We'll help you map out a plan to fix the problems in your trading and get you to the next level. Answer a few questions on our application and then choose a time that works for you.
BOOK YOUR FREE STRATEGY SESSION NOW >>Choosing which parameters to use when you put a strategy live is never a simple task.
As you know, we usually use the sets of inputs that produce the best results in testing. The problem is that those results are from the past, so we have no certainty that the same inputs will work just as well in the future.
In today's video, you’ll learn about a method of selecting inputs for live trading that will help you minimize the impact of probability on the live performances of your systems.
By watching the video you'll discover:
- Why it's important to minimize the impact of probability on the live performance of your systems
- How to allocate capital so that you can handle different scenarios
- How to combine different timeframes and input sets to further reduce the impact of probability on your portfolio
Enjoy the video!
Hey everybody and welcome back! I'm one of the coaches at Unger Academy and today we’re going to be talking about the selection of inputs and in particular how to select the set of parameters of a system to go live. I want to do this by taking inspiration from the real experience of a member of our group, the One Year Target, that has been using this methodology for some time with great success and has achieved great results.
As mentioned before, we’re going to be talking about the selection of the inputs of a system and I want to do it by starting from the experience of Giuseppe, one of our traders and one of the members of the One Year Target group who has been successfully using this methodology for quite some time.
What do we see on these charts? Well we see the VXX, an ETN derived from volatility, which means that it that derives its price from the trend of the Vix. On these different charts we can see the same system. It is a system that has already been used for a long time, so it has been validated and so we can be fairly sure about its robustness.
However, before we go live with a system we need to decide what set of inputs to use in live trading. I decided to go back in time to 2019. So, we have two years of out-of-sample (which we don't see on these charts above) to decide which set of parameters to use as if we wanted to go live.
The system produces more than satisfactory equity lines. Every set of inputs maximizes a specific result.
We have an input set that maximizes, for example, the return on the drawdown. Let's try and see what results we get. Yeah, definitely a very good equity line!
We could choose an input set that maximizes the net profit, for example. Maybe it isn't the best objective to use. Let's have a look. Also in this case, I imagine to see a very interesting equity line.
Or still within the same metric, we may find a set of parameters that maybe isn't the first but gives me a much higher net profit. You see, this second set of parameters has a lower return on drawdown but produces results that are certainly more interesting from an economic viewpoint, so it could be worth the while to use this set of parameters. Let's look at the equity line.
So, this is to say that it is not so easy to select which inputs to use in live trading, and often we have to make a choice. A choice that inevitably has a component of probability. Let me explain myself better.
Are we so sure that the set of parameters that we've just chosen and that has given excellent results so far, is the set of parameters that will give us the best and most stable results in the future, let's say in the next two years? How many probabilities are there that we'll actually be able to identify the best set of parameters to use in the future given all these input permutations?
This is exactly what I mean when I talk about probability, namely what has been the best set of parameters up to now will most likely not be the best set of parameters for the future, because randomness just plays an important role in this case. So, given a stable set of parameters to work with, finding the best combination for the future is pure luck.
So how can we minimize the impact of probability on this type of procedure? When it's possible, and here I refer to what Giuseppe actually does during live trading, a possible solution can be to use more than one set of parameters by investing the capital on different versions of the same system in order not to bet on a single horse that we hope will be the winner.
So, Giuseppe uses this system in real time and he uses it not only on one timeframe but on different timeframes: 15, 30, 60, 90 minutes. And for each timeframe he uses 4 different sets of inputs, still picking the inputs from an area of stability and selecting sets of parameters that obviously must be different from each other. This is to minimize the impact of probability on the future trend of the system.
So, I created a portfolio where the same system works in two different versions on 60 minutes and two other different versions on 30 minutes. So they use completely different sets of parameters. This is to see the out-of-sample performance that this portfolio could make.
So, let's run the backtest. The capital per single system is one million dollars only because due to the various reverse splits that have been done on this instrument it wouldn't otherwise be possible to backtest the system in the past.
So, let's leave the economic values aside for the moment and focus on the trend of the equity line. Obviously, you can trade in real time with a few thousand dollars, therefore with a capital that's truly within everyone's reach.
Returning instead to our equity line, you’ll see that in out-of- sample the portfolio of systems has continued to perform quite well. There's a certain regularity in the out-of-sample performance even if with a higher volatility. You can see that the equity line is probably more irregular than in the past, but this is normal because we are comparing an out-of-sample period to an in-sample period.
Having said that, it's interesting to see how the individual versions of our system performed. The first version in out-of-sample has done certainly very well. You see that the equity line continues to grow with regularity.
So we would have been lucky to pick that version. Now let's go and look at the second version. Well, here you see that the second version isn't as good. It's definitely positive but not as good as the first one.
Let's take a look at the third version. The third version isn't bad. It has continued to perform well even though in the last period we can see that it has been sideways.
And, finally, here's the fourth version of the system, which is probably the worst of all. But why? Well, because in the last period, namely in the last year or so, it hasn't produced any profit, okay?
Let's try to take a closer look at it. Here, this is the out-of-sample trend. But perhaps by next year it could go back up and give a greater economic result than previous versions.
Let's now look at the overall portfolio. Let's zoom in on the out-of-sample portion. Here it is. You can see that, overall, the systems portfolio has done quite well in out-of-sample.
It wasn't the best solution but it would have been pure luck to be able to select the set of parameters that gave the best out-of-sample results. So, this allows us to minimize the impact of probability in the selection of the parameters to trade our system live.
So, to put it in a nutshell: given a reliable, robust system, with a stable set of parameters over time, the best set of parameters to use in live trading is a mix of different sets of parameters, because this allows us to minimize the impact of probability in the selection of inputs.
So, in my opinion, it's very smart to do something like this. Obviously where we're allowed to do it. For example, I'm referring to the world of cryptocurrencies, where there's a lot of granularity, to the world of Forex, to stocks, as in this case. Applying this to futures is honestly a bit more complicated, but we know that today there are micro contracts and therefore the same version, for example, of a system on the Mini S&P could be traded in three different versions using micro contracts.
So, I realize that there are pros and cons and this methodology certainly requires more work in terms of monitoring the progress of the system, and it consumes more resources in the VPS, but it's the only way to minimize the impact of probability in the selection of the parameters to use in live trading.
Ok everyone, I really hope you found this topic interesting and I'm really curious to know if any of you are already using this approach in your live trading. So let us know, by writing in the comments below and we will answer everyone!
And if you are interested in our approach to trading and want to start investing in the markets in a systematic way, then I recommend you click on the link below. It will take you to a page where you'll find very useful resources including a presentation by Andrea Unger, our founder and the only four-time world champion of trading with real money who uses the Unger Method, which is a systematic method to invest in the markets. And hey, also much more.
And, finally, before saying goodbye, I invite you to leave us a “Like" if you liked the video, to subscribe to our channel, if you haven't already done so, and click on the notification bell to stay updated on the release of all our new videos coming out.
And with that, I will see you on our coming videos! Until then, bye-bye!
We'll help you map out a plan to fix the problems in your trading and get you to the next level. Answer a few questions on our application and then choose a time that works for you.
BOOK YOUR FREE STRATEGY SESSION NOW >>Hi, I'm Francesco Placci, a professional trader since 2005 thanks to the systematic approach to the markets.
My skills range from trading on index futures to bonds, from stocks to commodities, with a particular focus on volatility and options, which I consider to be among the most versatile and fascinating instruments available to traders.
After an experience with leading Italian credit institutions where I learned the basics of institutional finance, I became a successful independent trader, with great personal satisfaction.
Founder of Algoritmica.pro, in 2019 I joined Unger Academy as head of Research and Development.
