Transcription
This $42 book made me an accidental profitable trader. Now, I'm not trying to sell you this book, but I need to thank a huge part of my profits to it. It cost me $42 and it has made me hundreds of thousands. And to be honest, I've never given it the credit it deserves.
Now, what is the actual book I'm talking about? Well, the book is Evidence-Based Technical Analysis by David Arson. I stumbled across it from some online mentions when I was still back using discretionary approaches. Now, if you're not too sure what discretionary means, ultimately that was me doing, you know, support, resistance, feeling the market. Uh, and when I finished reading this book, it completely opened my eyes to one key thing, and that is evidence-based trading. And that's the main topic for this particular video.
To give you a bit of a backstory on me, I'm in, and I've been trading for about 5 years and a half. And primarily now, I am a fully automated, so algorithmic trader, uh, as well as, you know, using systematic trading methods. And a lot of this video is going to go behind the scenes of what actually systematic and algorithmic trading is. And ultimately, it goes back to being evidence-based and not relying on your intuition or gut feelings.
Now, for me to really get into this presentation, let me just make my camera camera a bit smaller here. Uh, here we go. So, you can kind of see the full screen. Let me zoom in. Uh, so you can see everything.
Well, before this book, I was trading like most people. I was drawing lines on a chart, following expert opinions. And by experts, I normally mean like gurus, you know, kind of selling entry and exit signals. We all go through it. It's the first year or the first couple of months into trading. We all stumble across those options and you know pay whatever money to think that someone else is going to be able to make money for us. And ultimately I was trusting my gut when it felt right. And looking back on this realistically this was just gambling. It wasn't really trading. I had no probability. I had no edge. I was simply following people and following my intuition. And when you think about it, it's a very silly thing because back then, right? I was trading for maybe a couple of months. What experience do I have trading markets? What intuition have I, you know, garnered over those years of trading? I've only been trading for three months or something. So, there was no reason that I should have been doing this approach at all.
Now, to save you from reading the full 500 plus pages as well as the cheeky two $42, but I do recommend you pick up this book if you are interested. There's a lot more that is behind this book than just this video, but I'm going to give you the four key takeaways that might transform your trading as it did for me.
Now, number one is subjective analysis. So, subjective analysis is anything that can't be defined and tested. So a really easy example of this is you'll see a lot of traders go, uh, you know, the order flow right now feels very bullish, but when they actually try and define what that means, they don't really have anything behind it. Uh, and this leaves us with meaningless claims that give you the illusion of conveying cognitive content. And this is a big factor of what the book talks about. It talks about the limitations of subjective analysis and the reason that a lot of it is just biases coming into play that then give us the illusion that we actually have some sort of reliable edge here.
Now instead, now I focus on objective analysis where I can confirm or deny a trading idea with real data and that's the key point of really anything around systematic or algorithmic trading. It's no more this chart looks bullish or I feel the market is about to break out. Instead, it is about saying, you know, there's a 70% chance based on five years of data that prices will be higher the day after touching the 200 EMA or MA. So, this is a very concrete proof thing and something that can be tested. For this example, this was just something I was pulling out of thin air, but you could obviously look up the data behind this and actually see what that percentage actually is because there's data you can pull upon and check and that there is no subjective analysis going on here. This was ultimately the most instrumental thing in reshaping the way I look at trading.
Now, number two, as I kind of touched on a bit before, was recognizing biases. And I think this is another really key part of his book, and he does an amazing job of breaking it down. And it is quite eye-opening when you go over the various biases that can affect your trading every single day.
Now, to give you a couple of my favorite ones, or I think the most important you should really know is number one, selection bias. So, selection bias is cherrypicking trades that support your beliefs. This is very common when people do manual backtesting. When people look over historical charts and say, "Hey, you know, let me just draw a chart out here." Uh, and they go, "Hey, if I had entered on the breakout of this, then this would have been a great trade. I would have made x money, x amount of R." That's great, but that's a single trade. Unless you have hundreds or thousands of trades that have been completely operated in the exact same way, you're going to be ultimately falling into this bias, which is selection bias.
The next bias is look ahead bias. So, this is accidentally using information that wasn't available at the time of the trade. For example, um, this kind of goes into this example as well here of this particular trade. Obviously, you didn't know if this trade was going to work out. And I have seen traders by examining them, um, completely ignore trades that didn't work and they just wouldn't put it in their backtest if they were doing it manually. This is an obvious example of look ahead bias. Another one actually that's a bit more sneaky is let's say we're doing a long only strategy on Nvidia and we have already looked at the Nvidia chart and we've seen, hey, every time the Nvidia chart touches the 200, uh, MA, it instantly bounces and makes a ton of money. That's technically a look ahead bias because you wouldn't have ever known, um, that Nvidia would have continued that strong of a momentum over that period of time. A more realistic test there would be, hey, let's test that exact same strategy so when price touches the 200 MA, but on, you know, 100 tickers or something like that. That would have less of a look ahead bias because you haven't looked through all of those 100 tickers to actually know if they are all, you know, long only performing trades.
The next is an optimism bias. So this is overestimating our abilities and the likelihood of favorable outcomes. Uh, I'll pull upon a personal story here. Um, as I mentioned before, you know, in my first three months or six months of trading, I thought, of course, I can use intuition to make trades and make smart decisions. But this is hugely optimism bias. I haven't had years of trading experience. I hadn't seen multiple market cycles. I hadn't take hundreds of trades. Um, but I was still, you know, overly confident thinking, hey, I could make money trading like this. Um, and just using my pure gut feelings and no proof, no nothing. Um, and this kind of falls back into this particular bias. I was guilty obviously of using all of these in my first year and I think most people are. I think it would be quite surprising if you already know these and if you're in your first year and you're watching this video and getting ahead of the curve by understanding these biases, then it will probably help you a lot more because, uh, if I knew these in my first year, uh, probably I could have saved some money in my first year for sure.
Now here are actually all of them from the book. He does a great breakdown, uh, of kind of each one and you can see how it goes into the illusion of validity. Meaning that this is around subjective analysis or technical analysis and you have all these biases coming into play when you're just using pure intuition/your own human eyes with no particular criteria behind it. Uh, these are all the biases you can fall victim to.
The next one is understanding sample size. Now when testing trading strategies, statistical significance is everything. I actually wrote a thread on Twitter a while back around how the number one thing you need to understand as a trader is statistical significance. It really is that big of a thing and I know many people dismiss it, but let's just break down an example. A strategy with 10 winning trades isn't enough evidence. There's no way you could ever say that 10 trades is enough evidence for you to have statistical significance. If you're not too sure what that word means, don't worry. I'll break it down in just a second. You need a much larger sample size to prove anything. If you think about this in any other way, right? If you think about a poll, for example, so if people voting for, you know, three different, uh, decisions on a particular poll, if you only had 10 people respond to that poll, ultimately it doesn't give you that much proof of anything. It's only 10 people. But if you had a thousand people respond to that poll or 100,000 people responding to that poll, that is a lot more clear proof to show what the, you know, average person is voting for on that particular poll.
To also imagine this actually in a practical example, let's look over a coin toss. Imagine you flipped a fair coin. So it's heads and tails three times and you get heads every single time. If we are using the same logic that many traders do, you might assume that the coin always lands on heads because, hey, it's landed on heads three times. This is, you know, my kind of sample. That means the next one should land on heads as well. But that just isn't the case. And it's simply because of a small sample bias. The truth only emerges after hundreds or thousands of flips where the probability converges to 50%. And we have the benefit of most people will know that obviously a coin toss is 50/50. But if you had never heard of a coin toss before, you had never heard of any statistics behind it, you could have fooled victim to the same bias here that many traders are falling victim to with their own strategies. Because obviously you don't know what your expected value or your expected win rate should be for a strategy when you've only traded it 10 times or you've never done a solid backtest on thousands of trades. This exact thing applies to trading and a handful of successful trades prove nothing overall. You really need a large sample for determining a real edge.
Now, number four is rigorous testing. Without rock solid testing, you'll never know which strategy ideas will succeed in live markets from the 90% that fail. I had a question from a trader recently and he asked me, you know, how many ideas do I normally test before I had a strategy that actually works on live markets? And to be truthful, I've tested hundreds of strategies that have completely failed, but I could have been more calculated. There are markets that are more inefficient where you can actually do very simple ideas and still make a solid return, but you are taking on other risks. An easy example of this is crypto. You're taking on brokerage risk. You're taking on other environmental risks when you're actually deploying capital on crypto. And that's why there can be easier edges there. Now, I will say crypto has obviously improved compared to, you know, like 2019, 2018 where there was a lot more frequent inefficiencies that were easy to take advantage of. But still, at least to most of the popular markets, crypto is a one where you can find a lot of inefficiencies.
Now, in the book, Austin introduces you into some different testing techniques. Uh, the main ones I'll say is out of sample testing. So, this is where you're separating your testing data from your strategy development data. If you're not too sure what that actually means, let me kind of draw you a quick visual. So, let's say we have, um, this much data. We're basically splitting that data in half. And half of it is going to be for optimizing and half of it is going to be for actually testing. Um, what I mean by optimizing is this half over here, we could test different parameter values. So let's say we had a very simple crossover strategy. We could test the 50. Uh, we could test the 60, etc. We can optimize as much as we want. Really doesn't matter overall because this is our in-sample data. Then when we have found the most profitable, let's say, um, EMA value here on out of sample, on, sorry, on in-sample data. So that's the data you're optimizing over. We can then test that parameter value on out of sample data. And then we can see if the performance still holds up. This is a very simple example. There's already, you know, some flaws in this, but hopefully that kind of demonstrates the idea of in-sample and out of sample data.
Now, one key thing here is let's say we chose 60 and that was the best performing, um, parameter value for our in-sample data and when we tested it on out of sample data, it performed horribly. So, it didn't hold up at all. If you then go back to your in-sample data and choose a different value and then apply it to the same out of sample data, it's no longer out of sample data. So those are things you do have to be careful of because you can't be just repeatedly testing on out of sample data as that's ultimately then the same thing you're doing with your in-sample data.
Now the second technique is randomization techniques. So this is like Monte Carlo simulations. If you don't know what that is, I actually have a video on the channel about Monte Carlos. It breaks it down really simply for you. So if you want to check that out after this video, feel free to. The next is also about properly penalizing strategy metrics to account for data mining bias. This is probably a bit too big of a topic to do in this particular video, but data mining is primarily where you're just overoptimizing on a ton of data and, uh, it causes data mining biases.
Now, all of this testing is the difference between watching most of your strategies die after a month or two versus seeing them maintain returns on live markets. I constantly get messages from traders saying, "Hey, you know, I've backtested a lot and then I deployed on live markets and it's completely underperformed for two months." That's quite a typical thing if you're not going through a rigorous testing sequence. And even when you are going through rigorous testing, your probability is not going to be 100% that every strategy is going to work. Personally for me, even with like all the knowledge I know about properly testing strategies, if I deploy a strategy after testing it on backtesting, I'd still say it's an uncertain if I know it if it will 100% work. But I do know over, you know, 10, 50 strategies, etc. that I'm deploying, a high degree of them will work because of my rigorous testing and also a lot of the common sense and lessons I've known about developing strategies.
Now, here's the crazy part about all of this is I was never intentionally meant to become this type of trader. When I started trading, I really did think that I was going to be, you know, an intuition-based trader. I thought that I was going to be one of those people that can read off, you know, levels on a chart and then, you know, make money every day like that. It was overall completely accidental that I actually moved to this more systematic algorithmic trading. At the end of the day, I just wanted to improve my trading and that was it. You know, we all just want to make money in the markets and that's the main goal. Instead, I found an entirely new approach that actually worked consistently for me and aligned with my strengths.
Now, I will say I've been coding since I was around 12, so I have always been in the realm of problem solving and statistics, etc. And that obviously gave me a big leg up when using this particular approach of systematic and algorithmic trading. But I also would strongly argue that most traders will have a better time in sense of returns being systematic at the beginning of their trading career compared to trying to use intuition and experience because ultimately at the beginning of your career, you have no experience. It's a very simple matter of the fact and if you can actually find a robust edge that you can then deploy and then over the years you can also integrate your experience. Some of the greatest traders I know are hybrid traders where they use data and they also use their experience because they've been trading for 5 to 10 years and they've seen multiple and multiple market cycles.
Now, I definitely can't claim that this book is flawless. Um, but it did open my eyes to new ways to tackle trading and I think that's what's most important and for that I'm forever grateful.
Now, if you want a shortcut to some of this, uh, and you want a more streamlined version, plus tons of code examples you can actually use to backtest as well as even automate your trading through common brokerages, you can check out the backtesting boot camp. I'll put it as the first link in the description. By the way, you have no coding experience is needed at all and it's helped a lot of traders. We have had over 100 traders go through it now. Uh, this was just one trader, Charles, uh, that did amazingly after taking the boot camp. But otherwise, feel free to check out the other videos on this channel. All of them are free and I'll try and give as much value as possible.