In this episode of 2Traders Podcast, Walter and Darren zoom in on what it means to trade multiple trading systems. Walter points out that it may be possible to determine the value of a good system rather quickly (this may surprise you).
According to Darren, accepting the “inevitable” is simply part of trading and you must determine how to deal with this (because it will happen one day for your system.)
Also, you will find out what your trading systems “superheroes” might be… Listen to this and more in this episode.
Download (Duration: 30:47 / 35.2 MB)
In this episode:
00:51 – bells and whistles
02:21 – reducing the damage
04:37 – commitment
07:29 – not very similar
10:42 – no longer applies
11:42 – drawdown versus broken
13:24 – gap decisions
15:55 – juggling
18:33 – diversification
20:25 – a two opposing systems
22:08 – the flip side
23:20 – a question of power
24:06 – a spec of dust
27:30 – float your boat
30:09 – design a system
Tweetables:
How do you know when it is a drawdown? [Click To Tweet].
Systematic trading relies on letting the system run for a long time. [Click To Tweet].
If the system really has an edge then, it should show up in a small sample. [Click To Tweet].
Download The Full Episode 82 Transcript Here
Darren: Those difficult decisions decide whether you become a really good trader or not, I believe. I do not think you can just systematically remove that just by having a really brilliant system…
Announcer: Two Traders, Darren and Walter, pull back the curtain on profitable trading systems, consistent money management, and profitable psychological triggers. Welcome to the Two Traders Podcast.
Walter: Welcome to the Two Traders Podcast. It’s Walter here. Hello, Darren.
Darren: Good evening, Walter.
Walter: We are going to talk about this idea of systems today and trading a lot of systems. I do not know why, I used to think that…When I first started trading, I thought that if you are a trader, you have this whole swaff of systems and that is how you approach things.
You know all these bells and whistles and monitors and everything is just firing off. It is really an exciting thing but I am curious, what do you think in terms of system? Is there other advantages or disadvantages? How do you view it in terms of trading multiple systems or should you just stick to one?
Darren: There’s some idea that I am playing with right at the moment and I know we touched on this the other week. I’ve always gone down that mentality that you should have like a core system and you might trade a few variations.
Generally, you worked on a system, perfect it and that is the system you trade. More and more, I am realizing that having other systems gives you some diversification. With any kind of investing, they say you should not put all your eggs in one basket.
At the moment, in that sort of exploration period, I am trying to get 2 systems to work really neatly together. For instance, when my main system is not making money or losing, the other system is making money and vice versa.
The idea here is I know that there is going to be periods where I just lose money and there is nothing I can do by it. It’s just particular market conditions do not work for my system and I think all systems have that to an extent.
I am just trying to find ways that reduce the damage of those. In the times when my primary systems is working really well, the other one just does not really generate any entry signals. I look at it from the entry point of view.
So, I am saying, “What is the distinguishing character of the market movement when I am not making money?” and “How can I design an entry system that will work then?” I bet if you can come up with some sort of a plan, then it can be something really good.
I haven’t got any results to back this up yet but I have noticed that a lot of traders are being around a long time and have been very successful to trade multiple systems so there is obviously something there.
Walter: Yeah. And, you obviously like the people who were into the automated stuff or even a lot of funds will do the same thing. They’ll do exactly what you say, Darren, which is you have some system. Maybe it works in a trending market, another one works in a directionless, choppy, consolidating market, maybe have a breakout system that works at the open or specific time of the day.
One thing that’s taken me a long time to get used to is this idea that I may have a signal that’s a buy. For example, right now, I have a trade that bought the NZD/USD. Bought NZD and sold the USD and at the same time, looking at the charts, I really want to sell the NZD/USD with another system.
It’s taken me a long time to get to the point where I can say “You know what? That actually make sense.” Just because you have this buy that was put in months ago for the kiwi and it was triggered the other day, like 2 days ago, does not mean that you cannot take a trend trade in the downward directions.
In other words, trade against this because it could be a situation where both trades work out. It could also be that only one works out but it could be… Let’s say, the buy trade on the kiwi goes to target 1 and then the rest of the trade at break even and then it comes back down and stops you at break even. Meanwhile, now you’ve got a sell signal to go with that downtrend so you are continuing on the downtrend. That sort of thing.
That was something that it is difficult if you feel super committed to your trade and you feel like you are doing the right thing. It is hard to actually take the other side and go, “You know what, I am actually going to take another account. I am going to take another trade here that is a little different”.
It is easier, I suppose, with different timeframe or something like that but sometimes, at least I have found out that it is difficult to get to that spot where you can actually do that.
Darren: Yeah. And, obviously, there is that sort of negative thing around looking at other systems, as well, where your system hop in. You need to be careful that that isn’t what is going on. You really need to be looking at a large sample size of data to see when and if you need to trade another system to diversify.
It is very difficult as well because, psychologically, when you are trading a particular approach, you are building yourself to believe in that particular system and, at the same time, take on a system which contradicts those beliefs. It is difficult to do because you say, “Well, this system isn’t perfect.” And then you start questioning, “Should I be trading it if it is not perfect?”
Again, you need to go back to that thing where you accept that no system is going to be perfect. It is unreasonable to accept any system is going to work in all market conditions. You are always going to have periods of drawdown.
There is always going to be variances in those and sometimes they are going to be a lot worse than what you tested. There is a lot of psychological issues with taking on another system.
Walter: Yeah, and performance, as well. Let’s say that you had a trend system and then reversion to the mean system and you are running them both at the same time. Like you say, sometimes you are going to be doing really, really well when the market is trending with your trend system and then maybe not so hot with the reversion to the mean system.
In a sense, you are waving the white flag and saying, “I am not going to have these huge chunks of profit that come into my account, not in my account, because it is going to be blunted by the other system”.
In exchange for that, I am going to smooth out my equity curve in theory. Like, have a shallower retracement but, in the end, what I am doing here is I am sacrificing some of those high peaks by creating shallower valleys, shallower drawdowns.
In theory, this is the way how it works. It is a little bit tricky because sometimes people, if they have a good month, they want to draw a straight line and go, “Okay, I made 12% this month. Let’s multiply that out, I’ll make 144% on the year” or whatever.
When you start adding more system, what you’ll notice happens, especially if they are not very similar at all is, it even things out a bit and sometimes that is difficult to take. I know that for most traders, the drawdown is really the uncle point where you get stopped as a trader.
When you give up or you change or you are throwing the towel, you take a break or you change things drastically just because of the drawdown. That could be good that you are blunting those drawdowns but, at the same time, you are also blunting your overall performance. In theory, that is what’s happening.
Darren: Yeah. And, I think also, you need to accept that sometimes systems will just stop working as well. Your system might be based on a particular market occurrence or movement that is happening but, for some reasons, because of — I do not know, something could change — it is going to stop working.
Maybe you trade a Friday close or something like that, you can find systems that work really well for 5, 10 years in a particular instrument and then will stop working. It gives you a little bit of showroom insurance as well going forward if you’ve got 1 or 2 or even more systems that have a positive expectancy.
It is a little bit naive as well to expect that you are only going to find one system that’s got a positive expectancy, if you understand the core elements that make a system profitable or not. You should really be able to not just make variations of that. You should be able to use some logic and build up a system based on those ideas.
Walter: Absolutely. The way I approach it is a little bit weird. Maybe it’s just that I want to have control but the way I approach it is I look at the charts and I say, “Okay, now, what sort of system should I be using here?” Because, some are going to work better than others.
If I look at a trending chart immediately, I am thinking and certain I’ve eliminated several types of trades because I am looking for a trendy type of trade. If the market is trending, hopefully, it will continue trending.
If it does not continue trending and I used the trending systems then, I am going to lose or at best break even on these trades. If I see a trending market and this is going to keep trending then, my trending system in theory should work quite well.
The same thing with the directionless market. If the market is directionless, then I am looking for those systems that are going to work in that style of chart, really, is the way I see it.
What’s interesting I think about what you just said is this idea that systems stopped working and I’ve always had a really hard time trying to distinguish between a system that is no longer working and a system that is just simply in a drawdown.
I do not really know. Statistically, I guess, there is something that you can do. Relying on parametric statistics and say, “Well, we know that this kind of a drawdown, this sort of a losing streak, is beyond say 3 standard deviations so it is very, very unlikely that this is still a part of this distribution of trades”.
So, we are looking at different distribution now. In other words, the system has changed, or the market has changed, and the system no longer apply as it did historically. You can do things like that but other than that I do not really know.
I mean, to me, it is just the trader decides. The trader says, “Look, this is too much. The drawdown is too great and something is wrong here. I’ve got to stop.” I do not really know what the solution is to that other than trying to use things like standard deviation and sequencing.
Like you’ll go, “With this sort of a win rate, I am likely to see 7 losing trades in a row. 1 out 10,000 times.” Then you see your 8th losing trade in a row and you’ll go, “Hold on here, something is wrong.”
Assuming, of course, that your stats are correct and that you are applying statistics in the correct manner. We’ve talked about this before and trading that may not be the case because you’ve got so much variability and so much volatility in the markets.
What are your thoughts on that? Presumably, if you have multiple systems at certain times, you can shut them off, turn them off, turn them away. So, how do you know when it is a drawdown versus it is broken?
Darren: That is the big nightmare, isn’t it? I believe that systematic trading relies on letting the system run for a long time and because you are going to get variants in your results, you need to let it play out in a long run.
In other words, just because this drawdown is worse than any in your testing period does not mean that the next period is not going to be better than any in your testing period.
There is that variants always and there can be just one particular period for a few months where the system performed really, really badly and that particular pattern of movement and entries may never happen.
You start in this dilemma of do you just buy the bullet and say, “I know that this has a good chance of being profitable over the long run” but you need to let it run over the long run and then just grind it out. Or, do you, like you say, at some point step in and say, “Right, this isn’t just good enough. I am going to have to stop it”.
I think you need to do that a little bit. There is nothing wrong to say, “Look, the system is not making money now. It is worse than my expectation were for the drawdown periods so I am just going to pause it for a bit and I am just going to perhaps trade in demo or just observe for a bit.”
There is nothing wrong with doing that. All trading comes down to those difficult decisions that cannot be predetermined. At some point, you’ve got to weigh up all the pros and con and make a gap decision.
Those happened all the time in entries with taking profit, with the position sizing, with whether you are going to stop trading a system or not. That is the difficult bid and it goes over all of trading.
Those difficult decisions decides whether you become a really good trader or not, I believe. I do not think that you can just systematically remove that just by having a really brilliant system.
Walter: Absolutely. I think you can sort of say, “Look, based on what I know, this system is really highly, highly unlikely. Let’s say, 1 in 10,000 chance of hitting 7 losers in a row”. If that is the case, what that means is, of course, you have a relatively high win rate for your system and then you’ll go “Alright”.
The other thing is, if I am risking x% per trade and this is the way I am going to use position sizing with this system, then knowing that I am unlikely to have 7 or more losers in a row. Then, I am unlikely to have a drawdown of x% — let’s call it, 16% — so, I am likely to have 16% drawdown.
Based on all this that I know, if I get to the point where I have that 7 or 8 losing trades in a row and I hit the 16, 17% drawdown, I could just have an automatic shut off and go, “Okay, that’s it. Move it to demo and see if it pulls out of this.”
Basically, it’s gone too far. The data that I have does not fit because either the market has changed, my system is no longer doing what it used to do historically or, I am not executing it correctly. System-error somewhere, something is going on here that just does not add up.
If you see those things in place beforehand because these are emotional decisions that people make. They want to take revenge and get their money back and all these things when they happen in real time.
It is a little bit easier if you plan it ahead and you’ll say, “Okay, this is what I am going to do. If this happens or this happens, we go to demo, pull the plug blah, blah, blah,” these sort of things.
Darren: Would you say that, in general, retail traders trade multiple systems or do they tend to focus on one? My feeling is that, generally, people will trade one system and then when they add another system, they’ll remove the first one so they’ll just switch systems. What is your experience with that?
Walter: The more experienced traders are better at juggling but, in the beginning, I think a lot of traders are like that. I do not know, it depends. For example, let’s say, you are 57 and you are retired. You and your wife were just trading, you do not have much going on.
Maybe you do have 4 or 5 systems running, maybe that is the way you do it. Let’s say you are 31, you’ve got a job and you work and out the door at 7 in the morning. You come home at 7:30 at night and you really do not want to sit in front of the screen from 8pm to 1 am trading the London market or whatever.
Are you really going to be able to juggle 4 or 5 systems? I suppose it depends on the timeframe. It is possible. I think you nailed it earlier when you were talking about having the perfect system.
When we first start trading, that is what we are seeking. We are seeking the perfect system and then, as we get experienced, we change and that is why diversification works.
Diversification of systems is going to beat diversification of markets. I would challenge anyone listening to this, go out and test this. Test this idea. Do not think — or maybe you should think, “Hey, you do not know what you are talking about, Walter.”
I challenge you to go out and test this. Test different systems on the same market versus the same system on multiple markets. Come back and let me know how it works. Generally speaking, as long as you have systems that are completely — I am not talking about 2 trend following system or 2 breakout system, I am talking about 2 very different systems. That is a better diversification than to go and trade multiple markets.
That is what I’ll encourage people to do if this is something that makes sense to you. If you are comfortable with your system and you want to add another one, try another that is really, really a different one because then you will be taking advantage of what we are talking about.
Darren: Let’s say for instance, we were trading in on 4-hour timeframe and we were using say, support and resistance. We were looking for breakout retest like a trendy support and resistance system. Where would we start?
Walter: Our friend Dennis does that. I’ll put a link to his account so you can see. He is really a good trader. He’s in France and he’s on that signal start web page or whatever. He is quite highly ranked. He was number 1 for a while but that is exactly how he trades. That is the reason why I mentioned this. He trades exactly what you just said.
Darren: Where would you then start to look for a system that gives diversification to that?
Walter: Great question! Let’s say you are trading the 4-hour trendy move where it breaks out, comes back down, touches and bounces off that and you expect it to continue on the direction of the trend. Alright, so what are you going to do that is going to trade different to that?
You could trade a 4-hour system where you are basically trading off of boxes or, maybe you have Bollinger Bands or something like that where you have some sort of envelope. You are waiting for the market to get to the extreme end of the envelope and then you sell it or buy it back to the median, to the middle, to the moving average in the middle.
That would be an example of a system where you can do that. Now, you do not have to have the Bollinger Bands or whatever. You can also just draw a box. You wait for the market to get into a 4-hour box and you wait for it to get to the edge of the box and when it prints a bullish candle at the bottom of the box, you buy.
You buy it to the middle or the other side of the box and if it gets to the top of the box and prints a bearish candle, you sell down to the middle of the other side of the box. That would be a very simple mean-reversion-style of trading that would probably be quite unrelated to a trending 4-hour system.
Darren: In essence, on one system you are buying in an uptrend and on the other, you are looking for price to get over extended and getting those retraces back as well.
Walter: Yeah. So, you are making money on the retraces whereas with the trendy, you are doing the opposite, aren’t you? You are waiting for the retrace to be “over.” According to your signal, the retrace is over and now it is going to keep exploding in that trend direction, whether it is down or up. It is the opposite.
Darren: When you actually break it down and think about it logically, it makes sense, really, that at least two systems that are opposing could work really well together. It is surprising that I haven’t really tried it before.
Walter: Yeah, but, you have traded different ways. I know that for a fact like you used to trade with the trend and stack the trades and occasionally just hit it out in the park.
Darren: Yeah. By switching one to the other, and I never really gotten to juggle together. It is quite difficult. You are expanding the amount of decisions and just the mindset that each system is going to have to deal with.
You are going to have to plan it and be organized and practice it, as well. Certainly, from a logical point of view, you can see that that is really a good way to trade.
Walter: Absolutely. The other thing I just want to mention because I think you probably have a good point because we talking about having enough trades, enough data so that you can have a good idea of what is going on, at least historically.
One thing I’ll mention about this because I don’t think it’s mentioned enough and I know it’s a contrarian to what’s most people will hear, what you’ll read in trading books or whatever. So I think maybe you’ll like this, Darren, because I know you are like me.
I mean, we tend to — maybe you more so than me — whenever I read something like in the business paper or whatever, I think why is that here? Why would they want us to think that?
I am always thinking the other side but what you’ll hear is they’ll say things like, “You need x number of trades to verify that your system is good enough or, you’ll need x number of trades to know if it really has an edge” or whatever.
They’ll throw numbers out there a 100, 30 o whatever and they will say that. But, here is the flip side to that, and this is like you can check this in science. You can go to your local university and ask the statistics professor or experimental design professor.
You can ask them this question which is, if you had two experiments — and we are talking about trading systems here so we are talking about backtesting but, in essence, they are experiments. There are two experiments, one experiment shows a huge effect.
In other words, the placebo versus the pill. The pill made a huge effect but they only had 10 subjects in the experiment. The other one shows a small to moderate effect but they had 10,000 participants in the study. Which one is more impressive?
There is always going to be that argument. “Well, you know we only had 10, maybe we just got lucky and we made a mistake here and even though it is showing in effect, we needed a larger sample”.”
Here is what they do not tell you. What they do not tell you is that when you have 10,000 subjects, you have 10,000 participants, you are almost guaranteed to get the result. It is almost guaranteed because it is a question of power.
Another example would be let’s say, you have a telescope and you are looking out at stars. If you look at through the telescope and you see this big, giant white thing out there and you go, “Whoa! That looks like the moon”. The reason you can see it is because the moon is so close, it is so big. You do not have to have a very good telescope to see it.
In fact, you can see it without a telescope. Here is the thing, what if you had a really, really, strong like the hubble telescope and you are looking far, far away. You see a spec out there and you think, “Wow! That looks like a planet. There might be a planet over there,” but you are not really sure.
Well, the thing is that could just be like a spec of dust or something on the lens. It could be lots of other things. It is really not a planet but we are able to possibly see it because the power, the strength of the telescope is so strong.
When you have a really strong effect, when you have a really strong trading system, when you have a system that really, really works, it’s got really positive expectancy. You are going to see it show up sooner than you would if you have a system that has a sort of moderate edge.
In other words, if you take 20 trades and you are seeing a huge profitable system here, it could be because you just got lucky on those 20 trades. That could be the case but, it could also be because that effect. That edge is so good that it’s shown up and you do not need to take 500 trades to see it.
I am not suggesting that you should only take 10 or 20 trades and then decide to trade the system live, that is not what I am saying. What I am saying is that the flip side to having x number of observations to be sure that something is true, is that the really powerful things will show up quickly. You do not need to have a powerful experiment or a powerful backtest to see it.
I think this is something that people do not think of. They do not realize this is the case. I just want people to keep that in mind too when you are looking at stuff. If you’ve got a good system, it will probably show up rather quickly. You will know right away. You do not have to go through to the whole thousand backtest trades to see it.
Darren: Yeah, I am with you now. For a minute there I was thinking, “Nah! What is he talking about? That is wrong.” But, I get your point. If the system is solid then a hundred trades should highly likely to give you a good result.
If you do a hundred trades and it is really terrible, the chances are it is not really going to get better over the long run.
Walter: Right, but you could find an edge if you keep going. Do you know what I mean?
Darren: Do you want to take 5,000 trades to make 3% profit? No, you won’t highly likely that over a hundred trades that you are making a decent profit. Okay, I see what you mean because that luck element place into it a lot.
I do a lot of backtesting at the start of this year trying to remove some of the discretion in my systems and I only found one period that was really, really terrible. Like 3 or 4 months in 2014, it was terrible.
Generally, I tried 4 or 5 different ways of managing my trades. Different exits, they were all putting in really nice, profitable periods. I agree with you on that. If the system really has an edge, then it will show up in a small sample as well.
Walter: Yeah, it is just power. The power of the backtest and your power goes up as you get more observations, that is all it is. Like you said, if you can keep going and you’ll probably find something.
You go 120 trades and you’ll go like, “Ah! I do not know about this.” If you keep going, if you plug along and do 900 trades well, maybe you will find something there. But, is that really going to be worth it?
Is that really something that you want to do? Like you say, is that worth the 3%? You are going to take 900 trades to make 3%, is that really going to float your boat? Alright, Darren, basically what we were saying here is it probably does make sense to diversify your systems. That would be my suggestion.
That is something that can possibly help. You’re probably not going to have the excellent months, or weeks, or days, or whatever that you would if you would just trade one but it reminds me of like, I have this system and I know it is a losing system.
People do not have to ride in to tell me, “Hey, Walter, that is a losing system.” Because, whenever I go to a casino or whatever, there is a system you can play on the roulette table and maybe I’ll put a link to it on the shownotes.
Basically, what you are doing is let’s say you are betting on black. About half the numbers you’ll win and then you put another bet on a third of the numbers and that third actually has mostly red numbers.
If you hit on black, it is a one to one payoff. You put 50 bucks down and a black number comes up, you get your 50 bucks back but then you put another bet. Let’s say you put 20 bucks on that third, now what happens on that third is most of those numbers are red.
Some of them are black but most of those numbers on that third are red. You actually will get paid if about half of the numbers come up which are the black numbers and then few of the red numbers too if those come up.
I think there’s 13 numbers on that strip and maybe 8 or 9 of them are red so if those come up then, you’d get paid 2 or 3 or whatever it is.
Darren: Three to one.
Walter: Yeah, 3 to 1. So, you get 60 bucks. The best case scenario would be that it lands on black number so you get 50 bucks. It’ll also happen to be one of the three black numbers in your other bet so then, you get 60 paid out for the 20 bet. You are actually making 110 but that rarely happens.
Instead, what happens is you get the black a lot and then, a few times, you’ll hit few of the red. You are just really able to keep playing. You are going to lose over the long run. There is no edge but what it does enable you to do is it keeps your bank rolling.
That is the same thing that happens when you have diversification of systems, isn’t it? It allows you to keep going even when you are going through a rough patch because one of those is likely to make a little bit of money for you whether it is the break-out system, or the trend following, or whatever.
I think it makes a lot of sense to diversify and as you get good at these stuff, you realize that there’s certain market where your system is not going to work well in. Now, let’s design a system that will work well in that style of market. That is really the key, I guess.
Darren: Exactly. I am focusing my efforts at the moment trying to work out a range-trading system to go with my break-out system. That looks pretty good so far.
Walter: Awesome! Sounds cool. Alright, thanks for your time, Darren, and we will see you next time.
Darren: Okay. Cheers, Walter!
EPISODE SHOWNOTES
Thomas
“…
I am always thinking the other side but what you’ll hear is they’ll say things like, “You need x number of trades to verify that your system is good enough or, you’ll need x number of trades to know if it really has an edge” or whatever.
They’ll throw numbers out there a 100, 30 o whatever and they will say that. But, here is the flip side to that, and this is like you can check this in science. You can go to your local university and ask the statistics professor or experimental design professor.
You can ask them this question which is, if you had two experiments — and we are talking about trading systems here so we are talking about backtesting but, in essence, they are experiments. There are two experiments, one experiment shows a huge effect.
In other words, the placebo versus the pill. The pill made a huge effect but they only had 10 subjects in the experiment. The other one shows a small to moderate effect but they had 10,000 participants in the study. Which one is more impressive?
…”
In the context of backtesting the problem is data snooping. Suppose a 10-person study where the researcher meets each participant and afterwards makes the hypothesis drinking coffee increases the chance of acne. The researcher then proceeds to use the same 10 people to support the hypothesis. The hypothesis may still very well be true but there’s no way that paper would get published if the researcher was honest about their methods. This is the typical problem with backtests, you make observations on the history, come up with a system, tweak parameters, iterate and then judge the system based on its performance on that same dataset. In this context more data is important because it’s not really an experiment it’s still just an observation.
Saying “you need x number of trades” is usually tied to a measure of complexity of the system. More data (or trades) allows for more complexity. Low complexity, high trade count and great performance provides confidence in the system when all the data has been burned. It’s analogous to machine learning, various ML methods have a complexity parameter. If you have lots of data, then increased complexity can lead to better more accurate models. Not enough data and too much complexity leads to overfitting. With trading systems it’s very easy to have good looking overfit backtests. Simplicity combined with a decent trade count reduces the chance the results are a result of curve fitting.
Now outside the context of backtesting and in the context of paper trading or out of sample tests on completely unobserved data. Data snooping is no longer an issue and what you’re saying is more accurate but that’s not the typical context of the “you need x number of trades” statements. The fact of the matter is I’d imagine there’s far more people getting burned by overfit systems then people not trading some amazing system because they don’t think it’s results are statistically significant.