System Trading Video by Nick Radge

Nick Radge explains systems trading and how to establish a robust strategy that will last over the long term. Nick uses a short term high frequency trading strategy and a longer term trend following strategy to manage his investments. He discusses both in this video.

Systems trading is predominantly all I’ve done since I started trading back in 1985. Many of you probably wouldn’t know, but that’s how I was actually introduced to trading. I looked over the shoulder of a guy doing a five and 10-day moving average crossover system on the spy and by that afternoon I’d gone down to the local Penfold’s shop and bought myself some graph paper and started doing it and it’s how I started trading. So obviously we’ve come a little bit of distance since then. But today a 100% of my trading is systematic. Personally, I currently run two models. One is a longer term trend following strategy here in Australia and the other is a higher frequency mean reversion strategy I trade in US equities, which I’ll show you a little bit of later on.

What is a trading system? I guess this is one of the fundamental issues I think with building a trading system, it’s actually a significantly broader foundation than what most people will think about. And this is very, very important and we’ll go through the attributes very shortly. But in my definition, a trading system is a broad yet well-defined set of guidelines and instructions, rules and procedures if you like that quantify the complete trading regime. Now I say the complete trading regime because it has to encompass both the qualitative and quantitative traits of trading itself, the qualitative traits are the emotional side of it all. I can give a person a fabulous trading system, but if they emotionally or psychologically can execute that strategy, then regardless of how good it is, you’re not going to be able to use it.

And that’s something I see a lot of, I have a lot of clients coming in and using the strategies that I’ve built and some of them walk away after three months, after four months and say, “I just don’t get it. It makes no sense to me.” And that’s simply a function of the strategy not suiting the personality of the individual. So a trading strategy has to encompass both of those and it will have moving parts and it will provide continual feedback. If we have a very specific look at what goes into a trading system, and remember we’re talking a complete trading system here. There are four broad categories, inputs, the model itself, the risk overlay, and then the outputs. And as you can see by the colors there you’ve got, some parts of this equation are dynamic and some parts are static.

Now the dynamic parts of a trading model are always moving. They’re either moving on a daily or a weekly or a monthly basis. So for example, markets can go through cycles of being very bullish or very bearish or very volatile and that can come right down to a daily, hourly and even a granular level of intro minute if you like. Same with the outputs. There’s going to be various outputs and feedback that your strategy gives you and that’s going to be dynamic. Sometimes you’re going to be feeling really good about it. Other times you’re going to be feeling a little bit negative about it. And we’ve got to take all this into account. So let’s take a look at some of these inputs to start with.

Price, it doesn’t matter if you’re a fundamental investor or not. Your P&L is going to be provided by a price movement. Even a value investor or a staunch fundamental investor relies on a price movement of some description to make a profit. That’s the bottom line so we can get rid of all the other junk if you like, and at the end of the day, price has to move from A to B to make a profit or a loss for that matter. So price always moving, moves differently, can move in a very volatile way, can move in slow grind if you like, can move in ranges, moves in trends, other data for example, that you can build into a trading system. We’ve mentioned some fundamentals. You can put that kind of data into a trading system. For example, a lot of people may build some kind of sentiment data into a system and obviously noise, which we’re all exposed to. The less noise we’re exposed to, the better. But at the end of the day we are always exposed to some kind of noise and that will be an input into your strategy.

You’ve probably been in a situation, some of you may have a trading system that you follow, but you might read on a forum or you might read on your Twitter feed or your Facebook feed that this is going on and that’s going on or you might listen to Alan Kohler on the news and he’s talking jibberish. But all of that builds up in your mind and can, ideally shouldn’t, but can influence the way your model reacts. So if we move along to the model itself and what actually goes into that. First and foremost, above absolutely everything else and above expectancy is your personality and your beliefs. This is absolutely essential.

Your beliefs will dictate every part of the trading process. If you believe that buying breakouts is a losing proposition, well that’s going to be very, very difficult for you to put a strategy into place that does that. Conversely, if you feel that you need to buy a deep or you need to buy weakness, then you’ve got to build that into your model. You need to find a strategy that suits your own personality. And that’s why so many people struggle with buying strategies off the shelf or using other people’s strategies. Objectives, not only the P&L that your wanting out of the whole equation, and let me say that if someone says to me, “Yeah, I’m looking to make a lot of money,” we’ll, I’ll say, “That’s a very one dimensional view of the world.” Because it’s not only just about making money, obviously that’s what we want to do, but it’s the ability to actually do that.

So for example, you’ve got three kids and you’re holding down a full time job. Well, being a spy scalper is not really going to do much good for you because really you can’t do that. You’ve got to be cognizant of what is going on in your life and your lifestyle, how much time you have to put towards it, how much skill you have to put towards it. So for example, I have people… I just had a gentleman the other day buy one of my turnkey trading strategies which we operate in AmiBroker. Now this guy has bought the strategy but he actually hasn’t bought AmiBroker yet, so he hasn’t thought ahead and realize that I’ve got to upskill and learn how to use AmiBroker before I can actually buy this. Put this trading system into place. So if you need to acquire skills in order to meet your objectives, you probably want to start doing that first.

You’ve got to look at what kind of instruments you actually want to trade. Do you want to trade to FX? Do you want to trade CFDs? Stocks? What do you want to do? I’m not a fan of someone coming to me and going, “I need to trade CFDs,” or, “I need to trade FX.” I don’t believe that’s a very valid demand or want. I think first and foremost, you’ve got to find a strategy that suits you and is going to make you money. What instrument do you you do that doesn’t really matter. At the end of the day, you can make money trading any kind of instrument, but you do eventually have to find an instrument that suits both your objectives and your beliefs and your personality and then go with that.

I spent the first 17 years of my life trading futures and foreign exchange, but in 2000, 2001 I became more and more interested in stocks. Maybe I was getting a little bit older, I don’t know, but I’ve predominantly traded stocks ever since and I’m happy to do that now, but everyone’s a little bit different. Then you’ve got your standard things going into your model, your set up, your entries, your exits and your execution rules. Very important. For example, my short term mean reversion system that we operate in the US market, it requires an API, so I had to have an API built out of Europe, so these are all the kinds of things you have to take into account. You might use a very basic trend-following model that doesn’t take too much time and then you’re not going to need an API or any fancy software to implement that kind of stuff, but you’ve got to build all that in depending on the kind of model that you’re wanting.

Risk is actually not a part of the model in my view. It’s an overlay that runs over the top. So position sizing is a core component of that, obviously. Clearly if you have a model with a negative expectancy, doesn’t matter what kind of position sizing you have in place, it’s not going to help you. You need to have a model that’s got a positive expectancy to start with. And from that you can amplify the results using different kinds of position sizing. I don’t think there’s a need to be completely super duper advanced in position sizing. Simple does tend to work reasonably well. We use some very, very basic position sizing models, equal dollar models, for example, equal percentage models and that’s part of our business model for our clients to simplify things. But we’ve also put them to the test and in most circumstances, simple comes out just as good as more advanced kind of things.

Portfolio construction is very, very important. You start moving into areas like sample bias and those kinds of things, depending on your portfolio. Some of our competitors have trading systems that they offer and there’s so many signals generated that it’s impossible to get a median return. I had one lady recently, she uses a competitor model and they get up to 90 signals generated per day in the US market. Now that’s just quite a ridiculous amount of money and that leads to biases or errors if you like, specifically sampled bias, which means if you’ve got a system that’s supposed to generate systematic signals, but you have to have some kind of discretionary overlay to pick those signals which to take, then you’re not helping yourself.

So the size of the portfolio that you’re trading is very, very important. You can use price limiters, you can use filters such a volume and turnover to reduce our large universe if you need be, but it’s very, very important to reduce those kind of external biases as much as possible. Obviously the outputs are dynamic. P&L, if your model is in sync with the market, then you’ll be making profits. If your model is out of sync with the market, then you’ll be in drawdown. Because a market and price action is dynamic and your model is static, then you will always at some stage be out of sync with the market. It is absolutely impossible to build a robust strategy, which we’ll talk about shortly, that looks to profit in every kind of market environment. That doesn’t exist, at least not in any robust format that I’ve ever come across and many professional traders who manage hundreds of millions of dollars.

You’ll notice what they do is they don’t look to find a model that trades in every market environment perfectly. They look for models that trade reasonably well all of the time and then they combine models and some of these traders have 10, 15, 20 models running at once. And the combination of all of those is what allows them to stay involved in the market without getting caught in the ups and downs of a single model being in and out of sync in the market. Model feedback is also very important. I listened to a great interview recently with Jerry Parker, who’s one of the original turtles. He manages several hundred million dollars and has got a good track record back to 1991 and it was interesting because a lot of people believe the turtle strategy’s dead and buried, but he still uses the exact same rules.

But what he’s done is listen to the feedback he’s model has given him over the years and he’s learned from that feedback and once he uses the same broad approach to buying and selling, he’s adjusted the parameters according to the model’s feedback. So to give you an idea, back in the day the turtles used a 20-day breakout. So what Jerry’s saying now is that he uses a much longer term break out. He didn’t exactly say what, but one would assume it could be 50 or a 100-day breakout, something like that. And as a result it keeps him in the market for longer and removes a lot of the noise that has taken over the market.

So a model itself is dynamic, albeit less dynamic than what price action is. So over time you’ll learn a little bit about the market more so than what you did before. You’ll learn a little bit more about your model and you’ll make small changes over time. A very good example, our core trend following strategy that we operate here in Australia, we found that we were in a drawdown even though the broader market was going up and we were trying to figure out what was going on. And it turned out that we were getting sucked into a lot of resource stocks that were very volatile and broadly trending down and we kept getting sucked into them and spat out, sucked in and spat out, which is death by a thousand cats if you like for a trend following model.

And with the advances of technology in the last few years, we were able to isolate resource stocks on their own and isolate industrial stocks on their own and have a look at them individually. And what we learned was that following trains in resource stocks, whilst profitable is probably not worth the effort. You kind of get an 8% annualized return with a max drawdown of about 26, 27% and in my view that’s not particularly good numbers to run by, but if you remove resource stocks completely, the numbers improved quite dramatically.

So we’re not really curve fitting or data mining because the model operates profitably in both parts of the portfolio. But even if we remove the big run up that we saw between 2004 and 2007 in the resource stocks, the model’s just not even worth trading outside of that. So you’ll get feedback, you’ll learn new things, not only about the model in the market but about yourself and then you’ll make these small granular changes to the model over time and there’s nothing wrong with that. And same with emotional feedback. You’ll learn from yourself, you’ll learn what kind of things affect you more, what emotional aspects of trading keeps you awake at night and what doesn’t. For example, our high frequency strategy that we trade in the US doesn’t use stops. Now for some people that’s just an absolute no-no.

With our longer term trend following strategy, there aren’t stops placed in the market, but there are stops based on closing prices. So we’re waiting for a closed price below a certain point and then we’ll exit the position on the next day. You will find there’s quite a bit of evidence out there that putting stops in the market is actually detrimental to performance. Robustness means a strategy that works reasonably well most of the time… Robustness is not a strategy that works exceptionally well here and there and that’s a trap for young traders and we’re going to talk a little bit about these because it is very, very important and I’m going to show you some evidence here of a very robust trader and strategy.

But I think a lot of amateur traders incorrectly believe that an equity curve of a professional trader is a 45 degree straight line and that’s not the case at all. If you are looking at a trading system that offers a 45 degree straight line equity curve, you’ve got to run for the hills because there’s going to be a lot of problems with that. The bumpiness in an equity curve is actually a sign of robustness and we’ll have a look at some examples here.

So your goal as a trader is to find something that works reasonably well most of the time rather than something that works exceptionally well just part of the time. Let me talk about this guy David Druz. He runs a firm called Tactical Investment Management in the US, he manages several hundred million dollars. He is a commodity trend following trader, and since 1981 he’s got a compound at annual return of 17% and you can see over the last 15 years, 17.4, over the last 10 years 15.6. Had a bit of a knock recently. So in his last five years, the annualized return has been 10.5 and last year he made 23.9. Now had you invested a hundred grand with him back in 1981, that hundred grand will be worth 18 million.

Now, what is a startling fact about these numbers is that these are net of a 25% performance fee. In other words, had you invested $100,000 into this strategy in 1981 and he hadn’t taken out his 25% performance fee on a compounded basis, your $100,000 would be today worth in excess of 30 million, maybe even closer to $35 million. So that’s truly astounding numbers and certainly better than Warren Buffett for that period of time.

Now, I think most of you would say that if this guy has been using the same system since 1981 which he has been and he’s trading the same markets which he has been, then you would make the assumption, this is a pretty robust strategy. I wouldn’t know too many people who wouldn’t want that kind of return profile, but here’s what I’m talking about, a strategy that works reasonably well most of the time. This is his monthly breakdown since 1981 and the yellow shaded months are losing months and as you can see, there are probably 50% losing months since 1981 and as you can see in the far right hand column there, the shaded annual is the annual return and you can see that he’s had a couple losing years, 1996 he lost 31%. In 1999 he lost 29%, 1999 he lost 25.74%, 2011 he lost 27.9%, so over the longer term you can see he is performance is exceptional, but in short windows of time, and a short window of time is actually measured in years he will have losing periods of time.

Now I don’t know too many mum and dad traders who will start a trading system and after a year when they’ve lost 27% will stay with it. I recently had an example. I recently received an email from a client of mine who started our trend following strategy in March. I’m not sure what drawdown he is in, I think only five or six percent or something like that. But our strategy switched off when the market started trending down.

Now he sent me an email the other day and let’s remember that he’s only been involved for six months and he basically said to me that it’s not good enough. There should be opportunities. We should be making money. What’s going on here? And that’s an example of somebody who is wanting something that makes money all of the time and is failing to look at the bigger picture. And that’s what so many amateur traders do. They just fail to look at the bigger picture. It’s like going to university, you’re wanting to become a doctor but you want to go to university for six months only. You don’t want to go for four years or six years and you certainly don’t want to be an intern for another four or five years after that. You want to go straight in and do brain surgery. It’s not going to happen.

And the same is true for trading. So a robust system is something that works reasonably well over the longer term but doesn’t necessarily work perfectly. Bumpiness in your equity curve is a solid sign of robustness. And you can see here with David Druz that he’s had his ups and downs, but over the longer term he’s certainly been well up and he certainly made a lot of money. When we’re assessing trading systems and this means whether or not you’re building your own or whether you’re looking to buy one or whatever it may be. Robustness tests in a very generic sense. First of all, there has to be a logical explanation as to why money is being made. A robustness test also has minimal degrees of freedom. Freedom is basically inputs or parameters into the system. The more complex a system is, that is the more inputs in parameters a system has then the higher the chance the system is going to break or the higher the chance is that you’re not going to understand why the system is making money.

And another robustness test that you should run is the strategy should operate reasonably well across a broad variety of different symbols. So let’s go on and have a look at these in more depth. First of all, a logical explanation. The top pain is a stock in the US called WBA. And if you have a look at the dates on that stock, you can see that’s a chart back to 1956 through to 1961 and then in the lower pane is the exact same stock and it’s a chart between 2011 and 2015. What’s the first thing that you see on those charts? It’s a trend, right?

So the market operates in two states and two states only. Markets operate in a state of momentum or trending or a state of flux or mean reverting if you like. And here we have a chart that or a stock that is showing a predominant trend back in the 1950s and a predominant trend as of today. And trend and momentum are a logical explanation as is mean reversion. Okay, let’s talk about those a little bit more. A trained or momentum occurs from the hurting principles, all right. People take prices a lot further than what fundamental value dictates. And we’ve seen that time and time again in every single market. You can see it right here in these two charts and you can see that we’re seeing trends in a stock 40, 50 years ago and we’re seeing the same kinds of trends today.

What I’m saying, this is a robust occurrence in the market, in other words, there are always trends somewhere, so an individual stock may not trend all of the time, but somewhere in a universe of stocks you will find some trends which you can exploit and that’s why David Druz’s strategy is so robust because he’s exploring trends, both up trends and down trends and all he’s got to do is stay around long enough and he knows a trend will come along.

Now trends have to occur. They can’t not occur. To say a trend doesn’t occur would suggest that a stock or an instrument or a symbol is fairly valued and they can be certain periods of time where that may happen. Especially you look at interest rates in the US at the moment that have been at 0% for so long, but one has to think eventually and God help us all if it doesn’t occur, but one has to think eventually that interest rates in the US will go up and we will start to see those interest rate products start a trend. So there can be extended periods of time of non trendiness but over the longer terms trends will show themselves. They have to.

If you think about the other state of market is mean reversion, who’s the world’s biggest mean reverting trader that you can think of? What about Warren Buffett? Is he not a mean reversion trader? Warren Buffett is a value investor. He is looking for stocks to fall to such undervalued levels that are being driven by the hood, i.e. the trend that he has to buy them because he knows eventually they’ll revert back to their true value. So essentially Warren buffet is actually a mean reversion trader, obviously very, very long term in nature, but essentially that’s what he’s doing. Obviously we do that ourselves in a much smaller timeframe. But essentially that’s what he’s doing. He is profiting from the concept of mean reversion. In his case it’s fundamental main reversion, but it’s still mean reversion all the same.

So one of the things that is very, very important to me is that you have to be able to explain in logical terms how and why your system is making money. So a trend following strategy uses a money management age. In other words, it will limit losses using a money management stop. And by limiting losses and allowing the profits to run, it creates a positive expectancy. Whereas a mean reverting strategy uses an entry edge. So there’s two kinds of systems there. You’ve got a money management system, which is generally found in trend following strategies or swing trading strategies. And then you’ve got a entry edge, which is where you’re relying on the market itself to give you a positive expected return. So for example, our mean reversion strategies buy stocks that have been sold off significantly in the scheme of their current price action.

And the probabilities suggest that they will bounce back at some stage and that’s when we’ll take a profit. So two traits of the market, only two traits, trend-following and mean reverting and your two trades when you combine that into a trading system is a money management edge, cut your losses, let your profits run, or a entry edge, which is where you’re finding a pattern that has some kind of high probability of generating a profit.

Let’s talk about minimal degrees of freedom. And this is a strategy from my book and Unholy Grails called the 20% flipper. Now the rules are very, very simple. If a stock moves 20% from a low point, you buy. And if a stock falls 20% from a high point, you exit the position. We filter stocks using a hundred-day moving average. Just very simple, but you can see there it’s a very simple kind of strategy and that’s the equity curve on the All Ordinaries back to 1995. Now that’s about a 19% annualized return, there’s nothing wrong with that. And you can see that’s a robust strategy because first and foremost, not too much can go wrong with it. A stock that’s going to go up 400% has to go up 20% first. So we’re simply jumping in when it does that. A stock that goes bankrupt has to fall 20% from a high point before it goes bankrupt. So we’re going to ensure we’re not going to be running stocks into the ground, and that’s the kind of equity curve you can expect.

Now remember I said earlier on a bumpy equity curve is a good sign of robustness. So here we know the strategy is robust. It’s only really got two rules and it’s made some very nice profits. So, that’s what we’re talking about by having a minimal degrees of freedom. I guess the next question is, well, how many degrees of freedom could you have? And I guess that’s how long is a piece of string? But really once you start getting above 10 degrees of freedom or 10 different inputs, you’re going to start running into curve fitting or data mining. And that’s when you’re starting to lose your robustness.

Now this next slide, and this is a real time example. A lady sent me these rules that she wanted programmed into a full working trading system. Interestingly enough, she’d paid $12,000 for a trading course and these are the rules that she came away with. Now if we think about what’s going on here, what is inherently happening here? Somebody who wants all these to align. You want a share price above a 30-week moving average. You want a golden cross with two moving averages on two different time durations. You want a rising momentum indicator from a historical lows. You want an up sloping trend line. You want this, you want that. You want a bit of everything. What this lady is wanting is perfection. That’s what she’s wanting and it doesn’t exist. She’s wanting something that, A, doesn’t exist and, B, something that doesn’t actually need to exist to make a lot of money.

And something like this, she’s not going to know why she’s making a buyer sole decision. She’s not going to go, “No. Why?” She’s actually making money. I can’t really tell if this is a trend following strategy or a main reversion strategy. I mean it looks like a trend following strategy, but my point is there’s a lot that can go wrong with this strategy. There’s a lot of filtering, there’s a lot of ways to remove bad trade and that’s basically her aim here to find perfection, to find the perfect trade to exclude bad trades. But unfortunately it’s not going to work like that in reality, and obviously there’s certain things in here that we can’t program because they’re discretionary calls, but for the vast majority of what it is here is just way too much and it’s simply not robust.

So when you come across a trading strategy that has lots of different inputs and lots of different roles, the chances are your data mining and your trying to strive for perfection rather than robustness. And that’s two very different core traits right there. So compare that to, “Hey, if it goes up 20% buy it. If it goes down 20% sell it.” That’s all there is to it. Some of you may have heard of Bill Dunn. Dunn Capital Management, go and have a look at his returns over the last 30, 40 years. He started back in 1974 and he uses a system very similar to this. I don’t know the exact inputs of that system, but he uses a percentage change from a high and a low to enter and exit.

Now when you put his performance against that of the S&P 500, the S&P 500 is like a flat line. He’s returns are just astronomical. Don’t get me wrong, he swings for the fence. He is looking to make massive returns and he does do that. I’m not saying it’s comfortable, but he’s got a very significant return profile over the longer term, the course that this particular lady did, we actually see a number of their clients come to us. Now so much so that just recently we had four of their clients come to us and the question is, you’ve just paid $14,000 for this course, why are you joining our service?

And the reason being is because they get fed so much information that they don’t actually know how to put it all together and use it. And you can see why, what’s going on right here. I mean, this lady would know if she’s coming or going or what the heck’s going on. There’s no structure to what they’re actually being taught. They’re being taught a lot, but there’s no actual structure to put it all together. Okay. Let’s move along to our last bit of robustness testing that we were talking about and we want a strategy that works reasonably well across a wide variety of symbols. Now, this is a strategy that came across my desk a few years ago. It only tried one instrument and that’s the gold ETF in the US. Now, if anyone is telling me they’ve built a trading strategy on one single instrument, I’m very, very, very skeptical that it’s probably being data mined or it’s just a random pattern that they’ve picked up. In other words, it’s not robust. I’m very skeptical of any service provider that will sell you a spy trading system or a British pound, US dollar trading system or an Aussie dollar trading system that doesn’t operate on other symbols and I’m going to show you why.

So this system came across my desk a few years ago now and the rules of this system are if the gold ETF, which is GDX closes 2% from the opening price, then buy that ETF on the close, hold it overnight and the next day exit that long position and immediately sell short and then close that short position at the close of that day. So pretty simple kind of rules. It fits the rules, simple rules criteria, which is good. And that’s the equity curve between 2006 and 2012. Now 2006 was when that ETF listed so we have no history prior to that. But you can see it’s a pretty smooth kind of equity curve making its way high there. Got a 13 or 14% annualized return. And I think this came across my desk in about 2012 which is why I stopped the back test right there.

Key number one, if this strategy rule that they’ve put there, that if a price closes 2% above the open, buy on close, sell short on the next open. Close at the end of the day. If that’s a valid rule and is robust, it should actually work across a more significant portfolio of individual instruments. So what I went and did and what I always do is I’ll go and test it on the S&P 500 so this is the exact same rules for the exact same period of time traded on the S&P 500 constituent lists. So this includes all stocks that have been delisted. So there’s no survivorship buyers. And as you can see, those rules don’t particularly work out that well.

In other words, the guy that’s found this little patent on this little ETF, he’s only found a random pattern. It is not robust in any way, shape or form. It might work for that ETF, for that period of time, but the chances of it continuing to work in the future are very, very small and I certainly wouldn’t be putting my money into that kind of a system. As you can see, it is not a robust pattern. it does not work very well. If it did work well, it would operate reasonably well across that whole S&P 500 constituent universe. And as you can see there, it doesn’t do that.

So let’s take that GDX system and actually now test it further forward on a out of sample data. Out of sample data is data that we didn’t have available at that particular period of time. And as you can see and as what was expected, the system degraded or the performance went down and we’ve had about a 10% drawdown, which is the biggest drawdown that strategy has had since release in 2006 and I would say potentially that drawdown will continue or that pattern has now been arbed out of the market and is not valid. So just be very, very wary of any system vendors selling you a system that is only operational on one particular market.

I know a couple of vendors here in Australia that do that and both of them have one, completely gone out of business, not surprising and the other one has completely changed all his models around and is now doing something completely different to what he was doing six years ago and the chances are in five years time that whole stuff is all going to fall over again and this comes from the want of having something that looks spectacular all of the time.

That model there is probably going to sell very, very well. If you find someone who’s wanting the Holy Grail, they’ll probably look and say, “Yep, I’ll buy that. That looks good.” But obviously the truth of a proper robust model doesn’t sell as well even though it will do them better over the longer term. Here’s some warning signs of what non robust strategies, what you’ve got to be careful of. We’ve talked about it before, single market system, if there’s only a small sample of trades. One of the biggest losses I’ve ever seen was a client. He came to me, this is going back about 15 years. He learned to code TradeStation, which at the time was what I was using and he sent me this trading system that had about 400 trades in it, tried to the S&P 500. 400 trades of about a four or five year period of time.

Seemed reasonably robust on the surface. He didn’t tell me any more about it than that. Anyway, in one night, in one night, he lost 60,000 US dollars out of his $50,000 trading account in one night. And I asked him what happened? I said, “What happened?” And it turned out that the model he had built was actually a compilation of about 15 different patterns, each of which had only occurred about 10 times in the past or 15 times in the past. So he had found these random patterns like we just saw here with this GDX system, but he’d found them in one market, specifically the S&P 500 and in fact, I’m sure some of you have read the Larry Williams book Longterm Secrets To Short Term Trading. And that is the absolute Bible of data mining and how not to build trading systems because there’s a very small sample of trades, there’s a heavy data mining in there, and the chances are that those patterns will not work into the future and in fact, most of them now don’t work whatsoever.

So you’ve got to have a big sample of trades. The more the better. So the more instruments you can have and the more trades you can have across all of those instruments, the better the chances are that you are actually onto something that works reasonably well and will continue to work into the future. As I said before, a sign of a robust strategy is actually a bumpy equity curve. I know everyone wants a nice smooth equity curve, but that’s not particularly realistic. Rather than try and build and optimize the system for a smooth equity curve, you’re better off having a very robust strategy with a bumpy equity curve and then add more strategies to it.

So if you can have three, four, five very robust strategies, each of which has bumpy equity curves, the chances are when you combine them all together, the equity curve will actually be a lot smoother. And that’s what diversification is all about. And if you go to someone like Salem Abraham or Toby Krabill, some of these guys have got 30-year track records, you’ll notice that they actually tried quite a number of different trading systems and some of them trading systems over different time frames. So what they’re doing is building robust trading systems that each have bumpy equity curves, but they’re overlying different strategies on top of each other and you get a much smoother equity growth and that’s what you kind of got to do.

And lastly, if someone doesn’t know why their strategy makes money, then you probably got to run for the hills. Always ask a vendor, why does your strategy make money? If they can’t answer you or they give you some mumbo jumbo, then the chances are they don’t actually know why it’s making money and they’ve probably just found some kind of anomaly in the market, a random occurrence that they’ve stumbled upon and the chances are that it’s probably not going to continue to work into the future.

What does a robust strategy look like? This is our high frequency strategy. This is a short term end of day mean reversion strategy that I personally trade in the US market. I don’t trade it in the Australian market. Now on the left hand side there, I’ve done a back test between 2006 and 2012 of the S&P 500 constituent list. So that’s all the S&P 500 stocks that exist today in every S&P 500 stock that existed in the past or was a part of the S&P 500 universe at some stage in the past. So we’ve completely removed survivorship bias. There is one entry rule. There is one exit rule and there’s 2,200 trades across those… Top of my head I think there’s 1200 stocks in that universe.

2,200 trades across 1200 stocks all being traded the exact same way. It’s a long only strategy, which means it’s traded through that GFC, which you can see there is the flat line. Then we’ve taken the exact same strategy. We have not made any adjustments to it in any way, shape or form, is the exact same rules, but what I’ve done in this case is I’ve applied it to the All Ordinaries and I’ve changed the date range as well. I’ve changed the dates back to 1995 to 2001. So again, it’s another six-year period of time. The same as the S&P 500. Same rules, it’s done 2,870 trades, so more or less the same kind of trades and you can see more or less, you’re getting the same equity growth. Your $100,000 is turned into $290,000. On the left hand side, our 100,000 in the S&P 500 has turned into 300,000 so more or less the same kind of equity growth, the same number of trades, even though we’re trading a completely different market and in a completely different period of time.

So that’s what you want to aim for. A simple set of rules that works reasonably well over a lot of different stocks over a broader period of time. It’s not perfectly linear. You can see there’s some bumps in those curves. You can see back on the S&P 500 there’s a period of time when there’s no trading. Believe it or not, a lot of people find it incredibly difficult not to trade, but this is a long only strategy and there are periods of times when you just shouldn’t trade. The last two months, I haven’t made a trade. And that’s because the market’s been going down and I’m happy to be sitting on the sidelines. And that is what helps our bottom line over the longer term, being able to sit on the sidelines. So, that is our US high frequency strategy, not appropriate for everybody, but it’s a pretty robust kind of robust strategy.

If you’re looking for some robust strategies, we offer high-end mentor course. Now this is a proper mentor course with one-on-one training. There’s six months of tutorials and you will learn high levels of AmiBroker programming, and then high levels of trading system design, build and implementation. Under our guidance you will then go away and build your own strategy from scratch. You will design your own strategy, you will build it, you will test it, and then you will have us looking over your shoulder the whole time to make sure you’re doing it properly and pointing you in the right direction.

Learn more about the Trading System Mentor Course here.