong with trading is an average spread of 2 pips on the EURUSD. There were 60 trades in the candidate strategy and each trade was for 1 standard lot. The cost is therefore $20 in spread per lot per trade * 60 trades = $1,200 in costs.

I would love to evaluate higher time frames. The problem is that we’re already limited to 60 trades in our profitable trade set. Cutting the potential trades in half or more eliminates the ability to extrapolate anything from the data. M30 is the end of the road for the strategy.

The total gross profit for the strategy was $2,510. After costs, that puts the profit at $1,310.

The real question now is to extrapolate the capital required to earn $1,310. The backtest traded 1 full standard lot per signal. There’s no hard and fast rule here. I would set the absolute minimum floor for the strategy at $5,000. That’s probably too low, but it could at least handle some drawdown. That puts the leverage at 20:1 and the theoretical profit for 2011 at 26.2%.

I personally wouldn’t consider trading full lots with less than a $10,000 starting balance. That puts the potential return at 13.1% and seems much healthier.

Also, remember that strategies can scale up and down. The percent return and leverage could remain unchanged if the lots traded and starting balance decrease proportionately. A $1,000 account trading mini lots accomplishes a 13.1% return of $131.00.

The next step is to evaluate the strategy on a walk forward basis, which I intend to do in the next few weeks.

Related articles:
Group trading strategy – the first in the series that explains the trading rules and genesis of the idea.
Time Frame – where I figured out problems related to the M1 charts

After-thoughts

This series eventually led to a profitable trading strategy. If you’d like to read through the journey, then I suggest reading the articles sequentially

Annoying Backtest Surprise

January 22, 2013 by Shaun Overton 2 Comments

The first backtest results are in from segmenting my price-SMA cross forex strategy. Looking back, I really should have pre-researched this idea before publicly moving ahead. Look before you leap!

I rediscovered the same irritating problem that wasted 2 months of my life back in March 2012. I’m more irritated about finding the same results again than anything. If you were able to trade crosses over the SMA 200 without the spread, the strategy makes gobs and gobs of money.

The idea led to using SMA 200 price crosses to trade with limit orders for free using BBBO (Best Bid Best Offer). That fell flat after orders received an 80% fill rate at multiple brokers. It needed 95%+ fill rates in order to work.

The strategy looks deceptively appealing. It’s completely useless.

Real world trading costs including the spread and slippage works out to something like 2 pips on EURUSD. The strategy traded 29,132 times on the M1 chart to earn a profit of $118,970. The problem is that the 2 pip spread cost would have cost $582,640. The pretty equity curve costs $4 for every $1 of profit. Not good.

The price crossed the 200 period simple moving average more than 29,000 times in 2011.

The SMA distance curve analysis didn’t turn out as I expected. There were far fewer trading opportunities away from the moving average than what I predicted – the SMA is stickier than I previously thought. No matter where I looked, the strategy always spent more money on trading costs than it earned. I posted all of the images here for the sake of thoroughness.

I’ll be moving this up to the M5 chart in the next round of tests. Everyone can say a loud, “I told you so!”

Trades entering at ±0.1% away from the SMA 200.	Trades entering at ±0.2% away from the SMA 200.
Trades entering at ±0.3% away from the SMA 200.	Trades entering at ±0.4% away from the SMA 200.
Trades entering at ±0.5% away from the SMA 200.	Trades entering at ±0.6% away from the SMA 200.

The next step is to take a quick look at the M5 charts to see if I can make the basic idea viable there.

After-thoughts

This series eventually led to a profitable trading strategy. If you’d like to read through the journey, then I suggest reading the articles sequentially

Time Frame

January 10, 2013 by Shaun Overton Leave a Comment

The feedback on the group trading strategy was phenomenal! I’ve never had so many emails with such a broad range of opinions. Thank you to everyone for your response and thoughts.

Numerous people pointed out an obvious oversight on my part: which timeframe will the strategy use? Before you have a heart attack, please hear me out. Notice on the graph (republished below) that the curve really does not kick outward until 0.3% or more. On the current EUR\USD price, that’s a movement of 39 pips (0.3% * 1.3000). I used the M1 chart to create the graph. The strategy, too, will use the M1 chart for its signals.

The graphs displays the percent distance of the price from the SMA against frequency (i.e., how often is the price this far from the 200 SMA?)

Another reason to consider the M1 chart is the number of trades per year. The average number of bars that the price spends above or below a 200 SMA is… 200 bars. This is based on research that I performed earlier this year. There are approximately 386,000 trading minutes per year. Divide that by an average of 200 M1 bars between crossings and you get a maximum of 1,930 trades per year (386,00/200).

I’m only looking at taking a percentage of those trades. If 20% of the crosses result in desirable opportunities, that is 386 trades per year. While active, it’s hardly crazy. 386 trades per year averages to ~1.5 trades per day.

Leave comments below if you see anything worth adding on the time frame.

After-thoughts

This series eventually led to a profitable trading strategy. If you’d like to read through the journey, then I suggest reading the articles sequentially

The initial strategy idea
Time frame
A research plan
An annoying surprise in the initial backtests
An attempt at range trading
Range trading results
The moving average envelope scalper

Group Trading Strategy

January 4, 2013 by Shaun Overton 42 Comments

Francis D. from Australia likes to bounce different EA ideas off of me. He mentioned in the latest emails a fear of getting whipsawed from a signal that is either long or short.

This type of problem occurs all the time. I first encountered it with a simple strategy that fades price crosses over the moving average. Whenever the price crosses and closes below the moving average, go long. Shorts follow the same rules. It’s the kind of stupid simple range trading strategy that I always advocate.

Buy when price crosses below the 200 SMA. Sell and go short when the price crosses back above

The example above highlights the same thing that Francis complains about. The price floats around the moving average without going anywhere.

Random Outcomes

Trades based on the price crossing above the SMA come out near breakeven. Winners occur approximately 66% of the time and are one third the size of losers. Such a strategy neither makes nor loses money when ignoring trading costs.

The vast majority of the winners in the strategy are teeny tiny. The strategy encounters its maximum opportunity whenever it is closest to the SMA. The further away it goes, the more likely it keeps going the wrong way.

Flipping the trades still yields a random outcome. The only difference is that winners drop to 33% accuracy, but the average winner is now 2:1.

The Birth of a Trading Strategy

I always wonder how a scaling strategy might affect the outcome. If the strategy is most likely to win when the opportunity is smallest, what happens when the strategy attempts to reduce its position size as the market moves adversely?

Alternatively, what happens if it takes a pass on all the small winners and scales into positions? Yes, it will scale into losers, but it should also make resulting winners bigger. The question then becomes how to decide the rate of scaling and when to bail out of losers.

Thank you, Francis! A new blog series is born. Now that I’ve decided to scale into trades, I need to choose how and when to do it.

Nothing insightful or special jumped to mind. I’m a visual person, so I spent the better part of this afternoon creating the little chart in Excel using NinjaTrader. What starts out as a simple project always grows on itself. It took nearly 4 hours to get the information and formatting correct.

The graphs displays the percent distance of the price from the SMA against frequency (i.e., how often is the price this far from the 200 SMA?)

I care about scaling into trades as the price moves further from the 200 SMA. My instinct from looking at the graph above says I should focus on the inflection point. The curve forms a nice bend around 0.3% away from the SMA. Maybe I can start buying at the inflection point until I get to 0.6% or so.

What do you think I should do?

After-thoughts

This series eventually led to a profitable trading strategy. If you’d like to read through the journey, then I suggest reading the articles sequentially

Selecting an appropriate time frame
A research plan
An annoying surprise in the initial backtests
An attempt at range trading
Range trading results
The moving average envelope scalper

Buy and Hold Gold

December 11, 2012 by Shaun Overton Leave a Comment

I came across a paper entitled A Quantitative Approach to Tactical Asset Allocation by Mebane Faber. This paper appears to have been extremely popular over the years. Reading through the contents and reviewing the charts, I loudly wondered if the strategy might apply to gold.

Rules for buying gold

The idea is very simple. Taken directly from the paper, the rules are:

BUY RULE
Buy when monthly price > 10-month SMA.

SELL RULE
Sell and move to cash when monthly price < 10-month SMA.

1. All entry and exit prices are on the day of the signal at the close. The model is only
updated once a month on the last day of the month. Price fluctuations during the rest of
the month are ignored.
2. All data series are total return series including dividends, updated monthly.
3. Cash returns are estimated with 90-day Treasury bills, and margin rates (for leveraged
models to be discussed later) are estimated with the broker call rate.
4. Taxes, commissions, and slippage are excluded.

Test results

The tests were done using Kinetick’s free end of day data. The data extends from August 4, 1997 until December 10, 2012.

One thing which drives me crazy about NinjaTrader is that the percent return calculations are so opaque. The actual numbers used a clearly wrong. The buy and hold return for gold is easy to calculate. The 1997 price was about $350 an ounce. Today, gold trades near $1,700. The buy and hold return should be in the neighborhood of 385%. NinjaTrader claims that the return is somewhere near 150%. I’ve gone through the charts to confirm that the trades are correct. You can download the gold strategy for NinjaTrader and verify the trades for yourself.

The percent return metric is consistent within the platform, so luckily the exact numbers are unimportant. The only thing that really matters is whether or not the strategy returns a bigger number than the buy and hold approach.

A chart of gold’s daily equity curve, extending back to June 1997.
The tactical gold strategy dramatically underperforms the buy and hold return.	Pursuing the opposite idea of buying on crosses underneath the SMA10 doesn’t help

The strategy severely underperforms the buy and hold return. The idea held in the paper itself is deeply flawed. Buy and hold is a silly concept. Most of the “returns” in the portfolio stem from long term inflation and the devaluing of the dollar. The prices of the securities rose because the dollar, which you can think of as the counter currency, declined enormously in value over this time period.

Additionally, the idea of the price crossing the moving average implies that the trend escapes the moving average. The MA, in effect, gets dragged along with the price. While this is true with a handful of monster trends, the way that price moves usually work is something along the lines of 10 steps forward, 9 steps backward. The results of the test indicate this.

If the proposed strategy returns less than buy and hold and the total buy and hold strategy returned 150%, then it stands to reason that the total buying return is trades taken by the strategy + trades not taken by the strategy. The test includes what are supposed to be trades not taken by the strategy – the reverse signals. Adding up the proposed strategy’s returns with the returns of the opposite signal only accounts for roughly one third (22% + 34%) of the 150% earned. Where did the rest of the money go?

Most of the move happens around the short term moving average. The majority of the move has already happened after the price to crosses and closes above the moving average. The instinct of many novice strategy developers is to move to intrabar signals. Intrabar trading ignores the basic problem; you cannot know whether or not the bar will close above the moving average.